Overview of Adaptive Designs Think What is Possible 2008 Rutgers - - PowerPoint PPT Presentation

Overview of Adaptive Designs… Think What is Possible

Jeff Maca, Ph.D.

• Sr. Associate Director, Biostatistics

Novartis Pharmacuticals

2008 Rutgers Biostatistics Day April 25, 2008

2 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Outline – Overview of Adaptive designs

• Motivation
• What can change?
• Sample size re-estimation
• Adaptive Dose Finding
• Seamless designs
• Conclusions

3 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Introduction and Motivation

Reducing time to market is/has/will be a top priority in pharmaceutical development

• Brings valuable medicines to patients sooner
• Allows companies to develop drugs more

efficiently Adaptive seamless designs can help reduce this development time

4 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Motivation

• An adaptive trial can plan at the design stage to

correct for incorrect assumptions

• Adaptive trials can merge what might be

considered two seperate trials into one trial

• Careful planning is necessity

Adaptive Designs: Using accumulating data to decide on how to modify aspects of the trial design, during the conduct of the trial and without violating the integrity of the trial

5 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

What can Change?

• Sample Size (sample size re-estimation)
• Can be on Blinded or Unblinded review of data
• Can be related to the hypothesis of interest
• Treatment arms (delete, add, change)
• Adaptive dose finding
• Adaptive Seamless Phase II/III trials
• Population of interest, testing strategies, etc…

Adaptive designs is a broad class of studies, and can be quite different from each other. Some Examples:

6 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Sample Size Re-estimation

• Number of patients in a clinical trial are such that

desired power is achieved.

• Required sample size depends on variability of

primary endpoint (and hypothesized treatment effect)

• Variability estimate may be uncertain for new

indications/some disease areas:

• increased risk of failure (too low sample size)
• unnecessary cost (too high sample size).
• Sample size re-estimation aims to correct for

the initial uncertainty of variability, and to maintain the desired power.

7 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Consequence of mis-specification: power loss

Influence of variability on power

Standard Deviation (SD) of primary endpoint

0.8 1.0 1.2 1.4 1.6 1.8 2.0 20 40 60 80 100

Initially assumed SD=1.0 90% power Actual SD=1.4 64% power

Power (%)

The loss of power if the standard deviation is larger than the pre-trial initial estimate. Risk increases from 10% to 36%

Power loss 26%

8 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Statistical methodology: typical study design without sample size re-estimation

A simple example Variability of primary endpoint: assumed/estimated standard deviation 1.0 unit n=150 patients are needed to achieve 90% power to detect a particular relevant difference

Active enrollment Final Analysis Control

9 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Statistical methodology: study design with sample size re-estimation

A simple example - improved Variability of primary endpoint could be much lower/higher than the initial guess  add an interim review to re-estimate variability  adjust sample size accordingly after interim review

Active enrollment Initially planned Final Analysis Control Interim review: Sample size Re-estimation

10 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Statistical methodology: interim review

At interim review: estimated standard deviation 1.4 units => would need n=300 patients to achieve 90% power Decision at interim review:

• No extra patients => power reduced to 64% (risk)
• Additional patients => adequate power, but cost/time

Active enrollment Initially planned Final Analysis Control Interim review: Sample size Re-estimation Final Analysis

11 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Blinded sample size re-estimation Sample size re-estimation are:

• Blinded
• r
• Unblinded
• Blinded: sample size re-estimation possible

without unblinding the study

• Generally more acceptable
• No DMC required
• No independent interim analysis team necessary
• Decisions on sample size re-estimation can be
• made by the trial team
• penly communicated

12 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Unblinded sample size re-estimation

• Unblinded: sample size re-estimation unblinds

the study

• More precise information on the variability of the

primary endpoint (and the treatment effect)

• Requires DMC and independent interim analysis team
• Some concerns on trial integrity:
• Potential biasing the trial if the investigators/patients think the

drug works better/worse than “expected”

