LAUSR

In Response: We would like to thank Hsieh et al for their interest in our article and for their thoughtful comments.1 First, they pointed out an important issue on our data that stroke in patients with prior use of oral anticoagulants (OAC) might be more frequent because of large artery atherosclerosis than in those without prior OAC, because of the higher proportion of patients with hypertension, hyperlipidemia, or coronary heart disease in the prior OAC group. We agree that such bias could affect the interpretation of our results. Among the analyzed 5645 patients presented in our article, subtypes of ischemic stroke by the Trial of ORG 10172 in Acute Stroke Treatment were available in 5527 patients (97.91%).2 The prevalence of large artery atherosclerosis was 1.28% (14/1090) in patients with prior OAC and 1.28% (57/4437) in those without (P=0.99, χ2 test). Cardioembolism was seen in 85.69% (934/1090) of the prior OAC group and in 84.02% (3728/4437) of the no prior OAC group (P=0.17). Although the distribution of stroke subtypes was similar between the 2 groups, the bias pointed out by Hsieh et al might have increased the risk of recurrent ischemic stroke in patients with prior OAC compared with those without prior OAC. Second, they suggested the influence of the timing of diagnosis of nonvalvular atrial fibrillation (NVAF) on the presented results,3 while they reported that whether NVAF was known before or diagnosed after stroke was not an independent predictor for a composite outcome of ischemic stroke, intracranial hemorrhage, or death.4 In our data, there was an imbalance in the timing of NVAF detection relative to the index event between the 2 groups. NVAF known before the index event was more frequent in patients with recurrent ischemic stroke (63.68%, 121/190) than in those without such events (56.85%, 3101/5455; P=0.06). However, as described in our article, the higher risks for recurrent ischemic stroke in patients with prior OAC than in those without were seen in multivariable Cox models including the timing of NVAF detection as a covariate. In subgroup analyses, the higher risk for recurrent ischemic stroke in patients with prior OAC than in those without was observed in patients with NVAF known before the index event (adjusted hazard ratio, 1.56 [95% CI, 1.03–2.36]) but was not observed in those with NVAF detected after the index event (adjusted hazard ratio, 0.93 [95% CI, 0.22–0.39]; P for interaction=0.49).