Catheter ablation is a well-established therapy for atrial fibrillation. However, some patients are not responders, and there exists currently no method to efficiently select non-responding patients.

The aim of this study is to look for predictive parameters of catheter ablation failure by analyzing the 12-lead surface electrocardiogram (ECG), combined with clinical parameters in patients undergoing the first ablation of persistent atrial fibrillation (PersAF).

Two hundred and fifty-eight patients with persistent AF (183 men; mean age 60,1±9,8years) hospitalized for first catheter ablation were retrospectively recruited. A standard 12-lead ECG was acquired before the ablation procedure for each patient. Clinical parameters (CHA₂DS₂-VASc, LA volume, cardiomyopathy, left common trunk of pulmonary vein) and electrical parameters (f-wave amplitude and frequency) were analyzed and integrated in univariate and multivariate prediction models using a logistic regression technique.

During an average follow-up of 13.2±4.7 months, 129 patients were in sinus rhythm at 1 year of follow-up after ablation. A feature selection step was performed to select the most informative outcome predictor. A multivariate analysis combining clinical and electrical parameters obtained an AUC of 67.3, with sensitivity of 49.7% and positive predictive value of 52.3%. This model performed better than clinical and electrical parameters analyzed separately.

Two clinical parameters are associated with failure: LA volume and CHA₂DS₂-VASc score in univariate analysis. Combination of clinical and electrical parameters allowed for a better selection of non-responding patients. CHA₂DS₂-VASc score showed a significant correlation with predictive ablation failure, but further studies are needed to confirm that.