Statistical power considerations in the use of win ratio in cardiovascular outcome trials

Wang, Bang; Zhou, Dejian; Zhang, Jialü; Kim, Yoonhee; Chen, Ling-Wan; Dunnmon, Preston; Bai, Steven; Liu, Ququan; Ishida, Eiji

doi:10.1016/j.cct.2022.107040

Cited by 3 publications

(2 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, imposing a margin is contrary to the principle of rank‐based statistical testing 3 . Indeed, simulation studies have shown a potential loss of statistical power if the chosen margin is substantial 17 …”

Section: Discussionmentioning

confidence: 99%

The win ratio method in heart failure trials: lessons learnt from EMPULSE

Collier

Ferreira

Collier

et al. 2023

European J of Heart Fail

View full text Add to dashboard Cite

Aims The EMPULSE trial evaluated the clinical benefit of empagliflozin versus placebo using the stratified win ratio approach in 530 patients with acute heart failure (HF) after initial stabilization. We aim to elucidate how this method works and what it means, thereby giving guidance for use of the win ratio in future trials. Methods and results The primary trial outcome is a hierarchical composite of death, number of HF events, time to first HF event, or a ≥5‐point difference in Kansas City Cardiomyopathy Questionnaire (KCCQ) total symptom score change at 90 days. In an overall (unstratified) analysis we show how comparison of all 265 x 265 patients pairs contribute to ‘wins’ for empagliflozin and placebo at all four levels of the hierarchy, leading to an unstratified win ratio of 1.38 (95% confidence interval [CI] 1.11–1.71; p = 0.0036). How such a win ratio should (and should not) be interpreted is then described. The more complex primary analysis using a stratified win ratio is then presented in detail leading to a very similar overall result. Win ratios for de novo acute HF and decompensated chronic HF patients were 1.29 and 1.39, respectively, their weighted combination yielding an overall stratified win ratio of 1.36 (95% CI 1.09–1.68) (p = 0.0054). Alternative ways of including HF events and KCCQ scores in the clinical hierarchy are presented, leading to recommendations for their use in future trials. Specifically, inclusion of both number of HF events and time‐to‐first HF event appears an unnecessary complication. Also, the use of a 5‐point margin for KCCQ score paired comparisons is not statistically necessary. Conclusions The EMPULSE trial findings illustrate how deaths, clinical events and patient‐reported outcomes can be integrated into a win ratio analysis strategy that yields clinically meaningful findings of patient benefit. This has implications for future trial designs that recognize the clinical priorities of patient evaluation and the need for efficient progress towards approval of new treatments.

show abstract

Section: Discussionmentioning

confidence: 99%

The win ratio method in heart failure trials: lessons learnt from EMPULSE

Collier

Ferreira

Collier

et al. 2023

European J of Heart Fail

View full text Add to dashboard Cite

show abstract

“…In a non‐stratified analysis, the impact of multiple outcomes on the win statistics can be complex. Wang et al 12 focused on the win ratio and pointed out that the first‐priority component (e.g., death) plays a dominant role, especially when that component has a large treatment effect and a high event rate; when adding a component to the composite endpoint, the performance of the win ratio depends on the treatment effect, the event rate, and the position of the component in the importance order. This finding on the win ratio generally applies to the other win statistics.…”

Section: The Stratified Win Statisticsmentioning

confidence: 99%

The stratified win statistics (win ratio, win odds, and net benefit)

Dong¹,

Hoaglin

Huang

et al. 2023

Pharmaceutical Statistics

View full text Add to dashboard Cite

The win odds and the net benefit are related directly to each other and indirectly, through ties, to the win ratio. These three win statistics test the same null hypothesis of equal win probabilities between two groups. They provide similar p-values and powers, because the Z-values of their statistical tests are approximately equal. Thus, they can complement one another to show the strength of a treatment effect. In this article, we show that the estimated variances of the win statistics are also directly related regardless of ties or indirectly related through ties. Since its introduction in 2018, the stratified win ratio has been applied in designs and analyses of clinical trials, including Phase III and Phase IV studies. This article generalizes the stratified method to the win odds and the net benefit. As a result, the relations of the three win statistics and the approximate equivalence of their statistical tests also hold for the stratified win statistics.

show abstract