Abstract:
|
Multi-stage adaptive designs offer the flexibility to make prospectively defined changes to aspects of trial conduct. However, if we decide to entertain notions of introducing such changes it behooves us to understand well the repercussions. At an interim analysis a trial designer might envisage one or more opportunities to drop arms, change randomization scheme, increase sample size, etc. When one or more of these adaptive components are being considered, natural questions arise: Does the proposed adaptation maintain trial integrity vis-à-vis control of type I error? How often and when do we expect a particular adaptation to be implemented? If a sample size re-estimation option is considered, we would like to characterize the distribution of the final sample size. When we cannot appeal to analytically derived solutions, simulation provides a means to answer these questions and gain insight into how the considered adaptions play out under controlled circumstances where the simulator controls the data generation. We will discuss aspects of planning and summarizing findings in the context of simulation reporting.
|