Perspectives on the Use of a Bayesian Approach in Designing Seamless Phase 2/3 Clinical Trials with Dose Selection, Early Stopping for Futility and Efficacy, or Sample Size Re-estimation.

Seamless Phase 2/3 clinical trial designs have become an increasingly attractive strategy to improve development efficiency by reducing the time between development phases and enabling the reuse of early-stage data in confirmatory analyses. In parallel, Bayesian methods have gained attention for their ability to support adaptive decision-making under uncertainty, particularly in complex trial designs. In this article, we discuss perspectives on the use of Bayesian approach in designing inferentially seamless Phase 2/3 trials that incorporate dose selection, early stopping for futility or efficacy, and sample size re-estimation. We outline key advantages and limitations of seamless approaches compared with separate Phase 2 and Phase 3 trials, and we contrast Bayesian and frequentist frameworks, including hybrid designs that combine Bayesian interim decision rules with frequentist final inference. Particular emphasis is placed on operating characteristics that are informative for trial planning and regulatory discussions, such as predictive probabilities of success, probabilities of false and correct conclusions, dose selection probabilities, and expected sample size. We propose an iterative design process that integrates prior internal and external evidence, multidisciplinary stakeholder input, and extensive simulation to calibrate decision rules. A detailed case study of an inferentially seamless Phase 2/3 trial illustrates how Bayesian posterior and predictive probabilities can guide interim decisions on dose selection, futility stopping, and adaptive sample size modification, while maintaining strong control of the family-wise error rate at the final analysis. We conclude that the use of Bayesian methods in designing seamless Phase 2/3 designs, when carefully planned and transparently evaluated through simulations, can provide a flexible and efficient framework for accelerating drug development while supporting robust and interpretable decision-making.