Nonlinear and Vessel-Specific Association Between Coronary Artery Calcium Burden and Coronary Computed Tomography Angiography Underestimation of Coronary Stenosis

Background: Coronary artery calcium (CAC) introduces substantial artifacts in coronary computed tomography angiography (CCTA). Although calcification is classically associated with overestimation of stenosis severity, underestimation may be even more clinically hazardous because underestimation can lead to inappropriate deferral of revascularization. Thus, this study aimed to investigate the association between CAC burden and CCTA underestimation, with emphasis on vessel-specific anatomical heterogeneity. Methods: We retrospectively evaluated 673 patients without known coronary artery disease who underwent both CCTA and invasive coronary angiography (ICA) within 90 days, yielding 2692 vessel observations. Underestimation was defined strictly at the vessel level as a CCTA stenosis grade lower than the corresponding ICA grade. Restricted cubic splines (RCS) were used to characterize the nonlinear association between CAC burden and the risk of underestimation. Within-patient clustering across the left main (LM), left anterior descending (LAD), left circumflex (LCX), and right coronary artery (RCA) was accounted for using generalized estimating equations (GEEs). Results: Any-vessel underestimation occurred in 389 patients (57.8%). Vessel-specific susceptibility varied markedly: LM, 3.0%; LAD, 27.9%; LCX, 29.3%; and RCA, 22.4%. RCS analysis demonstrated a significant nonlinear association between total CAC score and patient-level underestimation in the fully adjusted model (p-overall = 0.014; p-nonlinear = 0.004), with an upward risk trajectory observed as calcification increased relative to the median reference of 301. GEE modeling confirmed that patient-level CAC burden independently predicted vessel-level underestimation (CAC ≥400 vs. 0: odds ratio, 2.01; 95% confidence interval, 1.20–3.36; p = 0.008), with a significant interaction between CAC and vessel territory (p < 0.001). A sensitivity analysis restricted to patients without any CCTA overestimation (n = 155) confirmed a robust nonlinear association between total CAC score and pure underestimation risk (p-overall = 0.012; p-nonlinear = 0.005), suggesting that concurrent overestimation noise partially obscures this dynamic risk trajectory in the full cohort. Conclusions: The association between calcification burden and CCTA underestimation is nonlinear and strongly dependent on coronary territory. Since the RCS-derived odds ratio (OR) = 1, crossings depend on the median reference value (CAC = 301) and should be interpreted as markers of risk trajectory rather than absolute diagnostic cutoffs. Integrating total calcium burden with vessel-specific susceptibility may help mitigate the risk of missed severe stenosis on CCTA.