AI-derived ECG age gap: a novel predictor of mortality after CABG and PCI

Abstract

Abstract Background Risk stratification in patients undergoing coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI) is a cornerstone of clinical decision-making. Traditional predictors often fail to capture the full spectrum of mortality risk early and late after revascularisation procedures, especially in the elderly. Recent advances in Artificial Intelligence (AI) applied to 12-lead electrocardiograms (ECGs) enable a novel 'ECG age gap' (predicted ECG age minus chronological age), offering potential to refine mortality risk assessment beyond conventional tools. Purpose We evaluated whether the AI-derived ECG age gap independently predicts 120-day and 3-year mortality among patients undergoing CABG or PCI, adjusting for key demographic, clinical, and lifestyle factors. Methods A deep-learning model was developed to predict age from 12-lead ECGs (N=532,301; 500 Hz; 10-second duration). The network architecture and training process were as previously described, with modifications to the loss function ('dist loss') improving accuracy for extreme values in age prediction. Among 2,742 patients (1,244 CABG; 1,498 PCI), the ECG age gap was averaged over the six months preceding the intervention. Cox proportional hazards models (CPH) were constructed to assess the association between the ECG age gap and mortality at 120 days and 3 years. Models were adjusted for chronological age, sex, Charlson Comorbidity Index (CCI), obesity, and smoking (p<0.05). Results The performance of the model was determined using the mean absolute error. This metric showed that, on average, the predicted ages differed by 6.67 years from the actual ages. Mean chronological ages were 69.2±9.0 years (CABG) and 72.2 ± 10.8 years (PCI), with corresponding ECG ages of 69.1±9.5 and 71.3±10.1. Cox regression analyses demonstrated that each 1-year increment in the ECG age gap was significantly associated with increased mortality risk in both CABG and PCI cohorts (Hazard Ratios ranging from 1.04 to 1.06; p<0.05). An ECG age gap of +10 years raised mortality risk by 80% for CABG and 50% for PCI. These effects persisted in both the short-term (120 days) and long-term (3 years) analyses (Table 1, Figure 1). Conclusions In a large, real-world population, the ECG age gap emerged as a robust predictor of mortality following CABG and PCI, adding predictive value beyond traditional scores. Integrating AI-derived ECG biomarkers into routine clinical practice could personalize revascularisation decisions. In particular, the ECG age gap may help in deciding to what extent revascularisation remains appropriate in elderly or comorbid patients. Prospective trials are needed to translate these group-level observations into a clear cut-off for individualized clinical care and transform cardiovascular risk management.

Ähnliche Arbeiten