Triglyceride-glucose index and mortality in critically ill patients with cardiovascular-kidney-metabolic syndrome stage 4

Abstract

The triglyceride–glucose (TyG) index is a well-established surrogate marker of insulin resistance (IR). Previous studies have linked higher TyG levels to an increased risk of cardiovascular events in individuals with early-stage (0–3) Cardiovascular–Kidney–Metabolic Syndrome (CKM). However, its prognostic value in critically ill patients with CKM stage 4 remains unclear. In this retrospective study, we analyzed clinical data from critically ill CKM stage 4 patients in the MIMIC-IV database. The TyG index was calculated and patients were categorized into tertiles. Cox proportional hazards models and restricted cubic spline (RCS) analyses were used to evaluate the association between the TyG index and all-cause mortality. A total of 3,125 patients were included, of whom 65.22% were male.= The 1-year all-cause mortality was 34.18% overall (28.79% in Q1, 35.06% in Q2, and 38.68% in Q3; P < 0.05). In multivariable Cox regression, each 1–standard deviation increase in the TyG index was associated with a 23% higher risk of 1-year mortality (HR = 1.23, 95% CI 1.09–1.39). Compared with the Q1 group, patients in Q3 had a 30% higher 1-year mortality risk (HR = 1.30, 95% CI 1.08–1.55). RCS analysis showed a linear positive relationship between the TyG index and mortality (P for non-linearity > 0.05). These findings indicate that higher TyG levels are independently associated with increased short- and long-term mortality in critically ill patients with CKM stage 4. The TyG index may provide a simple indicator of acute metabolic disturbances and could support early risk stratification in this population. However, its role in clinical decision-making and potential utility in guiding interventions require confirmation in prospective studies.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-025-03061-4.

Introduction

CKM is a complex, multisystem condition characterized by the bidirectional interplay of metabolic dysfunction, cardiovascular disease (CVD), and chronic kidney disease (CKD). These components collectively accelerate one another’s progression and substantially increase hospitalization, adverse events, and mortality risk [1–3]. With the global rise in obesity, diabetes, hypertension, and population aging, the prevalence of CKM has grown steadily, posing a major public health challenge that requires urgent attention [4, 5].

In 2023, the American Heart Association (AHA) released an updated CKM classification system encompassing five stages (0–4), with CKM stage 4 representing the most severe form, defined by clinical CVD in the presence of metabolic dysregulation and/or CKD [6]. Epidemiological data suggest that CKM stage 4 affects approximately 9% of U.S. adults and is associated with markedly poor outcomes, particularly among patients requiring intensive care [7–9]. Early identification of high-risk individuals is therefore critical; however, reliable biomarkers capable of predicting prognosis in critically ill CKM stage 4 patients remain limited.

Insulin resistance (IR)—a central pathophysiological feature of CKM—promotes chronic inflammation, oxidative stress, endothelial dysfunction, atherosclerosis progression, and renal injury, ultimately exacerbating multiorgan deterioration [10–12]. The triglyceride-glucose (TyG) index, a validated surrogate marker of IR, has been consistently linked to incident CVD, CKD, metabolic syndrome, and increased mortality in populations with CKM stages 0–3 or in general cardiovascular/metabolic cohorts [13–16]. Elevated TyG levels reflect heightened metabolic stress and oxidative burden, both of which contribute to vascular damage and worse clinical outcomes [15, 17, 18].

Despite these findings, evidence regarding the prognostic value of the TyG index in CKM stage 4 is extremely limited, especially among critically ill patients. Previous studies have often relied on admission measurements or focused on earlier CKM stages, which may not accurately capture metabolic status during critical illness [14]. Furthermore, the dose–response relationship between TyG and mortality, and whether this relationship differs across subgroups (e.g., diabetes, hypertension), remains unclear. As a result, the clinical utility of the TyG index for early risk stratification in CKM stage 4 patients requiring ICU care is still uncertain.

