Multiplicity Concerns in Subgroup Analysis
View Presentation View Presentation
Christy Chuang-Stein, Pfizer   *David Li, Pfizer, Inc. 

Keywords: Subgroup analysis, Type I error rate, Bias, Adjusted effect size

Interpretation of subgroup analysis results is challenging, partly due to the issue of multiplicity. Several procedures are available to control multiplicity for pre-specified subgroup analyses. This talk will briefly review these procedures and will also propose a new approach for establishing an efficacy claim for the overall population and/or a targeted subgroup. The new approach takes advantage of the mathematical structure between the test statistics for the targeted subgroup and the overall population. If assumptions on the effect sizes in both the targeted subgroup and the complementary subgroup can be made, they can be incorporated into the new proposal.

If time permits, this talk will also address the need of correcting estimated effect sizes in subgroup analyses and discuss how to derive multiplicity-adjusted estimates for treatment effect.