Predicting decline in left ventricular function after new-onset left bundle branch block
- PMID: 41265768
- DOI: 10.1016/j.hrthm.2025.11.026
Predicting decline in left ventricular function after new-onset left bundle branch block
Abstract
Background: Left bundle branch block (LBBB) has been associated with an increased risk of incident left ventricular systolic dysfunction. Identifying high-risk patients early is challenging but important for timely management.
Objective: This study aimed to identify predictors of significant left ventricular ejection fraction (LVEF) decline in patients with LBBB using clinical, echocardiographic, and electrocardiographic (ECG) data.
Methods: A retrospective cohort of 769 patients with newly diagnosed LBBB and preserved LVEF (≥50%) was analyzed. Univariable and multivariable Cox proportional hazards regression analyses were used to identify predictors of ≥20% LVEF decline on follow-up echocardiography as the primary outcome. A risk prediction model was constructed using pooled estimates.
Results: Over a median follow-up of 4.9 years (interquartile range 2.3-8.4), 242 patients (31.5%) experienced an LVEF decline of ≥20%. In multivariable Cox regression, 2 artificial intelligence-derived ECG scores previously developed to detect LV systolic and diastolic dysfunction were independently associated with LVEF decline (hazard ratio 1.01, per 1% probability increase for low LVEF; P = .005; hazard ratio 1.25, per predicted diastolic dysfunction grade, P = .005, respectively). The final model demonstrated modest discriminative ability with a C-index of 0.615, improving to 0.65 after adjusting for baseline LVEF and further to 0.70 in a landmark analysis of 1-year follow-up.
Conclusion: Artificial intelligence-derived ECG markers of LV systolic and diastolic dysfunction independently predicted future LVEF decline in patients with new-onset LBBB. Echocardiographic parameters may also enhance risk stratification. This predictive framework could be used to support monitoring and early intervention strategies in patients with LBBB.
Keywords: Artificial intelligence; Cox regression; Electrocardiogram; Left bundle branch block; Left ventricular ejection fraction; Risk stratification.
Copyright © 2025 Heart Rhythm Society. Published by Elsevier Inc. All rights reserved.
Conflict of interest statement
Disclosures Drs Siontis, Attia, and Friedman are inventors of AI-ECG algorithms, which have been licensed by Mayo Clinic to Anumana Inc with potential for commercialization. Dr Siontis reports research funding from Anumana Inc not related to this work.
LinkOut - more resources
Full Text Sources
