Sinyavskaya L, Schnitzer M, Renoux C, et al. - Given inverse probability of censoring weights (IPCWs) may decrease selection bias because of informative censoring in longitudinal investigations, but the links between predictors and censoring may vary across treatment groups in studies with an active comparator, thus, to illustrate this, researchers herein used the clinical example of anticoagulation therapy with warfarin or a direct oral anticoagulant (DOAC) in atrial fibrillation. They included a cohort of people starting an oral anticoagulant during 2010-2016. The hazard ratio of the composite of stroke, major bleeding, myocardial infarction, or death related to continuous use of warfarin vs DOACs, was estimated. The hazard ratio related to continuous therapy with warfarin vs DOACs adjusted with exposure stratified IPCWs was estimated to be 1.26 (95% confidence interval: 1.20, 1.33). The estimated hazard ratio differed by 15 % in favor of DOACs, when exposure-unstratified IPCWs were used. Overall, it was inferred that misspecification of censoring weights as well as biased estimate on comparative effectiveness and safety both may result owing to not accounting for the different links between the predictors and informative censoring across exposure groups.