Bivariate network meta-analysis for surrogate endpoint evaluation

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Health Economics

Background:

• Surrogate endpoints are crucial for regulatory decisions, offering early insights into long-term clinical outcomes.
• Existing bivariate meta-analysis methods struggle with imperfect surrogate endpoints where treatment associations vary.
• A need exists for methods that can differentiate treatment effects and model varying surrogacy levels.

Purpose of the Study:

• To develop and evaluate novel bivariate network meta-analysis (bvNMA) methods.
• To enhance the prediction of treatment effects on final clinical outcomes using surrogate endpoints.
• To model treatment-specific surrogacy and identify reliable surrogate relationships.

Main Methods:

• Developed bivariate network meta-analysis (bvNMA) integrating surrogate and final outcome data from multiple trials.
• Estimated individual treatment contrast effects for both surrogate and final outcomes.
• Modeled trial-level and treatment-level surrogacy patterns to assess prediction accuracy.

Main Results:

• bvNMA methods effectively estimate treatment effects on both surrogate and final outcomes simultaneously.
• The methods allow for modeling of varying surrogacy strength across different treatment comparisons.
• bvNMA demonstrated improved prediction of final outcome treatment effects, particularly when surrogacy varied.

Conclusions:

• Bivariate network meta-analysis (bvNMA) offers a robust framework for evaluating surrogate endpoints in healthcare.
• bvNMA enhances prediction accuracy by accounting for treatment-specific surrogacy variations.
• These methods support more reliable regulatory decision-making and treatment effect predictions.