Design: Prospective observational cohort (MESA), 10-year follow-up
Population: 6,098 US adults aged 45–79 without atherosclerotic disease
Primary outcome: Fatal and non-fatal atherosclerotic cardiovascular events
Effect: C statistic change +0.02 (95% CI 0.01–0.03); categorical NRI 0.095

Using the Multi-Ethnic Study of Atherosclerosis, investigators followed 6,098 US adults aged 45 to 79 without atherosclerotic disease for 10 years and asked how much a coronary artery calcium score improved the PREVENT risk equations. The analysis, published in JAMA on 6 October, matters because the 2026 US dyslipidaemia guideline now suggests selective calcium scoring at a PREVENT 10-year risk of 3% to under 10%.

Across everyone, adding calcium nudged discrimination up only slightly (C statistic from 0.73, change +0.02) and improved reclassification modestly. The value was concentrated in the borderline group (3% to under 5%): there, 10-year event rates were 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.

In practice that spread is what changes a statin conversation. A borderline patient with a score of zero sits close to low risk; one with a score above 100 sits in the intermediate or high band.

This is an observational cohort from US sites. Indian patients were not represented, and PREVENT has not been validated in South Asian populations.

Reserve calcium scoring for patients whose statin decision is genuinely uncertain, rather than ordering it routinely.
In a borderline-risk patient, a score of zero supports deferring statins and re-assessing later.
A score of 100 or more in a borderline-risk patient places their observed risk in the intermediate or higher range.
Note that PREVENT has not been validated in South Asian populations, who were not part of this cohort.

Why it matters

It gives the new guideline's calcium-scoring window real numbers to talk patients through.

The statistics, in plain English

A C statistic measures how well a model ranks who will have an event; going from 0.73 to about 0.75 is a small gain. The net reclassification improvement of 0.095 means about one in ten people were moved into a more accurate risk band. The borderline-group event rates are crude observed rates, not model estimates.