Artificial intelligence-enabled coronary plaque quantification for personalized risk assessment and lipid-lowering therapy: Insights from the FISH&CHIPS study✰

Background: Coronary computed tomographic angiography (CCTA) is a guideline-endorsed tool to evaluate coronary artery disease (CAD) in symptomatic patients. Artificial intelligence enabled quantitative coronary plaque analysis on CCTA (AI-CPA) is a promising strategy for tailored management of atherosclerotic cardiovascular disease (ASCVD). Population-level data are needed on how CCTA-derived plaque analyses can inform lipid-lowering strategies for ASCVD risk reduction. Objectives: To model the utility and efficiency of a total plaque volume (TPV)-based risk staging system in guiding lipid-lowering therapy in patients undergoing clinically-indicated CCTAs for evaluation of stable suspected or known CAD. Methods: for DECIDE stages 1-4, respectively. The primary outcome was the estimated reduction in cardiovascular death or non-fatal MI, and the number needed to treat (NNT) based on AI-CPA, over 10 years. We modeled lipid-lowering therapy utilizing treat-to-target LDL-C goals of <100, <70, <55, and <40 mg/dL for DECIDE stages 1-4, respectively, to estimate risk reduction and NNT over 10 years. Results: The study population included 7899 total symptomatic participants undergoing CCTA and AI-CPA. Of these, 6054 patients had any plaque and were included in the final cohort; the mean age was 59.4 ± 11.7 years and 42.7% were women. Among the full cohort, the 10-year modeled relative risk reduction using a TPV treat-to-target LDL-C was 19.1% with NNT of 61. The 10-year relative risk reduction and NNT by DECIDE stages 1-4 was 1.5% (NNT = 1686), 18.2% (NNT = 59), 24.2% (NNT = 27), and 33.8% (NNT = 11), respectively. Conclusions: Quantitative TPV measured by AI-CPA identifies symptomatic patients at elevated long-term cardiovascular risk and may efficiently inform implementation of personalized lipid-lowering strategies to reduce cardiovascular events.