Event-specific win ratios and testing with terminal and non-terminal events

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Survival Analysis

Background:

• Clinical trials often use composite endpoints (e.g., death or non-fatal events), potentially omitting valuable data.
• The standard win ratio approach prioritizes events but may lack power if treatment effects are concentrated on non-terminal events.

Purpose of the Study:

• To propose novel statistical tests utilizing event-specific win ratios for terminal and non-terminal events.
• To enhance the power and utility of win ratio methods in clinical trial outcome analysis.

Main Methods:

• Developed event-specific win ratios calculated separately for terminal and non-terminal events.
• Formulated global tests (linear combination, maximum, and $$ tests) based on these event-specific ratios.

Main Results:

• Simulations demonstrated improved power for the new tests compared to the original win ratio and log-rank tests, especially with differential treatment effects.
• The proposed tests successfully rejected the null hypothesis in scenarios where traditional methods did not, as shown in the TOPCAT trial data.
• The maximum test offers desirable coherency, rejecting the global null if and only if a specific event type's null is rejected.

Conclusions:

• Event-specific win ratio-based tests, particularly the maximum test, provide a powerful and useful alternative for analyzing time-to-event outcomes in clinical trials with multiple event types.
• These new methods enhance the ability to detect treatment effects when they impact different event types differentially.