A Hybrid Bayesian/Frequentist Two-Stage Randomized Phase II Design

• Activity Number: 298
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 8:30 AM to 10:30 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Dawn Peters*+
• Companies: Oregon Health and Science University
• Keywords: Phase II ; Two-stage ; Bayesian ; Randomized

Abstract

There has been considerable discussion in the statistical community regarding the use of single-arm versus randomized two-arm designs for phase II studies assessing the promise of a new therapy. Points of contention include the extra resources needed for a two-arm study and the lack of a concurrent control in a one-arm study. In this paper we propose a hybrid Bayesian/frequentist two-stage partially randomized phase II design for assessing the efficacy of a new therapy. In the first (single-arm) stage, only the new therapy is provided to subjects and the decision to proceed to the second stage is based on Bayesian posterior probabilities that incorporate prior information on the new and standard therapies. At the second stage, subjects are randomized between the new and standard therapies and final inference in based on frequentist considerations only. Designs are constructed to satisfy Type I and Type II error constraints and optimality criteria are considered for design selection.