Identifying patients at low risk for atrial fibrillation after cryptogenic stroke: the PAVA score

Abstract

Abstract Aims Continuous cardiac monitoring with implantable loop recorders (ILRs) is recommended after cryptogenic stroke to detect occult atrial fibrillation (AF). We aimed to develop and validate a dedicated risk score for AF detection in patients undergoing ILR monitoring after cryptogenic stroke. Methods and results This multicentre study enrolled 1002 consecutive patients (mean age 66.0 ± 12.3 years, 57.9% male) who received an ILR after cryptogenic stroke. The primary outcome was a first episode of AF or atrial tachycardia (AT). A score was developed in a randomly split development cohort and validated internally, accounting for competing risks, missing data imputation, and hazard ratio-based weighting. Among the 668 patients in the development cohort, cumulative incidences of AF/AT at 1, 2, and 3 years were 20.3%, 28.9%, and 33.0%, respectively. Four independent predictors were identified: premature atrial contractions > 200/24 h on Holter-ECG (2 points), age >58–78 years (3 points) or > 78 years (4 points), significant valvular heart disease (1 point), and left atrial dilatation (2 points), forming the PAVA (premature atrial contractions, age, significant valvular heart disease, and left atrial dilatation) score. Discrimination was good in both the development and validation cohorts (c-index 0.73 and 0.79, respectively). The 3-year areas under the curve were 0.800 and 0.905, respectively. A PAVA score ≤ 2, observed in 23.6% of patients, yielded negative predictive values of 89.2% and 100% in the development and validation cohorts. ROC-curve analysis confirmed superior clinical utility over existing scores (Brown ESUS-AF [Brown embolic stroke of undetermined source-atrial fibrillation], HAVOC). Conclusion A PAVA score ≤ 2 demonstrates a high negative predictive value for AF/AT on ILR after cryptogenic stroke, offering a practical tool to optimize patient prioritization for ILR implantation.