Combining Insulin Resistance and Renal Function to Predict Cardiovascular Events: A Multicenter Cohort Study

Ge Zhang; Tianshu Gu; Kaisaierjiang Kadier; Zhijie Zhao; Qian Guo; Shanshan Cai; Wenming He; Fangkun Yang; Bo Yang; Chaoyang Yu

doi:10.15212/bioi-2026-0017

Article · 2026 Paper GEO V2

Combining Insulin Resistance and Renal Function to Predict Cardiovascular Events: A Multicenter Cohort Study

BIO Integration

Full Authors

Ge Zhang
Tianshu Gu
Kaisaierjiang Kadier
Zhijie Zhao
Qian Guo
Shanshan Cai
Wenming He
Fangkun Yang
Bo Yang
Chaoyang Yu

DOI / publisher Google Scholar search Cite

Official Abstract

Background

Insulin resistance (IR) and chronic kidney disease (CKD) are independent risk factors for cardiovascular disease (CVD). The triglyceride-glucose (TyG) index is recognized as a convenient marker for IR, while the estimated glomerular filtration rate based on both creatinine and cystatin C (eGFRcr-cys) is commonly used to evaluate kidney function. Therefore, the TyG:eGFRcr-cys ratio is introduced with the aim of obtaining a more effective predictor for CVD risk.

Method

This prospective cohort analysis included 261,865 UK Biobank participants with available information. Cox proportional hazards models were used to evaluate the associations of TyG, eGFRcr-cys, and the TyG:eGFRcr-cys ratio with incident CVD. Restricted cubic splines for model fitting with three knots placed at the 10th, 50th, 90th percentiles were used to determine the non-linear relationship between the TyG:eGFRcr-cys ratio and CVD. Furthermore, subgroup and sensitivity analyses were performed to illustrate the disparate associations across diverse groups and to reinforce the conclusions drawn, respectively. An independent single-center clinical cohort was assembled from the First Affiliated Hospital of Zhengzhou University for external validation by enrolling consecutive patients between 2018 and 2020 with baseline TyG index and eGFRcr-cys measurements and up to 4 years of follow-up. External validation with major adverse cardiovascular events was further performed as the outcome and observed consistent dose–response associations and robust risk stratification for the TyG:eGFRcr-cys ratio.

Result

Following extensive covariate adjustment, the hazard ratios [HRs] (95% confidence intervals [CIs]) for total CVD across increasing quartiles of the TyG:eGFRcr-cys ratio, with the lowest quartile (Q1) as the reference, were 1.15 (1.10–1.21) for Q2 versus Q1, 1.17 (1.12–1.23) for Q3 versus Q1, and 1.24 (1.18–1.30) for Q4 versus Q1, respectively, for the TyG:eGFRcr-cys ratio. In addition, each standard deviation increase in the TyG:eGFRcr-cys ratio was associated with a higher risk of CVD, coronary heart disease, and stroke, corresponding to estimated increases of 249% [HR, 3.49; 95% CI, 2.94–4.15], 250% [HR, 3.50; 95% CI, 2.92–4.20], and 165% [HR, 2.65; 95% CI, 1.80–3.90], respectively, in the fully adjusted models. The non-linear relationship between the TyG:eGFRcr-cys ratio and CVD (P-value for the overall model < 0.001 and P-value for the non-linear model < 0.001) indicated that the CVD risk increased as the TyG:eGFRcr-cys ratio increased. Subgroup analysis results suggested that these associations were more pronounced in White individuals and females.

Conclusion

As the TyG:eGFRcr-cys ratio increased, the risk of CVD also increased with a higher likelihood in females and White individuals.

Text reproduced verbatim from Version of Record (source); DOI; CC-BY-4.0.

Research Question

Is an index combining the triglyceride–glucose marker with creatinine–cystatin-C-based estimated renal function associated with incident cardiovascular events, and is the association reproduced in an independent clinical cohort?

Author Evidence Summary

The study analyzed baseline metabolic and renal markers in UK Biobank and examined an independent hospital cohort. Higher ratio quartiles were associated with overall CVD and CHD; the fully adjusted highest-versus-lowest-quartile stroke association was not statistically significant. The external complete-case analysis reproduced a ratio–MACE association but was small and imprecise. The findings support further evaluation of a complementary cardiometabolic–renal marker. They do not establish causal mechanisms, a universal treatment threshold, improved decisions, or incremental prediction beyond established risk scores.