To address this knowledge gap, we conducted a retrospective cohort study using the MIMIC-IV database to evaluate the association between the TyG index measured within the first 24 h of ICU admission and short-term (30-day) and long-term (365-day) all-cause mortality among critically ill patients with CKM stage 4. We further examined potential nonlinear relationships using restricted cubic spline (RCS) models and explored heterogeneity across clinically relevant subgroups. This study aims to provide new evidence on the association between the TyG index and mortality in critically ill patients with CKM stage 4 and to explore whether the TyG index may have potential as a simple, rapid bedside marker for early risk stratification in this high-risk population.

Methods

Data source

This retrospective cohort study used data from the MIMIC-IV 3.1 database, which is jointly maintained by the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center [19]. The database includes de-identified clinical data from patients hospitalized at Beth Israel Deaconess Medical Center between 2008 and 2022. Since all data were de-identified, the study was exempt from obtaining informed consent from the patients (Database Usage Qualification ID: 70985586).

Study design and population

We selected adult patients from the MIMIC-IV 3.1 database (2008–2022) who met the following criteria: (1) first ICU admission for ≥ 24 h; (2) age ≥ 18 years; (3) first recorded triglyceride (TG) and glucose (Glu) measurements within 24 h of ICU admission; (4) met the 2023 AHA definition of CKM stage 4, defined as the presence of both: (i) metabolic risk or CKD [BMI ≥ 25 kg/m², or ≥ 3 components of AHA/NHLBI metabolic syndrome, or type 2 diabetes mellitus (T2DM), or eGFR < 60 mL·min⁻¹·1.73 m⁻²/UACR ≥ 30 mg/g], and (ii) clinical cardiovascular disease (CVD), including coronary artery disease, myocardial infarction, heart failure, stroke, peripheral artery disease, or atrial fibrillation, as per ICD-9/10 diagnosis codes. We excluded patients with duplicate ICU admissions or missing TG/Glu data. A total of 3,125 patients were included in the analysis and categorized into groups based on the first TyG index tertiles after ICU admission (Fig. 1).

Fig. 1.

Flowchart of patient selection. From the MIMIC-IV database (version 3.1), adult patients (≥ 18 years) with CKM stage 4 who were admitted to the ICU were initially identified (n = 12,272). Patients were excluded if they had multiple ICU admissions (only the first ICU stay was retained), an ICU stay of less than 24 h, or missing triglyceride, glucose, or BMI measurements within the first 24 h after ICU admission (n = 9,147). The final analytic cohort consisted of 3,125 patients, who were categorized into three groups according to TyG index tertiles: Q1 (n = 1,042), Q2 (n = 1,041), and Q3 (n = 1,042)

TyG index calculation and grouping

The TyG index was calculated using the formula:

TyG = ln [fasting TG (mg/dL) × fasting glucose (mg/dL) / 2] [20].

Patients were categorized into three groups based on the tertiles of the TyG index.

