|
Activity Number:
|
12
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, August 8, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #301570 |
|
Title:
|
Optimized Two-stage Adaptive Design
|
|
Author(s):
|
Keaven M. Anderson*+ and Qing Liu
|
|
Companies:
|
Merck & Co., Inc. and J&J Pharmaceutical Research and Development
|
|
Address:
|
BL3-2, West Point, PA, 19486-0004,
|
|
Keywords:
|
adaptive design ; group sequential design ; optimal design ; clinical trials
|
|
Abstract:
|
Proschan and Hunsberger proposed an adaptive two-stage design with the sample size and critical value for the second stage determined by the observed treatment difference and critical value from the first stage. A conditional error function maps the observed critical value in stage 1 to a required critical value for stage 2 to maintain the overall Type I error. The observed treatment difference and desired power are then used to determine the second stage sample size. Liu and Chi modified this procedure by substituting a minimum treatment difference of interest for the observed stage 1 treatment difference. We select an optimized Liu and Chi design by defining a flexible conditional error function family, assuming a prior distribution for the true treatment difference and selecting a loss function (e.g., expected sample size). Examples will be given comparing these optimal two-stage adaptive designs to optimal two (or more)-stage group sequential designs. Cases will be presented where the optimal two-stage design from either class is slightly better. In general, the optimal two-stage adaptive and group sequential designs are comparable.
|
