Group-Sequential Procedures for Multiple Endpoints with Adaptive Allocation of Recycled Significance Levels

• Activity Number: 534
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Dong Xi*+ and Ajit C Tamhane
• Companies: Northwestern University and Northwestern University
• Keywords: Closed procedure ; error spending functions ; familywise error rate ; O'Brien-Fleming boundary ; Pocock boundary ; Sequentially rejective Bonferroni procedure

Abstract

Bretz et al. (2009) and Burman et al. (2009) proposed a graphical approach for testing hypotheses on multiple endpoints by recycling the significance levels from rejected hypotheses to unrejected hypotheses. Maurer and Bretz (2013) extended this graphical approach to constructing group sequential procedures (GSPs). The extension requires a scheme to allocate the recycled significance level to stages of the GSPs for unrejected hypotheses. Maurer and Bretz's scheme allocates the recycled significance level to all stages. Ye et al. (2013) proposed a similar scheme and another scheme which allocates the recycled significance level only to the final stage. We refer to these as non-adaptive allocation schemes and analyze their drawbacks. We propose an adaptive allocation scheme which recycles the significance level only to the current and future stages. It enhances overall performance while strongly controlling the familywise error rate. Simulation studies to compare powers of alternative schemes are carried out and illustrative examples are given.