We welcome the data provided by Levin et al. Indeed, a strong motivation for the publication of our single-center experience1 was to encourage others to do the same and to open a dialogue on this important and as yet unanswered question.
As Levin and colleagues' data suggest, there was no hemorrhage risk in warfarin-treated patients, while our data suggest a high risk; therein lies quite a range of possibilities. To conclude anything based on either article would be flawed; both are limited by small sample sizes and are open to multiple biases (type I and type II errors). For example, the small number of patients in their registry who were taking warfarin and were treated with tPA could suggest that warfarin-treated patients were preferentially selected for intravenous thrombolysis in mild cases and treated with endovascular approaches if moderate or severe deficits were noted. An analysis accounting for National Institutes of Health Stroke Scale score would be helpful but difficult to determine in an administrative data set. As you rightly point out and we stated in our discussion, additional confounders including age and stroke subtype need to be included in large multivariable models. Thus, the final answer will not come from small series like ours but a large detailed database such as the Get With The Guidelines stroke registry.
Prabhakaran S, Chong JY. Warfarin Therapy Does Not Increase Risk of Symptomatic Intracerebral Hemorrhage in Eligible Patients After Intravenous Thrombolysis—Reply. Arch Neurol. 2011;68(1):135-138. doi:10.1001/archneurol.2010.324