Comparing predictive risk to actual presence of coronary atherosclerosis on coronary computed tomography angiography

Background: There is limited data showing the predictive accuracy of traditional cardiovascular risk scores (CVRS) to predict asymptomatic coronary artery disease (CAD) determined by coronary computed tomography angiography (CCTA). Methods: Asymptomatic individuals without known CAD undergoing a screening CCTA and sufficient data to calculate their CVRS, were extracted retrospectively. Atherosclerosis was extracted using natural language processing of the CCTA report, including the coronary artery calcium score (CACS) and the extent and severity of CAD. Absence of atherosclerosis was defined as both zero plaque and zero CACS, and atherosclerosis was defined as low, moderate, or extensive by location and extent of plaque-burden. CVRS was categorized as high (>15 %), moderate (10-15 %), low (1-9 %) and "zero" (<1 %) risk. Results: 828 individuals (median age 58.6, IQR = 52.0, 65.3 years, 57 % male) met inclusion criteria, and a zero, low, moderate, and high CVRS was identified in 13, 483, 113 and 219 individuals (8 %, 49 %, 74 %, 66 % male), respectively. Predominantly low plaque-burden atherosclerosis was detected in 548 scans (67 % male). However, of the 137 males and 68 females with extensive atherosclerosis, 47 (34 %) and 38 (56 %) respectively had low CVRS classification. Overall, 23 % of males and 31 % of females had CAD predicted by CVRS (Monte Carlo: females, p = 0.024; males, p < 0.001), but there was little to no agreement between CVRS and atherosclerosis burden (Cohen's kappa: males, kappa = 0.149; females, kappa = 0.096). Conclusions: In asymptomatic individuals without known CAD, a low CVRS does not exclude extensive CAD. Newer tools incorporating additional markers may be helpful in risk prediction in such individuals.