Bayesian model averaging for randomized dose optimization trials in multiple indications.

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Conventional oncology dose-finding trials face challenges with targeted agents, often leading to toxicity without improved efficacy.
• Optimizing doses for targeted agents in multi-indication proof-of-concept trials is complex due to low prevalence and the need for indication-specific dose-response characterization.

Purpose of the Study:

• To propose a novel Bayesian model averaging approach using robust mixture priors (rBMA).
• To identify the recommended Phase III dose in randomized dose optimization studies conducted simultaneously across multiple indications.
• To offer an alternative to the "more is better" paradigm in oncology dose finding.

Main Methods:

• Developed a Bayesian model averaging approach with robust mixture priors (rBMA).
• Applied the approach to randomized dose optimization studies conducted simultaneously in multiple indications.
• Conducted systematic simulation studies to evaluate performance.

Main Results:

• The proposed rBMA approach improves the accuracy of dose recommendations compared to independent indication-specific strategies.
• The model effectively learns across indications, enhancing dose-finding precision.
• Simulation studies confirmed the approach's performance in making correct dose recommendations.

Conclusions:

• The rBMA approach provides a more accurate and efficient method for dose optimization in multi-indication oncology trials.
• This Bayesian strategy offers a viable alternative to conventional dose-finding paradigms for targeted agents.
• Cross-indication learning enhances the reliability of recommended Phase III doses.