Wide Variation in Reported Rates of Stroke Across Cohorts of Patients With Atrial Fibrillation

BACKGROUND: Oral anticoagulants decrease ischemic stroke rates in patients with atrial fibrillation (AF) but increase the risk of bleeding. For the average patient with AF, the threshold of annual ischemic stroke rate where the benefit of anticoagulation outweighs the bleeding risk (net clinical benefit) has been shown to be approximate to 1% to 2%. Guideline recommendations for oral anticoagulants in AF are based on the CHA(2)DS(2)-VASc stroke risk point scores, assuming that those scores translate to fixed stroke rates. However, the relationship between stroke point scores and annual stroke rates may vary substantially across populations. We sought to comprehensively assess the reported rates of stroke in patients with AF and the relationship of stroke rates to stroke risk point scores. METHODS: A systematic review of cohort studies and randomized controlled trials enrolled patients with nonvalvular AF not treated with oral anticoagulants. RESULTS: Of the 3552 studies screened, we identified 34 studies eligible for analysis. Overall stroke rates in cohort studies were highly heterogeneous (Q=5706.54, P<0.001; I-2 = 99.6%) and ranged from 0.45% to 9.28% per year, despite being of similar objective study quality. The mean North American stroke rate was less than one-third that of the mean European stroke rate (P<0.0001). However, a random effects regression indicated that between-study variability was not significantly accounted for by cohort region, prospective versus retrospective design, calendar year of study, or outcome event cluster. At a CHA(2)DS(2)-VASc score of 1, 76% of cohorts reported ischemic stroke rates <1% per year and only 18% of cohorts reported a stroke rate >2% per year. At a CHA(2)DS(2)-VASc score of 2, 27% of cohorts reported stroke rates below 1% per year, 40% reported stroke rates between 1 and 2% per year, and 33% reported stroke rates >2% per year. CONCLUSIONS: Substantial variation exists across cohorts in overall stroke rates and rates corresponding to CHA(2)DS(2)-VASc point scores. These variations can affect the point score threshold for recommending oral anticoagulants in AF. The majority of cohorts did not observe stroke rates that would indicate a clear expected net clinical benefit for anticoagulating AF patients with CHA(2)DS(2)-VASc scores of 1 or 2.