Comparison of balance ability of minimization, simple and stratified block randomization from true data of a clinical trial

Objective: Simple randomization, stratified block randomization and minimization are three most commonly used allocation methods in clinical trials. Our objective was to make a comparison of their balance ability from the true data of a clinical trial with a serials observational point from 50 to 1350 patients in 552 combined strata. Study design: We created random sequences by SAS plan program, and then allocated patients by MINI (Minimization), SB4 (Stratified Block randomization with block length of 4) and SR (Simple Randomization) when the sample size reach 50, 100, 300, 500, 800, 1000 and 1350 from a real clinical trial cohort and repeated 1000 times, standardized difference was used to evaluate the balance of baseline covariates. Results: When the number of subjects was small, minimization had the advantages against other two allocation methods. Stratified randomization achieved a slightly better balance result than simple randomization in given conditions. With the increasing number of patients, the difference between simple randomization and minimization or stratified block randomization decreased. Conclusions: We recommend minimization in small sample clinical trials when large number of combined strata exist, but prefer simple randomization when the sample size is large enough.