Clustering of Cardiovascular Risk Factors in Midlife Forms Trajectories That Affect Heart Aging Biomarker and Incidence of Cardiovascular Disease

Objective To (i) identify multivariable longitudinal mid-life patterns of systolic and diastolic blood pressure, body mass index and non-HDL cholesterol and (ii) assess whether ECG-derived heart age acceleration (δ-age) mediates their associations with incident myocardial infarction (MI), stroke and all-cause mortality. Patients and Methods We analysed 5244 participants from the observational cohort study, Tromsø Study, who were free of MI or stroke at baseline (Tromsø7, 2015-2016) who attended at least three surveys (Tromsø3-7, 1986-2016). Age-standardised risk factor z-scores were modelled with quadratic B-splines and grouped using K-means clustering to identify joint multivariable life-course risk profiles. δ-age was derived from 12-lead ECGs using a validated neural network. Natural-effects models decomposed hazard ratios (HR) between clusters and clinical events (follow-up to 2021) into direct and δ-age mediated components. Results Four distinct clusters were identified: stable low-risk (cluster 1, 30.6%), dyslipidemia (cluster 2, 25.6%), hypertension (cluster 3, 28.1%) and high-risk trajectory (cluster 4, 15.7%). Each 1-SD increment in δ-age was associated with 27-30% of excess MI, stroke and death independent of age and sex. Clusters 3 and 4 carried the greatest excess risk of clinical events, whereas mediation by δ-age was modest (4–8% of total effect for MI and stroke) and not statistically significant. Conclusion Joint multivariable life-course clustering of conventional risk factors identified clinically distinct long-term cardiovascular risk phenotypes. Although δ-age adds independent prognostic information, it explained little of the excess risk associated with chronic hypertension, dyslipidemia or obesity, indicating it functions primarily as a parallel marker not a causal mediator.