ABSTRACT
Randomization is a fundamental element of any randomized controlled clinical trial involving treatment comparison. Randomization balances known and unknown risk factors, provides the basis for comparative inferential analysis, and justifies the use of re-randomization tests for reliable distribution-free analysis. Various randomization designs are available. This chapter provides an overview of three important classes of randomization: 1) restricted randomization designs that aim at balancing treatment allocation throughout the trial; 2) stratified and covariate-adaptive randomization designs that aim at balancing treatment allocation overall in the trial and across selected covariates; and 3) outcome-adaptive randomization designs that aim at skewing allocation in favor of the putatively most effective treatment(s) while maintaining the randomized nature of the experiment. We describe these procedures, present some new simulation studies, and where possible make generalizations and recommendations for practice.
