Bayesian Study Designs In Early-Phase Oncology Trials

The landscape of early-phase oncology drug development is shifting toward more sophisticated and biologically informed study designs. Traditional dose-escalation methods, such as the 3+3 design, are increasingly recognized as insufficient for optimizing novel therapies, particularly immunotherapies, targeted agents, and cell and gene therapies. These approaches often fail to properly capture delayed toxicities or integrate complex pharmacological data, limiting their utility in selecting a biologically relevant dose. In response, regulatory initiatives like the FDA’s Project Optimus advocate for identifying optimal biological doses rather than relying solely on maximum tolerated doses (MTD).

Bayesian adaptive designs offer a compelling alternative, providing flexibility to escalate, de-escalate, or re-escalate dosing based on accumulating data. These designs enable multidimensional decision-making, improved dose-response characterization, and more efficient use of patient data. However, their success depends on thoughtful implementation, including prespecified criteria for integrating diverse data types and extended observation windows.