A Simulation Study of Bayesian Response-Adaptive Randomization in Clinical Trials

• Activity Number: 408
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Xiaoyan Yuan*+
• Companies: MD Anderson Cancer Center
• Keywords: Bayesian adaptive randomizaition ; clinical trials ; RPW ; SMLE

Abstract

This simulation study has two stages. Stage I is called in-trial stage with n patients, while Stage II, the out-of-trial stage, has (N-n) patients. During stage 1, patients are randomized to 2 treatment arms. And before stage 2, superior arm is selected out by estimation of response rate. And (N-n) patients in stage 2 are all assigned to the superior arm. Four different randomization methods: FR,RPW,SMLE and BAR are investigated for stage 1. Performance of these randomizations are evaluated by allocation of in-trial patients in each arm (n1:n2), in-trial response rate of n patients (y/n), total response rate of N patients (Y/N) and percentage of choosing correct arm (correct×100%) for Stage II. The comparisons are under conditions of that both n and N are fixed and of that only N is fixed. The results show that BAR achieves slightly larger in-trial response rate and total response rate than FR. When only N is fixed, BAR improves the percentage of choosing correct arm relative to FR.