Group Sequential Designs for the Supremum Weighted Log-Rank Test for Survival Data with Time-Varying Treatment Effects

• Activity Number: 171
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Qiang Zhang*+ and Junfeng Liu and Michael Kosorok
• Companies: American College of Radiology and GCE Solutions and University of North Carolina at Chapel Hill
• Keywords: Group sequential design ; survival data ; weighted log-rank test ; time varying

Abstract

Group sequential designs based on the log-rank test under the proportional hazard assumption are widely used in late phase clinical trials. However, for transitory treatment effects with non-proportional hazards, the usual logrank test is less than optimal and statistical power can be greatly compromised. One promising approach, a fixed sample size design based on the supremum weighted log-rank test, has been developed and shown to be more powerful under such conditions. In this research, we investigate group sequential designs for the supremum weighted log-rank test utilizing established results for the interim statistics, including independent increments properties. Statistical monitoring boundaries are derived using established theory for the supremum Brownian motion. We study algorithms for the protection of type I error rates and statistical power under various alternative hypotheses. Monte Carlo simulations show the design has desired properties under the null and is more powerful for time dependent treatment effects. A clinical trial example is used to illustrate how to use the proposed approach to design and monitor survival data with time varying treatment effects.