A Comprehensive Model to Predict Atrial Fibrillation in Cryptogenic Stroke: The Decryptoring Score

Abstract

Objetive: Cryptogenic stroke (CS) represents up to 30% of ischemic strokes (IS). Since atrial fibrillation (AF) can be detected in up to 30% of CS, there is a clinical need for estimating the probability of underlying AF in CS to guide the optimal secondary prevention strategy. The aim of the study was to develop the first comprehensive predictive score including clinical conditions, biomarkers, and left atrial strain (LAS), to predict AF detection in this setting.

Methods: Sixty-three consecutive patients with IS or transient ischemic attack with ABCD2 scale ≥ 4 of unknown etiology were prospectively recruited. Clinical, laboratory, and echocardiographic variables were collected. All patients underwent 15 days wearable Holter-ECG monitoring. Main objective was the Decryptoring score creation to predict AF in CS. Score variables were selected by a univariate analysis and, thereafter, score points were derived according to a multivariant analysis.

Results: AF was detected in 15 patients (24%). Age > 75 (9 points), hypertension (1 point), Troponin T > 40 ng/L (8.5 points), NTproBNP > 200 pg/ml (0.5 points), LAS reservoir < 25.3% (24.5 points) and LAS conduct < 10.4% (0.5 points) were included in the score. The rate of AF detection was 0% among patients with a score of < 10 and 80% among patients with a score > 35. The comparison of the predictive validity between the proposed score and AF-ESUS score resulted in an AUC of 0.94 for Decryptoring score and of 0.65 for the AF-ESUS score(p < 0.001).

Conclusion: This novel score offers an accurate AF prediction in patients with CS; however these results will require validation in an independent cohort using this model before they may be translated into clinical practice.

Keywords: Atrial fibrillation; Cryptogenic stroke; Left atrial strain; Natriuretic peptides.