A Multidomain Atrial Damage Index for Atrial Fibrillation Recurrence

Leire Moriones1, Blas Echebarria2, Javier Ibero3, Ignacio Garcia-Bolao3, Susana Ravassa4, Pablo Lamata5, Jean R. Bragard6
1Universidad de Navarra, 2Universitat Politècnica de Catalunya, 3Clinica Universidad de Navarra, 4CIMA, Universidad de Navarra, 5King's College London, 6University of Navarra


Abstract

Background and Aim: AF recurrence after ablation is a significant clinical challenge driven by heterogeneous atrial remodeling not fully captured by single-domain predictors. Current models often evaluate structural or electrical parameters in isolation, limiting predictive accuracy and interpretability. We propose a multidomain Atrial Damage Index (ADI 4D) to provide a structured clinical phenotyping framework and evaluate its association with one-year AF recurrence.

Methods: In a retrospective cohort of 143 patients undergoing first-time AF ablation, ADI 4D (composite score 0–8) integrated four domains: electrical substrate (Q2 healthy tissue >0.5 mV via high-density mapping), geometric remodeling (CT-derived volume tertiles), metabolic status (diabetes), and temporal status (AF type). Patients were stratified into three clinical stages representing increasing combined remodeling burden: Stage 1 (0–2), Stage 2 (3–4), and Stage 3 (>= 5).

Results: Recurrence exhibited a non-monotonic distribution across stages: 26.2% in Stage 1 (n=17/65), 50.9% in Stage 2 (n=28/55), and 34.8% in Stage 3 (n=8/23). While overall predictive performance was modest (ROC-AUC 0.58; OR 1.23, 95% CI 1.02–1.49), ADI 4D successfully identified highly vulnerable substrates. Notably, the intermediate E1+G2 phenotype—characterized by early electrical alteration with advanced geometric dilation—yielded an 80% observed recurrence rate (n=5).

Conclusions: The ADI 4D enables interpretable, multidomain phenotyping of AF patients. Our findings reveal a non-linear risk structure driven by complex domain interactions rather than isolated structural burden. This proof-of-concept system highlights specific high-risk intermediate phenotypes that warrant further validation in larger prospective multi-center cohorts.