Using Simulation to Aid in the Development of a Phase II Bayesian Adaptive Design

• Activity Number: 386
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Committee on Applied Statisticians
• Author(s): Elizabeth Krachey*+ and Jennifer Gauvin and Theresa C. Ashton and Ohad Amit
• Companies: GlaxoSmithKline and GlaxoSmithKline and GlaxoSmithKline and GlaxoSmithKline
• Keywords: Bayesian ; Adaptive ; Simulation

Abstract

We developed a phase II Bayesian adaptive design to compare multiple doses of a novel two-drug combination therapy to the single-agent therapies. The design incorporates multiple interim analyses for futility and efficacy, decision rules that incorporated multiple treatment comparisons, and adaptive allocation allowing randomization to favor assigning patients to more efficacious treatment arms as more trial data is accrued. We discuss how extensive simulation work was essential to developing the design and understanding key operating characteristics such as type I error, power, choice of prior, sample size determination, adaptation rules, and early and final decision rules. We highlight that examining particular simulated trials outcomes allowed us to 'see' hypothetical trials within the context of the larger simulation results, ultimately helping establish the final design.