Two-to-one (2:1) randomization has become a common feature in clinical trial design, particularly in oncology, often under the belief that it helps recruit more patients by increasing the chances of receiving the experimental treatment. But according to Freidlin and Korn in their recent JCO Oncology Practice article, this practice may be more problematic than promising.
On the surface, 2:1 randomization sounds appealing: more patients on the experimental arm means potentially more patients benefitting from a new therapy. It’s often pitched as an ethical advantage and a recruitment booster. However, the evidence supporting these claims is thin. There’s no strong causal link between 2:1 designs and higher accrual rates, and even if such a boost occurs, it’s frequently offset by the increased sample size required, up to 12.5% more participants, just to maintain statistical power.
From a statistical efficiency standpoint, 1:1 randomization remains superior. It maximizes the power of the study with the fewest patients, minimizing both cost and risk. In contrast, 2:1 designs expose a larger number of patients to potentially ineffective or toxic therapies. The authors highlight troubling examples where experimental arms with poor outcomes, such as in trials of PI3K inhibitors or bevacizumab plus lomustine, resulted in hundreds more patients facing unnecessary harm due to skewed randomization.
Ethically, the issue is just as complex. While it may feel “fairer” to give more patients access to a promising new treatment, doing so in the absence of confirmed efficacy undermines the very principle of clinical equipoise. Worse, patients assigned to the control arm might drop out at higher rates, feeling shortchanged by the uneven odds, introducing bias and affecting trial validity.
Operational challenges also abound. Stratified randomization, crucial for balancing prognostic factors, becomes messier under 2:1 designs. Minimization algorithms and block randomization require extra adjustments, increasing complexity without a clear payoff.
So, when does a 2:1 design make sense? According to the authors, only in limited early-phase settings, especially small phase II trials, where understanding the safety or pharmacokinetics of a novel therapy crucial. In those cases, the added exposure to the experimental drug may provide valuable data to inform further development. But beyond that, the rationale wears thin.
The takeaway is clear: While 2:1 randomization may be well-intentioned, its routine use lacks justification in most settings. Trial designers should default to 1:1 unless a specific, data-driven reason suggests otherwise. Efficiency, ethics, and scientific integrity all favor balance over bias.