Data collection

Clinical data for this study were derived from the MIMIC-IV 3.1 database. We extracted the relevant variables using SQL queries executed through Navicat Premium (version 16.1.15). To ensure consistency across patients, all measurements were restricted to values recorded within the first 24 h of ICU admission. Demographic information included age, sex, race, and BMI. Race information was obtained from the structured demographic tables in MIMIC-IV, which include the predefined categories White, Black/African American, Asian, Hispanic/Latino, and Other. In this cohort, the Asian and Hispanic groups had very small sample sizes and sparse event counts, which produced unstable estimates in multivariable and subgroup analyses. To maintain statistical stability and avoid sparse-data bias, these low-frequency categories were combined into a single “Other” group, consistent with practices used in prior MIMIC-based studies. Therefore, the final analyses included three race groups: White, Black, and Other. We also collected vital signs—heart rate, systolic and diastolic blood pressure, and body temperature—and a wide range of laboratory parameters, such as fasting glucose, triglycerides, creatinine, blood urea nitrogen, major electrolytes (sodium, potassium, calcium, and chloride), anion gap, hemoglobin, white blood cell count, red blood cell count, and platelet count. Comorbidities including diabetes mellitus, hypertension, atrial fibrillation, chronic kidney disease (CKD), heart failure, myocardial infarction, ischemic heart disease, cerebrovascular disease, and peripheral vascular disease were identified using ICD-9 and ICD-10 codes documented in the MIMIC-IV database. This coding-based approach reflects the standard method used in most studies based on MIMIC data. CKM stage 4 was defined in accordance with the 2023 American Heart Association (AHA) Scientific Statement on Cardiovascular–Kidney–Metabolic Health. As the MIMIC-IV database does not include a predefined CKM variable, CKM stage 4 was operationally identified by requiring that patients simultaneously met both components: (1) metabolic dysfunction or CKD (based on documented clinical diagnoses such as elevated BMI, metabolic syndrome components, type 2 diabetes mellitus, or ICD-coded CKD), and (2) established cardiovascular disease, identified using ICD-9/10 codes for conditions such as coronary artery disease, myocardial infarction, heart failure, atrial fibrillation, stroke, or peripheral arterial disease. This diagnostic-code–based approach is widely used in MIMIC-based research because it relies on clinician-verified diagnoses and avoids misclassification that may result from missing or non–time-aligned laboratory values. The severity of illness was quantified using the Sequential Organ Failure Assessment (SOFA) score and the Acute Physiology Score III (APS III). These scores are automatically generated within MIMIC-IV based on established algorithms that incorporate routinely collected clinical and laboratory data.We also recorded major interventions such as renal replacement therapy and mechanical ventilation during the ICU stay. Outcome measures included ICU length of stay, follow-up time, and all-cause mortality at 30, 90, and 365 days.

All laboratory measurements were extracted from the structured laboratory tables (labevents) of the MIMIC-IV database. To ensure consistency and minimize temporal variability, only the first recorded laboratory values within the initial 24 h after ICU admission were included in our analysis. These values are directly captured from the electronic health records and represent standardized measurements performed by certified hospital laboratories.

Follow-up endpoints

The primary endpoint was 365-day all-cause mortality. Secondary endpoints included 30-day and 90-day all-cause mortality. Mortality data were obtained from the Social Security Death Index linked to the MIMIC-IV database, with follow-up data available through October 1, 2024 (all data were de-identified, with the cutoff date based on official guidelines). For deceased patients, follow-up ended at the time of death.

Statistical analysis

Patients were grouped into three categories based on the TyG index tertiles (Q1–Q3). Continuous variables were tested for normality using the Shapiro-Wilk test. Normally distributed data were expressed as mean ± standard deviation (SD) and compared between groups using one-way analysis of variance (ANOVA). Non-normally distributed data were described as median (interquartile range) and compared using the Kruskal-Wallis H test. Categorical variables were expressed as count (percentage) and assessed for group differences using the χ² test or Fisher’s exact test.

The cumulative incidence of primary and secondary outcomes was presented using Kaplan-Meier curves, and differences between groups were compared using the log-rank test.

Cox proportional hazards models were used to assess the independent association between the TyG index and all-cause mortality. First, univariate analysis was performed, and variables with P < 0.05 were included in the final multivariate model along with clinical baseline variables such as gender and race. The TyG index was entered both as a categorical variable (using Q1 as the reference) and as a continuous variable (per 1 standard deviation increase).

To examine the nonlinear dose-response relationship between the TyG index and 30-day and 1-year mortality risk, RCS regression was used, and hazard ratios (HR) and 95% confidence intervals (CI) were calculated.

