Functional win-fractions regression models for composite outcomes

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Statistical Modeling

Background:

• Win ratios are common for evaluating interventions with composite outcomes in clinical trials.
• Adjusting win ratios for covariates improves treatment effect estimation.
• Traditional parametric models struggle with complex, non-linear covariate effects on composite outcomes.

Purpose of the Study:

• To introduce a flexible win fraction regression framework using splines to model complex covariate effects.
• To develop an efficient computational algorithm for inference using the generalized method of moments (GMM).
• To provide a large-sample test for assessing the significance of functional covariates.

Main Methods:

• Development of a win fraction regression framework utilizing B-splines.
• Application of the moment condition model framework and GMM for inference.
• Derivation of a large-sample test based on the asymptotic distribution of spline coefficients.

Main Results:

• The proposed spline-based framework adaptively assesses the extent and nature of covariate effects.
• Simulations demonstrate favorable operating characteristics compared to existing methods.
• The methodology shows practical utility in analyzing composite time-to-event data.

Conclusions:

• The flexible win fraction regression framework effectively captures complex, non-linear covariate effects.
• This approach enhances the accuracy and precision of treatment effect estimates in clinical trials.
• The method offers a valuable tool for analyzing composite time-to-event outcomes in cardiovascular and breast cancer trials.