Background/Aims: Baseline balance is key for stroke RCTs. On average RCTs achieve balance in baseline covariates and desired treatment allocation ratios, but not necessarily in individual trials. Covariate adaptive randomisation methods, such as CS-MSB, interfere in randomisation when baseline covariate imbalance between groups exists. However, asides from CSSize-MSB, current methods do not protect treatment allocation ratios. Permuted block methods are designed to address treatment allocation imbalance. Thus, integrating these principles into MSB-based algorithms may combine advantages from both approaches. This study aimed to investigate performance of integrated Permuted Block CS-MSB randomisation for stroke clinical trials.
Methods: The design and execution of in silico full factorial experiments followed best practice guidelines for evaluating medical statistical methods using simulation. Data from four hyperacute stroke clinical trials were used to investigate performance of the Permuted Block CS-MSB algorithm compared to CSSize-MSB and CS-MSB algorithms for various sample sizes, covariate numbers and covariate types (1,000 simulations per factorial cell; 4,725,000 simulations total). Performance measures included overall baseline covariate imbalance, treatment allocation imbalance and interference rate.
Results: The Permuted Block CS-MSB algorithm demonstrated better performance across measures compared to CSSize-MSB and CS-MSB algorithms. Results were consistent across clinical trial datasets of various sample sizes and baseline covariate numbers and types.
Conclusion: The Permuted Block CS-MSB adaptive randomisation algorithm demonstrated superior achievement of the desired treatment allocation ratio compared to existing CSSize-MSB and CS-MSB algorithms in hyperacute stroke clinical trials, without compromising balancing of baseline covariates or protection of treatment assignment randomness.