Practical guidance for Win Statistics and Tournament Methods for multifaceted outcomes in stroke research: Review and recommendations.

Abstract

Background: Most stroke trials conventionally use outcome measures that reflect a single point of concern, such as functional outcomes or neurological improvement. Such a focus has a limited capacity to reflect the holistic nature of clinical reality, where multiple facets of health are of importance. Other clinical areas are increasingly embracing alternative approaches that better reflect the multifaceted nature of health outcomes, driven by the development of Win Statistics methodology.

Methods: Win Statistics compare the outcomes for all possible pairs of patients from the treatment versus control group and summarize the treatment effect as a combination of proportions of such comparisons where the person from the treatment group has a better/worse outcome when compared to the person in the control group. In this context, "better outcome" is defined holistically, based on information collected across multiple health facets (e.g. mortality, symptom reduction, and quality of life). Such approaches are highly applicable to stroke research. They are a direct extension of "Tournament Methods" that are already used to analyze modified Rankin Scale data, and several stroke trials have already adopted this new approach to better reflect the multifaceted nature of stroke. Despite this, there is little guidance available to stroke researchers who wish to make use of the approach.

Results: This article provides a review of how to effectively use Tournament Methods to reflect the multifaceted nature of stroke. We review how Tournament Methods may be used to estimate treatment effects that simultaneously consider multiple clinical outcomes. We describe methods that have been used within stroke to combine multiple clinical outcomes for Tournament Methods analysis, provide an overview of the effect size measure estimates commonly used to summarize treatment effects, and discuss both available statistical methods and software for the analysis, visualization, and reporting of Win Statistics. We provide practical recommendations for conducting and reporting statistical analyses under the Tournament Methods approach using published stroke trials as illustrative examples. We further present newly developed point-and-click software that provides easy access to statistical codes for conducting Tournament Methods analyses and facilitates the use of our recommendations. This software can be downloaded from https://github.com/HannahJohns/compare-wins.

Conclusion: Tournament Methods are flexible and increasingly used within stroke research. We advocate for a principled approach to their use in the design and analysis of stroke trials to better reflect the multifaceted nature of health care.