Beyond randomization

Conventional opinion states that randomization is the appropriate way to allocate participants to treatments in biomedical studies. The most frequently cited justification is that it is the only method that guarantees balance between treatment groups with respect to all prognostic factors, whether measured or not. Here it is shown by simple arguments that this justification is false and misleading. Design-adaptive allocations are an alternative to randomization that are easy to implement, and virtually guarantee better balance than randomization, for both measured and unmeasured factors. The fraction of studies that will exhibit severe imbalance under randomization is not always trivial. For this reason, in small studies or studies with subgroup analyses or many prognostic factors, design-adaptive allocation is an attractive alternative to randomization. These considerations are particularly relevant to complementary and alternative medicine studies, where resources are relatively scarce, and otherwise underpowered studies might lead to premature termination of promising research paths.