Approximating win-loss probabilities based on the overall and event-free survival functions

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• The win ratio is increasingly used for hierarchical composite endpoints in clinical trials.
• Meta-analysis of win ratios is challenging due to the lack of reported win ratio measures in prior studies.
• Subject-level data is often unavailable for meta-analytic approaches.

Purpose of the Study:

• To develop a method for approximating the win ratio in meta-analysis when subject-level data is absent.
• To enable the use of win ratio meta-analysis with commonly reported summary statistics.
• To provide a practical tool for researchers conducting meta-analyses on composite endpoints.

Main Methods:

• Approximation of the win ratio using component-specific Kaplan-Meier curves.
• Inference of between-component association (cross ratio) from summary event counts and rates.
• Validation through simulations and case studies using real-world data.

Main Results:

• The proposed method accurately approximates win-loss probabilities compared to raw data-based estimates.
• Kaplan-Meier curves and summary event data are sufficient for win ratio approximation.
• The methodology demonstrates feasibility and accuracy in diverse scenarios.

Conclusions:

• A novel methodology allows for win ratio approximation in meta-analysis, overcoming common data limitations.
• This approach expands the utility of win ratio for synthesizing evidence from multiple studies.
• The winkm R package facilitates the application of this method in biostatistical research.