Novel insights into cardiovascular ageing in the general population using machine learning

Abstract

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): British Heart Foundation (RG/19/6/34387, RE/18/4/34215). Background Population ageing is a global trend and places an increased burden on healthcare resources, predominantly through cardiovascular morbidity and mortality. Ageing can be detected in the cardiovascular system as a decline in structure and function through cardiac magnetic resonance (CMR) imaging. This decline occurs at both organ and cellular levels, modulated by genetic, oxidative, inflammatory and metabolic stresses. Effect sizes and mechanisms of cardiovascular ageing are currently unknown; moreover, a robust biomarker is lacking. Purpose To develop a cardiovascular ageing model using CMR features to quantify an individual’s deviation ("age-delta") from healthy ageing and explore potential mechanisms contributing to the process. Methods We used data from the UK Biobank (UKB), a population-based cohort study of over 500,000 participants aged 40–69 recruited between 2006–2010. We studied 39,559 participants that had a CMR scan and used a machine learning model to estimate a "cardiovascular age" in 5065 healthy individuals from imaging traits including cardiac volumes, diastolic function, aortic distensibility and T1 mapping of fibrosis (Figure 1a). We applied this trained healthy ageing model to predict cardiovascular age in 34,147 new individuals (Figure 1b) and computed their age-deltas, which we used in a linear regression model to calculate the effect-sizes of selected diseases and lifestyle factors. We next performed a genome wide association study on 29,506 subjects to identify significant single nucleotide polymorphism (SNP) associations and computed a polygenic risk score in 373,948 independent genotyped participants of UKB to further explore phenotypic associations in a phenome wide association study. Results Hypertension (+1.58 years), diabetes (+0.74 years), smoking (+0.031 years/pack year) and alcohol (+0.015 years/gram per day) were significantly associated with adverse cardiovascular ageing (Figure 2a). Heart rate, blood pressure, inflammatory biomarkers and alkaline phosphatase were also significantly correlated with accelerated cardiovascular ageing, whereas telomere length, greater lung function, fat-free mass and basal metabolic rate were significantly correlated with attenuated ageing. GWAS revealed variants that contribute to myocardial contractility (Titin), arterial function (Elastin), inflammatory states (PLCE1, TREM2) and calcium signalling (MICU3) (Figure 2b). Conclusion Our study introduces a general population-derived CMR-based biomarker for cardiovascular ageing, which may in future be used to summarise an individual’s trajectory of ageing. The associations established with risk factors and physical measures could enable personalised cardioprotective interventions, complemented by novel therapies targeting mechanistic pathways discovered by our genetic associations.