Novel Calcium Score Accounting for Density and Location Ups MACE Prediction

A novel automated score that calculates both the dispersion and density of coronary artery calcium (CAC) seems to improve upon the predictive powers of the commonly used Agatston score for estimating major adverse cardiovascular events over time, new research suggests.

In a study of almost 1,000 patients who underwent cardiac computed tomography, an elevated CAC-dispersion and density (CAC-DAD) score was associated with more MACE events over 30 days than was an elevated Agatston score (74% vs 57%; P = 0.002). The data were published online last week in Circulation: Cardiovascular Imaging.

“CAC scoring has transformed the prevention of coronary artery disease in asymptomatic individuals, yet not all calcium scores are equal and patients with the same score may carry very different risks,” co-senior author Girish Dwivedi, MD, PhD (The University of Western Australia, Perth), told TCTMD in an email.

“Accurate assessment requires the analysis of individual plaque features, which is impractical for human readers but highly feasible using artificial intelligence (AI),” he explained. “Our method of CAC assessment (CAC-DAD score) enabled by an AI tool enables a more refined and personalized approach to cardiovascular risk prediction by addressing the limitations of conventional calcium scoring.”

Current clinical practice guidelines recommend measuring CAC as a noninvasive screening tool to assess cardiovascular risk (class IIa for patients at intermediate risk who are uncertain candidates for starting statin therapy), and it’s typically measured by the Agatston score in Hounsfield units. Previous efforts have attempted to overcome some known limitations with the Agatston score—including mitigating potential variability by CT scanner and adding the number of calcified vessels. The fully automated CAC-dispersion and density (CAC-DAD) score was developed in response to suggestions that factoring in CAC density and distribution could better predict the risk of adverse events.

Commenting for TCTMD, Matthew Budoff, MD (Harbor-UCLA Medical Center, Torrance, CA), said a new tool, or at least an updated score, is welcome. “We never thought [the Agatston score] was the best way to quantify the calcification,” he said. “If we can derive more information, we can better predict outcomes.”

While the evidence linking CAC density to outcomes is strong, Budoff continued, “the unique thing about this study is that nobody has put it into a new scoring system. . . . That conversion to an app, or that conversion to a program that can easily calculate it, just was the step that was missing. And it looks like they have accomplished that.”

MACE Prediction

In their study, led by Gavin Huangfu, MD (The University of Western Australia), the researchers included 961 patients (median age 67 years; 61% male) who underwent CT screening either as part of CVD assessment following symptoms (n = 338) or before elective noncardiac surgery (n = 623). The scans were analyzed by deep machine learning algorithms to automatically calculate both the Agatston (median 327) and CAC-DAD scores (median 2,176).

About half of patients had elevated CAC-DAD scores (≥ 2,050), and they were more likely to be older, male, and have hypertension and hyperlipidemia but less likely to use statins. Those with high CAC-DAD scores were significantly more likely to have elevated Agatston scores, but 3% of patients with an Agatston score of at least 400, and 20% of those with a score of at least 100, had a low CAC-DAD score. In total, 8% of patients had an Agatston score of 0 and all had corresponding 0 CAC-DAD scores.

Over a median follow-up period of 30 days, MACE (defined as nonfatal MI or cardiovascular mortality) were recorded in 61 patients. When patients were stratified by CAC-DAD score quartiles, those with the highest score had a significantly higher probability of MACE compared with those with the lowest (10.4% vs 3.3%; P = 0.002). This relationship was not observed by quartiles of the Agatston score.

Advancing medicine means challenging the status quo. Girish Dwivedi

On multivariate analysis, the CAC-DAD score (HR 2.57; 95% CI 1.43-4.61), along with age, statin use, and diabetes predicted MACE, but not the Agatston score. Adding the CAC-DAD score in a predictive model comprising both demographic factors and the Agatston score improved its discriminatory power (C statistic 0.61 to 0.66; P = 0.008).

“Advancing medicine means challenging the status quo,” Dwivedi said. “While several methods exist to measure coronary calcification, the Agatston score remains the clinical standard and most widely used tool. Demonstrating that the CAC-DAD score offers incremental or superior accuracy could transform how we assess and manage patient risk, potentially reshaping cardiovascular care.”

Validation Next

Tim Leiner, MD, PhD (Mayo Clinic, Rochester, MN), who was not involved in the study, told TCTMD he’s excited about the new score but would like to see it validated prospectively, especially in low- to intermediate-risk patients. “This is a, let’s say, relatively small study,” he said, adding that some patients were at intermediate-to-high risk “where calcium scoring is not typically used.”

If further analysis “yields the same signal, you could start looking into how we can further refine it, by, for instance, trying to come up with a vendor-agnostic score,” Leiner said.

Budoff agreed, suggesting using either the MESA or Framingham datasets to validate the novel score.

Dwivedi confirmed that his team has already begun validating the CAC-DAD score in “larger international cohorts.”

One potential issue Budoff foresees, however, is the difference in magnitude between the Agatston and CAC-DAD scores. “The biggest problem we’re going to have is that the scoring system is on a much different scale,” Budoff said. “That going to be a bit of a problem [for] people who are looking to compare their old score to their new score, because their score is going to go from 400 to 2,000 just because of the scoring system.”

While serial calcium scanning isn’t widely advocated, this likely won’t affect too many people, but it’s something for physicians to be aware of should the paradigm shift, he advised.