A modular framework for early-phase seamless oncology trials.

Area of Science:

• Clinical trial design
• Oncology research
• Biostatistics

Background:

• Phase I oncology trials increasingly incorporate multiple objectives, including dose estimation and efficacy assessment.
• Seamless trial designs integrate dose-finding and efficacy evaluation, posing challenges in assessing operating characteristics like probability of identifying acceptable doses.
• Key challenges include determining the likelihood of finding a safe and effective dose and the average number of patients assigned to such doses.

Purpose of the Study:

• To propose a novel modular framework for designing and evaluating seamless oncology trials.
• To address the complexity of modern phase I/II oncology trial designs.
• To facilitate the rigorous assessment of trial operating characteristics.

Main Methods:

• Developed a modular framework with sequential modules for dose assignment and dose-response evaluation.
• Incorporated existing phase I/II designs and introduced a new Bayesian isotonic regression module for dose-response assessment.
• Created a freely available R package, seamlesssim, for numerical estimation of trial operating characteristics via simulation.

Main Results:

• The seamlesssim R package enables simulation-based estimation of operating characteristics for modular seamless oncology trials.
• Demonstrated the utility of the framework in evaluating and comparing different candidate trial designs.
• The framework allows for more rigorous performance assessment and better sample size justification.

Conclusions:

• The proposed modular design framework and simulator enhance the selection of higher-quality seamless oncology trial designs.
• Clinical trialists can compare designs, assess performance, and justify sample sizes more effectively.
• This approach contributes to more efficient and robust oncology drug development.