Perspectives of Safety Issues in Drug Development

Industry Statistical Perspective Timothy Costigan, Ph.D.

Wei Shen, Ph.D.Eli Lilly and CompanyIndianapolis, Indiana

2003 FDA/Industry Statistics Workshop

Abstract

We review the objectives and requirements of safety assessment through the phases of drug development. We summarize recent regulatory guidelines and initiatives. We discuss how to communicate results of safety analyses through efficient use of graphical displays and data reduction. We also discuss multiplicity issues relating to safety data.

Overview of Presentation

I. Background and General Considerations

II. Recent Regulatory Guidances and Initiatives

III. Presentation of Integrated Safety Data – Synthesis, Data Reduction, Graphical Displays

IV. Multiplicity Issues in the Interpretation of Safety Data

V. Appendix: Assessing safety during the Phases of Drug Development

I. Background and General Considerations

Risk Assessment and Management - Context

Safety Data Collected in Clinical Trials

Objectives of Safety Data Analysis

Major Safety Issues in Drug Development

Adequacy of Long Term Exposure and Breadth of Effects Studied

Risk Benefit Interpretation of Safety and Efficacy – General Safety Effects

Risk Assessment and Management

Risk assessment is the process of identifying, estimating, and evaluating the nature and severity of risks associated with a product (draft FDA concept paper 2003). Http://www.fda.gov/cder/meeting/riskManageII.htm• Occurs throughout a product’s lifecycle• Comprehensive description of safety required by the Food,

Drug and Cosmetic Act• Emphasis on a Management Plan

Safety Data Collected in Clinical Trials

Standard Safety Data (Gait et al, 2000 DIA J, see appendix)

Exposure and reasons for discontinuation

Adverse events (SAE, DCAE, TEAE)

Clinical laboratory measurements

Vital signs

ECGs

Special Safety Data

Based on indication, class of medication, findings

Objectives of Safety Data Analysis

Identify and understand safety issues as early as possible

Identify risk factors related to increased toxicity and lack of efficacy

Consequences

Allow clinicians to assess the risk/benefit of therapies for classes of patients

Patient safety maintained overall and within special populations

Responsibilities and Goals

Regulatory Agency and Sponsor(s) Joint Responsibility

Drugs used in clinical trials are safe

Marketed drugs are safe and efficacious

Avoid product recalls and major label changes whichcould have been foreseen

Goal: each patient receives the therapy relative to efficacy and dose with the best risk/benefit profile

Adhere to the Declaration of Helsinki

Some Major Safety Issues in Drug Development

Hepatotoxicity

QT interval prolongation – surrogate for a major safety issue

Adverse events with etiology indicative of toxicity

Adequacy of Long Term Exposure and Breadth of Effects Studied

It is important to have adequate safety data in clinical studies of sufficient duration prior to approval

Guidance – minimum required prior to submission

Extension active treatment phases

Continued long term safety assessment during regulatory review, post marketing clinical study commitments

Breadth of effects studied overall and in special populations– Clinical pharmacology package is a major component (see Appendix)

Risk Benefit Interpretation of Safety and Efficacy

Safety and efficacy need to be considered jointly in a risk benefit analysis, particularly when comparing therapies

Adherence for all communications to physicians

Lack of efficacy after switching from an effective therapy is often a greater general safety concern than direct side effects, particularly when comparing therapies

Risk Management may address

Managing side effects

Appropriately monitoring patients who switch therapies

II. Recent Regulatory Safety Guidance and Initiatives

Current - Final

Assessment of hepatotoxicity

Proposed - Draft

Assessment of QT interval prolongation

Risk management plan

Assessment of Hepatotoxicity

FDA PhRMA Sponsored Workshop• Feb 12-13, 2001

Http://www.fda.gov/cder/livertoxThree White papers

Pre and Non ClinicalClinicalPost Marketing Considerations

Model for standardization of analysis

Clinical interpretation of liver enzymes discussed

Special laboratory analysis (as opposed to standard)

platelets, neutrophils, immunogenicity, immunotoxicity

Liver Signals

Hy’s Law

Individual Studies

Cumulative Meta Analysis

Integration of Studies with > 2 weeks exposure• By Study Design

– Placebo Controlled– Active Control– Open Label

• By Duration of Exposure• By frequency of Evaluation

Assessment of QT Interval Prolongation

Assessed in clinical pharmacology and clinical studies

New draft guidance: perform a large clinical pharmacology study to assess Qt prolongation very early in product development

Http://www.fda.gov./cder/calendar/meeting/qt4jam.pdf

January 2003 workshop, “The Clinical Evaluation of QT Interval Prolongation and Proarrhythmic Potential for Non-antiarrhythmic Drugs” which was sponsored by DIA, in collaboration with NASPE, FDA, and Health Canada.

