Association of hepatic steatosis with MRI-defined renal steatosis in type 2 diabetes: a cross-sectional imaging and clinical modeling study

Background Fatty Kidney Disease (FKD) is increasingly recognized as a distinct clinical entity, often coexisting with metabolic dysfunction-associated fatty liver disease (MAFLD) in type 2 diabetes mellitus (T2DM). However, the strength of their association and the metabolic factors underlying this association remain incompletely characterized. We examined the cross-sectional association between MRI-derived hepatic and renal fat and developed a clinical model for classifying an MRI-defined renal steatosis phenotype. Methods This single-center cross-sectional study included 257 adults with T2DM. Hepatic fat fraction (HFF) and renal fat fraction (RFF) were quantified using 3.0-T Dixon MRI. RFF ≥ 4% was used as an exploratory definition of renal steatosis, with sensitivity analyses at thresholds of ≥3% and ≥5%. Associations were assessed using correlation, multivariable regression, and restricted cubic splines. A classification model was developed using LASSO-assisted variable selection and logistic regression, with internal validation by 1,000 bootstrap resamples. XGBoost, SHAP, and exploratory cross-sectional indirect-effect analyses were also performed. Results HFF showed the strongest correlation with RFF (r = 0.534, p < 0.001). MRI-defined fatty liver, fasting C-peptide (FCP), and high-sensitivity C-reactive protein (hsCRP) were retained in the final model. The apparent AUC was 0.858, and the optimism-corrected AUC was 0.833. At RFF thresholds of ≥3% and ≥5%, the AUCs were 0.827 and 0.784, respectively. Systolic blood pressure, uric acid, HDL-C, and hsCRP showed significant indirect effects, accounting for 4.62–15.16% of the observed HFF–RFF association. Conclusion Hepatic steatosis was strongly associated with MRI-defined renal steatosis in individuals with T2DM. The three-variable model demonstrated good internally validated discrimination, with a bootstrap-corrected AUC of 0.833, although external validation remains necessary. Exploratory indirect-effect analyses suggested that inflammatory and metabolic factors may partially account for the observed association. These findings may inform the future development of risk-classification strategies and provide a basis for future longitudinal and mechanistic studies.