Abstract
BACKGROUND: We hypothesized that quantification of coronary atherosclerotic plaque burden by artificial intelligence-guided quantitative computed tomography can identify patients who derive outcome benefit from lipid-lowering medication (LLM). METHODS: In this observational cohort study, consecutive symptomatic patients undergoing coronary computed tomography angiography for suspected coronary artery disease (CAD) were assessed for percent atheroma volume (PAV) by artificial intelligence-guided quantitative computed tomography. The use of LLM was assessed based on drug purchase registry data within 6 months after coronary computed tomography angiography. Patients were followed for the composite of death, myocardial infarction, and unstable angina for a median of 6.9 years. RESULTS: Among 2269 patients (median age, 63 years; 42% men), 1261 (56%) patients used LLM after coronary computed tomography angiography, and 255 (11%) experienced the composite end point during follow-up. The median PAV was 6.6% among users and 1.4% among nonusers of LLM (P<0.001). Adapting the previously proposed CAD stages for artificial intelligence-guided quantitative computed tomography, the use of LLM (versus no use) was associated with improved outcomes among the 910 patients with PAV >5% (annual event rate, 2.62% versus 4.14%; adjusted P=0.002), even in the absence of obstructive CAD, but not among the 1359 patients with PAV ≤5% (annual event rate, 0.94% versus 0.65%; adjusted P=0.717). An adjusted Cox regression analysis, including interaction between PAV and LLM, suggested a PAV threshold between 4% and 10% for gaining prognostic benefit from LLM. CONCLUSIONS: In symptomatic patients with suspected CAD, LLM after coronary computed tomography angiography was associated with a lower rate of adverse events during long-term follow-up among those with PAV >5%, even in the absence of obstructive CAD. The quantification of coronary atherosclerotic plaque burden is a potential marker to guide preventive lipid-lowering therapy.