Large sample inference for a win ratio analysis of a composite outcome based on prioritized components

Area of Science:

• Biostatistics
• Clinical Trials
• Statistical Inference

Background:

• Composite outcomes are frequently used in clinical trials, particularly for multiple time-to-event endpoints.
• The conventional method of analyzing the time to the first event has significant limitations.
• Alternative methods like the proportion in favor of treatment and the win ratio have been proposed.

Purpose of the Study:

• To develop novel statistical tests and confidence intervals for composite outcomes using prioritized components.
• To provide a general non-parametric inference framework for comparing treatments.
• To extend these methods for stratified analyses and multi-group comparisons.

Main Methods:

• Utilized the large sample distribution of multivariate multi-sample U-statistics.
• Constructed significance tests and confidence intervals for composite outcomes.
• Applied a non-parametric approach for general inference.

Main Results:

• The proposed methods offer general inference for both the proportion in favor of treatment and the win ratio.
• The approach is adaptable to stratified analyses and comparisons involving more than two groups.
• Demonstrated utility with time-to-event data from a clinical trial.

Conclusions:

• The developed non-parametric methods provide a flexible and robust framework for analyzing composite outcomes in clinical trials.
• These methods address limitations of standard approaches for multiple time-to-event data.
• The approach facilitates more comprehensive treatment effect evaluation.