Bayesian adaptive design for concurrent trials involving biologically related diseases

Area of Science:

• Biostatistics
• Clinical Trial Design
• Bayesian Inference

Background:

• Clinical programs often involve multiple trials for related diseases.
• Demonstrating superiority to a control in each trial is a common goal.
• Existing methods for information borrowing can be limited, especially with diverse data types.

Purpose of the Study:

• To develop a robust Bayesian design method for concurrent clinical trials in related diseases.
• To enable effective information borrowing on treatment effectiveness across trials.
• To handle diverse data types for endpoints in clinical trials.

Main Methods:

• A Bayesian design using correlated mixture priors and Bayesian model averaging.
• Construction of mixture priors using pessimistic and enthusiastic predictions.
• Elicitation of mixture weights across all possible prior configurations for each disease.
• A novel approach for information borrowing accommodating different data types.

Main Results:

• The proposed Bayesian design framework demonstrates favorable operating characteristics via simulation.
• It outperforms traditional Bayesian hierarchical models for information borrowing with diverse endpoints.
• The method provides a robust framework for information sharing across related disease trials.

Conclusions:

• The developed Bayesian design method offers a superior approach for clinical programs with multiple related disease trials.
• It effectively facilitates information borrowing, improving the assessment of treatment effectiveness.
• This framework is particularly advantageous when trial endpoints involve different data types.