A roadmap to using randomization in clinical trials

Abstract Background Randomization is the foundation of any clinical trial involving treatment comparison. It helps mitigate selection bias, promotes similarity groups with respect to important known and unknown confounders, contributes validity statistical tests. Various restricted randomization procedures different probabilistic structures properties are available. The goal this paper present a systematic roadmap for choice application procedure in trial. Methods We survey available sequential allocation subjects randomized, comparative, parallel group equal (1:1) allocation. explore these procedures, including balance/randomness tradeoff, type I error rate power. perform head-to-head comparisons through simulation under various experimental scenarios, cases when common model assumptions violated. also provide some real-life examples illustrate thinking process selecting implementation practice. Results Restricted targeting 1:1 vary degree they induce, more importantly, terms efficiency inference violated (e.g. outcomes affected by linear time trend; measurement distribution misspecified; or bias introduced experiment). Some robust than others. Covariate-adjusted analysis may be essential ensure results. Special considerations required very small sample size. Conclusions design, data analytic technique (parametric nonparametric), strategy (randomization-based population model-based) all considerations. Randomization-based tests valid alternatives likelihood-based should considered frequently investigators.