Defining estimand for the win ratio: Separate the true effect from censoring

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• The win ratio is increasingly used for hierarchical composite endpoints in clinical trials.
• Current applications often lack attention to the estimand, hindering interpretation and cross-trial comparisons.

Purpose of the Study:

• To advocate for defining the estimand as a crucial first step in win ratio analysis.
• To identify and address the elusiveness of the win ratio estimand, stemming from its time-dependency.

Main Methods:

• Summarizing two statistical approaches from literature: a nonparametric method with pre-specified time frames and a semiparametric method assuming a constant win ratio.
• Highlighting publicly available software and real-world examples for these methods.

Main Results:

• The win ratio's estimand is intrinsically dependent on the comparison's time frame.
• Unspecified time frames lead to haphazard censoring, complicating interpretation.

Conclusions:

• Clearly articulating the estimand is essential for robust win ratio analysis.
• Addressing challenges like estimand construction and inference with intercurrent events is necessary for future research.