Bayesian hierarchical classification and information sharing for clinical trials with subgroups and binary outcomes.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Modeling

Background:

• Bayesian hierarchical models facilitate information sharing across subgroups in clinical trials.
• Traditional models pool all subgroups, which can cause bias or inefficiency when subgroup differences are substantial.

Purpose of the Study:

• To propose a novel hierarchical Bayesian classification and information sharing (BaCIS) model.
• To address limitations of traditional models in multigroup phase II clinical trials with binary outcomes.

Main Methods:

• Developed the BaCIS model, incorporating subgroup classification into hierarchical modeling.
• Classified subgroups into two clusters based on outcomes, enabling cluster-specific information sharing.
• Applied the model to the design and analysis of multigroup clinical trials.

Main Results:

• The BaCIS model demonstrated improved operating characteristics compared to traditional hierarchical models.
• Effective information sharing within identified clusters, mitigating bias and enhancing efficiency.
• Robust performance across various simulated scenarios for binary outcomes.

Conclusions:

• The BaCIS model offers a superior approach for designing and analyzing multigroup phase II clinical trials.
• Subgroup classification within a Bayesian framework effectively balances information sharing and subgroup specificity.
• This method enhances statistical power and reduces bias in trials with heterogeneous subgroups.