Subgroup analyses were conducted to evaluate the heterogeneity of the TyG index’s association with 1-year all-cause mortality across 16 prespecified variables, including age (≤ 65 vs. >65 years), gender (male vs. female), BMI (< 30 vs. ≥30), race (White, Black, Other), and comorbidities (diabetes, hypertension, atrial fibrillation, chronic kidney disease, heart failure, myocardial infarction, ischemic heart disease, cerebrovascular accidents, and peripheral vascular disease). Stratified Cox models were constructed for each subgroup, and interaction P-values (P_for_interaction) were calculated using likelihood ratio tests, with P < 0.05 considered statistically significant for subgroup differences.

Multicollinearity was assessed using variance inflation factors (VIF), with a VIF < 5 indicating no significant multicollinearity.

All statistical analyses were performed using R version 4.4.3 software, and P < 0.05 was considered statistically significant. Sensitivity analyses are detailed in the supplementary materials. Key R packages included survival, rms, foreign, tableone, and ggplot2.

Missing data handling and sensitivity analyses

To address missing values in covariates, we first examined the missingness pattern of all clinical variables. Variables with more than 20% missing values were excluded to avoid instability and potential bias (Supplementary Table 1) For variables with less than 20% missingness, multiple imputation was performed using the random forest imputation method implemented in the mice package in R, which has been widely validated for mixed clinical datasets. To further test the robustness of our findings, we conducted two sensitivity analyses. First, we excluded participants with any missing key variables and re-estimated the association between the TyG index and mortality using fully adjusted Cox regression models. Second, we reclassified the TyG index into quartiles and repeated the primary analyses under this alternative categorization to evaluate whether the imputation strategy introduced potential bias.

Results

Baseline characteristics

A total of 3,125 participants were included in this study, with 65.22% being male. The median age of the cohort was 70 years (range: 61–78). Based on the TyG index tertiles, the population was divided into three groups: Q1 (n = 1,042), Q2 (n = 1,041), and Q3 (n = 1,042). The baseline characteristics are shown in Table 1. Notably, higher TyG levels were associated with younger age and higher BMI. Additionally, heart rate and body temperature increased, while systolic blood pressure decreased with higher TyG levels. Furthermore, the prevalence of diabetes, myocardial infarction, and ischemic heart disease was higher in the high TyG group, while the prevalence of hypertension, atrial fibrillation, and cerebrovascular accidents decreased.

Table 1.

