Comparative Discriminative Performance of the HbA1c/HDL‐C Ratio, Stress Hyperglycemia Ratio, and ZJU Index for HOMA‐IR‐Defined Insulin Resistance in Patients With Type 2 Diabetes
Zhangna Zheng, Qianyi Li, Huijuan ZhangBackground
Routine assessment of insulin resistance (IR) remains challenging because fasting insulin measurements are not routinely available in many clinical settings. Several readily available metabolic indices, including the HbA1c/HDL‐C ratio, stress hyperglycemia ratio (SHR), and Zhejiang University (ZJU) index, have been associated with IR or related metabolic disturbances. However, their relative discriminative performance for HOMA‐IR‐defined IR in patients with established Type 2 diabetes has not been systematically compared. We aimed to compare their discriminative performance for HOMA‐IR‐defined IR and to assess whether ZJU provides incremental discrimination beyond conventional clinical and metabolic characteristics.
Methods
This was a single‐center retrospective cross‐sectional study. We included 1000 hospitalized adults with established T2DM. HOMA‐IR was calculated from fasting plasma glucose and fasting insulin, and IR was operationally defined as HOMA‐IR ≥ 2.69. Multivariable logistic regression and restricted cubic spline analyses evaluated associations and nonlinear associations. Receiver operating characteristic analyses with paired DeLong tests compared discriminative performance. Additional analyses included TyG and TyG‐BMI, comparisons with individual ZJU components, nested‐model analysis, decision‐curve analysis, bootstrap internal validation, and medication‐adjusted sensitivity analyses.
Results
All three indices were significantly associated with HOMA‐IR‐defined IR after multivariable adjustment. Compared with the lowest quartile, the adjusted odds ratios in the highest quartile were 3.16 (95% CI 1.95–5.18) for the HbA1c/HDL‐C ratio, 6.86 (4.11–11.84) for SHR, and 18.36 (10.21–34.79) for ZJU. Significant nonlinear associations were observed for all three indices. Among the evaluated composite indices, ZJU had the highest AUC (0.777, 95% CI 0.742–0.812), followed by TyG (0.749, 0.712–0.785), TyG‐BMI (0.733, 0.695–0.770), SHR (0.698, 0.659–0.738), and HbA1c/HDL‐C (0.615, 0.573–0.657). ZJU did not significantly outperform FPG alone (AUC 0.766; p = 0.509). Adding ZJU to the basic clinical model increased the AUC from 0.658 to 0.780 (ΔAUC = 0.123; p < 0.001), whereas adding ZJU to a more comprehensive model including its major metabolic components did not improve discrimination (0.806 vs. 0.807; p = 0.983).
Conclusions
In hospitalized patients with established T2DM, ZJU demonstrated moderate discrimination for HOMA‐IR‐defined IR and provided incremental discrimination beyond basic clinical characteristics.