Assessment of QT Interval Prolongation

Industry working group reviewing draft guidance

Recent advisory committee for two products

Midwestern Biompharm Workshop, May 2003 ‘Analysis of QT/QTc Interval Data’, Marilyn Agin chair

Shah A and Hajian G (2003), ‘A maximum likelihood approach to estimating the QT correction factor using mixed effects model’, Statis Med, 1901-1909

Draft QT Guidance: Summary

Super-therapeutic doses in healthy volunteers, above highest possible attainable exposure

Evaluation by clinical experts rather than computer readings

Evaluation at times covering maximum concentration of agent and metabolites

Inclusion of a positive control for assay sensitivity

Clinical relevant changes described (< 5 ms increase not associated with Torsades de Pointes)

Prominence given to outlier analysis and concentration effects

Upper 90% confidence interval of change versus placebo < 10ms

Draft QT Guidance: Technical Issues

QT interval must be corrected for heart rate changes

Standard corrections: Fridericia

Bazzett (for historical purposes)

Individual corrections

Population based corrections

Model based ANCOVA correction (Dmitrienko A, Smith B, “Repeated measures models analysis in the analysis of QT interval. Pharmaceutical Statistics, 2003, to appear)

Averaging using at least 3 measurements recommended

Issues: study timing, measurement methodology, positive control

Risk Management Plan

Proactively identify safety issues as soon as possible

Labeling (USPI and sometimes PPI) and monitoring spontaneous adverse events is usually adequate

Key additional component: Physician and patient education

Superiority over relying mainly on black box warnings

Metrics to evaluate performance

Sometimes establishes additional studies

Clinical Studies, Analyses and Components of Plan

Post marketing studies to investigate rare events

Post marketing studies to assess safety in special populations not adequately studied in clinical studies

Interaction studies to assess effects of administering contraindicated medications concurrently - clinical pharmacology studies with healthy volunteers

Assessment of potential impact of off-label use

Product naming, tamper proof packaging, counterfeit protection

III. Presentation of Integrated Safety Data

Context of Presentation

Transparency, Consistency, Detail

Data Reduction and Synthesis

Graphical displays

Descriptive versus inferential statistics

Context of Presentation of Safety Data

Degree of detail, sophistication of analysis and language used should be determined by target audience

Regulatory Authorities for Approval

Clinical Investigator Brochures

Physician Education

Patient Education

Information Based Medicine

Communication of risks is an important aspect of management

Data Reduction and Synthesis

Treatment Emergent Adverse Events Signal Detection (See Appendix: MedDRA) – large integrated database

Incidence > 2% for treated patients and > 1% greater for treated patients than for placebo patients

Simple and transparent

Consistency with other TEAE signal detection criteria (relatedness, statistical significance)

Adverse events within pre-specified clinically relevant clusters

Incidence of selected serious events per patient year exposure across studies of various designs

Supportive Analyses of Selected Target

Further focus on selected target (for selected audiences)

Related to study drug

Severity - % patients with severe targeted event

% targeted event classified as severe

Incidence and prevalence by visit

Concomitant medications taken for event

Presentation of Safety Data: Graphics

Typically, safety data are summarized in tabular forms by therapy groups

Individual listing by subject

One-page patient summary

Graphical presentation of safety data can add value • Kaplan-Meier curves commonly used to describe time to event

data

Onset of Adverse Events

05

101520253035404550

0 to 3 3 to 6 6 to 9 9 to 12

Placebo

Low dose

High dose

Onset of AE by time of dose (hours)

• High dose is associated with an early onset of AE

% o

f O

ccu

rren

ces

Prevalence and Incidence of AE by Visit

0

1

2

3

4

5

6

7

Placebo LowDose

HighDose

Placebo LowDose

HighDose

Placebo LowDose

HighDose

Visit 3(4 weeks) Visit 4(8 weeks) Visit 5(12 weeks)

Per

cent

Persistent OccurrenceFirst Occurrence

• AE decreases in frequency by visit, particularly for new occurrence.

Blood Pressure Change Over Time

90

100

110

120

130

140

0 6 12 18 24 30 36 42 48

Time relative to dosing (h)

Mea

n A

mbu

lato

ry s

ysto

lic

BP

(mm

Hg)

Placebo

Treatment

• Treatment decreases SBP, effect lasts 12 hours.

