Bayesian phase II clinical trial design with noncompliance

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Noncompliance is frequent in early-phase clinical trials, potentially biasing intent-to-treat effect estimates and leading to erroneous conclusions.
• Robust methods are needed to address noncompliance in adaptive trial designs.

Purpose of the Study:

• To propose a novel Bayesian approach for sequentially monitoring Phase II randomized clinical trials.
• To account for noncompliance information within the principal stratification framework.
• To estimate the complier average causal effect (CACE) for efficacy and toxicity outcomes adaptively.

Main Methods:

• Utilized the principal stratification framework to define relevant causal estimands.
• Employed Bayesian additive regression trees (BART) for covariate selection and CACE estimation.
• Implemented an adaptive decision rule for early trial termination based on estimated CACE.

Main Results:

• The proposed Bayesian sequential monitoring design effectively handles noncompliance in Phase II trials.
• Simulation studies demonstrated the design's excellent performance in providing unbiased CACE estimates.
• Adaptive decisions based on CACE improved the efficiency and reliability of trial monitoring.

Conclusions:

• The developed Bayesian approach offers a robust framework for managing noncompliance in adaptive Phase II clinical trials.
• This method enhances the accuracy of treatment effect estimation and supports informed decisions regarding trial continuation or termination.
• The findings suggest improved reliability and efficiency for early-phase clinical trial designs facing noncompliance issues.