Biomarker-Based Model for Prediction of Ischemic Stroke in Patients With Atrial Fibrillation

BACKGROUND In patients with atrial fibrillation (AF) the risk of ischemic stroke is central to recommendations for stroke-prevention treatment. OBJECTIVES The authors evaluated the biomarker-based Age, Biomarkers, Clinical history (ABC)-AF-stroke risk score and developed a modified ABC-AF-istroke risk score for prediction of respectively total and ischemic stroke in patients with AF. METHODS In 26,452 patients with AF assigned to direct oral anticoagulants (DOACs) or warfarin, information on age, clinical history of stroke, and levels of N-terminal pro B-type natriuretic peptide and troponin were used for calculation of the ABC-AF-stroke score and the modified ABC-AF-istroke score. RESULTS During follow-up, there were 756 cases with stroke or systemic embolism (SEE) including 534 with ischemic stroke/SEE. The discrimination of total stroke/SEE was superior for the ABC-AF-stroke score, C-index (0.667 95% CI: 0.648-0.687), compared with 0.632 (95% CI: 0.612-0.652) for the ATRIA (Anticoagulation and Risk Factors in Atrial Fibrillation) and 0.614 (95% CI: 0.594-0.633) for the CHA2DS2-VASc score (P < 0.001 for both). The results were similar for ischemic stroke/SEE with C-index for ABC-AF-istroke 0.677 (95% CI: 0.654-0.700) compared with 0.642 (95% CI: 0.618-0.666) for the ATRIA and 0.624 (95% CI: 0.601-0.647) for the CHA2DS2-VASc score (P < 0.001 for both). The ABC-AF-stroke scores showed good calibration for total and ischemic stroke. Results were consistent in relevant subgroups. Decision curve analyses showed a net benefit concerning stroke-prevention decision thresholds. CONCLUSIONS The biomarker-based ABC-AF risk scores for the risk of total and ischemic stroke were well calibrated, showed better discrimination than clinical risk scores in predicting total and ischemic stroke, and provided meaningful decision support for stroke-prevention treatments in patients with AF.