Torn et al1 conclude that the target international normalized ratio (INR) for atrial fibrillation (AF) should be 3.0 to 3.4, rather than the standard INR range of 2.0 to 3.0,2 based on a Leiden Thrombosis Centre cohort observed from 1994 to 1998. We believe that their data do not support their conclusion.
Precisely comparing the efficacy of adjacent INR intervals demands large numbers of relevant outcome events. Even if one accepts all the events reported by Torn et al1 as relevant, the total number is too small. The confidence intervals of crude incidence rates across INR intervals from 2.0 through 3.5 highly overlap. We found no assertion that an INR range of 3.0 to 3.4 was statistically significantly different from lower INR intervals. Furthermore, as the authors acknowledge, patients with AF are anticoagulated to prevent ischemic strokes. Torn et al1 report 15 ischemic strokes, reflecting an extraordinarily low rate of 0.4% per year—far too few events to sustain the analysis. All major bleeding events were included, even though only intracranial hemorrhages (ICHs) have an impact similar to ischemic strokes, which anticoagulants are used to prevent. Indeed, ICHs account for 90% of all fatal and disabling events among anticoagulation-related hemorrhages in AF.3 Only 21 ICHs were observed—again, too few to sustain their analysis. Torn et al1 report that when outcomes are limited to life-threatening events, an INR from 2.0 to 3.0 is, in fact, optimal. Their report also fails to address potential confounding across INR categories.