ABSTRACT
With increasing interest in personalized medicine over the last years, study designs allowing to demonstrate efficacy in particular subgroups of the overall patient population become more important. Adaptive enrichment designs provide the possibility to both selecting the target population with the most promising treatment benefit and testing for efficacy within a single trial. Here, the target population is selected in a prespecified interim analysis. So far, it has not been very well investigated how timing of the interim analysis should be chosen.
We investigate the impact of the interim analysis timing on power for the situation of a normally distributed outcome considering two different classes of selection rules. The interim selection is based either on the estimated effect difference between subgroup and total population or on absolute effect estimates. In this article, we demonstrate that there are indeed scenarios in which the timing of the interim analysis has a large impact on power. However, no universally applicable timing with favorable performance exist since power depends on treatment effects, subgroup prevalence, and especially the applied selection rule. Instead, the operating characteristics should be investigated for the specific scenario at hand to determine the most appropriate timing.
Acknowledgments
We thank the reviewers for their helpful comments.
Funding
We gratefully acknowledge the support for this work by the program “Mathematics for innovations in industry and services” of the German Federal Ministry of Education and Research (BMBF) under grant 05M13VHC.