External validation of the Antwerp score: a novel score for predicting responders to catheter ablation of atrial fibrillation in patients with depressed left ventricular function

Abstract

Abstract Introduction Little exists to identify patients with heart failure (HF) and atrial fibrillation (AF), who will benefit from pulmonary vein isolation (PVI). The Antwerp score was proposed as a new prediction model to identify patients with recovery left ventricular ejection fraction (LVEF) after PVI for AF. Purpose To externally validate the Antwerp score in a large cohort of patients undergoing PVI. Methods Patients with reduced LVEF (< 50%) scheduled to undergo PVI at our hospital were retrospectively analyzed. These were then classified as responders or non-responders to PVI according to the "Universal definition of HF". For each patient, the Antwerp score was calculated, and the performance of the score in predicting LVEF recovery after PVI was evaluated using receiver operating characteristic (ROC) curves and calibration plots. Results Between May 2010 and January 2022, 1665 patients underwent CA for AF at our institution. Of these, 208 patients (median age 63 [IQR 54-69] years, 19% women) had reduced LVEF at presentation and were included in the final analysis (27% with paroxysmal and 73% with persistent AF). The median LVEF at baseline was 43% [IQR 38-46]. At 30 [IQR 20-34] months follow-up after the index procedure, 161 patients (77%) were responders, and 47 patients (23%) were non-responders. The median time to LVEF recovery in responders was 8 [IQR 3-22] months. Between responders and non-responders, differences in AF type were not significant (p = 0.99), while QRS width, absence of known HF etiology and LAVI were significant (p<0.001, p<0.001 and p=0.03). After adjusting for difference in baseline characteristics and procedural outcomes between responders and non-responders, only a QRS duration<120 ms (p=0.002) and absence of known etiology (p<0.001) remained statistically significant with LVEF recovery (Figure). The Antwerp score had an area under the curve (AUC) of 0.75 (95% CI 0.67-0.83). A model containing QRS<120 ms and absence of known etiology performed similarly to the complete Antwerp model (Likelihood Ratio p=0.77). Conclusion The Antwerp score has a moderate ability to identify patients with the recovery of LVEF after PVI in an external population, showing modest rule-in and rule-out capabilities. QRS width and absence of known HF etiology were confirmed as excellent predictors of LVEF recovery.Figure