Accuracy of incidental visual coronary artery calcium assessment compared with dedicated coronary artery calcium scoring

Introduction: CAC can be detected on routine chest computed tomography (CT) scans and may contribute to CVD risk estimation, but the accuracy of visual CAC scoring may be affected by the specialty of the interpreting radiologist and/or the use of contrast. Methods: The accuracy of visual CAC estimation on non-gated CT scans was evaluated at UT Southwestern Medical Center (UTSW) and Parkland Health and Hospital System (PHHS). All adults who underwent CAC scanning and a non-gated CT scan within 6 months were identified and the scores from the two CTs were compared overall and stratified by type of reader and whether contrast was used. Visual CAC categories of none, small, moderate, and large were compared to CAC ​= ​0, 1–99, 100–399, and ≥400, respectively. Results: From 2016 to 2021, 934 patients (mean age 60 ​± ​12 ​y, 43% male, 61% White, 34% Black, 24% Hispanic, 54% from PHHS) had both CT scans. Of these, 441 (47%) had no CAC, 278 (30%) small, 147 (16%) moderate, and 66 (7%) large CAC on non-gated CT. Visual CAC estimates were highly correlated with CAC scores (Kendalls tau-b ​= ​0.76, p ​< ​0.0001). Among those with no visual CAC, 76% had CAC ​= ​0 (72% of contrast-enhanced vs 85% of non-contrast scans, 88% of scans interpreted by CT radiologist vs 78% of those interpreted by other radiologist). In those with moderate-to-large visual CAC, 99% had CAC >0 and 88% had CAC ≥100, including 89% of those with contrast, 90% of those without contrast, 80% of those read by a CT radiologist, and 88% of those read by a non-CT radiologist. Discussion: Visual CAC estimates on non-gated CT scans are concordant with Agatston score categories from cardiac CT scans. A lack of visual CAC on non-gated CT scans may not be sufficient to “de-risk” patients, particularly for contrast-enhanced scans and those read by non-CT radiologists. However, the presence of moderate-to-large CAC, including on contrasted scans and regardless of radiologist type, is highly predictive of CAC and may be used to identify high-risk patients for prevention interventions.