AI-ECG age predicts carotid atherosclerotic plaque volume and progression

AI derived biological age from surface ECGs (AI ECG age) has shown prognostic value beyond chronological age. We hypothesized that AI ECG age reflects atherosclerotic burden as indicated by carotid plaque volume (PV).

We retrospectively analyzed 101 patients with cardiovascular disease or ≥1 risk factor from a prospective single-center cohort (NCT01895725) on carotid atherosclerosis progression. Carotid plaque volume (PV) was measured by standardized 3D ultrasound at baseline and at ∼12-month intervals (median follow-up 1091 days). AI ECG age was derived from standard 10-s 12-lead ECGs using a validated deep neural network. Associations between AI ECG age, Δage (AI ECG age - chronological age), and PV were assessed by correlation, regression, and ROC analyses. At baseline, 101 patients had both ECG and 3D ultrasound; follow-up PV was available for 95, 88, and 80 patients at 1, 2 and 3 years, respectively.

AI ECG age and chronological age correlated with PV (r = 0.54 and r = 0.48, both p < 0.001). In multivariable linear regression, AI ECG age was independently associated with PV (β = 6.95, 95 %CI: 2.88-11.01, p = 0.001), whereas chronological age was not (p = 0.120). Adding Δ age to a model with age, sex, lipid and inflammatory markers improved AUC from 0.77 to 0.82 and enhanced net reclassification (NRI = 0.48, p = 0.017). AI-ECG age predicted PV progression over time (β = 1.83, 95 %CI: 0.42 to 4.09, p = 0.042), independent of chronological age.

AI ECG age correlates more closely with carotid plaque burden than chronological age. Its divergence from chronological age independently predicts plaque progression, supporting AI ECG age as an accessible adjunct for vascular risk assessment.