Cite this paper

Plain citation

Ge Zhang, Tianshu Gu, Kaisaierjiang Kadier, Zhijie Zhao, Qian Guo, Shanshan Cai, Wenming He, Fangkun Yang, Bo Yang, Chaoyang Yu. Combining Insulin Resistance and Renal Function to Predict Cardiovascular Events: A Multicenter Cohort Study. BIO Integration. 2026. https://doi.org/10.15212/bioi-2026-0017

Author-controlled citation; verify against the publisher version of record.

BibTeX RIS / EndNote CSL JSON DOI

Research & review context

Research fields: Cardiovascular epidemiology
Cardiometabolic–renal markers
Review topics: Combined metabolic and renal risk assessment
External association-replication studies
Scientific role: Prospective observational biomarker-association study with independent clinical replication and exploratory network analysis.
Evidence position: Human cohort association evidence, distinct from incremental prediction validation, causal inference, or treatment effects.
Related concepts: TyG:eGFRcr-cys
MACE
UK Biobank
Possible citation contexts: Reviews evaluating integrated metabolic–renal markers for incident cardiovascular outcomes.
Discussions distinguishing independent association replication from validated clinical prediction and incremental decision value.

Possible literature-review contexts, not evidence of existing citations. Read the findings and Evidence Scope before citing the original article.

Questions & evidence · Evidence limits · Original article & citation downloads

Evidence Scale

TyG:eGFRcr-cys and cardiovascular events: prospective UK Biobank association with independent clinical replication

A higher TyG:eGFRcr-cys ratio was associated with incident CVD, particularly CHD, in a prospective UK Biobank analysis. A smaller independent clinical cohort supported an association with MACE. The study did not evaluate whether the ratio improves discrimination or reclassification beyond established cardiovascular risk factors or scores.

Primary cohort: UK Biobank analytic sample: 261,865 participants
External source cohort: 1,798 initially assembled clinical participants
External fully adjusted analysis: 328 complete cases, including 88 MACE outcomes

Counting note

Primary and external cohort counts are kept separate. The external complete-case group is a subset of the initially assembled hospital cohort. No synthetic total combines the populations or adds the subset to its parent cohort. External validation here means association replication, not an independently calibrated clinical prediction tool.

Key Findings

KF1 The study combined a UK Biobank prospective analysis with an independent clinical association-replication cohort.

The primary and external populations are distinct and are reported separately. The external fully adjusted complete-case analysis represents a subset of the initially assembled clinical cohort.

Evidence for KF1
Measure	Value
UK Biobank analytic participants	261,865
External cohort initially assembled	1,798
External complete-case analysis	328 participants, including 88 MACE outcomes
External recruitment and follow-up	2018–2020 enrollment; up to 4 years of follow-up

Source locator: Methods: Study design and population, Primary cohort and External validation cohort; Results: Population characteristics; Table 1; Figure 5

Permalink to KF1

KF2 Higher TyG:eGFRcr-cys quartiles were associated with higher incident CVD hazards in UK Biobank.

Fully adjusted Model 3; the lowest ratio quartile is the reference. Hazard ratios describe observational associations, not causal effects or a validated clinical risk score.

Evidence for KF2
Measure	Value
Q2 versus Q1 CVD HR	1.15; 95% CI 1.10–1.21
Q3 versus Q1 CVD HR	1.17; 95% CI 1.12–1.23
Q4 versus Q1 CVD HR	1.24; 95% CI 1.18–1.30

Source locator: Abstract: Result; Results: Associations between the TyG index and the TyG:eGFRcr-cys ratio with CVD; Table 2, CVD quartile rows, Model 3

Permalink to KF2

KF3 The highest-quartile association was clearer for CHD than for stroke.

Same highest-versus-lowest ratio-quartile contrast in fully adjusted Model 3. The stroke confidence interval includes the null; other exposure contrasts should not be treated as interchangeable with this quartile comparison.

Evidence for KF3
Measure	Value
Q4 versus Q1 CHD HR	1.30; 95% CI 1.23–1.37
Q4 versus Q1 stroke HR	1.05; 95% CI 0.96–1.16
Stroke Q4 versus Q1 P value	0.29

Source locator: Results: Associations between the TyG index and the TyG:eGFRcr-cys ratio with CVD; Table 2, CHD and STROKE quartile-4 rows, Model 3

Permalink to KF3

KF4 The ratio–MACE association was also reported in the external clinical complete-case cohort, with substantial uncertainty in effect magnitude.

External Cox Model 3; per-standard-deviation contrast, distinct from the UK Biobank quartile comparison. This is independent association replication, not proof of incremental predictive utility or calibrated individual risk.

