Power priors based on multiple historical studies for binary outcomes.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacology

Background:

• Historical clinical trial data is valuable but often limited and heterogeneous, especially for rare or resistant infections.
• Integrating this data into new trials is challenging but crucial for efficient drug development.
• Existing methods may not optimally leverage available historical information.

Purpose of the Study:

• To develop and evaluate an adaptive empirical Bayes method for incorporating historical clinical trial data.
• To compare the proposed method against traditional approaches like the meta-analytic prior.
• To enhance the design and analysis of new clinical trials, particularly for challenging indications.

Main Methods:

• Utilized a multiple power prior approach for adaptive weighting of historical studies.
• Employed empirical Bayes to determine the optimal weight for each historical study.
• Considered fully Bayesian methods alongside the proposed empirical Bayes approach.
• Applied the methods to real-world antibiotic clinical trial data.

Main Results:

• The proposed empirical Bayes adaptive method demonstrated strong performance in operating characteristic analyses.
• The method effectively adapts the weight of historical studies based on data similarity.
• Achieved favorable comparisons against alternative methods, including the meta-analytic prior.
• Showcased the utility of the approach in a binomial setting.

Conclusions:

• The adaptive empirical Bayes method offers a robust and flexible way to incorporate historical data in clinical trials.
• This approach can improve the efficiency and reliability of trial analysis, especially when historical data is heterogeneous.
• The findings support the use of adaptive methods for optimizing the integration of past trial information into future studies.