Reducing Alpha Adjustment When Tests Are Structurally Correlated

• Activity Number: 372
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jonathan Siegel*
• Companies: Bayer HealthCare Pharmaceuticals
• Keywords: Multiple testing ; Correlated test ; Multi-arm trial ; Shared control arm ; subgroup ; PFS and OS

Abstract

Pharmaceutical clinical trials with multiple comparisons require adjustment for familywise false positive error. Positively correlated tests have less alpha inflation than independent tests. Dmitrienko, Tamhane, and Bretz (2010) outline general test procedures which can take advantage of alpha inflation reduction resulting from positive correlation. In many cases, the design itself results in structural correlation implicit in design assumptions. A customized adjustment taking into account the specific design context can result in greater power and/or reduced sample sizes. This presentation outlines the approach, briefly covers background theory, and discusses and provides theory-derived and/or simulation results for examples including a multi-arm trial with a common control arm, a trial with an overall population and a subgroup, and correlation between PFS and OS under a correlated bivariate exponential model.