Balancing Treatment Allocation Over Continuous Covariates: A New Randomization Procedure

• Activity Number: 35
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yanqing Hu*+ and Feifang Hu
• Companies: West Virginia University and University of Virginia
• Keywords: covariates ; continuous ; randomization ; distribution-free ; maximum-imbalance

Abstract

Balancing treatment allocation over covariates is essential in clinical trials. Although randomization for balancing over discrete covariates has been well studied, only a few papers have been published on how to balance over continuous covariates. To take advantage of the existing methods, a continuous covariate usually needs to be transformed to a discrete one by splitting its range into several categories. Such practice may lead to loss of information, and is susceptible to mis-specification of covariate distribution. We propose a distribution-free randomization design, which defines the imbalance measure by the maximum assignment difference when all possible divisions of the covariate range are considered. Besides, we developed an efficient algorithm to implement the new procedure. To compare the proposed design with other procedures we use two criteria: the overall assignment difference between the two treatment groups, and the Kolmogorov-Smirnov distance between the empirical distributions of covariate values in the two groups. By simulation studies the new design shows advantages over discretization and a few other distribution-free procedures.