Calcium score sharpened PREVENT risk mainly at borderline risk

Design

Prospective observational cohort (MESA), 10-year follow-up

Population

6098 US adults aged 45-79 without ASCVD

Primary outcome

Prediction of fatal and non-fatal ASCVD (C statistic, reclassification)

Effect

C statistic +0.02 (95% CI 0.01-0.03); NRI 0.095

This analysis of the Multi-Ethnic Study of Atherosclerosis followed 6098 US adults aged 45 to 79 without cardiovascular disease for 10 years. It asked how much a coronary artery calcium (CAC) score adds to the PREVENT atherosclerotic cardiovascular disease (ASCVD) equations, which the 2026 US dyslipidaemia guideline uses to decide who gets a calcium scan. Published in JAMA on 6 October 2026.

Across everyone, the gain was small: the C statistic rose from 0.73 by 0.02, and net reclassification improved by about 0.10. The value lay in the borderline group (3% to under 5% 10-year risk), where events occurred in 1.9% with a score of zero, 3.9% with 1 to 99, 7.4% with 100 to 299 and 14.3% with 300 or more.

So a calcium scan is not worth ordering for everyone, but in the uncertain middle it can move a patient clearly up or down. A score of zero in a borderline patient supports holding off a statin; a high score supports starting one.

This is an observational US cohort, and PREVENT has not been validated in Indian populations, where risk tends to be underestimated at a given profile.

• Reserve calcium scoring for adults whose 10-year PREVENT risk is roughly 3% to under 10%
• Treat a score of 300 or more in a borderline-risk patient as a strong reason to discuss a statin
• Treat a score of zero as reassuring but not permanent; rescanning in 5 or more years is reasonable
• Remember that a zero score does not cover smokers, diabetes or a strong family history
• Use the score to support shared decisions with patients unsure about a statin

Why it matters

It defines where calcium scoring earns its cost under the new US guideline, and where it adds little.

Don't overread it

Better prediction is not proof that scan-guided treatment improves outcomes.

The statistics, in plain English

The C statistic measures how well a model ranks who will have an event; a rise of 0.02 is real but small. Net reclassification of about 0.10 means roughly one in ten people moved to a more correct risk group. The event rates by score are what make the scan useful in borderline patients.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.