<scp>DOD</scp> ‐ <scp>SSR</scp> : An Adaptive Seamless Phase <scp>II</scp> / <scp>III</scp> Design With Dose Optimization Decision and Sample Size Re‐estimation

The adoption of seamless Phase II/III designs has grown in popularity as a strategy to potentially accelerate the drug development. Making well-informed decisions regarding the drug's potential and addressing important clinical inquiries at the conclusion of the exploratory phase has become a critical step. In response to the increased emphasis on dose optimization, it becomes logical to integrate treatment arm/dose selections into Phase II and implement corresponding design adjustments. Within this framework, employing a fixed sample size presents challenges due to limited information availability before the trial planning and elevated development risks. Furthermore, practical and feasibility considerations have led to the increased utilization of surrogate endpoints for making interim decisions. In this study, we introduce a novel framework for a seamless Phase II/III design involving multiple treatment arms, leveraging Bayesian predictive probability of success (PPoS) for both treatment arm selection and interim sample size re-estimation (SSR) using surrogate endpoints. The proposed design demonstrates improved performance, including a higher likelihood of selecting favorable treatment arm, increased overall statistical power, and reduced average event sizes and trial durations compared to traditional separate Phase II and III designs, as well as other seamless Phase II/III designs without SSR or of which treatment arm selection is based on conditional power. We also showcase the implementation of the proposed design through a case study in non-small cell lung cancer (NSCLC).