In a single-center retrospective study of 381 patients with coronary artery disease undergoing CCTA, AI-quantified total noncalcified plaque burden predicted major adverse cardiovascular events, with 67 patients (17.6%) experiencing MACE over 18 months. The study found total noncalcified percent atheroma volume >4.68% was the strongest predictor (HR 5.073, 95% CI 2.930-8.786), while adding CT-derived fractional flow reserve did not meaningfully improve risk prediction.

Why It Matters To Your Practice

• AI-based plaque quantification may help identify higher-risk CAD patients beyond stenosis-focused reads alone.

• Noncalcified plaque burden appears more prognostically useful than adding CT-FFR in this dataset.

• Risk cutoffs such as total noncalcified percent atheroma volume >4.68% could support more standardized reporting and follow-up decisions.

Clinical Implications

• Consider that AI-enabled CCTA analysis may shift attention toward plaque composition and total burden, not just luminal narrowing.

• Patients with higher noncalcified plaque burden may warrant closer surveillance, intensified preventive therapy, or cardiology follow-up.

• CT-FFR may still have selective value, but this study suggests it did not add clinically meaningful prognostic lift for overall MACE prediction.

Insights

• All six total plaque parameters were significantly associated with MACE in univariate analyses.

• Among target vessels, RCA plaque burden showed the greatest prognostic weight, though this was exploratory.

• MACEs included death, myocardial infarction, heart failure death, malignant arrhythmia, coronary revascularization, and rehospitalization for angina exacerbation.

The Bottom Line

• For clinicians tracking how AI may affect practice, this study suggests automated CCTA plaque quantification can add actionable prognostic information in CAD.

• The clearest signal was noncalcified plaque burden, while CT-FFR did not significantly improve model performance in this cohort.

• Because the study was retrospective and single-center, use these thresholds cautiously until validated externally.