Bayesian meta-analytical methods to incorporate multiple surrogate endpoints in drug development process.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Epidemiology

Background:

• Evaluating surrogate endpoints is crucial for efficient clinical trials.
• Existing bivariate meta-analysis methods predict treatment effects but have limitations with multiple surrogates.
• Uncertainty in surrogate endpoint effect estimates impacts prediction accuracy for final outcomes.

Purpose of the Study:

• To extend multivariate meta-analytic models for incorporating multiple surrogate endpoints.
• To reduce uncertainty in predicting treatment effects on final outcomes.
• To develop a Bayesian framework for flexible modeling of multiple surrogate endpoints.

Main Methods:

• Developed Bayesian multivariate meta-analytic models with a product of normal univariate distributions for between-study variability.
• Proposed two models: one with an unstructured covariance matrix and another with a structured matrix.
• Incorporated individual-level associations using Prentice's criteria to inform within-study correlations.

Main Results:

• The proposed multivariate models effectively incorporate multiple surrogate endpoints.
• The Bayesian framework allows flexible modeling of complex relationships between outcomes.
• Demonstrated application in relapsing-remitting multiple sclerosis, using relapse rate and MRI lesions as surrogates for disability worsening.

Conclusions:

• Multivariate meta-analysis offers enhanced prediction of treatment effects for final outcomes using multiple surrogates.
• The developed Bayesian approach provides a flexible and robust framework for surrogate endpoint evaluation.
• These methods can improve the efficiency and reduce uncertainty in clinical trial evaluations.