QTc Change vs Plasma Concentration

-60

-40

-20

0

20

40

0 50 100 150 200 250 300 350 400 450 500

Serum Concentration (ng/mL)

QTC

Cha

nge

(mse

c)

• Lack of association between QTc change and serum concentration

Advantage of Graphical Analysis

Visual simplicity

Describe the entire distribution of data, including outliers - Box plots

Graphical summary of safety data is complimentary to tables and listings – use to illustrate most important features of data (observed patterns, lack of pattern)

Descriptive versus Inferential Statistics

Always present easily interpretable clinical summaries based on descriptive statistics (with basic statistical analyses such as Fisher’s exact test if appropriate)

More sophisticated inferential statistical analyses are sometimes warranted (with appropriately large sample size - reduce type II error rate)

Special safety studies

Newly proposed QT pharmacology study

Special laboratory analysis

IV. Multiplicity Issues Interpretation of Safety Data

Correction for multiplicity philosophy and strategy for efficacy and safety outcomes

QTc Example, Hypothetical example

Gate keeping strategy

Assessment of safety in subgroups

Multiplicity issues in meta analysis

Multiplicity adjustments for tolerance and adverse events

Identifying Optimal Treated Population

References

Multiplicity adjustment philosophy

Dunnett C and Goldsmith C “When and how to do multiple comparisons”, in Statistics in the Pharmaceutical industry, Buncher and Tsai Eds, 1981, Marcel Decker

Efficacy – often Safety – usually not

Still need to ask the question

Efficacy – always Safety – never Not correct

When multiplicity adjustments are made for safety outcomes present both unadjusted and adjusted p-values – aids in synthesis (type II error still primary concern)

When?

Usually not for standard safety data

More frequently for special safety data – pre-specified hypotheses in special safety studies – with adequate sample size to control type II error

Determining factors

Active versus placebo controlled

Whether non-inferiority design

Nature of outcome

Phase of Study

Hypothetical example

Special safety study undertaken due to non-clinical finding

Regulatory agency and sponsor design a large study

Primary endpoint main analysis (p>.5)

Primary endpoint supportive analysis (p>.3)

Secondary endpoints 2, 3 main analysis (p>.2)

Secondary endpoint 2 supportive analysis (p>.3)

Secondary endpoint 3 supportive analysis (p=.045)

Gate keeping strategy

Dmintrinko A, Offen W and Westfall P (2003) ‘Gate keeping strategies for clinical trials that do not require all primary effects to be significant’, Statist Med, 2387-2400

Specify a hierarchy of outcomes in the protocol and use a stage-wise testing strategy which controls the false error rate by the closed testing principle

Main use is assessing secondary efficacy outcomes

Is applicable in some special safety scenarios

QT example

Many ways of analyzing one outcome; QTc

Multiplicity adjustments probably not appropriate, but multiple testing can result in false positives

Synthesis is still needed

Pre-specify a primary correction

Present results based on all corrections in the same table

Emphasize pre-specified primary correction and the method that fits the data best (Correlation with RR closest to 0)

Give less emphasis to Bazzett (historical correction) – explain outliers due to large RR changes

Assessment of Safety in Subgroups

The correlation coefficients for test statistics for a single outcome in overlapping high (low) risk subgroups depends only on the sample sizes of the subgroups and the sample size of their intersection

Moreover in these situations multivariate probabilities can be easily calculated

Consequently model based multiplicity adjustments can be obtained to complement adjustments based on resampling

Multiplicity Considerations in Meta Analysis

Separate meta analyses should be performed based on study design and study duration

Consistency of results across studies should be examined

When meta analysis is performed in which not all studies contain all doses then a conservative test procedure based on direct comparisons is obtained when one includes study in the model and uses Dunnett’s test in SAS

Multiplicity adjustments for tolerance and adverse events

Novel Approaches for analyzing Clinical Safety/ Adverse event Data, Midwestern Biompharm Workshop, May 20, 2003 Devan Mehrotra organizer

“Multiplicity considerations in evaluating safety in Clinical Trials”, Joe Heyse and Devan Mehrotra

Active control setting

1. SAE

2. Pre-specified TEAE with expected high incidence

3. Remaining (multiplicity adjustment based on resampling and false discovery rate)

Identifying the optimal treated population

Therapy A and Therapy B are only treatments of a disease

Populations A Therapy A superior

B Therapy B superior

E Both equally efficacious

N Neither superior to placebo

Goal: Subscribe therapy A to all population A patients and to no population B or N patients

Identifying the optimal treated population: strategies

Competitive non-inferiority for efficacy and superiority for selected TEAE

Clinically relevant efficacy overall

Clinically relevant efficacy in selected subgroups

Diabetes, hypertension, hyperlipedemia

Various indicators of less severe disease

Competitive non-inferiority for efficacy and superiority for selected TEAE in selected subgroups

Conclusion

Recent guidelines will help ensure patient safety, especially with respect to hepatotoxicity and QT interval prolongation

Risk management occurs throughout a product’s life cycle and involves the effective communication of risks

