Abstract Body (Do not enter title and authors here): Introduction
We evaluated the ability of AI-guided quantitative CT plaque quantification (AI-QCT) to risk stratify patients alongside the ESC Risk Factor-weighted Clinical Likelihood (RF-CL) of obstructive CAD.
Methods
The CONFIRM2 is a multicenter, international cohort study, that included symptomatic patients with clinically indicated coronary CT angiography (CCTA) and follow-up for MACE. Across the entire coronary tree CAD was quantified by AI-QCT software and 24 variables at a patient, vessel and plaque level were derived. Multivariable modeling identified the non-calcified plaque volume and quantitative diameter stenosis as strongest predictors and these were combined for statistical modeling as ‘AI-QCT’.
The RF-CL classifies patients as very-low (0-5%), low (5-15%), and moderate (15-50%) likelihood of obstructive CAD (>50% stenosis), based on age, sex, symptoms, and risk factors.
The primary MACE endpoint was: all-cause death, myocardial infarction (MI), stroke, congestive heart failure, late revascularization, and hospitalization for unstable angina. The secondary endpoint was: death and MI.
Results
A total of 6,550 patients (58.5 ± 12.1 years, 48.4% male) were included and followed for 4.4 ± 1.9 years. MACE and MI/death events occurred in 326 (5.0%) and 119 (1.8%) patients, respectively. Patients were at very-low, low, and moderate likelihood for obstructive CAD (no patients had a higher likelihood) by the RF-CL in 35.7%, 42.7%, and 21.7%, and true obstructive CAD rates were 5.8%, 14.2%, and 25.7%, respectively. The addition of AI-QCT to the RF-CL improved the prediction of MACE: AUC 0.75 (95% CI 0.72-0.78) vs 0.62 (95% CI 0.59-0.65), p<0.001, and MI/death: AUC 0.71 (95% CI 0.66-0.76) vs 0.60, (95% CI 0.55-0.65), p<0.001. Within the very low, low, and moderate risk groups, AI-QCT was significantly associated with MACE: AUC 0.71 (95% CI 0.64-0.78), 0.73 (95% CI 0.68-0.78), and 0.71 (95% CI 0.66-0.76), respectively, and also with MI/death (Table 2). Risk separation by AI-QCT within the 3 RF-CL groups is shown in Figure 1.
Conclusion
Current symptomatic patients predominantly fall into low or very-low likelihood for obstructive stenosis, resulting in low yield of imaging focusing solely on severe stenosis. However, AI-guided quantification of (non-obstructive) plaque provides incremental prognostic information for MACE and MI/death, and classifies risk within the three RF-CL groups, therefore allowing early optimization of anti-atherosclerotic therapies.