Bayesian design and analysis of composite endpoints in clinical trials with multiple dependent binary outcomes

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Modeling

Background:

• Clinical trials often involve multiple primary endpoints, which can necessitate unmanageably large patient populations.
• Composite endpoints, combining several binary events, can reduce trial size but face challenges in balancing clinical importance and event frequency.

Purpose of the Study:

• To develop a statistical procedure for creating composite endpoints that addresses the heterogeneity and dependency of multiple primary outcomes.
• To propose a weighting method that accounts for the varying clinical importance of different events within a composite endpoint.

Main Methods:

• Utilizing a Bayesian approach with multinomial distribution and Dirichlet priors to model multiple endpoints.
• Applying a Bayesian test of noninferiority for calculating weighting parameters, ensuring effectiveness in small clinical trials.
• Developing composite endpoints for superiority hypothesis testing in single-arm and two-arm clinical trials, with endpoints following a beta distribution.

Main Results:

• The proposed weighting method compensates for differing clinical importance and event frequencies among endpoints.
• The Bayesian framework effectively handles the mutual dependency of primary endpoints.
• The technique provides a statistically sound procedure for constructing composite endpoints suitable for various trial designs.

Conclusions:

• The developed statistical procedure offers a robust method for creating composite endpoints in clinical research.
• This approach facilitates more efficient and informative clinical trials by managing multiple primary outcomes effectively.
• The Bayesian weighting strategy enhances the reliability of hypothesis testing, particularly in smaller patient cohorts.