New Covariate-Adaptive Randomization Procedure: Balancing Continuous and Discrete Covariates Simultaneously

• Activity Number: 341 - Topics in Adaptive Designs: Sample Size, Randomization and Related Topics
• Type: Contributed
• Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yifan Zhou* and feifang Hu
• Companies: and George Washington University
• Keywords: covariate-adaptive randomization; clinical trial; balancing covariates; Mahalanobis distance; Markov chain; within-stratum imbalances

Abstract

In clinical trials, covariate-adaptive randomization balances treatment allocation over essential covariates. Most of the well-developed procedures only consider discrete covariates. Continuous covariates are categorized to qualitative variables, which brings information loss. In addition, most of traditional procedures do not work well with large number of covariates. In this paper, we propose a new sequential procedure balancing both continuous and discrete covariates by introducing a new imbalance score, which is a weighted sum of Mahalanobis distance and squared within-stratum imbalances. Theoretical properties of the new procedure are obtained. Advantages of the new procedure are demonstrated by simulation studies and a real clinical trial data analysis.