The win odds: statistical inference and regression

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Statistical Inference

Background:

• Generalized pairwise comparisons and win statistics (win ratio, win odds, net benefit) are valuable for analyzing composite outcomes in clinical trials.
• A key limitation of these statistics is the inability to adjust for covariates beyond stratified analysis.
• The win odds statistic, which accounts for ties, has gained attention as an alternative to the win ratio.

Approach:

• This work reviews and synthesizes information on win odds to clarify their statistical inferences.
• Alternative variance estimators using exact permutation and bootstrap methods are presented.
• Statistical inference is explored via the probabilistic index.

Key Points:

• The study extends multiple-covariate regression probabilistic index models to the win odds for univariate outcomes.
• Regression models are applied to data from the CHARM trial for illustration.
• The enhanced win odds methodology provides a more robust approach to analyzing clinical trial data with covariates.

Conclusions:

• The developed methods offer improved statistical inference for win odds in clinical trials.
• Covariate adjustment is achieved through extended regression models, overcoming limitations of stratified analysis.
• These advancements contribute to more comprehensive and accurate interpretation of composite outcomes in clinical research.