# Development and validation of a predictive model for left ventricular diastolic improvement following catheter ablation in atrial fibrillation with diastolic dysfunction: a retrospective analysis

## TL;DR

This study developed a model to predict improvement in heart function after ablation in patients with atrial fibrillation and diastolic dysfunction.

## Contribution

A novel predictive model for post-ablation diastolic improvement in AF patients with LVDD is introduced.

## Key findings

- 49.6% of patients showed improvement in LVDD after ablation.
- The model achieved an AUC of 0.786 in training and 0.756 in validation.
- The model includes predictors like LVDD grade, LVMI, LVEF, and stroke history.

## Abstract

The impact of catheter ablation on left ventricular diastolic function in patients with atrial fibrillation (AF) and left ventricular diastolic dysfunction (LVDD) remains poorly characterised. There are no robust tools available to predict improvement in LVDD post-ablation.

The retrospective study enrolled 141 patients diagnosed with AF and LVDD undergoing ablation. LVDD improvement was defined as a reduction of at least one grade in LVDD grade (LVDDG) within 12 ± 3 months after ablation. The predictive model was constructed using a logistic regression with 10-fold validation with stratification. The performance of the model was evaluated using ROC analysis (AUC), calibration plots, and decision curve analysis (DCA). An exploratory sensitivity analysis applying the updated 2025 ASE recommendations for LVDD assessment was additionally performed.

49.6% (70/141) of patients showed improvement in LVDD. The predictive model incorporated four independent predictors: higher baseline LVDDG, lower left ventricular mass index (LVMI), lower left ventricular ejection fraction (LVEF), and absence of a history of stroke. The AUC was 0.786 (95% CI 0.760–0.809) in the training cohort and 0.756 (95% CI: 0.677–0.835) in the validation cohort, with acceptable calibration and net clinical benefit across a restricted range of threshold probabilities. In the sensitivity analysis using the redefined LVDD endpoint, discrimination remained comparable (AUC 0.783, 95% CI 0.703–0.856).

A novel model integrating LVDDG, LVMI, LVEF and stroke history effectively predicts post-ablation LVDD improvement in AF patients with LVDD, aiding patient selection for ablation and guiding adjunctive therapies.