Baseline characteristics by TyG index quartiles

Characteristic	Overall (N = 3,125)	Quartile 1 (n = 1042)	Quartile 2 (n = 1041)	Quartile 3 (n = 1042)	p value
TyG _Index	9.11 (8.65–9.73)	8.46 (8.22–8.65)	9.11 (8.97–9.29)	10.00 (9.73–10.47)	< 0.001
Demographics
Age, year	70.00 (61.00–78.00)	73.00 (64.00–82.00)	71.00 (62.00–79.00)	66.00 (58.00–74.00)	< 0.001
≤ 65	1,129.00 (36.13%)	303.00 (29.08%)	349.00 (33.53%)	477.00 (45.78%)
> 65	1,996.00 (63.87%)	739.00 (70.92%)	692.00 (66.47%)	565.00 (54.22%)
Male sex	2,038.00 (65.22%)	691.00 (66.31%)	661.00 (63.50%)	686.00 (65.83%)	0.352
BMI, kg/m2	30.86 (27.71–35.59)	29.64 (27.03–33.72)	30.89 (27.73–35.40)	32.26 (28.58–37.58)	< 0.001
Race					< 0.001
White	1,926.00 (61.63%)	643.00 (61.71%)	665.00 (63.88%)	618.00 (59.31%)
Black	315.00 (10.08%)	133.00 (12.76%)	85.00 (8.17%)	97.00 (9.31%)
Other	884.00 (28.29%)	266.00 (25.53%)	291.00 (27.95%)	327.00 (31.38%)
Vital signs on ICU admission
Heart rate, beats/min	86.00 (73.00–101.00)	83.00 (71.00–96.00)	86.00 (73.00–101.00)	89.50 (77.00–106.00)	< 0.001
Systolic BP, mmHg	127.00 (109.00–145.00)	128.00 (110.00–147.00)	128.00 (111.00–147.00)	125.00 (107.00–143.00)	0.014
Diastolic BP, mmHg	71.00 (59.00–84.00)	72.00 (60.00–85.00)	71.00 (59.00–85.00)	69.00 (58.00–82.00)	0.016
Temperature, °F	98.30 (97.70–98.80)	98.20 (97.70–98.70)	98.30 (97.70–98.80)	98.40 (97.80–99.00)	< 0.001
Laboratory parameters
Fasting glucose (mg/dL)	140.00 (111.00–188.00)	112.00 (96.00–133.00)	139.00 (116.00–173.00)	197.00 (152.00–274.00)	< 0.001
Triglycerides (mg/dL)	127.00 (88.00–192.00)	80.00 (64.00–98.00)	130.00 (106.00–157.00)	234.00 (173.00–341.00)	< 0.001
Creatinine (mg/dL)	1.10 (0.90–1.80)	1.00 (0.80–1.50)	1.10 (0.90–1.70)	1.30 (0.90–2.20)	< 0.001
Blood urea nitrogen (mg/dL)	22.00 (15.00–36.00)	20.00 (14.00–30.00)	21.00 (15.00–35.00)	26.00 (17.00–46.00)	< 0.001
Sodium (mmol/L)	138.00 (135.00–141.00)	139.00 (136.00–141.00)	139.00 (136.00–141.00)	138.00 (135.00–141.00)	< 0.001
Potassium (mmol/L)	4.20 (3.80–4.60)	4.10 (3.70–4.50)	4.10 (3.80–4.60)	4.30 (3.90–4.90)	< 0.001
Hemoglobin (g/dL)	11.40 (9.70–13.10)	11.60 (9.90–13.10)	11.50 (9.60–13.30)	11.30 (9.60–13.00)	0.055
White blood cells (×109/L)	11.00 (8.10–15.40)	9.80 (7.40–13.10)	11.30 (8.40–15.40)	12.40 (9.10–17.70)	< 0.001
Platelet count (×109/L)	201.00 (148.00–261.00)	192.00 (147.00–250.00)	207.00 (150.00–269.00)	205.00 (148.00–268.00)	0.006
Red blood cells (×10¹²/L)	3.87 (3.24–4.43)	3.88 (3.32–4.40)	3.87 (3.23–4.45)	3.86 (3.19–4.42)	0.604
Calcium (mmol/L)	8.60 (8.00–9.00)	8.60 (8.10–9.00)	8.60 (8.00–9.00)	8.40 (7.90–8.90)	< 0.001
Chloride (mmol/L)	103.00 (99.00–107.00)	104.00 (100.00–107.00)	103.00 (100.00–107.00)	102.00 (98.00–106.00)	< 0.001
Mean corpuscular volume (fL)	91.00 (87.00–96.00)	91.00 (88.00–96.00)	92.00 (87.00–96.00)	91.00 (87.00–95.00)	0.005
Red cell distribution width (%)	14.40 (13.40–15.90)	14.30 (13.40–15.90)	14.40 (13.50–15.80)	14.40 (13.50–15.90)	0.189
Anion gap (mmol/L)	14.00 (12.00–17.00)	14.00 (11.00–16.00)	14.00 (12.00–17.00)	15.50 (13.00–19.00)	< 0.001
International normalized ratio (INR)	1.30 (1.10–1.50)	1.30 (1.10–1.55)	1.30 (1.10–1.50)	1.20 (1.10–1.50)	0.088
Comorbidities
Diabetes mellitus	1,446.00 (46.27%)	312.00 (29.94%)	456.00 (43.80%)	678.00 (65.07%)	< 0.001
Hypertension	1,628.00 (52.10%)	591.00 (56.72%)	544.00 (52.26%)	493.00 (47.31%)	< 0.001
Atrial fibrillation	1,612.00 (51.58%)	582.00 (55.85%)	536.00 (51.49%)	494.00 (47.41%)	< 0.001
Chronic kidney disease	884.00 (28.29%)	270.00 (25.91%)	300.00 (28.82%)	314.00 (30.13%)	0.091
Heart failure	1,347.00 (43.10%)	445.00 (42.71%)	449.00 (43.13%)	453.00 (43.47%)	0.939
Myocardial infarction	643.00 (20.58%)	175.00 (16.79%)	227.00 (21.81%)	241.00 (23.13%)	< 0.001
Ischemic heart disease	1,737.00 (55.58%)	542.00 (52.02%)	592.00 (56.87%)	603.00 (57.87%)	0.016
Cerebrovascular accident	610.00 (19.52%)	239.00 (22.94%)	203.00 (19.50%)	168.00 (16.12%)	< 0.001
Peripheral vascular disease	200.00 (6.40%)	76.00 (7.29%)	65.00 (6.24%)	59.00 (5.66%)	0.304
Severity scores
SOFA score	5.00 (2.00–8.00)	4.00 (2.00–6.00)	4.00 (2.00–7.00)	6.00 (3.00–9.00)	< 0.001
APS III score	44.00 (32.00–60.00)	39.00 (29.00–53.00)	44.00 (32.00–58.00)	51.00 (37.00–69.00)	< 0.001
Interventions
Renal replacement therapy	76.00 (2.43%)	21.00 (2.02%)	17.00 (1.63%)	38.00 (3.65%)	0.007
Mechanical ventilation	145.00 (4.64%)	47.00 (4.51%)	55.00 (5.28%)	43.00 (4.13%)	0.442
Outcomes
Follow-up time, days	365.00 (55.99–365.00)	365.00 (154.15–365.00)	365.00 (62.58–365.00)	365.00 (24.41–365.00)	< 0.001
ICU length of stay, days	4.60 (2.17–9.97)	3.63 (1.94–6.84)	4.32 (2.15–8.88)	6.66 (2.79–14.24)	< 0.001
One-year mortality rate	1,068.00 (34.18%)	300.00 (28.79%)	365.00 (35.06%)	403.00 (38.68%)	< 0.001

