A dynamic power prior for borrowing historical data in noninferiority trials with binary endpoint

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Modeling

Background:

• Traditional frequentist methods for noninferiority testing use fixed margins.
• Bayesian approaches can enhance efficiency by incorporating historical control data.
• Developing appropriate informative priors for Bayesian methods is challenging.

Purpose of the Study:

• To propose a hybrid Bayesian approach for testing noninferiority hypotheses with binary endpoints.
• To introduce a dynamic P value-based power prior for adjusting historical information.
• To improve the control of Type I error rates in noninferiority trials.

Main Methods:

• A hybrid Bayesian framework is developed for noninferiority hypothesis testing.
• A dynamic P value-based power prior parameter is proposed to manage information heterogeneity.
• An adjusted alpha level is introduced to control Type I error.
• Simulations compare the proposed method with test-then-pool and hierarchical modeling.

Main Results:

• The proposed hybrid Bayesian method demonstrates improved efficiency.
• The dynamic power prior effectively adjusts information borrowing from historical data.
• The adjusted alpha level provides better control over Type I error.
• The method is robust in simulations and illustrated with vaccine trial data.

Conclusions:

• The hybrid Bayesian approach offers a flexible and efficient alternative for noninferiority testing.
• The dynamic power prior parameter is a valuable tool for leveraging historical data.
• This method enhances statistical power and error control in clinical trials.