AHA Scientific Sessions 2026 Abstract · Nov 2, 2026

Artificial Intelligence-Assisted Opportunistic Computed Tomography for Steatotic Liver Disease Detection and Association with Cardiovascular and Mortality Outcomes in Adults with Cardiometabolic Risk

By Derry L, Rajasekaran V, Aali A et al.

Introduction: Metabolic dysfunction-associated steatotic liver disease (MASLD), the most common steatotic liver disease (SLD), is a sentinel feature of cardiovascular-kidney-metabolic (CKM) syndrome and is underrecognized. Cardiovascular disease is the leading cause of death in this population. Abdominal CT visualizes the liver, enabling artificial intelligence (AI)-assisted opportunistic SLD detection. Whether incidentally identified, uncoded SLD carries the same cardiovascular risk as clinically recognized SLD is unknown. Hypothesis: Clinically recognized (CR; ICD-coded) SLD represents a more advanced cardiometabolic phenotype than clinically uncoded (CU; incidentally identified) SLD and carries higher all-cause mortality and 3-point major adverse cardiovascular events (MACE). Methods: Adults with ≥1 cardiometabolic risk factor undergoing abdominal CT at a multi-site academic center across two data extraction windows (1/2013-9/2018 and 1/2021-6/2024) were included; exclusions were other chronic liver disease, alcohol use disorder, prior liver transplant, and myocardial infarction (MI)/stroke within 30 days of CT. Steatosis was ascertained by ICD coding, radiology report documentation, or liver attenuation ≤80 Hounsfield Units using an open-source AI tool (Comp2Comp). Patients with SLD were classified as CR (ICD-coded within 30 days of CT) or CU (radiology report or AI-detected). Primary outcomes were all-cause mortality and 3-point MACE (cardiovascular death, nonfatal MI, nonfatal stroke). Modified Poisson with robust standard errors estimated event rates per 10,000 person-years (PY) and adjusted relative risks (RR); models adjusted for demographics and comorbidities. Results: Of 9,395 at-risk adults, 553 had CR and 1,808 had CU SLD. CR patients were younger (mean 59.6 vs 64.5 years), more racially/ethnically diverse (Hispanic/Latino 18.4% vs 12.4%), and had a heavier cardiometabolic burden (type 2 diabetes 32.7% vs 22.3%; obesity 41.8% vs 22.1%). All-cause mortality occurred in 14.1% of CR and 9.6% of CU SLD; rates per 10,000 PY were 1,473 (95% CI 1,178-1,843) vs 835 (708-985); adjusted RR 1.70 (1.31-2.21; p<0.0001). MACE occurred in 10.7% of CR and 8.9% of CU SLD; rates per 10,000 PY were 1,084 (835-1,407) vs 715 (597-858); adjusted RR 1.18 (0.87-1.59; p=0.29). Median follow-up 2.2 years (IQR 0.85-3.41). Conclusion: Clinically recognized SLD carried substantially higher all-cause mortality than uncoded SLD, while MACE rates were similar after adjustment for cardiometabolic comorbidities. Cardiovascular risk in SLD appears independent of clinical recognition, yet current diagnostic pathways capture only a more advanced, higher-comorbidity phenotype. AI-assisted opportunistic CT detection may surface a large undiagnosed population earlier, when primary prevention can most modify cardiovascular outcomes.

doi.org/10.1161/circ.154.suppl_1.9297Populations

AgingResearch.News