*Data are median [Q1–Q3] or n (%); p values from Kruskal-Wallis test (continuous) or χ² test (categorical).

Regarding hematological variables, blood glucose, triglycerides, creatinine, blood urea nitrogen, potassium, white blood cell count, and anion gap all significantly increased with higher TyG levels, while calcium and chloride levels significantly decreased. As the TyG index increased, disease severity scores also increased (SOFA score: Q1 = 4, Q2 = 4, Q3 = 6; APS III score: Q1 = 39, Q2 = 44, Q3 = 51), ICU length of stay was longer (Q1 = 3.63 days, Q2 = 4.32 days, Q3 = 6.66 days), and the one-year mortality rate showed an upward trend (Q1: 28.79%, Q2: 35.06%, Q3: 38.68%).

Main results

The Kaplan-Meier survival curves, stratified by TyG index tertiles, are shown in Fig. 2. During the 1-year follow-up of critically ill patients with CKM stage 4, mortality rates between the groups consistently showed statistical significance (log-rank P < 0.0001). The 30-day mortality rate in the Q3 group was significantly higher than in the Q1 and Q2 groups (log-rank P < 0.0001, Fig. 2A), and this survival disadvantage persisted at both the 90-day (Fig. 2B) and 365-day follow-up endpoints (Fig. 2C).

Fig. 2.

