Predictive value of TyG-BMI, CTI, and SII in identifying metabolic dysfunction-associated steatotic liver disease among patients with type 2 diabetes mellitus.

BACKGROUND: Metabolic dysfunction-associated steatotic liver disease (MASLD) has become a major comorbidity in patients with type 2 diabetes mellitus (T2DM). This study aimed to evaluate the diagnostic performance of the triglyceride-glucose-body mass index (TyG-BMI), the C-reactive protein-triglyceride-glucose index (CTI), and the systemic immune-inflammation index (SII), individually and in combination, for identifying MASLD in patients with T2DM.

METHODS: This retrospective cross-sectional study enrolled 796 hospitalized patients with T2DM from January 2020 to January 2025. Sample size was determined based on available eligible cases during the study period. Data distribution was assessed using the Kolmogorov-Smirnov test. There were 280 MASLD patients and 516 non-MASLD patients. We collected anthropometric, biochemical, and inflammatory parameters. We performed correlation and multivariate logistic regression analyses to find independently associated factors. We evaluated diagnostic performance using receiver operating characteristic (ROC) curves, the area under the curve (AUC), and the Youden index. We also used net reclassification improvement (NRI) and integrated discrimination improvement (IDI) to assess predictive value. External validation was performed using an independent cohort of 394 patients. Key limitations include the retrospective design, potential misclassification due to ultrasound-based diagnosis, and residual confounding.

RESULTS: MASLD patients exhibited significantly higher BMI, triglycerides, low-density lipoprotein cholesterol (LDL-C), TyG-BMI, CTI, and SII levels than non-MASLD patients (all p < 0.05). TyG-BMI, CTI, and SII were independent MASLD predictors. The combined model (TyG-BMI + CTI + SII) had the highest discriminative power, outperforming each single indicator. Adding CTI and SII to TyG-BMI significantly improved classification (NRI = 0.312, IDI = 0.143, both p < 0.05). External validation indicated strong performance (AUC = 0.940, sensitivity = 84.6%, specificity = 90.1%). Machine learning validation using a random forest algorithm further indicated the high predictive accuracy of the model, achieving an AUC of 0.950 in the development cohort and maintaining a robust AUC of 0.885 in the external validation cohort.

CONCLUSION: TyG-BMI, CTI, and SII are independently associated with MASLD in T2DM patients. Their combined use significantly improves early identification accuracy and could serve as a convenient, cost-effective screening tool in clinical practice.