Flexible Bayesian subgroup analysis in early and confirmatory trials

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Bayesian Inference

Background:

• Subgroup analysis is crucial in clinical trials for assessing treatment effect consistency.
• Challenges exist in interpreting subgroup results, potentially leading to flawed conclusions or decisions.
• Reliable subgroup effect estimation is vital for early-phase trials (e.g., basket trials) to guide go/no-go decisions.

Purpose of the Study:

• To review recent Bayesian methods for subgroup analysis designed to correct bias.
• To evaluate the performance of novel Bayesian hierarchical models in phase II basket trials.
• To demonstrate the application of flexible Bayesian models in a phase III oncology trial for regulatory decision support.

Main Methods:

• Review of recent Bayesian subgroup analysis methodologies.
• Simulation studies comparing Bayesian hierarchical models in a phase II basket trial setting.
• Application of Bayesian models to a phase III oncology trial case study.

Main Results:

• Comparison of average total sample size and frequentist operating characteristics (power, familywise type I error rate).
• Assessment of model estimate precision via average relative bias and credible interval width.
• Demonstration of bias correction and improved estimation accuracy with Bayesian approaches.

Conclusions:

• Bayesian hierarchical models offer effective bias correction for subgroup analysis.
• These methods provide reliable treatment effect estimates crucial for both early-phase and confirmatory clinical trials.
• Recommendations are provided for implementing Bayesian subgroup analysis in regulatory and early-phase trial settings.