Dose--schedule finding in phase I/II clinical trials using a Bayesian isotonic transformation

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Traditional dose-finding trials assume simple toxicity-dose relationships.
• Dose-schedule-finding trials present more complex, matrix-ordered toxicity probabilities.
• Optimizing both efficacy and safety requires advanced statistical methods.

Purpose of the Study:

• To develop a statistical framework for dose-schedule-finding trials.
• To model unordered toxicity and efficacy probabilities simultaneously.
• To create an algorithm for optimal patient allocation in these trials.

Main Methods:

• Proposed a Bayesian hierarchical model for joint toxicity and efficacy probability modeling.
• Applied Bayesian isotonic transformation to enforce matrix-order constraints on toxicity probabilities.
• Developed a sequential patient allocation algorithm based on posterior distributions.

Main Results:

• Demonstrated the necessity of statistical modeling for dose-schedule-finding success.
• The proposed Bayesian method effectively handles matrix-ordered toxicity.
• Illustrated the methodology with a leukemia clinical trial application.

Conclusions:

• The developed Bayesian approach provides a robust framework for dose-schedule-finding trials.
• The method allows for simultaneous optimization of efficacy and safety.
• This approach enhances clinical trial design for complex oncology treatments.