Bayesian dose-finding in phase I/II clinical trials using toxicity and efficacy odds ratios

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Pharmacology

Background:

• Phase I/II clinical trials require careful dose-finding to balance treatment efficacy and toxicity.
• Traditional methods often analyze toxicity and efficacy separately, potentially missing crucial correlations.
• A need exists for adaptive designs that can dynamically adjust dosage based on observed bivariate outcomes.

Purpose of the Study:

• To propose a novel Bayesian adaptive design for dose-finding in phase I/II clinical trials.
• To incorporate both toxicity and efficacy bivariate outcomes into the dose-selection process.
• To model the correlation between toxicity and efficacy without assuming a specific dose-response curve.

Main Methods:

• A Bayesian adaptive design framework was developed for joint modeling of bivariate binary data (toxicity and efficacy).
• Novel prior distributions using logit transformations were introduced to ensure monotonic dose-toxicity constraints and correlate outcomes.
• Dosage adjustments for subsequent patient cohorts were based on posterior toxicity and efficacy probabilities via odds ratio criteria.

Main Results:

• Simulation studies demonstrated favorable operating characteristics for the proposed Bayesian design.
• The toxicity-efficacy odds ratio trade-off approach effectively identified desirable dose levels.
• The design successfully treated most simulated patients at optimal or near-optimal doses across various scenarios.

Conclusions:

• The proposed Bayesian adaptive design offers an effective strategy for dose-finding in early-phase clinical trials.
• Jointly modeling toxicity and efficacy improves the precision and safety of dose selection.
• The method is applicable to real-world trial designs, as illustrated by a breast cancer oncology study example.