Operating characteristics of unequal allocation ratios in platform trials with the staggered addition of drugs using binary endpoints

Abstract

Background

One recommendation for the allocation ratio between multiple drugs and a shared placebo control group in platform trials (PTs) is to use a $\sqrt{k}$:1 allocation ratio for the placebo group relative to the drug group, where $k$ is the number of drug groups with ongoing patient enrollment during the trials. However, the practical utility of such unequal allocation ratios in PTs lacks adequate study.

Methods

We compared the performances of equal and unequal allocation ratios through simulations to imitate practical PTs using only concurrent controls and binary endpoints for hospitalized patients with infectious diseases. The operating characteristics, including the type I error rate, power of hypothesis testing, and total sample size, were evaluated.

Results

In PTs, using an unequal allocation ratio (i) results in a considerable augmentation of the total sample size and prolongs the study duration when monthly patient enrollment is low, but (ii) the target power of hypothesis testing is often preserved compared to an equal allocation ratio, even when we incorrectly specify the drug and placebo group mortality rates assumed in the sample size calculation. The average power increase using an unequal allocation ratio relative to the equal allocation ratio per 100-patient increase in the placebo group was approximately 1.9 % in the selected scenarios of our simulation studies.

Conclusion

The results of the current study highlight the quantitative advantages and disadvantages of using unequal allocation ratios in PTs using only concurrent controls under the specific conditions assumed in our simulations and analyses.