early phase clinical trials greg pond ph.d., p.stat. escarpment cancer research institute department...
TRANSCRIPT
Early Phase Clinical Trials
Greg PondPhD PStat
Escarpment Cancer Research InstituteDepartment of OncologyDepartment of Clinical Epidemiology and Biostatistics
6 November 2014
Learning Objectives
1 Understand clinical questions of interest
2 Understand statistical questions of interest
3 Understand different phases (12others)
4 Able to discuss designs for early phase clinical trials
My Background (biases)
bullDepartment of Oncology (associate appointment with Clin Epi amp Biostats)
bull13+ years in cancer clinical trials
bullMinimal involvement with other diseases
bullInteraction (questions) are good
Your Background
bullDoes anyone have experience designing an early phase trial
bullHow many have clinical trials experience What sort
bullWho is clinician statistician researcher other
Thought Experiment
bullYou (or collaborator) has new drug which showed astonishing activity in animals lab experiments Ultimately you want to apply for regulatory approval
bullYou need to convince financier to fund you to get to regulatory approval What processes will you use
Standard Drug Development Paradigm
bullPhase I ndash safety Find safe dose
bullPhase II ndash preliminary efficacy Is it active
bullPhase III ndash efficacy Is it superior to standard
bullPhase IV ndash effectiveness Is it beneficial in the real-world
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Learning Objectives
1 Understand clinical questions of interest
2 Understand statistical questions of interest
3 Understand different phases (12others)
4 Able to discuss designs for early phase clinical trials
My Background (biases)
bullDepartment of Oncology (associate appointment with Clin Epi amp Biostats)
bull13+ years in cancer clinical trials
bullMinimal involvement with other diseases
bullInteraction (questions) are good
Your Background
bullDoes anyone have experience designing an early phase trial
bullHow many have clinical trials experience What sort
bullWho is clinician statistician researcher other
Thought Experiment
bullYou (or collaborator) has new drug which showed astonishing activity in animals lab experiments Ultimately you want to apply for regulatory approval
bullYou need to convince financier to fund you to get to regulatory approval What processes will you use
Standard Drug Development Paradigm
bullPhase I ndash safety Find safe dose
bullPhase II ndash preliminary efficacy Is it active
bullPhase III ndash efficacy Is it superior to standard
bullPhase IV ndash effectiveness Is it beneficial in the real-world
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
My Background (biases)
bullDepartment of Oncology (associate appointment with Clin Epi amp Biostats)
bull13+ years in cancer clinical trials
bullMinimal involvement with other diseases
bullInteraction (questions) are good
Your Background
bullDoes anyone have experience designing an early phase trial
bullHow many have clinical trials experience What sort
bullWho is clinician statistician researcher other
Thought Experiment
bullYou (or collaborator) has new drug which showed astonishing activity in animals lab experiments Ultimately you want to apply for regulatory approval
bullYou need to convince financier to fund you to get to regulatory approval What processes will you use
Standard Drug Development Paradigm
bullPhase I ndash safety Find safe dose
bullPhase II ndash preliminary efficacy Is it active
bullPhase III ndash efficacy Is it superior to standard
bullPhase IV ndash effectiveness Is it beneficial in the real-world
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Your Background
bullDoes anyone have experience designing an early phase trial
bullHow many have clinical trials experience What sort
bullWho is clinician statistician researcher other
Thought Experiment
bullYou (or collaborator) has new drug which showed astonishing activity in animals lab experiments Ultimately you want to apply for regulatory approval
bullYou need to convince financier to fund you to get to regulatory approval What processes will you use
Standard Drug Development Paradigm
bullPhase I ndash safety Find safe dose
bullPhase II ndash preliminary efficacy Is it active
bullPhase III ndash efficacy Is it superior to standard
bullPhase IV ndash effectiveness Is it beneficial in the real-world
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Thought Experiment
bullYou (or collaborator) has new drug which showed astonishing activity in animals lab experiments Ultimately you want to apply for regulatory approval
bullYou need to convince financier to fund you to get to regulatory approval What processes will you use
Standard Drug Development Paradigm
bullPhase I ndash safety Find safe dose
bullPhase II ndash preliminary efficacy Is it active
bullPhase III ndash efficacy Is it superior to standard
bullPhase IV ndash effectiveness Is it beneficial in the real-world
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Standard Drug Development Paradigm