Evidence for KF4
Measure	Value
TyG:eGFRcr-cys HR per 1-SD increase	6.04; 95% CI 2.12–17.19
External analysis scale	328 complete cases; 88 MACE outcomes

Source locator: Methods: External validation statistical analysis; Results: External validation; Figure 5D and caption; Methods: External validation cohort

Permalink to KF4

KF5 Dynamic network biomarker analysis identified a possible transition-related pattern across ratio-based risk strata.

Participants were ordered by between-person exposure quartiles. The exploratory network pattern is not longitudinal demonstration of an individual patient entering a pre-disease state.

Evidence for KF5
Measure	Value
Risk strata	Four ratio quartiles, Q1–Q4
Composite-index pattern	Highest reported value in Q3
Interpretation	Exploratory network instability across exposure strata; not a proven mechanism or clinical early-warning threshold

Source locator: Methods: Dynamic network biomarker analysis; Results: Dynamic network biomarker analysis suggested a potential transition stage during CVD risk progression; Figure 6B–E

Permalink to KF5

Study Design & Analytical Framework

Study profile

Study Design: Prospective observational UK Biobank analysis with Cox models, restricted cubic splines, subgroup/sensitivity analyses, independent single-center clinical association replication, and exploratory network analysis across ratio quartiles.
Evidence Type: Longitudinal observational associations and independent clinical association replication
Population: UK Biobank adults without baseline CVD; a separately assembled Zhengzhou University vascular-surgery clinical cohort with complete cases used for the fully adjusted external analysis.
Primary Endpoint: Incident fatal or nonfatal CHD or stroke in the primary cohort; MACE defined as fatal or nonfatal CHD or stroke in the external cohort.
Secondary Endpoint: CHD and stroke separately; dose–response patterns and exploratory network instability across exposure-defined strata.
External dataset evaluation: Yes

Data modalities

Baseline blood triglycerides and glucose
Creatinine and cystatin-C-based estimated GFR
Demographic, lifestyle, medication, and clinical covariates
Hospital and death-registry outcome codes
External clinical cohort follow-up

What This Study Adds

Examines a combined metabolic–renal marker rather than treating its metabolic and renal components as unrelated domains.
Separates prospective primary-cohort associations from an independent, smaller hospital-cohort analysis.
Reports outcome-specific associations and explicitly leaves incremental predictive value beyond established risk scores unresolved.

Evidence Scope

Supports

Prospective observational associations between the baseline ratio and incident CVD/CHD in the analyzed UK Biobank population.
Independent association replication for MACE in a small, complete-case hospital cohort.
Exploratory network analysis and further research into integrated cardiometabolic–renal markers.

Does Not Establish

A causal effect of the ratio on cardiovascular disease or a confirmed biological mechanism.
A statistically significant highest-versus-lowest-quartile stroke association in fully adjusted Model 3.
An independently calibrated individual-risk calculator, improved C-statistic/NRI/IDI, or superiority to established risk scores.
A universal clinical cut-off, treatment efficacy, or benefit from targeting the ratio.
That exposure-quartile network differences demonstrate within-person disease transitions.

Limitations

The observational design cannot establish causation and residual confounding remains possible.
UK Biobank selection and demographic composition limit generalizability.
The external cohort was from one clinical center; only a complete-case subset entered the fully adjusted analysis, limiting precision and representativeness.
The source reports different results for continuous, quartile, and spline contrasts; these should not be collapsed into one uniform association.
Incremental discrimination, calibration, reclassification, and clinical decision benefit beyond established risk scores were not established.
Exploratory network comparisons between exposure quartiles cannot demonstrate longitudinal within-person phase transitions.

When This Study Is Useful to Cite

CU1: What prospective evidence links a combined metabolic–renal ratio with cardiovascular outcomes?
UK Biobank observational associations, preserving null or weaker outcome-specific contrasts.
Evidence: KF1, KF2, KF3
CU2: Was the association examined in an independent clinical population?
Small complete-case external-cohort association replication, without demonstrated incremental predictive utility.
Evidence: KF4

What This Study Should Not Be Cited to Claim

A causal effect of the ratio on cardiovascular disease or a confirmed biological mechanism.
A statistically significant highest-versus-lowest-quartile stroke association in fully adjusted Model 3.
An independently calibrated individual-risk calculator, improved C-statistic/NRI/IDI, or superiority to established risk scores.
A universal clinical cut-off, treatment efficacy, or benefit from targeting the ratio.
That exposure-quartile network differences demonstrate within-person disease transitions.