Special safety data should be analyzed and interpreted differently than standard safety data, including multiplicity considerations

Joint assessment of safety and efficacy: references

Bryant J and Day R (1995) ‘Incorporating toxicity considerations into the design of two-stage phase II trials, Biometrics, 1372-1383

Jennison C and Turnbull B (1993), ‘Group sequential tests for bivariate response: interim analyses of clinical trials with both safety and efficacy endpoints, Biometrics, 741-752

Letierce A, Tubert-Bitter P, Kramar A and Maccario J (2003) ‘Two treatment comparison based on joint toxicity and efficacy ordered alternatives in cancer trials’ Statist Med, 859-868

Identifying the optimal treated population: references

Bristol D, ‘p-value adjustments for subgroup analyses’, J Biopharm Stat, 1997, 313-321.

Byar D and Corle R,’Selecting optimal treatments in clinical trials using covariate information’, J Chron Dis, 1977, 445-459

Gail M and Simon R, ‘Testing for quantitative interactions between treatment effects and patient subsets’ JASA, 1985, 361-372

Koch GG, Stuart A and Gansky MS, “Statistical considerations for multiplicity in confirmatory protocols, DIA J, 1996, 523-533

Pocock S, Assmann S, Enos L and Kasten L (2002) ‘Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems, Stat Med, 2917-30

V. Appendix: Assessing Safety during the phases of development

Non-clinical and Toxicology Studies

Clinical Pharmacology Studies

Clinical Studies

Post Approval Studies

Interconnectedness of phases

Predominance of Safety

Pre-clinical and Toxicology Studies

Animal and Laboratory Studies

Target Organ Toxicity

Liver is a major target organ

Target reaction

QT prolongation – HERG assay study

Consequences of findings

End development, clinical hold, design special safely study, implement special monitoring (additional laboratory measurements, post marketing study/surveillance)

Standard Clinical Pharmacology Studies

PK and PD Studies

characterize exposure (Cmax, tmax, half life, accumulation)

Food effect study

Alcohol effect study

PK in special populations – hepatic, renal, elderly

Interaction of therapy with concomitant medications which inhibit or induce enzymes involved in metabolism

Objective - characterize maximum exposure (increased toxicity)

Objective - characterize minimum exposure (lack of efficacy)

Clinical Pharmacology Studies with Safety Outcomes

Effects on safety outcome (adverse event, laboratory analyte,vital sign)

Aspirin interaction study– Bleeding time- Healthy volunteers

Antihypertensive medication interaction study – Blood pressure outliers

Exercise tolerance – Patients with stable angina

Clinical Pharmacology Studies under Extreme Conditions

Super-therapeutic dose –signal detection

Healthy volunteers, no need to match target disease population

Interpret in relation to phase 2/3 studies

Additional clinical pharmacology studies under less extreme conditions may be useful

Clinical Studies Prior to ApprovalReference

Gait JE, Smith S and Brown S, “Evaluation of safety data from controlled clinical trials: the clinical principles explained”, DIA Journal, Vol 34, 273-287, 2000.

Disposition

Exposure

Adverse events: SAE, discontinuations due to AE, treatment emergent adverse events (TEAE)

Laboratory safety data

Safety profile for patients with targeted TEAE versus without

Adverse Events: MedDRA (new standard)

MedDRA ,a hierarchical dictionary which combines features of several dictionaries, is superior to COSTART which has inadequate specificity – actual term, LT, PT, HLT, HLGT, SOC

Requires adequate site training for actual term recording

Review of coding hierarchy relevant to each therapeutic area (TA) for subtleties in hierarchy

Sometimes coding is too specific for required purpose and TA specific clustering required

Regularly updated – logistical challenges (integration)

A listing of all TEAE sorted by preferred term is helpful (in addition to sorted by patient)

Adverse Event Analysis References

1996 FDA/Industry Workshop, 1996 Biopharmaceutical Report, Vol 4, No 3

1996 Biopharmaceutical Report, Vol 4, No 2, B Northington, ‘A review of issues in the collection and reporting of adverse events’ and L Tremmel ‘Describing Risk in Long-Term Clinical Trials’

O'Neil R “Statistical analysis of adverse events from clinical trials, with emphasis to serious adverse events’ Drug Inf J, 1987, 9-20

Post Approval DataTypes of Studies

Spontaneous Adverse Event Reporting

Case Series Analysis, Registry for rare events

Case Crossover Study -Within patient analysis based on exposure status while experiencing a targeted event relative to overall extent of exposure (Encyclopedia of Biostatistics, Wiley; Pharmacoepidemiology, Overview)

Post marketing regulatory commitments

dictated from results of earlier phases

Other post-marketing studies - useful safety assessment in real world clinical practice (even when not primary objective)