From Phase Ib/II to Seamless Phase II/III: A Simulation-Based Discussion of Design Strategies for Dose Optimization...

The emergence of targeted therapies and immunotherapies has challenged the traditional "more is better" paradigm underlying maximum tolerated dose (MTD)-based trials, as dose escalation beyond receptor saturation often yields limited additional efficacy while substantially increasing toxicity. In response, the U.S. Food and Drug Administration's Project Optimus emphasizes dose optimization (DO) prior to confirmatory studies to ensure a favorable benefit-risk profile. Motivated by this shift, we systematically evaluate strategies for incorporating DO at different stages of oncology drug development. We compare three approaches: (1) DO combined with proof-of-concept in a seamless Phase Ib/II study prior to a registrational trial; (2) DO conducted in an operational seamless Phase II/III design; and (3) DO embedded within an inferential seamless Phase II/III design. Simulation studies were performed across a range of efficacy-toxicity scenarios to assess dose-selection accuracy, statistical operating characteristics, expected sample size, and overall study duration. Under the simulated settings, both the Phase Ib/II DO strategy and the inferential seamless Phase II/III design reduced total development time compared with the operational seamless approach. The inferential seamless design achieved the lowest expected sample size and the shortest average trial duration while maintaining high power under the primary simulated scenarios. The comparative performance of the three strategies also depended on how well the short-term dose-selection endpoints predicted long-term survival. Overall, DO can be effectively implemented at multiple stages of clinical development. The optimal strategy should balance statistical rigor, development speed, practical feasibility, and the suitability of the endpoints used for dose selection.