Artificial intelligence-enhanced electrocardiography derived body mass index as a predictor of future cardiometabolic disease

Abstract Background Obesity, a key risk factor for cardiometabolic diseases, is poorly evaluated by body mass index (BMI), which doesn't account for visceral fat and fat distribution. Given that obesity-related cardiac changes can manifest in surface ECG changes, we hypothesised that an AI-ECG model could predict BMI, and the discrepancy between AI-predicted and actual BMI (delta-BMI) could indicate cardiometabolic health. We also sought to understand the biological mechanisms contributing to the AI-ECG-derived delta-BMI. Methods The AI-ECG model was developed using a data split of 60/10/30% in a USA secondary care dataset of 512,950 ECGs and externally validated in the UK Biobank (N = 42,386), employing a residual neural network architecture. Model performance was evaluated using Mean Absolute Error (MAE), Pearson correlation coefficient (r), and coefficient of determination (R2). The study aimed at identifying incident cardiometabolic diseases, using Cox regression analyses adjusted for BMI, age, and sex. To assess the underlying biological associations of delta-BMI, phenome-wide, genome-wide, metabolome-wide, and proteome-wide association studies were performed. Results In the test set, the model achieved an MAE of 3.95 (3.93–3.97), r of 0.65 (0.65-0.66), and R2 of 0.43 (0.42-0.43). In the UKB, the model achieved an MAE of 2.94 (2.91–2.96), r of 0.62 (0.62–0.63) and R2 of 0.39 (0.38-0.40). The top tertile of delta-BMI showed an increased risk of future cardiometabolic disease (test set: HR 1.15, p<0.001; UKB: HR 1.58, p<0.001) and diabetes mellitus (test set: HR 1.25, p<0.001; UKB: HR 2.28, p<0.001) after adjusting for covariates including measured BMI. Phenotypic profiling highlighted associations between delta-BMI and cardiometabolic diseases, anthropometric measures of truncal obesity, and pericardial fat mass. Metabolic and proteomic profiling associates delta-BMI positively with valine, lipids in small HDL, syntaxin-3, and carnosine dipeptidase 1, and inversely with glutamine, glycine, colipase, and adiponectin. A genome-wide association study revealed associations with regulators of cardiovascular/metabolic traits. These results highlight delta-BMI's potential as a non-invasive, ECG-based biomarker for cardiometabolic risk stratification. Conclusion Our study demonstrates the effectiveness of an AI-enhanced ECG model in accurately predicting BMI, validated across diverse populations. The introduction of delta-BMI as a predictor of cardiometabolic disease offers valuable additional prognostic insights beyond traditional BMI assessments. Our exploration of biological pathways, spanning phenotypic, genotypic, metabolomic, and proteomic associations, contributes to a comprehensive understanding of the association between delta-BMI and cardiometabolic disease. The clinical implications could extend to enhanced screening strategies, personalised risk assessment, and motivation for lifestyle interventions.