Diabetes mellitus results from metabolic dysregulation and is increasingly becoming a global health burden. Ectopic fat accumulation is strongly linked to insulin resistance, the precursor of type 2 diabetes. The Hepatic Steatosis Index (HSI) is a non-invasive measure for predicting hepatic steatosis, while the Visceral Adiposity Index (VAI) reflects visceral adipose tissue activity through lipid profiles and body metrics. Most studies have evaluated HSI and VAI as independent indicators of metabolic disorders, yet few have examined their relationship with outcomes such as inadequate HbA1c control. This study aims to develop a combined Adiposity–Steatosis model (VAI–HSI) to predict poor glycemic control in patients with type 2 diabetes mellitus (T2DM)

A descriptive cross-sectional analytical study was conducted on 77 patients aged over 18 years diagnosed with T2DM at the University Medical Center Ho Chi Minh City Branch 2, following ADA 2024 criteria. Data were analyzed using Stata 17.0 with Spearman correlation, linear regression, logistic regression, nonlinear modeling, and ROC curve analysis.

Participants were mostly older adults (mean age > 63 years), had a long disease duration, and poor glycemic control (median HbA1c > 7,2%). Multivariable linear regression identified three statistically significant independent predictors of HbA1c levels: sex, diabetes duration, and HSI. The study demonstrated a nonlinear relationship between fat indices and glycemic control through Restricted Cubic Spline (RCS) analysis. For VAI, the risk of poor HbA1c control did not increase linearly but peaked in the third quartile (Q3: 3,5 – 4,7) with an OR = 4,64 (p < 0,05). Conversely, HSI showed a clearer upward trend in risk at higher values, particularly in the highest quartile (Q4 > 38,3), where the risk increased 4,58-fold compared with Q1. We developed and compared predictive models for poor glycemic control using fat indices and baseline patient characteristics. The base model (Age + Sex + Diabetes duration) had an AUC of 0,70. Adding both VAI and HSI increased the AUC to 0,76 and the lowest AIC, indicating optimal model fit. Importantly, Net Reclassification Improvement (NRI) analysis showed that integrating VAI and HSI correctly reclassified 51% of patients (p < 0,05), while the Integrated Discrimination Improvement (IDI) increased by 8,5% (p < 0,05), reflecting enhanced discrimination between good and poor glycemic control.

These findings establish a strong scientific basis for recommending the routine use of the VAI–HSI pair in clinical practice to optimize glycemic management in individuals with type 2 diabetes.