|
Abstract:
|
Multi arm clinical trials often include interim analyses where arms may be dropped for futility, safety, or ethical reasons. The arm dropping may be fully pre-specified or due to unanticipated interim results or circumstances. An arm being dropped gives rise to (at least) two potential adaptations for the trial: modifying the final sample size and/or modifying the allocation ratio for the remaining arms. For example, if an arm is dropped, should the study size be decreased to account for the dropped arm or maintained at the planned size? Further, should the allocation ratio be modified to maximize power of a dose-placebo comparison and/or modified to maintain the original percentage of allocation to control group? In this paper, we discuss how the combination of what appear to be straightforward adaptations may lead to complex consequences in terms of power and Type I error. We will provide practical guidance on dealing with such questions in arm dropping trials that are not extensively covered in the literature. Trials with fixed sample size designs and trials that include a potential sample size re-estimation will be examined.
|
