Prognostic value of AI-derived epicardial adipose tissue beyond coronary artery calcium on chest CT in an asymptomatic screening population.

BACKGROUND: The incremental prognostic value of epicardial adipose tissue (EAT) quantified on nongated, noncontrast chest CT beyond conventional risk stratification remains uncertain. We evaluated whether AI-derived EAT volume and attenuation improve prediction of major adverse cardiovascular events (MACE) beyond clinical risk factors and coronary artery calcium (CAC) in an asymptomatic screening population. METHODS: We retrospectively analyzed asymptomatic individuals who underwent nongated, noncontrast chest CT at two screening centers between January 2007 and December 2014. EAT volume and attenuation were automatically quantified using an AI-based algorithm. MACE was defined as a composite of myocardial infarction, ischemic stroke, or cardiovascular death. Associations with MACE were assessed using multivariable Cox proportional hazards models adjusting for clinical risk factors and CAC. Incremental prognostic value was evaluated using changes in discrimination. RESULTS: Of the 5342 included individuals, the median EAT volume was 133.0 ​mL (IQR, 103.0-166.0 ​mL), and median EAT attenuation was -76.5 HU (IQR, -80.2 to -71.2 HU). Over a median follow-up of 7.5 years, higher EAT volume was independently associated with an increased risk of MACE (hazard ratio per 10 ​mL, 1.04; 95% CI, 1.00-1.08; p ​= ​0.031), whereas EAT attenuation was not. Addition of EAT volume to a model including clinical risk factors and CAC did not significantly improve discrimination (C-index, 0.756 vs. 0.757; p ​= ​0.729). CONCLUSIONS: AI-derived EAT volume on screening chest CT was independently associated with MACE in asymptomatic individuals. However, its incremental prognostic value beyond clinical risk factors and CAC was limited.