Automated abdominal aortic calcification scoring from routine bone density scans is independently linked to future atherosclerotic cardiovascular disease events, according to researchers. The study, which appears in the March edition of JACC: Advances, found that individuals with moderate or high machine learning–derived abdominal aortic calcification scores (ML-AAC24) had substantially higher risks of developing a first atherosclerotic cardiovascular disease (ASCVD) event. In contrast to participants with low scores, those with moderate calcification had an 80% higher risk of ASCVD, while those with high scores had nearly a threefold increase in risk. Higher relative hazard “Compared to individuals with low ML-AAC24, those with moderate and high ML-AAC24 would be at a substantially higher relative hazard for an incident (first-ever) ASCVD event in the next 5 years,” the authors wrote. “ML-AAC24 extent was also the second most important contributor when considering incident ASCVD.” Led by Marc Sim, PhD, from Edith Cowan University in Perth, Australia, the research team began by validating ML-AAC24 using lateral spine images generated during standard DXA scans, emphasizing the opportunity to extract cardiovascular risk data from an existing imaging workflow. While abdominal aortic calcification is a recognized marker of subclinical atherosclerosis, the team noted that its clinical uptake had been limited by the need for specialist expertise and labor-intensive scoring. UK Biobank study The research team also looked at the DXA images taken from 53,611 participants in the UK Biobank Imaging Study (mean age 65 years; 52% women). After eliminating patients with prevalent cardiovascular disease or missing data, 50,923 participants were followed for incident ASCVD events, including myocardial infarction and ischemic stroke. Further findings revealed that event rates rose progressively with an increasing ML-AAC24 burden. Over 4.1 years, ASCVD occurred in 1.7% of individuals with low scores, compared with 3.9% and 6.8% among those with moderate and high calcification. These relationships held after multivariable adjustment and were observed across coronary artery disease, myocardial infarction and stroke outcomes. Strong and robust gradient “We observed a strong and robust gradient of risk for incident ASCVD with increasing ML-AAC24 extent, independent of standard modifiable and nonmodifiable CVD risk factors,” the researchers noted. “This supports ML-AAC24 being a novel and clinically important marker for identifying individuals with high ASCVD risk.” The researchers added that the addition of ML-AAC24 to established risk models modestly improved predictive performance, supporting its potential role as a cardiovascular “risk enhancer,” analogous to coronary artery calcium scoring. The investigators noted that DXA-based assessment offers practical advantages, including minimal radiation exposure and broad clinical availability. “Assessment of ML-AAC24 using widely available bone density machines can identify high-risk individuals with subclinical ASCVD,” the authors concluded. “This provides a promising opportunity to alter their trajectory of disease.” Source: Sim M, Webster J, Smith C, et al. Automated abdominal aortic calcification scores and atherosclerotic cardiovascular disease in the UK Biobank Imaging Study. JACC Adv. 2026;5:102570. Image Credit: Rasi – stock.adobe.com