A closer look at combining data among a small number of binomial experiments

Area of Science:

• Biostatistics
• Clinical Trial Design
• Regulatory Science

Background:

• Disagreements often arise in regulatory settings regarding the use of subjective prior information in clinical trials.
• Data-based priors offer a more agreeable alternative, but assigning appropriate weight to this prior data remains a challenge.

Purpose of the Study:

• To introduce and evaluate a hierarchical model for objectively linking prior data with current clinical trial data.
• To demonstrate how hierarchical models can facilitate agreement on the use and weighting of prior information.

Main Methods:

• Utilized a hierarchical Bayesian modeling approach.
• Illustrated the methodology with examples combining two binomial experiments.
• Compared finite mixture models with continuous mixture models.
• Presented an example involving the combination of three concurrent studies.

Main Results:

• Hierarchical models provide a relatively objective method for assigning weight to prior data.
• The model's effect on rate estimation, data combination, and hypothesis testing was demonstrated.
• Explained the phenomenon of reduced precision when combining data.
• Finite mixture models performed comparably to more complex continuous mixture models.

Conclusions:

• Hierarchical models offer a robust and objective framework for incorporating prior data in clinical trials.
• This approach can enhance consensus between regulatory bodies and trial sponsors.
• Simpler finite mixture models are effective alternatives to computationally intensive methods.