BACKGROUND The prognostic efficacy of quantitative platelet activity in atrial fibrillation (AF) remains unclear.

OBJECTIVE The purpose of this study was to evaluate the platelet count (PLT) as a prognostic indicator in patients with nonvalvular AF.

METHODS Data on 10,978 patients with nonvalvular AF were retrieved from a prospective registry of a single medical center in Korea. Cumulative risk for stroke and bleeding events were compared between patients with normal PLT (n = 8322), mild thrombocytopenia (n = 1791), and moderate to severe thrombocytopenia (n = 865) after propensity score matching. Prediction models for stroke were derived by conventional risk factors (model 1) and by combining PLT with model 1 (model 2), and model performance was assessed by area under the receiver operator characteristics curve (AUC).

RESULTS During the follow-up period, 7.3%, 7.0%, and 4.5% had stroke and 7.6%, 10.8%, and 17.2% had bleeding events in the normal PLT, mild, and moderate to severe thrombocytopenia groups, respectively. Compared to the normal PLT group, the moderate to severe thrombocytopenia group showed a lower risk of stroke (hazard ratio [HR] 0.57; 95% confidence interval [CI] 0.40-0.80; P = .002). A reverse relationship was found between PLT and bleeding risk (moderate to severe thrombocytopenia: HR 2.19; 95% CI 1.77-2.70; P <.001; mild thrombocytopenia: HR 1.43; 95% CI 1.18-1.73; P <.001). Compared to model 1, model 2 showed significant improvement in risk prediction (AUC 0.628 vs 0.644; P <.001).

CONCLUSION A lower PLT was associated with a lower risk of stroke and a higher risk of bleeding events. PLT combined with conventional risk factors showed significant improvement in prediction for stroke.