DiabetesLiver score: A non-invasive algorithm for advanced liver fibrosis and liver-related outcomes in type 2 diabetes mellitus population.

Background: This study aimed to develop and validate a non-invasive model for screening advanced liver fibrosis and predicting liver-related outcomes in patients with type 2 diabetes mellitus (T2DM).

Methods: This study included patients with T2DM from five tertiary hospitals for the development and internal validation of a non-invasive model. Advanced liver fibrosis was defined as a liver stiffness measurement ≥12 kPa. An external validation cohort was obtained from the National Health and Nutrition Examination Survey (NHANES), and the model's predictive performance for hepatocellular carcinoma (HCC) and liver-related mortality was assessed in the UK Biobank.

Findings: In total, 28,197 patients with T2DM were enrolled. In the derivation cohort (n = 1,129), waist circumference, alanine aminotransferase, aspartate aminotransferase, platelet count, and albumin were identified as independent risk factors for advanced fibrosis and were fit to develop the "DiabetesLiver score." The area under the curve (AUC) was 0.835 (95% confidence interval [CI]: 0.781-0.890), significantly higher than the AUCs of non-invasive tests (all p < 0.01). It maintained high AUCs of 0.870 and 0.823 in the internal validation (n = 1,000), and NHANES cross-sectional (n = 1,432) cohorts, respectively. A dual cutoff of 2.39 and 3.99 with sensitivity ≥90% and specificity ≥90%, respectively, was used to classify patients into low-, middle-, and high-risk groups. In the UK Biobank cohort (n = 24,636), the high-risk group had an elevated risk of liver-related outcomes.

Conclusions: The DiabetesLiver score demonstrated good performance in identifying advanced liver fibrosis and the development of liver-related events in the T2DM population.

Funding: National Natural Science Foundation.