Evidence Matrix

Row ID	Component	Context	Finding	Evidence Level	Scope	Source Locator	Evidence refs
evidence-bioi0017-1	The study combined a UK Biobank prospective analysis with an independent clinical association-replication cohort.	The primary and external populations are distinct and are reported separately. The external fully adjusted complete-case analysis represents a subset of the initially assembled clinical cohort.	UK Biobank analytic participants: 261,865; External cohort initially assembled: 1,798; External complete-case analysis: 328 participants, including 88 MACE outcomes; External recruitment and follow-up: 2018–2020 enrollment; up to 4 years of follow-up	Prospective observational cohort association	The primary and external populations are distinct and are reported separately. The external fully adjusted complete-case analysis represents a subset of the initially assembled clinical cohort.	Methods: Study design and population, Primary cohort and External validation cohort; Results: Population characteristics; Table 1; Figure 5	KF1
evidence-bioi0017-2	Higher TyG:eGFRcr-cys quartiles were associated with higher incident CVD hazards in UK Biobank.	Fully adjusted Model 3; the lowest ratio quartile is the reference. Hazard ratios describe observational associations, not causal effects or a validated clinical risk score.	Q2 versus Q1 CVD HR: 1.15; 95% CI 1.10–1.21; Q3 versus Q1 CVD HR: 1.17; 95% CI 1.12–1.23; Q4 versus Q1 CVD HR: 1.24; 95% CI 1.18–1.30	Prospective observational cohort association	Fully adjusted Model 3; the lowest ratio quartile is the reference. Hazard ratios describe observational associations, not causal effects or a validated clinical risk score.	Abstract: Result; Results: Associations between the TyG index and the TyG:eGFRcr-cys ratio with CVD; Table 2, CVD quartile rows, Model 3	KF2
evidence-bioi0017-3	The highest-quartile association was clearer for CHD than for stroke.	Same highest-versus-lowest ratio-quartile contrast in fully adjusted Model 3. The stroke confidence interval includes the null; other exposure contrasts should not be treated as interchangeable with this quartile comparison.	Q4 versus Q1 CHD HR: 1.30; 95% CI 1.23–1.37; Q4 versus Q1 stroke HR: 1.05; 95% CI 0.96–1.16; Stroke Q4 versus Q1 P value: 0.29	Prospective observational cohort association	Same highest-versus-lowest ratio-quartile contrast in fully adjusted Model 3. The stroke confidence interval includes the null; other exposure contrasts should not be treated as interchangeable with this quartile comparison.	Results: Associations between the TyG index and the TyG:eGFRcr-cys ratio with CVD; Table 2, CHD and STROKE quartile-4 rows, Model 3	KF3
evidence-bioi0017-4	The ratio–MACE association was also reported in the external clinical complete-case cohort, with substantial uncertainty in effect magnitude.	External Cox Model 3; per-standard-deviation contrast, distinct from the UK Biobank quartile comparison. This is independent association replication, not proof of incremental predictive utility or calibrated individual risk.	TyG:eGFRcr-cys HR per 1-SD increase: 6.04; 95% CI 2.12–17.19; External analysis scale: 328 complete cases; 88 MACE outcomes	Prospective observational cohort association	External Cox Model 3; per-standard-deviation contrast, distinct from the UK Biobank quartile comparison. This is independent association replication, not proof of incremental predictive utility or calibrated individual risk.	Methods: External validation statistical analysis; Results: External validation; Figure 5D and caption; Methods: External validation cohort	KF4
evidence-bioi0017-5	Dynamic network biomarker analysis identified a possible transition-related pattern across ratio-based risk strata.	Participants were ordered by between-person exposure quartiles. The exploratory network pattern is not longitudinal demonstration of an individual patient entering a pre-disease state.	Risk strata: Four ratio quartiles, Q1–Q4; Composite-index pattern: Highest reported value in Q3; Interpretation: Exploratory network instability across exposure strata; not a proven mechanism or clinical early-warning threshold	Prospective observational cohort association	Participants were ordered by between-person exposure quartiles. The exploratory network pattern is not longitudinal demonstration of an individual patient entering a pre-disease state.	Methods: Dynamic network biomarker analysis; Results: Dynamic network biomarker analysis suggested a potential transition stage during CVD risk progression; Figure 6B–E	KF5

Q&A

How many participants contributed to the analyses?

The UK Biobank analytic cohort contained 261,865 participants. The external hospital cohort initially contained 1,798 participants, but the fully adjusted external analysis used 328 complete cases, including 88 MACE outcomes. These populations and subsets should not be added into a synthetic study total.

Evidence: KF1

What does the TyG:eGFRcr-cys ratio combine?

It divides the triglyceride–glucose index by estimated GFR derived using both creatinine and cystatin C. It is intended to summarize metabolic and renal risk-marker information, rather than directly measure a causal process.

Evidence: KF1, KF2

What was the main prospective CVD association?