• “Backward calculation“ based on adjusted sample size may

give hint on treatment effect

 Integrate as part of flexible/group sequential design

13 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Sample Size Re-estimation

• Sample size re-estimation can not be used in all

clinical trials

• There must be a quick readout of the primary endpoint

compared to enrollment time in order to estimate the variability during enrollment

• Logistics and drug supply may in some cases prevent

use of sample size re-estimation

14 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Prior to study the true position of dose response curve is unknown Response Dose

Initial doses

In the adaptive dose finding approach, a small number of patients on many initial doses are used to

• utline the unknown

dose-response. As the dose response emerges more patients are allocated to doses (including new doses) within the dose- range

• f interest. In addition

the number of patients allocated to „non- informative‟ doses („wasted doses‟) is decreased.

Adaptive dose finding

X = Mean dose response after a pre-defined number of patients

X X X X X X

Region of interest

X X X X X

• Overview

15 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless designs Primary objective – combine “dose selection” and “confirmation” into one trial

• Although dose is most common phase IIb objective, other

choices could be made, e.g. population

• After dose selection, only change is to new enrollments

(patients are generally not re-randomized)

• Patients on terminated treatment groups could be

followed

• All data from the chosen group and comparator is used in

the final analysis. Appropriate statistical methods must be used

16 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless Designs

Dose A Dose B Dose C Placebo Dose A Dose B Dose C Placebo

• < white space >

Phase II Phase III Stage A (learning) Phase B (confirming) Time

17 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless designs

Statistical methodology for Adaptive Seamless Designs must account for potential biases and statistical issues

• Selection bias (multiplicity)
• Multiple looks at the data (interim analysis)
• Combination of data from independent stages
• Closed testing procedure and Bonferroni adjustment

are two possible methods

18 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless designs

With the added flexibility of seamless designs, comes added complexity.

• Careful consideration should be given to the feasibility for

a seamless design for the project.

• Not all projects can use seamless development
• Even if two programs can use seamless development,
• ne might be better suited than the other
• Many characteristics add or subtract to the feasibility

19 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless designs Enrollment vs. Endpoint

• The length of time needed to make a decision

relative to the time of enrollment must be small

• Otherwise enrollment must be paused
• Endpoint must be well known and accepted
• If the goal of Phase II is to determine the endpoint for

registration, seamless development would be difficult

• If surrogate marker will be used for dose

selection, it must be accepted, validated and well understood

20 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless designs

Clinical Development Time

• There will usually be two pivotal trials for

registration

• Entire program must be completed in shorter

timelines, not just the adaptive trial

21 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Adaptive Seamless designs Logistical considerations

• Helpful if final product is available for adaptive

trial (otherwise bioequivalence study is needed)

• Decision process, and personnel must be

carefully planned and pre-specified

22 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

Conclusions

• Adaptive seamless designs have an ability to improve the

development process by reducing timelines for approval

• Statistical methods are available to account for adaptive

trial designs

• Extra planning is necessary to implement an adaptive

seamless design protocol

• Benefits should be carefully weighed against the

challenges of such designs before implementation

Think what is possible (and think early and often)

23 | Overview of Adaptive Designs…Think What is Possible/ Biostatistics Day/ April 25, 2008

References

• Friede, T. and Kieser, M. (2006). Sample size recalculation in

internal pilot study designs: A review. Biometrical Journal 48 (2006) 4, 537–555.

• Proschan, M. (2005). Two-stage sample size reestimation based
• n a nuisance parameter: A review. Journal of Biopharmaceutical

Statistics 15 , 559-574.

• EMEA/CMPH (draft 23/3/2006). Reflection Paper on

Methodological Issues in Confirmatory Clinical Trials with Flexible Design and Analysis Plan. Check the adaptive designs page on the CIS Statistical Methodology homepage for a complete and current list. Note: „sample size re-estimation“ is also called: „sample size re- assessment“, „sample size re-calculation“, „sample size review“, „internal pilot study design“, etc.