Artificial intelligence derived ECG-age as a biomarker of mortality: a systematic review and meta-analysis

Electrocardiogram (ECG)-age, an age estimate derived from ECGs using artificial intelligence (AI), has emerged as a potential biomarker of cardiovascular ageing. The difference between ECG-age and chronological age, known as delta-age, has been proposed as a predictor of adverse outcomes, but its prognostic value has not yet been systematically established.

To assess the prognostic value of AI-derived ECG-age and delta-age in predicting mortality.

We conducted a systematic review and meta-analysis of retrospective cohort studies reporting associations between ECG-age or delta-age and clinical outcomes. Databases searched included LILACS, Scielo, MEDLINE, ScienceDirect, EMBASE, CENTRAL, CINAHL, Web of Science and Scopus, from inception to 1 May 2025. Eligible studies included participants aged ≥16 years and reported outcomes such as all-cause mortality, cardiovascular mortality, stroke, or atrial fibrillation. Two reviewers independently screened articles and extracted data. A random-effects meta-analysis (Hartung-Knapp method) was performed. The meta-analysis was registered in PROSPERO (CRD420251042467).

Ten retrospective cohort studies were included in the qualitative synthesis, with prognostic analysis populations ranging from 2,183 to 234,036 individuals. AI-derived ECG-age was calculated from 12-lead ECGs using deep learning models. Most studies utilized ECG waveforms, and only one analyzed ECG images. Across studies, a higher delta-age (≥6–10 years) was consistently associated with an increased risk of adverse outcomes, while four studies reported a statistically significant protective association in lower delta-age groups. Most studies adjusted for age, sex, and comorbidities. Seven studies assessed the association between delta-age and mortality; three of these included external validation cohorts, resulting in 11 datasets analysed. The pooled hazard ratio (HR) for all-cause mortality in individuals with higher delta-age was 1.83 (95% CI: 1.45–2.32), with substantial heterogeneity (I² = 82.9%).

AI-derived ECG-age, particularly when significantly higher than chronological age, is associated with an increased risk of mortality. These findings support its potential role as a novel, non-invasive biomarker of cardiovascular ageing. Prospective studies are needed to confirm its clinical utility and to establish its role in individual risk stratification.

Meta-analysis of all-cause mortality