What is it about?
As the win ratio statistic is currently defined, a pairwise comparison in the calculation of win ratio is indeterminate when a data point is missing and, therefore, we either move down the hierarchy or assign a score of zero to the comparison. We point out that under such a definition there is no way to define the underlying estimand for the win ratio statistic so that it is free of parameters governing missing data, even if missing is completely at random. Since parameters governing missing data is seldom of direct clinical interest, the clinical interpretation of win ratio is challenging when missing data is present.
Featured Image
Why is it important?
Our paper questions the clinical interpretability of the win ratio statistic when there is missing data. This is an important question given the ubiquity of missing data.
Perspectives
I hope this article serves as a word of caution when the win ratio is being considered for the primary endpoint. Defining the corresponding estimand is far from straightforward. The win ratio was originally introduced to address the limitation that the Finkelstein–Schoenfeld test lacks an associated effect size. However, it remains doubtful whether this goal has truly been achieved.
Heng Li
Read the Original
This page is a summary of: The Elusiveness of the Win Ratio Parameter in the Presence of Missing Data, Therapeutic Innovation & Regulatory Science, March 2024, Springer Science + Business Media,
DOI: 10.1007/s43441-024-00645-2.
You can read the full text:
Contributors
The following have contributed to this page
