Consecutive acute ischemic stroke patients without permanent atrial fibrillation were enrolled in a derivation sample (n = 294) or a validation sample (n = 155). We developed a score for predicting PAF by independent risk factors derived from a logistic regression analysis of the derivation cohort and validated the score in the external cohort.
Multivariate analysis in the derivation cohort identified 3 variables independently associated with PAF. We calculated a score from these variables (history of arrhythmia or antiarrhythmic agent use [yes, 3 points], left atrial dilation [≥40 mm, 1 point], brain natriuretic peptide [BNP, ≥50 pg/mL, 1 point; ≥90 pg/mL, 2 points; ≥150 pg/ml, 3 points], total score, 0-7). The iPAB score (identified by past history of arrhythmia or antiarrhythmic agent use, atrial dilation, and BNP elevation) predicted PAF in the derivation (c statistic, .90) and validation (.94) cohorts at levels statistically superior to other biomarkers and clinical scores. For a total score 2 or more, the sensitivity and specificity were 93% and 71%, respectively. For a total score of 4 or more, the corresponding values were 60% and 95%.
Our prospective study suggests that this simple risk score superior to other scores help clinicians predict PAF or identify good candidates for further evaluation to detect PAF.