Obesity-stratified differences in the fasting insulin-attributable proportion of the TyG-CI and MASLD association

Abstract Background Dysregulation of lipid metabolism contributes to metabolic dysfunction-associated steatotic liver disease (MASLD), characterized by systemic visceral lipid overload. The nonlinear dose-response relationship, racial heterogeneity, and insulin-mediated mechanisms of the triglyceride-glucose-conicity index (TyG-CI)—a glycolipid composite biomarker integrating glucose-lipid metabolism with abdominal obesity—remain unclear, particularly regarding systemic visceral lipid overload in visceral adiposity. We hypothesized, based on cross-sectional associational patterns, that the relative contribution of insulin resistance may differ across obesity strata-a hypothesis that requires longitudinal validation. Methods This cross-sectional study included 13,682 nondiabetic Japanese adults and 2,985 US adults. We evaluated TyG-CI associations with prevalent MASLD, hepatic lipid accumulation, and liver fibrosis using multivariable regression. Restricted cubic splines (RCS) and piecewise regression characterized nonlinear relationships, and bootstrap-based decomposition analysis quantified the proportion of the total association statistically accounted for by fasting insulin in the NHANES cohort. Results Higher TyG-CI was independently associated with increased MASLD odds (Japanese: OR = 3.23; US: OR = 1.93) and greater severity of steatosis/fibrosis. Nonlinear dose-response relationships showed population-level differences in statistical inflection points; below these thresholds, the association was markedly stronger in the Japanese cohort (OR: 7.25 vs. 4.14). These differences may reflect ethnic, age, adiposity, and outcome-ascertainment heterogeneity combined. In cross-sectional decomposition analysis, 21.10% of the total TyG-CI–MASLD association was statistically accounted for by fasting insulin. The proportion attributable to insulin was lower in centrally obese participants (18.33%) than in those with normal waist circumference (26.78%), representing a relative reduction of 31.6%. TyG-CI outperformed TyG and CI alone in MASLD discrimination. Conclusion TyG-CI demonstrates population-level differences in statistical inflection points across cohorts and obesity-stratified differences in the fasting insulin-attributable proportion of the cross-sectional association with MASLD. The stronger insulin-associated association in normal waist circumference individuals may help identify occult metabolic phenotypes. These cross-sectional, hypothesis-generating findings suggest TyG-CI as a preliminary screening-stratification candidate for prevalent MASLD. Longitudinal cohorts are required to validate adiposity-stratified patterns and underlying biological mechanisms.