Artificial intelligence has demonstrated significant clinical value in cardiovascular medicine through PARAMOUNT, the first randomized controlled trial comparing AI-enabled coronary plaque analysis against conventional stress testing. The study found that AI-guided care substantially improves diagnostic certainty and more than doubles the initiation of preventive lipid-lowering therapy in patients with chest pain.

The late-breaking findings, presented at the American Heart Association Scientific Sessions 2026, represent a potential paradigm shift from functional ischemia testing toward anatomic plaque quantification for evaluating suspected coronary artery disease.

Trial Design and Primary Outcomes

PARAMOUNT enrolled 530 symptomatic outpatients across 21 sites between January 2025 and February 2026, with 523 participants included in the final analysis. The study population had a median age of 57, median baseline LDL cholesterol of 127 mg/dL, and notably included 57% women—addressing historical gender disparities in cardiovascular research.

The trial compared AI-based quantitative coronary CT analysis (AI-QCT) from Cleerly against usual stress testing in patients with suspected coronary artery disease, elevated LDL cholesterol, and no prior CAD history. At 180 days, the AI-QCT arm achieved a win ratio of 2.00 on the hierarchical composite primary endpoint (95% CI 1.37-2.56, p<0.0001), with plaque-based care showing better outcomes in 46.7% of unmatched comparisons versus 23.4% for usual care.

According to co-principal investigator Todd C. Villines, MD, Donald C. Barnes Professor of Cardiology at the University of Virginia, improved diagnostic certainty about coronary artery disease as the cause of chest pain was the largest contributor to the primary endpoint, with a win difference of 16.6 percentage points.

Substantial Increase in Preventive Therapy

The most clinically significant finding was the substantial increase in lipid-lowering therapy utilization. The AI-QCT arm increased therapy use from 31% at baseline to 75% post-testing, compared to 32% to 47% in the usual care arm—a 44-percentage-point difference in implementation of guideline-directed preventive care.

The AI-guided approach also appeared to improve patient selection for invasive procedures. Among patients undergoing invasive coronary angiography, 38% in the AI-QCT arm did not proceed to revascularization (PCI or CABG) versus 60% in the usual care arm, suggesting better identification of patients who truly require intervention.

Implications for Clinical Practice

PARAMOUNT provides Level 1 evidence supporting anatomic plaque quantification over functional testing for stable chest pain evaluation. The trial design—comparing AI-augmented imaging against functional testing rather than against standard CT reading—positions the technology as a care pathway intervention rather than merely an imaging enhancement.

Udo Hoffmann, MD, MPH, Chief Medical, Scientific, and AI Officer at Cleerly, noted that the results represent the first randomized evidence for AI-enabled coronary plaque analysis versus standard care. Cleerly's FDA-cleared AI solutions for coronary artery disease phenotyping were trained on millions of images from over 40,000 patients.

The results have not yet undergone peer review, with full publication expected in early 2027. The findings may accelerate adoption of AI-driven plaque assessment in cardiovascular care and establish a benchmark for competitors in the cardiac AI market to generate similar clinical validation.

Sources

  1. PARAMOUNT, the First Randomized Trial of AI-Enabled Plaque-Based Care, Meets Primary Endpoint in Late-Breaking Results at AHA 2026 — Business Wire