We read with great interest the article by Prabhakaran et al.1 We noted a significant imbalance in cardioembolic stroke between the groups (76.9% for the warfarin sodium group vs 38.3% for the controls) that was not included in the multivariable model because the overall effect of TOAST classification2 on the exploratory analysis failed to meet their prespecified inclusion criteria. Further, the small sample size of warfarin-treated patients (n = 13) precludes assessment of the association between international normalized ratio and outcome within a robust multivariable model. It is likely that the reported rate of post–tissue plasminogen activator (EPA) intracerebral hemorrhage associated with premorbid warfarin use was confounded by a higher propensity for hemorrhagic transformation in cardioembolic stroke.3,4