Abstract 4145964: Association of Echocardiographic Traits with Deep Neural Network-Predicted Age from ECGs

Background: Electrographic age (ECG-age), obtained by the estimation of age through deep neural networks, has been useful in predicting cardiovascular outcomes. Given the wide availability of echocardiograms and known associations of metrics with age, we sought to amplify the applicability of ECG-age, by exploring the association of echocardiographic traits with ECG-age. Research Questions/Hypothesis: We hypothesized that accelerated aging is associated with adverse cardiac remodelling measured by echocardiography. Conversely, decelerated aging is associated with improved levels of these echocardiographic traits. Goals/Aims: We aimed to ascertain the association of echocardiographic characteristics with ECG-age in the Framingham Heart Study (FHS). Methods/Approach: We obtained echocardiographic data and studied seven echocardiographic traits, including mean peak velocity E/A - ratio of early to late peak velocities, mean diastolic velocity time integral, left ventricular (LV) wall thickening, left atrial diameter, aortic root diameter LV fractional shortening, and LVmass. The predicted ECG-age was adjusted for chronological age, and the Δage was defined as the difference between the adjusted predicted ECG-age and chronological age. Participants were classified into accelerated, normal, or decelerated aging based on their Δage being within, higher, or lower than the mean absolute error. The association of each echocardiographic trait with Δage was tested using linear regression models, adjusting for age, sex, and cohort. In the secondary analysis, participants were classified into age groups using logistic regression models and adjusted for the same covariates as above. Results/Data: Our study included 9077 participants from the FHS (mean age 50±15 years, 53.6% women). The mean absolute Δage was 6.3 years. Accelerated aging was associated with increased LV wall thickness, left atrial diameter, aortic root diameter, and LV mass. Similarly, decelerated aging was associated with decreased LV wall thickness, left atrial diameter, aortic root diameter, and LV mass. Conclusion(s): Our study provides insight into the anatomical and functional factors underlying ECG-age. Both accelerated and decelerated aging assessed by ECG-age are associated with cardiac remodeling measured by echocardiography, and found changes in LV wall thickness, left atrial diameter, aortic root diameter, and LV mass that correlates with ECG-age.