Ultrasound-defined hepatic steatosis and clustering of cardiometabolic abnormalities: endpoint-dependent performance of seven metabolic indices in 356,197 adults

Hepatic steatosis commonly coexists with cardiometabolic abnormalities, but the extent to which an ultrasound finding identifies their clustering in routine health-examination populations remains unclear. We quantified the association between hepatic steatosis and cardiometabolic burden and compared seven metabolic indices across outcomes with different component domains. This cross-sectional analysis included 356,197 adults with available liver ultrasonography, age, and sex data. Person-level deduplication had been completed during source-database preparation before de-identification, leaving one examination record per individual. High cardiometabolic burden was defined as at least three of six examination-based abnormalities. We compared the triglyceride-glucose index, atherogenic index of plasma, triglyceride-to-high-density lipoprotein cholesterol ratio, cardiometabolic index, lipid accumulation product, visceral adiposity index, and waist-to-height ratio. Analyses included modified Poisson regression with robust variance, adjusted prevalence differences, logistic regression, common-complete flexible area under the receiver-operating characteristic curve (AUC) comparisons, systematic outcome-domain removal, comparisons with simple component models, and sex-, age-, and examination-year-stratified analyses. Among 356,197 adults, 84,384 (23.7%) had ultrasound-defined hepatic steatosis. High cardiometabolic burden was present in 64.1% of participants with steatosis and 14.5% of those without. The adjusted prevalence ratio was 3.46 (95% confidence interval [CI] 3.42–3.50), and the adjusted prevalence difference was 39.9 percentage points (95% CI 39.5–40.3). In the common-complete sample using flexible index terms, lipid accumulation product had the highest AUC for hepatic steatosis (0.868, 95% CI 0.867–0.869), followed by waist-to-height ratio and cardiometabolic index, which performed similarly (0.840 each). For burden excluding adiposity components, the triglyceride-glucose index ranked first (AUC 0.851), lipid accumulation product second (0.842), and waist-to-height ratio seventh (0.783). Atherogenic index of plasma and triglyceride-to-high-density lipoprotein cholesterol ratio had essentially identical flexible AUCs. Composite indices did not outperform models that flexibly included their corresponding component variables separately. Ultrasound-defined hepatic steatosis was associated with substantially greater concurrent cardiometabolic burden. The relative performance of metabolic indices depended on outcome composition and functional form. Composite indices offered parsimonious summaries but did not provide more discriminatory information than flexible models of their component measurements.