Abstract
BACKGROUND: Mild aortic stenosis (AS) is associated with adverse outcomes but is incompletely defined. OBJECTIVES: The purpose of this study was to examine the epidemiology of AV function measured without clinical indications. METHODS: We developed a deep learning model to measure aortic valve (AV) area, peak velocity, and mean gradient in velocity-encoded cardiac magnetic resonance imaging in 62,902 UK Biobank participants. Study findings were externally validated in NEDA (National Echo Database Australia), a clinical cohort of 365,870 people. RESULTS: From measuring reference ranges of AV function in a healthy subcohort (n = 41,859), we observed a natural boundary between normal and abnormal AV hemodynamics (>95th percentile) that we refer to as "mild AS(proposed)": peak velocity >1.65 m/s, mean gradient >4.9 mm Hg, or aortic valve area <2.1 cm(2) (men) or <1.7 cm(2) (women). In the full cohort, 3,676 (5.8%) participants met these novel criteria; the HR for a subsequent AV replacement for each severity category was 31.7 (mild AS(proposed)), 522.4 (moderate AS), and 3,057.4 (severe AS), all P < 0.001. Over a mean 3.9 years of follow-up, those with mild AS(proposed) also had a higher risk of atrial fibrillation (110 events; HR: 1.86; P = 1.4 × 10(-9)) and heart failure (70 events; HR: 2.37; P = 5.9 × 10(-11)) compared with those without AS. In NEDA, the 101,335 participants with mild AS(proposed) identified with echocardiography using the same cardiac magnetic resonance imaging-defined criteria had increased all-cause mortality (HR: 1.25; 95% CI: 1.24-1.27). CONCLUSIONS: We report a large-scale study of AV hemodynamics and identify a population threshold between normal and abnormal AV function. Mild AS, as defined by the proposed criteria, was linked to adverse outcomes in the UK Biobank and in NEDA.