A Parametric Multiple Comparison Procedure for Clinical Trials with Planned Evaluation of Treatment Effect in Pre-Defined Subgroups and Interim Analyzes

• Activity Number: 228 - Innovative Statistical Designs with Real Life Case Studies for New Paradigms in Oncology Drug Development
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Liang Fang* and Ron Yu and Zhishen Ye and Neby Bekele and Ming Lin
• Companies: MyoKardia and Gilead Sciences, Inc. and Gilead Sciences and Gilead Sciences and Gilead Sciences
• Keywords: multiple comparison procedure; multiplicity; precision medicine; adaptive design; biomarker; clinical trial

Abstract

Parametric multiple comparison procedures (MCPs), taking into account the correlations between test statistics, are expected to be more powerful than their counterpart non-parametric procedures, and can be a useful testing strategy for clinical trials with subgroups and interim analyses. This is because test statistics are inherently correlated between the interim and final analyses as well as between the subgroups and overall population. We derived a flexible parametric MCP based on an approximation of correlations between stratified log-rank test statistics and closure principle, referred to as Closed Testing Parametric Procedures (CTPPs). We compared the performance of CTPPs to other parametric and non-parametric MCPs. Simulation studies showed that the type I error rate of a CTPP was strongly controlled. Compared to weighted Bonferroni, group sequential Holm procedure, and Spiessens and Debois' method, the CTPP exhibited the greatest statistical power and the statistical power of the group sequential Holm procedure was only slightly lower.