bullPhase I ndash safety Find safe dose
bullPhase II ndash preliminary efficacy Is it active
bullPhase III ndash efficacy Is it superior to standard
bullPhase IV ndash effectiveness Is it beneficial in the real-world
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase I
bullFDA Definition Phase 1 includes the initial introduction of an investigational new drug into humans These studies are closely monitored and may be conducted in patients but are usually conducted in healthy volunteer subjects These studies are designed to determine the metabolic and pharmacologic actions of the drug in humans the side effects associated with increasing doses and if possible to gain early evidence on effectiveness During Phase 1 sufficient information about the drugs pharmacokinetics and pharmacological effects should be obtained to permit the design of well-controlled scientifically valid Phase 2 studies
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II
bullFDA Definition Phase 2 includes the early controlled clinical studies
conducted to obtain some preliminary data on the effectiveness of the drug for a particular indication or indications in patients with the disease or condition This phase of testing also helps determine the common short-term side effects and risks associated with the drug Phase 2 studies are typically well-controlled closely monitored and conducted in a relatively small number of patients usually involving several hundred people
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Clinical Trial Phases
bullNIH Definition PHASE I TRIALS Initial studies to determine the metabolism and pharmacologic actions of drugs in humans the side effects associated with increasing doses and to gain early evidence of effectiveness may include healthy participants andor patients
PHASE II TRIALS Controlled clinical studies conducted to evaluate the effectiveness of the drug for a particular indication or indications in patients with the disease or condition under study and to determine the common short-term side effects and risks
PHASE III TRIALS Expanded controlled and uncontrolled trials after preliminary evidence suggesting effectiveness of the drug has been obtained and are intended to gather additional information to evaluate the overall benefit-risk relationship of the drug and provide and adequate basis for physician labeling
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Clinical Trial 1bull What question should be answered in first clinical trial
bull Is it safe to administer Is there any biological activity How much should we give
bull Give to small number of humans in a controlled environment (why) and check (side) effects
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
TGN1412 for RAbull Injected to six volunteers at sub-clinical dose on 13 March 2006 10 min apart in London UK
bull All lsquoimmediatelyrsquo hospitalised six with multi-organ failure due to lsquocytokine stormrsquo 1 ballooned up like lsquothe elephant manrsquo
bull Volunteers given ₤2000 All survived but may have long-term immune system issues
bull The company TeGenero went bankrupt within the year
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Dose Levelsbull DL1 usually based on animal data (eg 110th LD50 in
mice)
bull Escalation of doses uses Fibonacci sequence (x2 x167 x15 x133 x133 hellip)
bull Oral medications based on availability of doses (ie if only come in 50 mg pills)
Penel N and Kramar A BMC Med Res Meth 12103 2012
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase I Cancer (3+3)bull Give to 3 patients
bull If 03 have dose limiting toxicity - escalate
bull If 13 has DLT expand to 3 more patients
bull If 16 has DLT ndash escalate
bull If ge23 or ge26 ndash reached maximum tolerated dose (MTD)
bull DLT serious or life-threatening AE occurring in first cycle ndash not standard
Hansen Graham Pond Siu Phase I Trial Design Is 3+3 the Best Cancer Control 21(3) 2014
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
DLTsbull Serious or life-threatening AE occurring in first
cycle ndash not standard
bull Grade of AE (1=mild 2 3 4 5=death) Grade 3+
bull Leukemia patients almost all have grade 4 hematological AE
bull Attributable to drug
bull Chronic grade 2 AE (fatigue nausea etc)
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase I Cancer (3+3)bull Recommended phase II dose is the dose below
MTD
bull Europe MTD=RP2D
bull DEFINE
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase I Non-cancerbull Give drug to paid volunteers
bull How much would you need to be paid to take a drug with a 33 chance of giving you a seriouslife-threatening AE
bull What is compensation if 33 of volunteers get SAE
bull Luckily most treatments expected to have fewnone AE
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase I Non-cancerbull How do you measure safetyoutcomes
bull Measure the effects on body (pharmacokinetics pharmacodynamics)
bull Ensure it is within acceptable limits
