Vessel-Specific Quantitative Plaque Characteristics on Coronary CT Angiography for Prediction of FFR-Defined Myocardial Ischemia

Objectives: To determine whether quantitative coronary computed tomography angiography (CCTA) plaque characteristics associated with fractional flow reserve (FFR)-defined ischemia vary among coronary arteries and whether they improve prediction when added to conventional clinical and stenosis information. Methods: This retrospective study included 204 patients and 289 vessels (161 left anterior descending [LAD], 55 left circumflex [LCX], and 73 right coronary artery [RCA] vessels) evaluated by CCTA and invasive FFR. Plaque characteristics were quantified using semi-automated software. Predictors were selected separately by vessel using the least absolute shrinkage and selection operator (LASSO) for LAD and RCA and Boruta-XGBoost for LCX. Model evaluation incorporated five-fold cross-validation and bootstrap resampling. Incremental performance was assessed by adding selected plaque features to Model 1, which comprised clinical variables and severe stenosis (≥70%). Results: FFR-defined ischemia (FFR ≤ 0.80) was identified in 90 LAD, 24 LCX, and 37 RCA vessels. Ischemic vessels had longer lesions, smaller minimum luminal area, greater stenosis, and higher plaque burden. Vessel-specific plaque models yielded areas under the receiver operating characteristic curve (AUCs) of 0.755 (95% CI: 0.682–0.826), 0.789 (95% CI: 0.656–0.905), and 0.775 (95% CI: 0.661–0.877) for LAD, LCX, and RCA, respectively. In a separate incremental analysis, compared with Model 1, Model 2 improved AUC in LAD (0.613 to 0.755; net reclassification improvement (NRI) = 0.736; integrated discrimination improvement (IDI) = 0.163) and RCA (0.489 to 0.767; NRI = 0.955; IDI = 0.276), with smaller gains in LCX (0.617 to 0.754; NRI = 0.651; IDI = 0.169). Conclusions: Quantitative CCTA plaque characteristics showed potentially vessel-dependent associations with FFR-defined ischemia. Incorporating selected plaque features into clinical and stenosis-based models improved discrimination in the LAD and RCA. These exploratory findings require validation in larger, independent cohorts.