A dose-response study of the association between MRI-derived liver proton density fat fraction and prevalent T2DM in postmenopausal women with abnormal glucose metabolism

Objective This study aimed to investigate the dose-response association between magnetic resonance imaging-derived hepatic proton density fat fraction (MRI-derived PDFF) and the type 2 diabetes mellitus (T2DM) prevalence in postmenopausal women with abnormal glucose metabolism, evaluate its utility in risk stratification, while examine the potential mediating role of liver enzymes and to explore the nonlinear relationship between MRI-derived PDFF and insulin resistance assessed by the Triglyceride-Glucose (TyG) index, in this specific population. Methods In this cross-sectional study, 96 postmenopausal women with abnormal glucose metabolism (46 with prediabetes, 50 with T2DM) were included. Hepatic and pancreatic fat fractions were measured by MRI, and visceral fat mass (VFM) was quantified by quantitative computed tomography. Inter-observer agreement was assessed using intraclass correlation coefficients. Multivariable logistic regression, with sequential adjustment for age, BMI, total cholesterol, and VFM, was used to assess the association between liver fat fraction and T2DM, treating hepatic proton density fat fraction of the left lobe (FF-LL) as both a continuous variable and in quartiles. The discriminative ability of FF-LL was evaluated by ROC analysis. Mediation analysis explored the potential role of liver enzymes, and restricted cubic splines examined the relationship between FF-LL and the TyG index. Results In postmenopausal women with abnormal glucose metabolism, those with T2DM demonstrated a more unfavorable metabolic profile compared to those with prediabetes. Inter-observer agreement was excellent for liver-related parameters and VFM (ICCs: 0.977–0.984), but moderate for pancreatic fat fraction (ICC range: 0.602–0.637). FF-LL was significantly higher in the T2DM group compared to the prediabetes group. ROC analysis demonstrated that FF-LL had good discriminative ability for T2DM (AUC = 0.816). The optimal cutoff value was 6.3%, yielding a sensitivity of 80.0% and a specificity of 87.0%. As a continuous variable, each unit increase in FF-LL was associated with a 15% to 20% higher prevalence of T2DM across adjusted models. A significant dose-response trend was observed across increasing quartiles of FF-LL. In addition, mediation analysis showed that both alanine aminotransferase (ALT) and the AST (aspartate aminotransferase)/ALT ratio partially mediated the association of FF-LL with T2DM in postmenopausal women, accounting for 22.0% and 24.9% of the total effect, respectively (both P = 0.040). Additionally, restricted cubic spline analysis revealed a significant nonlinear association between FF-LL and the TyG index (overall P < 0.001, P for nonlinearity = 0.007), characterized by a threshold effect with a steep increase in FF-LL when TyG exceeded approximately 8.75. Conclusion In postmenopausal women with abnormal glucose metabolism, FF-LL is an independent risk factor for T2DM, exhibiting a dose–response relationship partially mediated by liver enzymes. With an optimal cutoff of 6.3%, FF-LL demonstrates good discriminative ability and may serve as an effective biomarker for T2DM risk stratification. Additionally, a significant nonlinear association is observed between FF-LL and the TyG index in this population.