Subgroup analysis is a common practice in clinical research and drug evaluation. Subgroup analysis by gender, race and age are routinely required by FDA for each NDA submission. The purpose of subgroup analysis is primarily for assessing internal consistency of treatment effect, signaling high risk subpopulation, and generating hypothesis for further studies. The importance and limitation of subgroup analysis was discussed in several guidance.

The primary concerns for subgroup analysis are 1) the lack of sufficient pre-planning and multiplicity adjustment, which cause difficulty in result interpretation, and 2) small sample sizes in subgroups limit the ability in making inference in the size and consistency of treatment effects among the subgroups. In this talk, we will address issues related to the second concern by exploring the optimal sample size allocation among subgroups, taking into considerations of the prevalence, prior information, cost and time, as well as potential post marketing confirmation.