AI-based quantification of percent atheroma volume significantly improved the prediction of acute coronary syndrome on top of obstructive CAD (C-index 0.814 vs 0.790; p=0.010).
Cohort (n=2,007)
Does AI-based automated plaque quantification from CCTA improve the prediction of acute coronary syndrome on top of obstructive or ischemic CAD in symptomatic patients?
AI-based quantification of non-calcified and low-attenuation plaque burden from CCTA significantly improves the prediction of acute coronary syndrome beyond traditional assessment of obstructive or ischemic CAD.
Effect estimate: C-index 0.814
p-value: p=0.010
Abstract Background Artificial intelligence (AI)-based quantitative computed tomography (AI-QCT) is a novel tool for automated plaque characterization and quantification from coronary computed tomography angiography (CCTA). The prognostic value of various AI-QCT plaque types on top of the presence of obstructive or ischemic coronary artery disease (CAD) is unknown. Purpose To investigate the added prognostic value for acute coronary syndrome (ACS) of AI-QCT total plaque burden (percent atheroma volume (PAV) %), and its subcomponents non-calcified plaque burden (NCPB), low-attenuation plaque burden (LAP), and calcified plaque burden (CPB) (%) on top of obstructive or ischemic CAD. Methods A cohort of 2007 symptomatic patients having undergone CCTA and selective downstream PET perfusion for suspected CAD was analyzed. Patients with prior or early elective revascularization within 6 months from CCTA were excluded. Obstructive CAD was defined as ≥1 vessel with 50% visual stenosis, and ischemic CAD as absolute hyperemic myocardial blood flow ≤2.3 ml/g/min in ≥2 adjacent segments on 15O-H2O PET. Multivariable Cox regressions adjusted for clinical confounders (age, sex hypertension, typical angina) including each plaque type separately (per 1.0%) on top of obstructive or ischemic CAD, respectively, were performed. The amount of LAP was 1% and therefore pooled together with NCPB. Results Throughout a median follow-up of 7 years, 72/2007 patients experienced ACS (50 myocardial infarction, 22 unstable angina). On top of obstructive CAD, all plaque types were independent predictors of ACS (Figure 1A), however according to the C-indexes (p-value vs. obstructive CAD), only PAV (C-index=0.814, p=0.010) and NCPB+LAP (C-index=0.818, p=0.014), but not CPB (C-index=0.800, p=0.066) significantly improved the prediction of ACS on top of obstructive CAD (C-index=0.790) (Figure 2A). On top of ischemic CAD (1967 patients, 66 events), all plaque types were also independent predictors of ACS (Figure 1B). But, similarly, according to the C-indexes (p-value vs. ischemic CAD), only PAV (C-index=0.818, p=0.003), and NCPB+LAP (C-index=0.818, p=0.005), but not CPB (C-index=0.798, p=0.054) significantly improved the prediction of ACS on top of ischemic CAD (C-index=0.776) (Figure 2B). NCPB+LAP alone performed similarly as PAV on top of obstructive (p=0.554) and ischemic CAD (p=0.991) (Figure 2). Conclusions Among patients with native coronary arteries treated medically, AI-based plaque quantification significantly improved the prediction of ACS on top of obstructive or ischemic CAD. The risk of ACS was largely related to NCPB+LAP, whereas CPB did not significantly improve risk stratification.Adjusted HR for ACS per plaque typeC-indexes for ACS
Bär et al. (Tue,) conducted a cohort in Suspected coronary artery disease (n=2,007). AI-based quantitative computed tomography (AI-QCT) plaque quantification vs. Obstructive or ischemic CAD alone was evaluated on Acute coronary syndrome (ACS) (C-index 0.814, p=0.010). AI-based quantification of percent atheroma volume significantly improved the prediction of acute coronary syndrome on top of obstructive CAD (C-index 0.814 vs 0.790; p=0.010).