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase I Non-cancerbull Statistical tests on differences in variability
bull Differences in mean less important (why)
bull How do you define sample size
bull Often want to ensure biodistribution
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Cancer Phase I Ethicsbull Often start lsquotoo lowrsquo for safety reasons
bull Suboptimal dose reduces chance of response
bull Phase I response rates quoted as ~5
bull Patients do not enter these trials altruistically
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Ethicsbull Should you allow intrapatient dose escalation
bull Chance of response for patient
bull How to evaluate late AE
bull How likely is it for a patient with worsening disease to stay on treatment for gt2 cycles
bull Generally No
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Ethicsbull Average patient survival may be a few months
bull What if drug requires weekly IV injections
bull How to assess long-term toxicities
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Correct RP2Dbull Too low reduced chance of efficacy in later
trials
bull Too high excess toxicity dose reductions
bull Based on 3+3=6 patients
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
httpjoneslabchemwsuedu
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Modified Designsbull Rapid early escalation and intra-patient
escalation permitted
bull 1 pt dose level until grade 2 AE then 3+3
bull Reduce of patients in trial (slightly) at suboptimal dose More likely to give patients too toxic dose
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Model-Based Designsbull lsquoRecentrsquo innovation
bull Specify dose-toxicity relationship
bull Update based on data
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Example dose-toxicity curve
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Othersbull EWOC ndash curve based on probability of
observing a SAE
bull TITE-CRM ndash time to event CRM
bull Mixed effects proportional odds model ndash odds of severe toxicity per cycle
bull Some gain in assessing accurate dose May use more patients (pre-define limit on sample size)
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Hansen AR1 Graham DM Pond GR Siu LL Phase 1 trial design is 3 + 3 the best Cancer Control 2014 Jul21(3)200-8
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Secondary Outcomesbull Small numbers of patients
bull Want correlative (pKpDbiological effects)
bull Patients receive different doses
bull Many correlative outcomes are binaryordinal
bull Consider costbenefit
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Secondary Outcomesbull Does biology change pre- to post-treatment
bull Is patient healthy enough post-treatment
bull Ethics of paired biopsies
bull What is accuracy of assay
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Howeverhellipbull Some argue no reason to pursue development
of new drug if no marker of target
bull Molecularly targeted agents
bull Drug A targets marker X highly prevalent in cancer
bull If patient does not express X then drug useless
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Targeted Agentsbull If 80 of patients with X respond but only
10 of patients have X and 0 of patients without X respond SOC is 5 response
bull Phase III RCT needs gt2250 patients without accurate assay
bull Phase III RCT needs 16 lsquoenrichedrsquo patients with accurate assay
bull Regulatory bodies asking for proof of mechanisms of action
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase IIbull Is there any evidence of activity
bull Might this agent be potentially useful
bull Do not want to test if drug is lsquobetterrsquo than standard treatment (why not)
bull Proof of principle
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Clinical phase transition probabilities for investigational oncology compounds for the 20 largest firms by pharmaceutical sales (2005) by period during which compound first entered clinical testing
DiMasi J A and Grabowski H G JCO 200725209-216 copy2007 by American Society of Clinical Oncology
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II non-cancerbull Compare efficacy ndash often of a surrogate
outcome
bull Small randomised trial
bull Does blood pressure go down ndash surrogate for MI
bull Do MS patients improve activity levels walk ndash long-term abilities
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Placebo
5 mg
20 mg
10 mg
50 mg
R
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II non-cancerbull Size ndash few hundred
bull Multiple comparison adjustments
bull α=005 β=080
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II non-cancerbull Select optimal dose regimen
bull Highest dose may be unsafe not tolerable
bull Ideally want dose-response relationship
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II non-cancerbull Trial design standard statistical analyses
bull T-tests χ2 tests Wilcoxon rank-sum Fisherrsquos exact tests
bull Estimation of differences also important
