U-shaped relationship between coronary artery calcification and underestimation of moderate coronary stenosis by...

Coronary computed tomography angiography (CCTA) is crucial for diagnosing coronary artery disease (CAD). However, its diagnostic accuracy is frequently influenced by factors such as coronary artery calcification (CAC), which can lead to both overestimation and critically, underestimation of stenosis severity. Underestimation of stenosis, particularly in the clinically ambiguous 50-70% range, may delay necessary invasive coronary angiography (ICA). Previous studies have lacked a systematic understanding of the entire spectrum of CAC score (CS) in relation to the risk of CCTA underestimation in this critical 50-70% stenosis range. Therefore, this study aimed to: (I) quantify the non-linear relationship between CS and CCTA underestimation of moderate coronary stenosis; (II) identify optimal CS thresholds for risk stratification of underestimation; and (III) determine clinical and imaging predictors to improve CCTA diagnostic accuracy.

This retrospective study included 216 patients who underwent both CCTA and ICA. Patients with CCTA-reported 50-70% stenosis were classified into an underestimation group (n=63; ICA stenosis >70%) and a non-underestimation group (n=153; ICA stenosis ≤70%). Demographic, clinical, and laboratory data were systematically collected, and CS was quantified using the Agatston method. Logistic regression and restricted cubic spline (RCS) analyses evaluated the association between CS and CCTA underestimation, identified non-linear relationships, and determined optimal CS cutoffs. Diagnostic performance and subgroup analyses were also conducted.

The underestimation group exhibited significantly higher median CS (P=0.08), higher diabetes prevalence (P=0.007), elevated triglycerides (TG) (P=0.004) and fasting glucose (P=0.03), and lower high-density lipoprotein cholesterol (HDL-C) (P=0.004). RCS analysis revealed a significant U-shaped non-linear relationship between CS and underestimation risk (P-nonlinearity <0.05). Distinct cutoffs were identified: a lower threshold around 118 Agatston units (AU) and an upper threshold between 253 and 330 AU. Patients with CS <118 or >330 AU had the highest underestimation risk, whereas those with intermediate CS (118-330 AU) were prone to overestimation. In the fully adjusted model, higher CS remained an independent predictor of underestimation [adjusted odds ratio (OR) =1.002; 95% confidence interval (CI): 1.000-1.003, P=0.02]. Diagnostic models utilizing these thresholds demonstrated high sensitivity (82.5-87.3%) but limited specificity (24.8-30.7%). Subgroup analyses confirmed robustness, with pronounced effects in females, elderly patients, and those with calcified plaques (P=0.002).

Our study reveals a U-shaped non-linear relationship between CS and CCTA underestimation of moderate stenosis. Patients with CS <118 or >253-330 AU are at higher underestimation risks, while intermediate CS values are tilting towards overestimation. Despite a modest per-unit OR, the cumulative effect across clinically relevant CS ranges may hold clinical significance. Clinically, application of these CS thresholds offers a practical framework to identify patients requiring further ICA, minimizing missed diagnoses in high-risk groups and reducing unnecessary invasive procedures in intermediate-risk patients, thereby facilitating precise, individualized CAD management.