Kaplan-Meier survival curves for 30-day, 90-day, and 365-day all-cause mortality, stratified by TyG index tertiles. Panel A shows the 30-day all-cause mortality, Panel B shows the 90-day all-cause mortality, and Panel C shows the 365-day all-cause mortality. The mortality rate in the high TyG group (Q3) was consistently higher than in the low TyG group (Q1), with statistically significant differences observed in all comparisons (log-rank P < 0.0001 for all)

Table 2 presents the results of the association between the TyG index and the primary outcome using three progressively adjusted multivariable Cox models. The Variance Inflation Factor (VIF) analysis revealed VIF values below 5 for all covariates, suggesting no substantial multicollinearity (Supplementary Table 2). In the fully adjusted model (Adjusted II), each 1 standard deviation increase in the TyG index was associated with a 22% increase in the 30-day all-cause mortality risk (HR = 1.22, 95% CI 1.06–1.40), and a 23% increase in the 365-day all-cause mortality risk (HR = 1.23, 95% CI 1.09–1.39). Further stratification by TyG tertiles revealed a trend of increasing 30-day all-cause mortality risk with higher TyG levels (P for trend < 0.01). However, the difference between the Q2 group and Q1 was not statistically significant (Q2: HR = 1.13, 95% CI 0.93–1.39), while the difference between Q3 and Q1 was statistically significant (Q3: HR = 1.27, 95% CI 1.02–1.59). Similarly, the 365-day mortality risk also showed an increasing trend (P for trend < 0.01), with both Q2 and Q3 groups significantly higher compared to Q1 (Q2: HR = 1.28, 95% CI 1.05–1.45; Q3: HR = 1.30, 95% CI 1.08–1.55).

Table 2.

Cox regression analysis of TyG index quartiles and All-Cause mortality risk

TyG quartiles	Non-adjusted	Adjusted I	Adjusted II
Hazard ratios (HR) for 30-day all-cause mortality
	NO-adjusted	Adjusted I	Adjusted Ⅱ
TyG (per SD change)	1.31(1.18–1.46)***	1.48(1.33–1.66)***	1.22(1.06–1.40)**
Q1	1(ref)	1(ref)	1(ref)
Q2	1.2 (0.99–1.47)	1.23 (1.01–1.50)*	1.13 (0.93–1.39)
Q3	1.62 (1.34–1.95)***	1.79 (1.48–2.17)***	1.27 (1.02–1.59)*
P for trend	< 0.001	< 0.001	< 0.01
Hazard ratios (HR) for 365-day all-cause mortality
TyG (per SD change)	1.25(1.14–1.37)***	1.42(1.30–1.57)***	1.23(1.09–1.39)**
Q1	1(ref)	1(ref)	1(ref)
Q2	1.26 (1.08–1.47)***	1.29 (1.11–1.50)***	1.28 (1.05–1.45)**
Q3	1.47 (1.27–1.71)***	1.65 (1.42–1.92)***	1.30 (1.08–1.55)**
P for trend	< 0.01	< 0.001	< 0.01

*Adjusted I: adjusted for age, sex, BMI

†Adjusted II: further adjusted on the basis of Model I plus Calcium (CA), Potassium, Sodium, Chloride (CL), Blood Urea Nitrogen, Creatinine, Fasting Glucose, White Blood Cells, Platelet Count, Red Cell Distribution Width (RDW), Anion Gap (AG), Chronic Kidney Disease (CKD), Heart Failure (HF), Hyperlipidemia (HLD), Hypertension (HTN), Ischemic Heart Disease (IHD), Cerebrovascular Accident (CVA), Peripheral Vascular Disease (PVD), Diabetes mellitus (DM), Myocardial Infarction (MI) and International Normalized Ratio (INR)

*P < 0.05, **P < 0.01, ***P < 0.001.

To further explore the association between the TyG index and mortality, detailed subgroup analyses and interaction analyses were conducted. These analyses were stratified by several key variables, including age (≤ 65 vs. >65 years), gender (male vs. female), BMI (< 30 vs. ≥30), race (White, Black, Other), and the presence of comorbidities such as diabetes (DM), hypertension (HTN), atrial fibrillation (AF), chronic kidney disease (CKD), heart failure (HF), myocardial infarction (MI), ischemic heart disease (IHD), cerebrovascular accidents (CVA), and peripheral vascular disease (PVD). The results revealed significant interactions between the TyG index and mortality in the DM and HTN subgroups (P < 0.05). However, no significant interactions were observed in the other subgroups (interaction P > 0.05). In the subgroup analysis, the TyG index showed a significant positive correlation with both 30-day and 365-day all-cause mortality (Fig. 3).

