hypothesis testing and dynamic treatment regimes s.a. murphy schering-plough workshop may 2007...
Post on 20-Dec-2015
216 views
TRANSCRIPT
![Page 1: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/1.jpg)
Hypothesis Testing and Dynamic Treatment Regimes
S.A. Murphy
Schering-Plough Workshop
May 2007
![Page 2: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/2.jpg)
2
Collaborators
• Lacey Gunter
• A. John Rush
• Bibhas Chakraborty
![Page 3: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/3.jpg)
3
Outline
• Dynamic treatment regimes
• Constructing a dynamic treatment regime
• Non-regularity & an adaptive solution
• Example/Simulation Results.
![Page 4: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/4.jpg)
4
Dynamic treatment regimes are individually tailored treatments, with treatment type and dosage changing according to patient outcomes. Operationalize clinical practice.
k Stages for one individual
Observation available at jth stage
Action at jth stage (usually a treatment)
![Page 5: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/5.jpg)
5
Goal: Construct decision rules that input information available at each stage and output a recommended decision; these decision rules should lead to a maximal mean Y where Y is a function of
The dynamic treatment regime is the sequence of two decision rules:
k=2 Stages
![Page 6: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/6.jpg)
6
Data for Constructing the Dynamic Treatment Regime:
Subject data from sequential, multiple assignment, randomized trials. At each stage subjects are randomized among alternative options.
Aj is a randomized action with known randomization probability.
binary actions with P[Aj=1]=P[Aj=-1]=.5
![Page 7: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/7.jpg)
7
Sequential Multiple Assignment Randomization
Stage 1 Intermediate Outcome Stage 2
Relapse
Responder R Prevention
Low-levelMonitoring
Switch toTx C
Tx A
Nonresponder RAugment withTx D
R
Responder Relapse
R Prevention
Low-levelMonitoring
Tx B
Switch toTx C
Nonresponder R
Augment withTx D
![Page 8: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/8.jpg)
8
Sequential, Multiple Assignment Randomized Studies
• CATIE (2001) Treatment of Psychosis in Schizophrenia
• STAR*D (2003) Treatment of Depression• Tummarello (1997) Treatment of Small Cell Lung
Cancer (many, for many years, in this field)• Oslin (on-going) Treatment of Alcohol
Dependence• Pellman (on-going) Treatment of ADHD
![Page 9: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/9.jpg)
9
Two Levels of STAR*D (Tx-resistent Depression)
Stage 1 Stage 2Preference Treatment Intermediate Treatment
Action Outcome Action
Mirtazapine
Switch R Remission Continue on Current Tx
Nortriptyline
TranylcypromineLithium
Augment R Non-remission R
ThyroidMirtazapine + Venlafaxine
![Page 10: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/10.jpg)
10
Outline
• Dynamic treatment regimes
• Constructing a dynamic treatment regime
• Non-regularity & an adaptive solution
• Example/Simulation Results.
![Page 11: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/11.jpg)
11
A natural approach: Myopic Decisions
•Evaluate each stage of treatment in isolation; the dependent variable is 1 if remission in that stage, 0 otherwise.
•In stage 1 there are two treatment actions for those who prefer a switch in treatment (Mirtazapine or Nortriptyline) and two treatment actions for those who prefer an augment (Lithium or Thyroid).
•Compare the two switches in treatment according to the remission rate achieved by end of stage 1. Do the same for the two augments.
![Page 12: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/12.jpg)
12
Need an alternative
•This is not a good idea if we want to evaluate the sequence of treatments (e.g. adaptive treatment strategies).
• Some of the stage 1 non-remitters went on to have a remission in stage 2; these people have an dependent variable equal to 0 in the myopic analysis.
•the remission or lack of remission in stage 2 may be partially attributable to the stage 1 treatment.
•Patching together the separate analyses of the stages requires unnecessary causal assumptions.
![Page 13: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/13.jpg)
13
Need an alternative for the stage 1 dependent variable
•What should the value of the stage 1 dependent variable be for those that do not remit and move to stage 2?
•We should not use a stage 1 dependent variable of Y=1 for those people who remit in stage 2.
•We should not use an stage 1 dependent variable of Y=0 for those people who remit in stage 2.
•The dependent variable should be something in between.
![Page 14: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/14.jpg)
14
Regression-based methods for constructing decision rules
•Q-Learning (Watkins, 1989) (a popular method from computer science)
•Optimal nested structural mean model (Murphy, 2003; Robins, 2004; I like the term A-learning)
• When using linear models, the first method is an inefficient version of the second method when each stages’ covariates include the prior stages’ covariates and the actions are centered to have conditional mean zero.
![Page 15: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/15.jpg)
15
There is a regression for each stage.
A Simple Version of Q-Learning –
• Stage 2 regression: Regress Y on to obtain
• Stage 1 regression: Regress on to obtain
![Page 16: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/16.jpg)
16
for patients entering stage 2:
• is the estimated probability of remission in stage 2 as a function of variables that may include or be affected by stage 1 treatment.
• is the estimated probability of remission assuming the “best” treatment is provided at stage 2 (note max in formula).
• will be the dependent variable in the stage 1 regression for patients moving to stage 2
![Page 17: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/17.jpg)
17
A Simple Version of Q-Learning –
• Stage 2 regression, (using Y as dependent variable) yields
• Stage 1 regression, (using as dependent variable) yields
![Page 18: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/18.jpg)
18
Decision Rules:
![Page 19: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/19.jpg)
19
Outline
• Dynamic treatment regimes
• Constructing a dynamic treatment regime
• Non-regularity & an adaptive solution
• Example/Simulation Results.
