Statistical Properties of Covariate-Adaptive Randomization Procedures for Trials with Unequal Treatment Allocation Ratios

• Activity Number: 546
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Oleksandr Sverdlov*+ and Yiwei Zhang
• Companies: Novartis and Novartis
• Keywords: Allocation ratio preserving ; Randomization design ; Unequal allocation

Abstract

Many modern randomized clinical trials involve multiple treatment arms and use designs to achieve treatment groups of unequal size. Some rationales for using unequal allocation include better statistical efficiency of treatment comparisons, cost-efficiency in situations when some treatment protocols are more expensive than others, and potential ethical constraints. Recently, several designs expanding covariate-adaptive randomization procedures to trials with unequal allocation have been proposed in the literature. In this paper we consider several restricted randomization designs which are fully randomized and maintain throughout the trial the target allocation ratio (if necessary, within each stratum). We compare operating characteristics of these designs through a simulation study. We study such issues as balance, randomness of treatment assignments, variations in the allocation ratio, and type I error and power of statistical tests.