bull Some subgroup analyses performed though power is small ndash pre-define
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II non-cancerbull Secondary outcomes safety tolerability
treatment completion rate
bull Preliminary evidence of OS (or phase III outcome)
bull Use for designing phase III trial
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II Cancerbull Placebo generally considered unethical
bull Any disease which is terminal or requires intervention of some sort (schizophrenia HIV)
bull Best supportive care may be given for palliation
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II Cancerbull How would you measure if there is lsquoany
evidence of activityrsquo
bull Most cancer drugs historically fail and have substantial side effects
bull Remember 33 have seriouslife-threatening AE in 1st 28 days alone and ~5 respond at all
bull Thoughts
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II Cancerbull Give to small numbers of patients (single-arm)
bull If anyone responds =gt evidence of activity
bull Add a few more patients and estimate RR
bull Gehan - 1960
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase II Cancerbull Two stages formalize based on α and β
(Fleming 1982)
bull Optimal designs (Simon 1989)
bull Assuming drug is ineffective what is minimum number of patients needed
bull Reduces total sample size across all trials
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Design Issuesbull Some advances ndash some patients respond to std
bull Ethically can not give novel therapy with less chance of response
bull Novel therapies designed not to shrink tumour but only prevent it from growing
bull How to proceed
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Questionsbull Often novel therapies are given in addition to
standard of care
bull Outcome of response changed to time to progression
bull Randomization
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Histogram of number of randomized phase II cancer studies published from 1986 to 2002
Lee J J and Feng L JCO 2005234450-4457
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Randomized Phase II designsbull gt4x number of patients needed
bull Solutions look for massive effects inflate α and β
bull One suggestion is α=020 β=020 HR=06 ndash unrealistic
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Randomized Phase II Outcomesbull PFS instead of OS
bull PFS often a poor surrogate
bull May be difficult to run phase III if large differences observed
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Other designsbull lsquoPick-the-winnerrsquo
bull 21 allocation with control only to lsquoensure H0 is accuratersquo
bull No formal comparison different regimens
bull None widely accepted
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Statistical Design No Design categoriesemspemspemspemspSWE randomized phase II 13 106
emspemspemspemsp1-stage design 48 390
emspemspemspemspemspemspemspemsp1-sample binomial 14
emspemspemspemspemspemspemspemsp2-sample binomial 15
emspemspemspemspemspemspemspemsp2-sample continuous 12
emspemspemspemspemspemspemspemsp2-sample survival 7
emspemspemspemsp2-stage design 55 447
emspemspemspemspemspemspemspemspGehanrsquos 14
emspemspemspemspemspemspemspemspSimonrsquos optimal 14
emspemspemspemspemspemspemspemspSimonrsquos minimax 11
emspemspemspemspemspemspemspemspAd-hoc binomial 10
emspemspemspemspemspemspemspemspFlemingrsquos 3
emspemspemspemspemspemspemspemspMultinomial 2
emspemspemspemspemspemspemspemspLogistic 1
emspemspemspemsp3-stage design 4 33
emspemspemspemspemspemspemspemspEORTC binomial 3
emspemspemspemspemspemspemspemspEnsignrsquos binomial 1
emspemspemspemspOthers 3 24emspemspemspemspemspemspemspemspBayesian binomial 2
emspemspemspemspemspemspemspemspGroup sequential binomial 1
Design by type of primary end points
emspemspemspemspBinomial 102 829emspemspemspemspContinuous 12 98emspemspemspemspSurvival 7 57emspemspemspemspMultinomial 2 16Total 123 1000
Type of Statistical Design Used in Randomized Phase II Trials Among the Studies With Reported Statistical Designs (N = 123)
Lee J J and Feng L JCO 2005234450-4457
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
lsquoNovelrsquo designsbull Randomised discontinuation trials
bull Treat all patients with agent for 1 cycle
bull Keep treating patients with a response
bull Stop treating patients who are progressing
bull Randomize patients with stable disease to treatment versus placebo
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Summary of phase II trial designs
Rubinstein L et al Clin Cancer Res 2009151883-1890
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
RDT Trialsbull Primary analysis is comparison of randomized
patients
bull Re-challenge placebo patients who progress
bull Ethics
bull Can fail miserably if of patients with SD is small
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Crossoverbull Trial is novel therapy versus BSC
bull Patients who fail BSC lsquocrossoverrsquo to receive novel therapy
bull Ethically all pts receive trt
