Two Simple Insulin Resistance Scores Outperform TyG in Predicting Heart Events in Diabetic Heart Failure Patients

Patients who leave the hospital with both heart failure and type 2 diabetes face one of the most dangerous prognostic combinations in cardiovascular medicine, and clinicians have long lacked simple laboratory-based tools to identify who is most likely to deteriorate within months. A new multicenter retrospective cohort study from two tertiary hospitals in Zhejiang, China, published in Frontiers in Nutrition, suggests that two relatively obscure insulin resistance indices—SPISE-IR and METs-IR—may fill that gap, substantially outperforming the more widely studied triglyceride-glucose (TyG) index in predicting major adverse cardiovascular events (MACE) within 18 months of discharge.

The research team, led by investigators at the Second Affiliated Hospital of Zhejiang Chinese Medical University, enrolled 477 hospitalized patients with concurrent heart failure and type 2 diabetes between 2018 and 2024, applying unified inclusion criteria across both centers. The rationale rests on a growing body of evidence that insulin resistance is not merely a metabolic bystander in this population but a shared mechanistic driver of myocardial dysfunction, vascular injury, and inflammation. Because direct measurement of insulin sensitivity through clamp techniques is impractical in routine care, surrogate indices calculated from ordinary blood tests have become attractive candidates for bedside risk stratification.

Three indices were compared head-to-head. The TyG index, computed from fasting triglycerides and glucose, has accumulated considerable epidemiological support in general cardiovascular cohorts. METs-IR incorporates fasting glucose, triglycerides, body mass index, and high-density lipoprotein cholesterol into a weighted formula. SPISE-IR, the Single Point Insulin Sensitivity Estimator adapted for insulin resistance, draws on waist circumference, body mass index, triglycerides, high-density lipoprotein cholesterol, and fasting glucose to approximate whole-body insulin sensitivity from a single blood draw and anthropometric measurement. The question was whether any of these routinely available calculations could meaningfully sharpen prognostication beyond standard clinical variables in this high-risk population.

The primary endpoint was a composite of major adverse cardiovascular events—cardiovascular death, heart failure rehospitalization, and recurrent angina—within 18 months after discharge. During follow-up, 191 of the 477 patients, or 40.04 percent, experienced at least one such event, including three cardiovascular deaths. Notably, heart failure rehospitalization dominated the event profile, accounting for 138 of the 191 events, or 72.3 percent, underscoring that recurrent decompensation rather than sudden death drives the short-term burden in this cohort. Patients who experienced events were older, had lower left ventricular ejection fractions, worse renal function, and markedly higher NT-proBNP concentrations than those who remained event-free.

In fully adjusted multivariable Cox regression models, the divergence between the indices was striking. Each unit increase in SPISE-IR was associated with a hazard ratio of 9.31 for MACE (95 percent confidence interval 5.96 to 14.55, p less than 0.001), while METS-IR carried a hazard ratio of 1.07 per unit (95 percent confidence interval 1.05 to 1.09, p less than 0.001). The TyG index, by contrast, showed no statistically significant association (hazard ratio 1.19, 95 percent confidence interval 0.95 to 1.49, p equal to 0.127). Restricted cubic spline analyses revealed that both SPISE-IR and METs-IR bore nonlinear, dose-response relationships with event risk, meaning the hazard did not rise in a simple straight line but accelerated across certain ranges of the indices.

Predictive performance told a similar story. Areas under the receiver operating characteristic curve reached 0.790 for SPISE-IR and 0.801 for METs-IR, both comfortably in the range considered clinically useful, while TyG managed only 0.539—barely better than a coin flip. Precision-recall curves, decision curve analysis, and quartile-based Kaplan-Meier comparisons reinforced the pattern: patients in higher SPISE-IR and METs-IR quartiles experienced progressively higher cumulative event rates (log-rank p less than 0.001), whereas no significant between-quartile separation emerged for TyG. When either SPISE-IR or METs-IR was added to a baseline clinical model, both net reclassification improvement (NRI) and integrated discrimination improvement (IDI) rose significantly, with SPISE-IR delivering numerically larger gains on both metrics.

Subgroup analyses added nuance. The association between SPISE-IR and MACE remained consistent across categories of age, sex, smoking status, and hypertension, suggesting a robust signal that is not confined to a particular patient profile. METs-IR, however, exhibited effect modification—its association with events varied depending on age, sex, smoking status, and hypertension status—implying that clinicians applying this index may need to interpret values differently across patient strata. TyG remained largely non-significant throughout, failing to establish itself as a reliable prognostic marker in this specific population despite its popularity in broader cardiovascular epidemiology.

The investigators subjected their findings to several robustness checks. After excluding 24 recurrent-angina events to test a stricter MACE definition, SPISE-IR (hazard ratio 10.83, 95 percent confidence interval 6.71 to 17.45) and METs-IR (hazard ratio 1.07, 95 percent confidence interval 1.06 to 1.09) remained strongly associated with the endpoint, both at p less than 0.001. Internal validation through 1,000 bootstrap resamples produced optimism-corrected areas under the curve of 0.818 for a clinical model incorporating SPISE-IR and 0.806 for one incorporating METs-IR. Center-based external validation, using held-out data from the second hospital, yielded areas under the curve of 0.804 and 0.806 respectively, indicating that the models retained discriminative power outside the derivation setting.

Why might SPISE-IR and METs-IR outperform TyG in this population? The answer likely lies in what each formula captures. TyG reflects only the glucose-triglyceride axis, a narrow window on insulin action, whereas SPISE-IR and METs-IR integrate adiposity measures and high-density lipoprotein cholesterol—variables that encode the broader metabolic syndrome phenotype. In heart failure patients with type 2 diabetes, where cardiac dysfunction, volume status, renal impairment, and insulin resistance interact in complex feedback loops, a more comprehensive metabolic snapshot appears to carry more prognostic information. The nonlinear dose-response relationships further suggest threshold effects, where risk climbs disproportionately once insulin resistance surpasses certain levels.

The authors conclude that SPISE-IR and METs-IR, both calculable from routinely available clinical and laboratory data, may provide meaningful additional value for risk stratification in hospitalized patients with heart failure and type 2 diabetes. As retrospective cohort data from a single geographic region, the findings warrant prospective validation in diverse populations before widespread clinical adoption, but the effect sizes, validation statistics, and consistency across sensitivity analyses position these two indices as promising candidates for integration into discharge planning and follow-up intensity decisions for one of cardiology’s most vulnerable patient groups.

Subject of Research: Prognostic value of insulin resistance indices for short-term cardiovascular events in heart failure patients with type 2 diabetes

Article Title: Study on the correlation between SPISE-IR, METs-IR, TyG and short-term adverse cardiovascular events in patients with heart failure and diabetes mellitus

Article References: Deng, H., Zhu, Q., Lin, C., Qiu, W., Zhang, Y., & Hua, J. (2026). Study on the correlation between SPISE-IR, METs-IR, TyG and short-term adverse cardiovascular events in patients with heart failure and diabetes mellitus. Frontiers in Nutrition, 13, Article 1938459. https://doi.org/10.3389/fnut.2026.1938459

Image Credits: AI Generated

DOI: 10.3389/fnut.2026.1938459

Keywords: heart failure, type 2 diabetes, insulin resistance, SPISE-IR, METs-IR, TyG index, major adverse cardiovascular events, risk stratification, prognostic prediction, retrospective cohort study, NT-proBNP, rehospitalization