An automated scoring system for breast arterial calcification detected on routine screening mammograms may be independently associated with increased risk of cardiovascular disease and all-cause mortality in women, according to a study.

In the study, published in in JACC: Advances, investigators analyzed the mammograms of 18,092 women (mean age 56.8 ± 11.4 years) who underwent screening between 2007 and 2016 at the University of California, San Diego Health. Breast arterial calcification (BAC) was present in 4,223 (23%) of the women and was quantified using an artificial intelligence (AI) algorithm that generated a score from 0 to 100.

The study population included women with prevalent cardiovascular disease (CVD) risk factors: diabetes (13%), hypertension (36%), and hyperlipidemia (40%). BAC was more prevalent among women who were older, Black, Hispanic, diabetic, hypertensive; those with a history of atherosclerotic cardiovascular disease (ASCVD) or chronic kidney disease (CKD); and those receiving statins and/or antihypertensive medications. BAC was less prevalent in current smokers (3.17% vs 5.1%, P < .001).

This retrospective cohort study included women aged 40 to 90 years who underwent screening digital mammography. BAC was quantified using a proprietary AI algorithm (cmAngio, CureMetrix) that analyzed four standard mammogram views per patient and assigned a score of 0 to 100, with BAC presence defined as a mean score ≥ 5.

Clinical data and outcomes were collected from electronic health records using ICD-10 codes. The primary outcome was all-cause mortality, with a composite of acute myocardial infarction, heart failure, stroke, and mortality as the key secondary outcome.

Cox proportional hazards regression was used to assess associations between BAC (as binary, continuous, and quartile variables) and outcomes, adjusting for age, race/ethnicity, smoking, blood pressure, cholesterol, diabetes, and history of CVD or CKD.

Over a median follow-up of 4.8 years (interquartile range [IQR] = 4.2 years) for mortality and 4.3 years (IQR = 4.3 years) for the composite outcome, women with BAC had significantly higher rates of mortality (7.8% vs 2.3%, P < .001) and a composite CVD outcome including myocardial infarction, heart failure, stroke, and mortality (12.4% vs 4.3%, P < .001) compared with those without BAC.

In multivariable analysis adjusting for traditional cardiovascular risk factors, BAC presence was associated with increased risk of all-cause mortality (adjusted hazard ratio [HR] = 1.49, 95% confidence interval [CI] = 1.33–1.68, P < .001) and the composite CVD outcome (adjusted HR = 1.57, 95% CI = 1.42–1.74, P < .001).

Each 10-point increase in the automated BAC score was independently associated with a higher risk of mortality (adjusted HR = 1.08, 95% CI = 1.06–1.11, P < .001) and the composite outcome (adjusted HR = 1.08, 95% CI = 1.06–1.10, P < .001). There was a graded increase in risk across BAC score quartiles:

• First quartile [score 1 to 25]: adjusted HR = 1.22 (95% CI = 1.06–1.41, P = .006) for mortality, adjusted HR = 1.26 (95% CI = 1.11–1.43, P < .001) for composite outcome
• Second quartile [score 26 to 50]: adjusted HR = 1.44 (95% CI = 1.13–1.85, P = .004) for mortality, adjusted HR = 1.74 (95% CI = 1.42–2.13, P < .001) for composite outcome
• Third quartile [score 51 to 75]: adjusted HR = 1.69 (95% CI = 1.33–2.14, P < .001) for mortality, adjusted HR = 1.83 (95% CI = 1.49–2.25, P < .001) for composite outcome
• Fourth quartile [score 76 to 100]: adjusted HR = 2.27 (95% CI = 1.81–2.85, P < .001) for mortality, adjusted HR = 2.30 (95% CI = 1.88–2.82, P < .001) for composite outcome.

The association between BAC and outcomes remained significant after excluding women taking statins (n = 3,947) and those with baseline ASCVD (n = 758). BAC appeared most predictive among younger women aged 40 to 59 years (mortality adjusted HR = 1.51, 95% CI = 1.22–1.87; composite outcome adjusted HR = 1.52, 95% CI = 1.25–1.85).

The study had several limitations, including its single-center, retrospective design, reliance on ICD codes for outcomes, and lack of cause-specific mortality data. Additionally, most of the participants identified as White, potentially limiting generalizability to other populations.

Conflict of interest disclosures from the authors can be found in the study.