BayCAR: a Bayesian-based covariate-adaptive randomization method for multi-arm trials

Randomization is an essential component of a successful controlled clinical trial. Many randomization methods have been developed to balance the distributions of covariates across treatment arms to remove potential confounding effects. While the restricted randomization methods would not work well if the number of covariates is large, the theoretical base of the minimization methods needs more justifications. We propose a Bayesian covariate-adaptive randomization method that not only has meaningful interpretations on its adaptive randomization probability, but also achieves desirable marginal and overall balances for both categorical and continuous covariates, particularly when balancing a large number of covariates is necessary.