Regularized win ratio regression for variable selection and risk prediction, with an application to a cardiovascular...

Area of Science:

• Biostatistics
• Clinical Trials
• Health Outcomes Research

Background:

• The win ratio is valuable for analyzing hierarchical composite endpoints, prioritizing critical outcomes like mortality.
• Existing regression frameworks for the win ratio have limitations with high-dimensional datasets and numerous predictors.
• A robust variable selection method is needed for the win ratio framework.

Purpose of the Study:

• To propose an elastic net-type regularization approach for win ratio regression.
• To extend the proportional win-fractions (PW) model to handle high-dimensional data.
• To develop a scalable and robust method for variable selection in win ratio analysis.

Main Methods:

• An elastic net-type regularization approach is developed for win ratio regression.
• Subject-level cross-validation is used for optimized model selection.
• A generalized concordance index is defined for performance metrics.
• The wrnet R-package is implemented for practical application.

Main Results:

• Simulation studies show wrnet outperforms traditional Cox regression for time-to-first-event analysis.
• The method demonstrates superior predictive accuracy in the HF-ACTION trial data.
• wrnet effectively identifies prognostic variables in complex datasets.

Conclusions:

• The wrnet approach integrates the interpretability of the win ratio with the scalability of elastic net regularization.
• The wrnet R-package offers a user-friendly tool for applying these procedures.
• Future work may address non-proportionalities and nonlinearities in covariate effects.