Keywords: Win ratio, stratified win ratio, inverse variance weighted win ratio, unweighted win ratio, odds ratio, Mantel-Haenszel stratified odds ratio
Win ratio was first proposed by Pocock et al. (2012) to handle a composite endpoint by considering the clinical importance order and the relative timing of components. During the past few years, the win ratio has been applied to designs and analyses of some clinical trials. It is not uncommon that some clinical trials require a stratified analysis. In this talk, we present a stratified win ratio statistic, derive its estimate, and assess its statistical performance using simulation studies. We provide an estimate of the stratified win ratio illustrated with the generalized approach (Dong et al., 2016). Our simulations show that our proposed stratified win ratio performs similarly to the Mantel-Haenszel stratified odds ratio with asymptotic variance by Robins et al. (1986) for the situation when the win ratio reduces to the odds ratio, and our proposed stratified win ratio is preferred compared to the inverse variance weighted win ratio and unweighted win ratio particularly when the data are sparse. In addition, we suggest the inverse variance weighted approach (Cochran, 1954) to test the homogeneity of win ratios across strata as this method is used frequently in meta-analyses and a better test for the win ratio homogeneity is not available yet.