Predictive value of three insulin resistance surrogate indices for symptomatic hypoglycemia risk in hospitalized patients with type 2 diabetes: a retrospective cohort study

Inpatient hypoglycemia is a serious complication in hospitalized patients with type 2 diabetes mellitus (T2DM), associated with adverse outcomes. While insulin resistance (IR) is central to hyperglycemia, its relationship with hypoglycemia risk in acute care settings remains unclear. Convenient IR surrogate indices, such as the triglyceride-glucose index (TyG), triglyceride to high-density lipoprotein cholesterol ratio (TG/HDL-C), and metabolic syndrome insulin resistance index (METS-IR), warrant evaluation for hypoglycemia prediction. To evaluate and compare the predictive utility of three IR indices (TyG, TG/HDL-C, METS-IR) for in-hospital symptomatic hypoglycemia in T2DM patients. This retrospective cohort study included 660 hospitalized T2DM patients (April 2024–April 2025) with ≥8-day stay and valid CGM data. Symptomatic hypoglycemia was defined by CGM < 3.9 mmol/L, capillary glucose confirmation, and clinical symptoms. Kaplan-Meier cumulative incidence analysis, multivariable Cox regression, restricted cubic splines, time-dependent ROC, and decision curve analysis were used. During follow-up, 193 (29.2%) patients developed symptomatic hypoglycemia. TyG and METS-IR were inversely associated with symptomatic hypoglycemia risk (per 1-SD increase: HR = 0.680, 95%CI = 0.600–0.773; HR = 0.748, 95%CI = 0.680–0.823), with linear dose-response relationships. TG/HDL-C showed no significant association. TyG exhibited superior model fit (lower AIC), better predictive performance on hospital day 2 (AUC = 0.721), and higher net clinical benefit at 20%–30% risk thresholds. TyG may be more closely linked to symptomatic hypoglycemia risk in patients with good glycemic control, while METS-IR may show a stronger association in those with a longer diabetes duration. TyG and METS-IR are valid predictors of in-hospital symptomatic hypoglycemia in T2DM patients. TyG demonstrates optimal predictive performance and clinical utility, supporting its integration into early risk stratification for individualized glycemic management.