Associations between multiple metabolic biomarkers with steatotic liver disease subcategories: A 5-year Chinese cohort study

Abstract

The effectiveness of established biomarkers for non-alcoholic fatty liver disease (NAFLD) within the updated framework of steatotic liver disease (SLD) remains uncertain. This cohort study examines the association of four metabolic biomarkers-retinol-binding protein 4 (RBP-4), fibroblast growth factor 21 (FGF-21), adiponectin, and osteocalcin-with SLD and its subtypes: metabolic dysfunction-associated steatotic liver disease (MASLD) and metabolic dysfunction with alcohol-related liver disease (MetALD)/alcohol-related liver disease (ALD). Among 3,504 Chinese participants aged 55-70, 938 (26.8%) have developed SLD over 5 years, including 871 with MASLD and 67 with MetALD/ALD. The findings indicate that models incorporating RBP-4, FGF-21, adiponectin, and osteocalcin improve predictive accuracy for SLD beyond conventional models. Notably, adiponectin emerges as the most versatile marker, while elevated baseline levels of FGF-21 or RBP-4 indicate specific needs for metabolic or alcohol-related interventions, respectively, supporting tailored precision medicine strategies.

Keywords: ALD; MASLD; MetALD; NAFLD; adiponectin; cohort study; fibroblast growth factor 21; osteocalcin; retinol-binding protein 4.

Graphical abstract

Figure 1

Flowchart of participants SLD, steatotic liver disease; MASLD, metabolic dysfunction-associated steatotic liver disease; MetALD, metabolic dysfunction and alcohol-associated/related liver disease; ALD, alcohol-associated/related liver disease.

Figure 2

Correlations between metabolic biomarkers and clinical characteristics The standardized correlation coefficients between metabolic biomarkers and clinical characteristics were calculated by using linear regression. WC, FPG, 2hPG, HbA1c, FINS, HOMA-IR, SBP, DBP, TGs, HDL-C, ALT, AST, and GGT were log-transformed prior to analysis. Partial correlation coefficients between biomarkers with FPG, 2hPG, HbA1c, FINS, HOMA-IR, SBP, DBP, TGs, HDL-C, ALT, AST, and GGT calculated after adjusting for age and BMI. ∗∗∗, p < 0.001; ∗∗, p < 0.01; ∗, p < 0.05. BMI, body mass index; WC, waist circumference; FPG, fasting plasma glucose; 2hPG, 2-h postprandial plasma glucose; HbA1c, glycated hemoglobin; FINS, fasting insulin; HOMA-IR, homeostasis model assessment of insulin resistance; SBP, systolic blood pressure; DBP, diastolic blood pressure; TGs, triglycerides; HDL-C, high-density lipoprotein cholesterol; ALT, alanine aminotransferase; AST, aspartate transaminase; GGT, γ-glutamyltransferase; RBP-4, retinol-binding protein 4; FGF-21, fibroblast growth factor 21.

Figure 3

Associations between biomarkers and new-onset SLD (A) Incidence rate of SLD across tertiles of biomarkers. Data are represented as rate (95% CI). (B) The RRs for new-onset SLD were calculated after adjusting for age, BMI, WC, 2hPG, HbA1c, FINS, SBP, TGs, HDL-C, ALT, AST, and GGT by using modified Poisson regression models. Data are represented as RR (95% CI). See also Tables S1 and S2; Figures S1 and S2. SLD, steatotic liver disease; RBP-4, retinol-binding protein 4; FGF-21, fibroblast growth factor 21; T1–3, tertiles 1 to 3; RR, relative risk; CI, confidence interval.

Figure 4

Associations between biomarkers and new-onset SLD subcategories in males (A) The proportion of patients with MASLD and MetALD/ALD in all SLD patients. (B) The RRs for new-onset SLD subcategories were calculated after adjusting for age, BMI, WC, 2hPG, HbA1c, FINS, SBP, TGs, HDL-C, ALT, AST, and GGT by using modified Poisson regression models. Data are represented as RR (95% CI). Heterogeneity between the RRs of biomarkers for different subcategories of SLD was examined using the Q test. See also Tables S1 and S2; Figures S1 and S2. SLD, steatotic liver disease; RBP-4, retinol-binding protein 4; FGF-21, fibroblast growth factor 21; MASLD, metabolic dysfunction-associated steatotic liver disease; MetALD, metabolic dysfunction and alcohol-associated/related liver disease; ALD, alcohol-associated/related liver disease; RR, relative risk; CI, confidence interval.

Figure 5

Predictive performance of biomarker-based models The predictive performance of each of the four biomarkers was evaluated by comparing the differences in their AUC (delta AUC), NRI, and IDI (A). Delta AUC, NRI, and IDI > 0, = 0, or < 0 indicate the biomarkers had a better, similar, or worse performance than the reference in predicting the development of SLD, MASLD, or MetALD/ALD, respectively. Further, the RBP-4/FGF-21/adiponectin/osteocalcin models were constructed by including the RBP-4/FGF-21/adiponectin/osteocalcin combined with body mass index and waist circumference. The predictive performance of these biomarker-based models compared with traditional SLD predictive models, including FLI, HSI, and LFS, was evaluated by AUC (B), NRI (C), and IDI (C). Data are represented as NRI (95% CI) and IDI (95% CI) (C). Predictive performance for MetALD/ALD was evaluated exclusively in males due to the limited case number in females. See also Figure S3. ∗∗∗, p < 0.001; ∗∗, p < 0.01; ∗, p < 0.05. AUC, area under the curve; NRI, net reclassification index; IDI, integrated discrimination improvement; MASLD, metabolic dysfunction-associated steatotic liver disease; MetALD, metabolic dysfunction and alcohol-associated/related liver disease; ALD, alcohol-associated/related liver disease; RBP-4, retinol-binding protein 4; FGF-21, fibroblast growth factor 21; FLI, fatty liver index; HSI, hepatic steatosis index; LFS, liver fat score; CI, confidence interval.

Figure 6

Mediation analyses for MASLD (A) The structural equation model with significant standardized path coefficients. Coefficients were calculated with adjustments for age. Potential mediators that are significantly associated with both biomarker and outcome were considered to mediate the role of the biomarker in the development of outcome. (B) The mediation proportion of HOMA-IR and TGs was calculated using the following formula: 100% × (coefficient_{biomarkers-mediator} × coefficient_{mediator-outcome})/(coefficient_{biomarkers-outcome} + coefficient_{biomarkers-mediator} × coefficient_{mediator-outcome}). ∗∗∗, p < 0.001; ∗∗, p < 0.01; ∗, p < 0.05. MASLD, metabolic dysfunction-associated steatotic liver disease; RBP-4, retinol-binding protein 4; FGF-21, fibroblast growth factor 21; HOMA-IR, homeostasis model assessment of insulin resistance; TGs, triglycerides.