Optimal Patient Enrollment in Sequential Adaptive Clinical Trials with Binary Response

Adaptive clinical trials for new drugs or treatment options promise significant benefits to both trial sponsors and patients, but complicate resource allocation. We study the patient enrollment and trial termination decision in a Phase 3 group sequential adaptive trial, which allows for early trial termination due to established efficacy or anticipated futility. Our objective is to quantify the impact of interim analyses on key performance metrics, including the new drug's misclassification risk and time-to-market, and the firm's profit. To this end, we build a stochastic dynamic programming model that incorporates Bayesian updates on the drug's efficacy. The patient enrollment decision in this setting is characterized by endogenous uncertainty, and a trade-off between the incentive to establish the drug's efficacy early on (exploitation), due to a time-decreasing market revenue, and the benefit from collecting information on the drug's performance (exploration) prior to committing a large amount of resources. We characterize important structural properties of an optimal patient enrollment policy and perform a numerical study utilizing realistic data. Our results suggest that sequential adaptive clinical trials, even with a single interim analysis, outperform their traditional counterparts (i.e., fixed sample size trials) across all metrics, because patient enrollment is re-optimized adaptively, based on updated information on the drug's efficacy