A streamlined CMR-derived machine-learning model for estimating cardiovascular biological age: development and validation in the UK-Biobank and Multi-Ethnic Study of Atherosclerosis.
Authors
Affiliations (10)
Affiliations (10)
- School of Biomedical Engineering and Imaging Sciences, King's College London, London - United Kingdom.
- British Heart Foundation Centre of Excellence - King's College London United Kingdom.
- Institute of Management, Scuola Superiore Sant'Anna - Italy.
- Division of Cardiology, John Hopkins University School of Medicine, Baltimore, Maryland - United States of America.
- Department of Twin Research and Genetic Epidemiology, School of Life course and Population Sciences, King's College London, London - United Kingdom.
- School of Cardiovascular Medicine & Metabolic Medicine & Sciences, King's College London, London - United Kingdom.
- GKT School of Medical Education, King's College London, London - United Kingdom.
- Faculty of Public Health and Policy, The London School of Hygiene and Tropical Medicine, London - United Kingdom.
- Department of Radiology, Johns Hopkins Hospital, Baltimore, Maryland - United States of America.
- Center for Biosignal Research, University of California, San Francisco, California - United States of America.
Abstract
Current models predicting cardiovascular biological age rely on radiomics or complex large feature sets including T1 and strain. We developed and validated a machine learning-based cardiovascular biological age estimate (HeartAge) using cardiovascular-magnetic-resonance (CMR) phenotypes and assessed the prognostic value of its deviation from chronological age (HeartAge-gap) for cardiovascular outcomes and mortality. HeartAge was developed using gradient-boosting regression in 3,760 healthy UK-Biobank participants based on readily extractable CMR phenotypes. HeartAge-gap was defined as the difference between HeartAge and chronological age. The association of HeartAge-gap with prevalent cardiovascular conditions and composite cardiovascular outcome or all-cause mortality was tested in 31,784 UK-Biobank participants (64±7 years; 16,640 females) and validated in 897 Multi-Ethnic Study of Atherosclerosis (MESA) participants (60±10 years; 472 females) using logistic and Cox regression, respectively.Over a median 5.5-year follow-up (IQR:4.7-7.1), 2,316 (7.3%) and 363 (1.1%) participants experienced the composite cardiovascular outcome and all-cause mortality, respectively. Each one-year increase in HeartAge-gap, was associated with the composite cardiovascular outcome in females (HR:1.022, 95%CI:1.001-1.044, P=0.048) and males (HR:1.017, 95%CI:1.002-1.033, P=0.027) independently of chronological age and confounders including, body-mass-index, ischaemic heart disease, diabetes, and hypertension. In females only, increased HeartAge-gap predicted all-cause mortality (HR:1.061, 95%CI:1.007-1.118, P=0.027), regardless of chronological age. In female MESA participants only, increased HeartAge-gap predicted the cardiovascular outcome (HR:1.113, 95%CI:1.025-1.210, P=0.011) independently of chronological age and other confounders. A biologically older cardiovascular system was independently associated with adverse cardiovascular outcomes across both sexes. In females, advanced cardiovascular ageing also predicts all-cause mortality, irrespective of chronological age.