Simultaneous Inference for Joint Overall and Subgroup Confirmatory Hypothesis Testing

• Activity Number: 234
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Hui Wang*+ and Ying Lu and Ilana Belitskaya-Lévy and Gheorghe Doros and Robert A. Lew and Mei Chiung Shih and Lu Tian
• Companies: Palo Alto VA/CSPCC and Palo Alto VA CSPCC/Stanford University and Palo Alto VA CSPCC and Boston University School of Public Health and VA Cooperative Studies Program Coordinating Center and Stanford University School of Medicine and Stanford University
• Keywords: subgroup analysis ; clinical trial ; phase III ; confirmatory trial ; hypothesis testing

Abstract

In subgroup analysis of confirmatory clinical trials, the objective is to evaluate treatment effects in pre-specified patient subpopulations (such as genotype groups defined by genetic markers) in addition to standard analyses in the overall population. We propose a joint test for simultaneous inference in the overall and subgroup populations assuming certain orders of effects in the subgroups. Our test is based on the joint distributional space of marker positive and negative groups that can lead to definitions of three rejection regions 1. efficacy in subpopulation only; 2. efficacy in overall population; and 3. efficacy in both subpopulation and overall population. Our test achieves higher statistical power for the overall and subgroup effects while controlling the family-wise type I error rate than a class of multiple testing procedures proposed by Millen and Dmitrienko in 2011.The optimal power depends on the specific alternative hypothesis and the sample size. In conclusion, we propose a simple one-step procedure for subgroup analysis that can be used to improve the statistical design of Phase III trials with tailoring objectives.