Comparison of ATRIA and CHA2DS2-VASc risk stratification schemes for the prediction of stroke in the individual patient with atrial fibrillation and the impact on treatment decisions

Abstract

Purpose: To compare the predictive ability of the currently recommended CHA2DS2-VASc ischaemic stroke risk score with the new ATRIA stroke risk score in patients with atrial fibrillation (AF). Methods: Patients with AF, not using warfarin, were assembled from the Clinical Practice Research Datalink (CPRD) database. Patients were followed from date of AF diagnosis until occurrence of ischaemic stroke, prescription of warfarin, death or the end of study. Independent predictors of ischaemic stroke were identified with a Cox proportional hazard model by stepwise backward selection. The c-index assessed the discriminative ability of the risk schemes. Net reclassification improvement (NRI) assessed net correct risk reclassification using ATRIA versus CHA2DS2-VASc, using published point score cut-offs. As correct stroke risk thresholds for low/moderate/high risk, 1% and 2% per year were used. Results: We included 60,594 patients. The overall stroke rate was 2.45% per year. Age and previous stroke were the strongest predictors of ischaemic stroke. Other independent predictors were hypertension (HR 1.25 CI 95%, 1.15-1.35) and diabetes (HR 1.27 CI 95%, 1.14-1.41). Vascular disease and heart failure were not significant predictors. For the full point scores, the c-index was 0.71 (CI 95%, 0.70-0.72) for the ATRIA score and 0.69 (CI 95%, 0.68-0.70) for the CHA2DS2-VASc score. The NRI was 0.38 for ATRIA compared to the CHA2DS2- VASc-score, resulting entirely from downward reclassification (Figure). (Figure presented) Conclusion: The ATRIA score had better discriminative ability than CHA2DS2- VASc. The CHA2DS2-VASc-score assigns most AF patients to the moderate and high risk categories, which could lead to overtreatment. In this community-based, low-risk cohort, the ATRIA score correctly reclassified patients as lower risk.