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Crossoverbull Problem is study is really treat now or treat
later
bull All patients are progressing at start of study
bull Tumour grows and becomes more resistant
bull Cannot determine long-term OS advantage If trial is + and regulatory approval difficult for fundersinsurance (OHIP)
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Bayesian Response-Toxicity Curve
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Genetic Marker Designbull NSCLC pts primary outcome 8-week SDbull Genetic profiling improved primary outcome from 30
(expected) to 46 in BATTLE
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Bayesian Adaptive Designbull BATTLE randomize pts 1111 initially
bull As info increases allocate patients preferentially to treatmentmarker combination with best performance
bull Outcome is Prob(8 week SD|marker amp treatment)
bull Large complex need real-time datahistology
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Marker Designsbull I-SPY2 underway in breast cancer
bull Lots of information essentially many single-arm trials
bull May reject treatmentmarker combo with small numbers
bull Costs ~$27 million over 5 years for 800 patients
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Combined Phase Trialsbull Phase III trials
bull Include phase I patients in phase II analysis
bull Only pts treated at RP2D
bull Population must be the same thus may reduce s evaluable for phase I
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Combined Phase Trialsbull Phase IIIII
bull Do randomised phase II study using a surrogate endpoint
bull Stop at analysis 1 if lack of efficacy
bull Continue to phase III if some evidence
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Combined Phase Trialsbull Does not affect the α since interim analysis
based on different endpoint
bull However must be willing to commit resources for phase III study
bull Phase III tends to have large of institutions if stop at IA much work for little gain
bull 3-6 months to get study started
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Secondary Outcomesbull Safety toxicity cost QOL ease of
treatment
bull Many not known after phase I
bull May inform phase III design but may be added costs with no benefit if phase II fails
bull Can add support for regulatory approval of positive trials
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Basis of All Initial Approvals
Cytotoxic Drugs (n = 13 drugs in 34 indications)
Targeted Drugs (n = 11 drugs in 26 indications)
No No
Phase III data 30 88 20 77
Single-arm phase II data only
4dagger 12 4Dagger 15
Randomized phase II data only
0 0 2sect 8
Registration Information for US Food and Drug Administration (FDA) Oncology Drug Approvals for Solid Tumors 1998 Through 2008 Inclusive
Hui K Gan Axel Grothey Gregory R Pond Malcolm J Moore Lillian L Siu and Daniel Sargent dArr Randomized Phase II Trials Inevitable or InadvisableJCO 28(15) 2641-2647 2010
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase 0bull Looking for biological effects only
bull Give agent as a single-agent for 1 weekday
bull Evaluate biological markers only
bull Some then combine agent with standard in full trial
bull No clinical benefit
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Phase 0bull Preferentially given as part of phase I study
(can identify dose effect)
bull Not necessarily given to all patients in study
bull Occasionally given in isolation (give 1 month of treatment to 5-10 patients)
bull Recruitment is a major issue
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Additional Considerationsbull Benefit to patient is of concern
bull Considerable risk of harm
bull Small numbers of patients
bull Benefit is that may identify assay target
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Additional Considerationsbull Phase I and II are supposed to be lsquoquickrsquo
bull Can take 2-3 years (or more) each
bull What will landscape look like in 10 years (when applying for regulatory approval)
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Additional Considerationsbull Up to 1 year to get study activated after
protocol finalized
bull Research Ethics Boards
bull Legal contracts with all sites
bull Database data collection
bull Site scientific and finance reviews
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Data Collectionbull Up to 1000 data items collected trial
bull Average of lt20 used (lt100 trial)
bull Massive wastage
OrsquoLeary Seow Julian Levine Pond Data collection in cancer clinical trials Too much of a good thing Clin Trials 10(4)624-632 Aug 2013
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Data Collectionbull Clinical trials nurse collects databull Data manager enters databull Trial coordinator reviews data finds issuebull Coordinator sends DCF to data managerbull Data manager reviews with nursebull Data manager sends revised databull Trial coordinator re-reviews data bull Study investigator reviews and signs offbull Database is locked
