A Bayesian subgroup analysis using collections of ANOVA models.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Statistical Modeling

Background:

• Subgroup analysis is crucial for identifying heterogeneous treatment effects in clinical trials.
• Existing methods may lack robust statistical frameworks for identifying and validating subgroups.
• Parsimonious and interpretable subgroup structures are highly desirable for clinical application.

Purpose of the Study:

• To develop a Bayesian approach for subgroup analysis in two-arm clinical trials.
• To extend previous work by incorporating multiple covariates and a model selection framework.
• To identify subgroups with heterogeneous treatment effects using an objective Bayesian methodology.

Main Methods:

• Utilized analysis of variance (ANOVA) models with multiple categorical covariates.
• Employed a model selection approach to identify subgroups with differential treatment effects.
• Developed objective Bayesian priors and calculated multiplicity-adjusted posterior model probabilities.
• Implemented a structured algorithm for reporting subgroup effects based on posterior probabilities.

Main Results:

• The Bayesian approach effectively identifies subgroups with heterogeneous treatment effects.
• Simulation studies demonstrated favorable frequentist operating characteristics.
• The method provides multiplicity-adjusted posterior probabilities for model selection.
• The approach is illustrated with a real-world clinical trial data example.

Conclusions:

• The proposed Bayesian subgroup analysis offers a statistically rigorous framework for identifying treatment effect heterogeneity.
• The model selection approach facilitates the discovery of parsimonious and interpretable subgroups.
• The method is applicable to various clinical trial settings, particularly the 2x2 covariate case.