Tomographic Data from ISCHEMIA According to AI and Cardiovascular Risk

Coronary computed tomography angiography (CCTA) is an increasingly popular alternative to invasive diagnosis. This technique allows for the assessment of the severity of coronary artery disease (CAD) through specific protocols. Among its benefits there is the ability to obtain three-dimensional volumetric data and analyze the characteristics of atherosclerotic plaque, such as calcification, the presence of necrotic components, vascular remodeling, and high-risk phenotypes.

The ISCHEMIA study enrolled patients with moderate to severe ischemia and no left main coronary artery disease according to CCTA. The baseline study results revealed that 79% of patients had multivessel involvement, and 86.8% had obstructive involvement (≥50%) of the left anterior descending artery, according to the affected area.

The objective of the study presented by Nurmohamed et al. was to investigate whether the tomographic characteristics of coronary plaque were independently associated with myocardial infarction or cardiovascular death (the primary endpoint), as well as with the composite of major cardiovascular events (MACE).

The tomographic data for ISCHEMIA were collected through core-lab analysis. CCTA interpretation was conducted using semi-automatic analysis by artificial intelligence (AI-QCT, Cleerly) and neural networks (VGG19 network, 3D U-Net, and VGG Network Variant).

Read also: Definite Pacemaker Implantation Predictors in TAVR High Implant Using Cusp Overlap.

Three models were used for the multivariate analysis: model 1, with clinical variables; model 2, which added vessel involvement determined by AI-QCT; and model 3, which included plaque variables. The study population consisted of 3711 patients with an average age of 64 years; 79% were men, 13% were active smokers, and 41% were diabetic. The average follow-up was 3.3 years, during which 374 patients suffered a myocardial infarction or cardiovascular death.

The mean percentage atheroma volume (PAV) was 494 mm³, and the noncalcified plaque volume (NCPV) was 292 mm³. According to AI tomographic assessment, 77% of patients had ≥50% stenosis in at least one vessel, 25% had two-vessel disease, and 14% had three-vessel or left main coronary artery disease.

The parameters most associated with the primary endpoint were PAV (hazard ratio [HR] 1.60; 95% confidence interval [CI] 1.42–1.81; P <0.001) and diffuse disease (HR 1.47; 95% CI 1.32-1.65; P <0.001). After adjusting for clinical characteristics, the presence of single-vessel disease determined by AI resulted in an aHR of 1.12 (95% CI 0.83–1.53) for the primary endpoint, while two-vessel involvement showed a risk of aHR of 1.57 (95% CI 1.15–2.15), and three-vessel disease an aHR of 2.17 (95% CI 1.56–3.02; P = .001).

Read also: Do Leadless Pacemakers After TAVR Offer Benefits?

Subsequently, researchers assessed the prognostic value of the models at 6 months, showing that the clinical model had an area under the curve (AUC) of 0.637 (95% CI 0.592–0.682), which improved to 0.670 with the addition of AI-QCT (95% CI 0.627–0.714). Adding plaque characteristic data did not significantly change the discriminative value, with an AUC of 0.688 (95% CI 0.641–0.735).

Conclusions

Atherosclerotic burden, as measured by plaque volume and the number of vessels involved, was independently associated with an increased risk of cardiovascular death or myocardial infarction. The incorporation of parameters obtained by artificial intelligence partially improved the diagnostic performance according to the analyzed models.

Original Title: Atherosclerosis quantification and cardiovascular risk: the ISCHEMIA trial. 

Reference: Nurmohamed NS, Min JK, Anthopolos R, et al. Atherosclerosis quantification and cardiovascular risk: the ISCHEMIA trial. Eur Heart J. Published online August 5, 2024. doi:10.1093/eurheartj/ehae471.