A phase II, seamless single-arm to two-arm Bayesian design for a time-to-event endpoint

Phase II oncology trials often face challenges such as patient recruitment difficulties and limited resources, leading to the use of single-arm designs without concurrent controls. This study introduces a novel two-stage Bayesian design bridging single-arm and randomized two-arm trials for time-to-event endpoints. The design incorporates an initial experimental treatment stage with futility monitoring, followed by a randomized second stage comparing experimental and control arms, contingent on first-stage progression. The methodology enables comprehensive efficacy analysis by combining data from both stages. The design parameters are optimized to minimize the expected sample size under the null hypothesis of no treatment effect. Extensive simulation studies demonstrate consistently lower expected sample sizes than conventional two-arm trials utilizing log-rank tests. Furthermore, the design achieves reduced maximum sample sizes relative to the sequential implementation of separate single-arm and two-arm trials across various scenarios. An illustrative application in extensive-stage small cell lung cancer demonstrates practical implementation considerations. As a resource-efficient approach, this design streamlines phase II evaluations and expedites clinical decision-making for evaluating experimental oncology treatments.