In fully adjusted Model 3, the highest ratio quartile had a CVD HR of 1.24 compared with the lowest quartile, with a 95% CI of 1.18–1.30. This is an observational hazard ratio.

Evidence: KF2

Were all cardiovascular outcomes significantly associated in the quartile comparison?

No. The highest-versus-lowest-quartile CHD HR was 1.30, while the stroke HR was 1.05 with a 95% CI of 0.96–1.16 and P=0.29. Outcome and exposure-contrast details must be preserved.

Evidence: KF3

What does external validation mean here?

It means that the ratio–MACE association was examined in an independent hospital cohort. The fully adjusted complete-case analysis was small and the confidence interval was wide; this does not establish a calibrated prediction model or improved clinical decision-making.

Evidence: KF4

Does the network analysis establish a pre-disease transition?

No. The analysis compared exposure-defined groups and was exploratory. It does not show that an individual crosses a validated early-warning threshold over time.

Evidence: KF5

Can the ratio replace established cardiovascular risk scores?

The article states that incremental discrimination and reclassification beyond established risk factors or scores were not evaluated. The ratio remains a candidate complementary marker requiring additional evaluation.

Evidence: KF2, KF4

Concepts & Entities

Markers

Triglyceride–glucose indexTyG:eGFRcr-cys ratioCreatinine–cystatin-C-based estimated GFR

Outcomes

Incident cardiovascular diseaseCoronary heart diseaseStrokeMajor adverse cardiovascular events

Methods

UK BiobankProspective cohortCox proportional-hazards modelIndependent association replicationDynamic network biomarkers

Related Research

The cholesterol, high-density lipoprotein, and glucose index and its derivatives related to obesity and inflammation as novel metabolic markers for predicting new-onset cardiovascular-liver-metabolic multimorbidity: A prospective cohort study
A separate CHARLS analysis of metabolic indices addresses a different surrogate-defined multimorbidity endpoint; it is related marker research rather than replication of this ratio. · DOI: 10.1111/jdi.70462
AI hybrid survival assessment for advanced heart failure patients with renal dysfunction
A distinct heart-failure prognosis study illustrates a different clinical outcome and model-development setting; its performance does not validate this ratio. · DOI: 10.1038/s41467-024-50415-9

Publication & Provenance

DOI: https://doi.org/10.15212/bioi-2026-0017
Publisher: https://bio-integration.org/10-15212-bioi-2026-0017/
Doi Url: https://doi.org/10.15212/bioi-2026-0017
Version Of Record: BIO Integration 2026; volume 7, pages 1–17; published online 15 June 2026
License: Creative Commons Attribution 4.0 International (CC BY 4.0)
License Url: https://creativecommons.org/licenses/by/4.0/
Data Source: UK Biobank application 102158; independent First Affiliated Hospital of Zhengzhou University vascular-surgery cohort.
External Cohort Counting: 1,798 initially assembled; 328 complete cases and 88 MACE events for the fully adjusted external analysis.
Supporting Material Url: https://bio-integration.org/wp-content/uploads/2026/06/bioi20260017_Supplemental.zip
Supporting Material Note: Official supplementary link identified; direct retrieval returned HTTP 406. The claims on this page are supported by the accessible official full-text Methods, Results, Table 2, and figure captions.
Source Note Table3: The official HTML Table 3 former-smoker confidence interval is malformed; that subgroup estimate is not used on this evidence page.
Source Note Subgroups: Subgroup descriptions are not promoted to causal or mechanistic claims, and demographic subgroup estimates are not used for clinical recommendations.
Author Identity: Ge Zhang is first and corresponding author, affiliated with Zhengzhou University; the article gives the same public correspondence identity and the DOI appears in ORCID 0000-0002-3116-3246.
Evidence Basis: Official publisher full-text HTML read through the web tool, Crossref metadata, and public ORCID works. Private raw web-tool text and source hashes retain the reviewed version; no full-text PDF was uploaded.

This is an original evidence synthesis linked to the published article. Cite the original article for cohort findings. Cohort/subset counts, outcome-specific results, and the distinction between association replication and clinical prediction are preserved.

Study-level evidence table

Study-level author-maintained evidence summary. Not participant-level data and not a ready-to-pool meta-analysis dataset. For formal citation, use the original article DOI.

Download study-level evidence CSV (UTF-8)

Formula safety: cells starting with =, +, -, @ (after leading whitespace), or tab/CR/LF are prefixed with an apostrophe. Treat all cells as text; remove only that safety prefix when recovering the original string.

Verify publication identity

These records identify the same original work; database inclusion is not evidence of scientific validity or clinical benefit.

OpenAlex record

All Publications Homepage