The study, led by Madhulika L. Mahashabde and Sai Karthik Guniganti of the Department of General Medicine at Dr. D.Y. Patil Vidyapeeth in Pune, enrolled 110 adults with type 2 diabetes and compared four candidate markers head to head. The prespecified composite outcome captured the presence of at least one of the three major microvascular complications of diabetes: diabetic retinopathy, diabetic kidney disease and diabetic peripheral neuropathy. Among the participants, 64, or 58.2 percent, had at least one such complication. The researchers then asked a deceptively simple question with profound clinical implications: which of these four blood-based numbers best separates patients with microvascular damage from those without it?

The answer, measured by the area under the receiver operating characteristic curve, or AUC, was remarkable. The triglyceride-glucose index, known as TyG, achieved an AUC of 0.993 with a 95 percent confidence interval stretching from 0.984 to 1.000, a figure so close to perfect discrimination that the authors themselves flag it as unusually high. Glycated haemoglobin, the gold-standard marker of long-term blood sugar control, reached 0.971. The triglyceride-to-HDL-cholesterol ratio followed at 0.959. The homeostatic model assessment of insulin resistance, HOMA-IR, which requires a fasting insulin measurement and has long been considered the reference standard for quantifying insulin resistance in research settings, trailed at 0.810.

Statistical rigour underpinned these comparisons. The team used DeLong confidence intervals for each AUC and then applied DeLong’s test for all six pairwise comparisons, adjusting the resulting p-values with the Holm-Bonferroni method to control the family-wise error rate. After this adjustment, TyG, TG/HDL-C and HbA1c each outperformed HOMA-IR with adjusted p-values below 0.001, while no statistically significant differences separated the three leading markers from one another. In other words, the two insulin-free surrogates, calculated from fasting biochemistry that costs a fraction of an insulin assay, performed at least as well as the established glycemic marker and clearly better than the insulin-based index they were designed to replace.

The biological logic behind these indices is worth unpacking. Insulin resistance drives a characteristic lipid pattern: the liver pumps out more very-low-density lipoprotein particles, triglycerides rise, and high-density lipoprotein cholesterol falls. The TyG index captures this by combining fasting triglycerides with fasting plasma glucose in a logarithmic formula, reflecting the fact that insulin-resistant tissue allows both circulating fats and sugars to accumulate. The TG/HDL-C ratio distils the same physiology into a single fraction. Because insulin resistance is increasingly recognised as a contributor to microvascular complications, including retinopathy, nephropathy and neuropathy, these surrogates may be flagging the underlying metabolic milieu that damages small vessels, rather than merely the degree of glycemic excursion that HbA1c measures.

Exploratory cut-offs emerged from the analysis using the Youden index, which maximises the sum of sensitivity and specificity. For TyG, the cut-off was 9.23 or above, at which the index showed 93.8 percent sensitivity and 100 percent specificity for identifying prevalent microvascular complications. For TG/HDL-C, the exploratory threshold was 3.40 or above. These figures, if they hold up under external validation, would mean a single fasting blood panel could triage patients for complication screening without waiting for an insulin assay or even a glycated haemoglobin result.

Beyond discrimination, the researchers examined whether these markers add genuine incremental value to existing models. Multivariable associations were assessed using Firth’s penalised-likelihood regression, a technique designed for small samples where conventional maximum-likelihood logistic regression can produce unstable or infinitely large coefficient estimates. Each one-standard-deviation rise in TyG and TG/HDL-C remained significantly associated with the composite outcome after Firth penalisation, with p-values of 0.002 and 0.003 respectively. Calibration was assessed using the intercept, slope and Brier score, and decision curve analysis explored the net benefit of using each marker at various threshold probabilities. Adding TyG to a base model containing HbA1c and HOMA-IR yielded an integrated discrimination improvement, or IDI, of 0.219, indicating that TyG carries information about microvascular complications that the established markers do not fully capture.

The authors are careful, and rightly so, about the limits of what this study can claim. The unusually high AUCs, particularly TyG’s 0.993, may reflect the disease spectrum of a tertiary-care referral population, where patients tend to have more advanced disease and more pronounced metabolic derangement than the general diabetic population. The researchers explicitly state that these figures should be regarded as upper bounds rather than realistic expectations of performance in community settings. The cross-sectional design means the study identifies prevalent complications at a single point in time; it makes no claim of predictive or screening performance, and the exploratory cut-offs require prospective, multicentre external validation before any triage use.

India carries a disproportionate share of the global diabetes burden, and microvascular complications, including blindness, kidney failure requiring dialysis and disabling neuropathy, impose enormous human and economic costs. If the Pune findings replicate in larger, multicentre, prospective cohorts, the implications for resource-constrained health systems could be transformative. A fasting blood panel already drawn for routine lipid and glucose monitoring could double as a complication triage tool, flagging patients who need retinal imaging, urinary albumin-to-creatinine ratio testing or nerve conduction studies, without adding a single new test to the workflow.

For now, the study stands as a provocative proof of concept from Western India: two numbers calculated from tests already sitting in every clinic’s biochemistry report, the triglyceride-glucose index and the triglyceride-to-HDL-cholesterol ratio, matched glycated haemoglobin and beat the insulin-based HOMA-IR in identifying the microvascular fingerprints of type 2 diabetes. The road from exploratory cut-offs to validated triage tools runs through prospective validation, but the destination, a cheaper, faster and more accessible way to spot diabetes’ most feared complications, is one worth watching closely.

Subject of Research: Triglyceride–glucose index and triglyceride/HDL-cholesterol ratio versus HbA1c and HOMA-IR for identifying microvascular complications in type 2 diabetes: a cross-sectional comparative diagnostic accuracy study from Western India

Article Title: Triglyceride–glucose index and triglyceride/HDL-cholesterol ratio versus HbA1c and HOMA-IR for identifying microvascular complications in type 2 diabetes: a cross-sectional comparative diagnostic accuracy study from Western India

Article References: Mahashabde, M. L., & Guniganti, S. K. (2026). Triglyceride–glucose index and triglyceride/HDL-cholesterol ratio versus HbA1c and HOMA-IR for identifying microvascular complications in type 2 diabetes: a cross-sectional comparative diagnostic accuracy study from Western India. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02674-4

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02674-4

Keywords: Type 2 Diabetes, Insulin Resistance, TyG Index, HbA1c, Diagnostic Accuracy, Diabetic Complications, Clinical Research, Endocrinology, Triglyceride, glucose, index, HDL-cholesterol