![Page 20: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/20.jpg)
20
Non-regularity
![Page 21: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/21.jpg)
21
Non-regularity
![Page 22: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/22.jpg)
22
Non-regularity –
• Replace hard-max
• by soft-max
![Page 23: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/23.jpg)
23
A Soft-Max Solution
![Page 24: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/24.jpg)
24
Distributions for Soft-Max
![Page 25: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/25.jpg)
25
To conduct inference concerning β1
• Set
• Stage 1 regression: Use least squares with outcome,
and covariates to obtain
![Page 26: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/26.jpg)
26
Interpretation of λ
Future treatments are assigned with equal probability, λ=0
Optimal future treatment is assigned, λ=∞
Future treatment =1 is assigned with probability
Estimator of Stage 1 Treatment Effect when
![Page 27: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/27.jpg)
27
Proposal
![Page 28: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/28.jpg)
28
Proposal
![Page 29: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/29.jpg)
29
Outline
• Dynamic treatment regimes
• Constructing a dynamic treatment regime
• Non-regularity & an adaptive solution
• Example/Simulation Results.
![Page 30: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/30.jpg)
30
STAR*D
• Regression at stage 1:
•
•S1= ((1-Aug), Aug, Aug*Qids)
•X1 is a vector of variables available at or prior to stage 1, Aug is 1 if patient preference is augment and 0 otherwise
• We are interested in the β1 coefficients as these are used to form the decision rule at stage 1.
®T1S01+¯T1 S1A1
S01 = (1;X 1)
![Page 31: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/31.jpg)
31
STAR*D
Decision Rule at stage 1:
•If patient prefers a Switch then
•if offer Mirtazapine, otherwise offer Nortriptyline.
•If patient prefers an Augment then
•if offer Lithium, otherwise offer Thyroid Hormone.
![Page 32: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/32.jpg)
32
Stage 1 Augment Treatments
bbb
![Page 33: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/33.jpg)
33
Decision Rules from Soft-Max Q-Learning Y=1 if remission or sufficient response to move to follow-up, Y=0 otherwise
Stage 1 Stage 2 Switch MIRT = NTP (225)
QIDS < 11 LI = THY (45) Augment QIDS ≥ 11 LI < THY (88)
TCP<VEN+MIRT(104)
= means not significant in two sided test at .05 level
< means significant in two sided test at .05 level
![Page 34: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/34.jpg)
34
Simulation
![Page 35: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/35.jpg)
35
P[β2TS2=0]=1 β1(∞)=β1(0)=0
Test Statistic Nominal Type 1 based on Error=.05 .045
.039
.025*
(1)Nonregularity results in low Type 1 error
(2) Adaptation due to use of is useful.
![Page 36: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/36.jpg)
36
P[β2TS2=0]=1 β1(∞)=β1(0)=.1
Test Statistic Power based on
.15
.13
.09
(1)The low Type 1 error rate translates into low power
![Page 37: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/37.jpg)
37
Test Statistic Power based on
.05
.11
.12
(1) Averaging over the future is not a panacea
P[β2TS2=0]=0 β1(∞)=.125, β1(0)=0
![Page 38: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/38.jpg)
38
Test Statistic Type 1 Error=.05 based on
.57
.16
.05
(1) Insufficient adaptation in “small” samples.
P[β2TS2=0]=.25 β1(∞)=0, β1(0)=-.25
![Page 39: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/39.jpg)
39
Discussion
• We replace the test statistic based on an estimator of a non-regular parameter by an adaptive test statistic.
• This is work in progress—limited theoretical results are available.
• The use of the bootstrap does not allow to increase too fast.
![Page 40: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/40.jpg)
40
Discussion
• Robins (2004) proposes several conservative confidence intervals for β1.
• Ideally to decide if the stage 1 treatments are equivalent, we would evaluate whether the choice of stage 1 treatment influences the mean outcome resulting from the use of the dynamic treatment regime. We did not do this here.
• Constructing “evidence-based” regimes is of great interest in clinical research and there is much to be done by statisticians.
![Page 41: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/41.jpg)
41
This seminar can be found at:http://www.stat.lsa.umich.edu/~samurphy/
seminars/Harvard0507.ppt
Email me with questions or if you would like a copy!
![Page 42: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/42.jpg)
42
STAR*D
• Regression at stage 2: α2TS2' + β2S2A2
• S2' =(1,X2, (1-Aug)*A1, Aug*A1, Aug*A1*Qids),
•(X2 is a vector of variables available at or prior to stage 2)
• S1 = 1
• Decision rule: Choose TCP if , otherwise offer Mirtazapine + Venlafaxine XR
![Page 43: Hypothesis Testing and Dynamic Treatment Regimes S.A. Murphy Schering-Plough Workshop May 2007 TexPoint fonts used in EMF. Read the TexPoint manual before](https://reader036.vdocument.in/reader036/viewer/2022062516/56649d445503460f94a20dfd/html5/thumbnails/43.jpg)
43
Switch -.11(.07) -1.6
Augment .47(.25) 1.9
Augment*QIDS2 -.04(.02) -2.3
^̄(s:e:) z statistic
Stage 1 Coefficients
^̄11 =^̄12 =
^̄13 =