AI-Driven Quantitative Coronary CT Angiography in Suspected Coronary Artery Disease: Multicenter CONFIRM2 Registry.
Level 3 - non-randomized controlled study
Multicenter prospective/observational cohort study evaluating prognostic performance
PubMed 41906604 · doi:10.1016/j.jacadv.2026.102618
What was done
In the international, multicenter CONFIRM2 registry, investigators assessed the prognostic value of artificial intelligence-guided quantitative coronary computed tomography angiography (AI-QCT) in patients with clinically indicated CCTA for suspected coronary artery disease (CAD), excluding asymptomatic patients or those with prior CAD. AI-QCT software derived 24 variables across the coronary artery tree (including percent luminal narrowing, remodeling index, and plaque volume/composition). Patients were followed for a median of 4.27 years for primary major adverse cardiovascular events (MACE: all-cause death, myocardial infarction, stroke, heart failure, late revascularization, and unstable angina hospitalization) and secondary MACE (all-cause death and myocardial infarction).
What was found
Among 3,551 patients (mean age 58.8 ± 12.5 years, 50.5% male), 167 (4.7%) primary MACE events occurred over follow-up. Multivariable analysis identified diameter stenosis (HR 1.25, 95% CI: 1.18–1.32 per 10% increase) and noncalcified plaque volume (HR 1.07, 95% CI: 1.03–1.11 per 50 mm³) as the sole independent predictors of MACE. Adding these AI-QCT parameters improved model discrimination (AUC) from 0.63 (95% CI: 0.58–0.67) to 0.76 (95% CI: 0.77–0.80, P < 0.001) over clinical likelihood models, from 0.67 to 0.77 (P < 0.001) over traditional risk factors with age and sex, and from 0.63 to 0.75 (P < 0.001) over ASCVD risk scores, with similar improvements for death/MI.
Why it matters
Automated AI-driven quantification of coronary stenosis and noncalcified plaque volume provides incremental prognostic information beyond established clinical risk scores, facilitating standardized CAD risk stratification from routine CCTA scans.
Limits
As an observational cohort, unmeasured confounding and variation in downstream clinical management after CCTA cannot be excluded. The total number of events was relatively small (167 events, 4.7% event rate), and the primary endpoint incorporated soft outcomes such as revascularization and hospitalization alongside hard cardiovascular endpoints.
Cited by
- supports Cleerly's AI-enabled coronary computed tomography angiography (CCTA) analysis can predict heart attacks up to five years in advance by quantifying soft coronary plaque.