bull All at gt$50 hour
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Data Collectionbull Is it critical to the study
bull How likely will it be for planning future study
bull Data clarification is importantWeight=110 (kglbs)What date is 11614 How important is family history if half patients
answer=UNK
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Privacybull Large restrictions on data captured due to
privacy
bull Date of birth diagnosis ndash restricted to year
bull Genetic markers
bull Want tissue sample sent to third party for QA review ndash did you specify in the consent
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Consent Formbull Legal document describing procedures
goals information needed options for patient
bull Very detailed (I have seen gt30 page consents)
bull Sub-consents for biomarker research
bull Signed by patient personnel (not in position of authority)
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Regulatory Burdenbull 296 distinct processes required to activate a phase III trial
bull Dilts DM Sandler AB Cheng S et al (2009) Steps and time to process clinical trials at the Cancer Therapy Evaluation Program J Clin Oncol 271761ndash1766
Area Processing Step Decision Point Loop Stopping Point
Concept (total) 106 21 5 6
emspemspemspemspCRM 76 13 2 4
emspemspemspemspTask force 71 19 4 5
Protocol 88 17 8 4
PMB 10 6 2 1
CDE 20 4 1 0
CIRB 34 9 5 2
Final review 16 3 3 0
Total CRM 244 52 21 11
Total task force 239 58 23 12
Comparison of Works Steps Decision Points Loops and Stopping Points by Major Stage of Development
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Dilts D M et al JCO 2009271761-1766
Median of gt600 days from concept to activation
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Procedure CostLife-Year Saved
Clinical trials regulations $2700000
Hemodialysis29 $43000-$104000
Statins for heart disease (moderate- to high-risk patients)30 $19000-$25000
Colorectal cancer screening by colonoscopy32 $14000
Adjuvant trastuzumab breast cancer31 $20000
Bevacizumab advanced nonndashsmall-cell lung cancer33 $380000
Paclitaxelcisplatin for advanced ovarian cancer34 $26000
David J Stewart Simon N Whitney Razelle Kurzrock Equipoise Lost Ethics Costs and the Regulation of Cancer Clinical Research JCO 28(17) 2925-2935 2010
Costs per Year of Life Gained by Selected Interventions
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
DSMBbull Debate about if they are needed
bull Phase I probably not
bull Cancer phase II ndash depends on design (randomized or single-arm of patients)
bull Non-cancer phase II - probably
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
DSMBbull 1-2 sites investigators know each patient
bull DSMB knows just numbers
bull Steering committee should meet before every stagedose escalation decision etc
bull Larger studiesmore sites may need independent review
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Final thoughtsbull Early phase studies often have lsquosimplersquo
designs
bull More complex logistical practical issues than late stage designs
bull These must be incorporated in analysis design interpretation inference
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-
Final thoughtsbull Must as a biostatistician have understanding
of non-statistical aspects and
bull Must be able to communicate effectively with non-statisticians (clinicians nurses CRA ethics regulatory pathologists radiologists lab technicians scientists IT geneticists hellip)
- Early Phase Clinical Trials
- Learning Objectives
- My Background (biases)
- Your Background
- Thought Experiment
- Standard Drug Development Paradigm
- Phase I
- Phase II
- Clinical Trial Phases
- Clinical Trial 1
- TGN1412 for RA
- Dose Levels
- Phase I Cancer (3+3)
- DLTs
- Slide 15
- Phase I Non-cancer
- Slide 17
- Slide 18
- Cancer Phase I Ethics
- Ethics
- Slide 21
- Correct RP2D
- PowerPoint Presentation
- Slide 24
- Modified Designs
- Model-Based Designs
- Example dose-toxicity curve
- Garrett-Mayer E Understanding the Continual Reassessment Method for Dose Finding Studies An Overview for Non-Statisticians (March 2005) Johns Hopkins University Dept of Biostatistics Working Papers Working Paper 74
- Others
- Slide 30
- Secondary Outcomes
- Slide 32
- Howeverhellip
- Targeted Agents
- Slide 35
- Slide 36
- Phase II non-cancer
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Phase II Cancer
- Slide 44
- Slide 45
- Slide 46
- Design Issues
- Questions
- Slide 49
- Randomized Phase II designs
- Randomized Phase II Outcomes
- Other designs
- Slide 53
- lsquoNovelrsquo designs
- Slide 55
- RDT Trials
- Crossover
- Slide 58
- Bayesian Response-Toxicity Curve
- Genetic Marker Design
- Bayesian Adaptive Design
- Marker Designs
- Combined Phase Trials
- Slide 64
- Slide 65
- Slide 66
- Slide 67
- Phase 0
- Slide 69
- Additional Considerations
- Slide 71
- Slide 72
- Data Collection
- Slide 74
- Slide 75
- Privacy
- Consent Form
- Regulatory Burden
- Slide 79
- Slide 80
- DSMB
- Slide 82
- Final thoughts
- Slide 84
-