A Bayesian Dynamic Model-Based Adaptive Design for Oncology Dose Optimization in Phase I/II Clinical Trials

Area of Science:

• Oncology Drug Development
• Clinical Trial Design
• Biostatistics

Background:

• The traditional "more is better" chemotherapy dosing paradigm is outdated with new targeted therapies and antibody-drug conjugates (ADCs).
• The US Food and Drug Administration (FDA) launched Project Optimus to address dose optimization and selection in oncology drug development.
• Early-phase oncology trials face challenges due to data variability and rigid parametric models.

Purpose of the Study:

• To develop a novel adaptive clinical trial design for optimizing dose selection in early-phase oncology studies.
• To simultaneously incorporate both toxicity and efficacy data for robust optimal dose (OD) identification.
• To enhance the design's adaptability for delayed toxicity and efficacy outcomes.

Main Methods:

• Utilized Bayesian dynamic models to leverage information across different dose levels with minimal assumptions.
• Employed Bayesian model averaging to manage uncertainties in dose-response relationships.
• Developed an adaptive design integrating toxicity and efficacy for optimal dose selection in Phase I/II trials.

Main Results:

• The proposed Bayesian adaptive design demonstrated desirable operating characteristics across various simulated scenarios.
• The method effectively integrates toxicity and efficacy data for robust optimal dose identification.
• The design proved adaptable to scenarios involving delayed toxicity and efficacy outcomes.

Conclusions:

• The proposed Bayesian adaptive design offers a robust and flexible approach to optimal dose selection in oncology drug development.
• This method addresses limitations of traditional dosing paradigms and supports Project Optimus' goals.
• The design provides a practical framework for Phase I/II oncology trials, as illustrated by a trial example.