Fig. 3.

Subgroup analysis of TyG index and all-cause mortality risk. Panel A presents the 30-day mortality risk stratified by key variables, and Panel B presents the 365-day mortality risk. Hazard ratios (HR) and 95% confidence intervals (CI) are shown for each subgroup. The P_for_interaction values highlight significant interactions (P_for_interaction < 0.05), indicating that the TyG index has differential predictive value in certain subgroups. Notably, a significant interaction was observed in patients with diabetes mellitus (DM) and hypertension (HTN), suggesting that the TyG index may be more predictive in these populations (P < 0.05)

This study also used RCS to explore the nonlinear relationship between the TyG index and both 30-day and 1-year all-cause mortality. The fully adjusted RCS curves (Fig. 4) indicated a linear trend between the TyG index and both 30-day and 1-year all-cause mortality (P for Nonlinear > 0.05).

Fig. 4.

RCS regression analysis of the TyG index and all-cause mortality. Panel A presents the association between the TyG index and 30-day all-cause mortality, and Panel B shows the association with 365-day all-cause mortality. Both curves were generated using the fully adjusted Cox proportional hazards model. The fully adjusted model incorporated demographic characteristics (age, sex, race, BMI), vital signs (heart rate, systolic and diastolic blood pressure, body temperature), laboratory variables (creatinine, blood urea nitrogen, sodium, potassium, calcium, chloride, anion gap, hemoglobin, white blood cell count, platelet count), comorbidities (diabetes mellitus, hypertension, atrial fibrillation, chronic kidney disease, heart failure, myocardial infarction, ischemic heart disease, cerebrovascular accident, peripheral vascular disease), severity scores (SOFA, APS III), and ICU interventions (renal replacement therapy, mechanical ventilation). The RCS models demonstrated a linear association between the TyG index and mortality, with no evidence of nonlinearity (P for nonlinearity = 0.289 for 30-day mortality; 0.145 for 365-day mortality)

To ensure the robustness of our results, several sensitivity analyses were conducted. First, to evaluate potential bias from missing data imputation, we excluded samples with missing key variables and performed fully adjusted multivariable Cox regression analysis. The results remained consistent, showing that a higher TyG index was associated with increased all-cause mortality risk at both 30 days and 1 year, with no significant changes (Supplementary Table 3). Secondly, we applied a quartile-based grouping method, categorizing the TyG index into four levels, and repeated the main analysis with this new classification. The results remained similar (Supplementary Table 4). These sensitivity analyses further validate the robustness of our findings.

Discussion

This study found that in critically ill patients with CKM stage 4, the first TyG index measurement within 24 h of ICU admission was significantly positively correlated with both 30-day and 365-day all-cause mortality. This relationship remained significant after adjusting for confounding factors. Additionally, RCS regression analysis revealed a linear dose-response relationship between the TyG index and all-cause mortality, with no threshold or plateau effect observed. Therefore, the first TyG index measurement within 24 h of ICU admission was independently associated with short- and long-term mortality in critically ill CKM stage 4 patients and may have potential value for risk stratification in this population.

CKM Stage 4 represents the final stage of severe decompensation in the cardiovascular, renal, and metabolic systems, with high inpatient mortality and a poor prognosis [1, 9]. IR, a key mechanism underlying CKM [21], plays a central role in the pathophysiological processes of the syndrome, accelerating its progression [22]. Specifically, IR leads to prolonged elevations in pro-inflammatory factors and oxidative stress markers, which directly damage both the cardiovascular and renal systems [23–25]. The resulting hyperglycemia further impairs endothelial function, accelerates atherosclerosis, and increases the risk of heart failure [26, 27].Moreover, IR activates the renin-angiotensin-aldosterone system (RAAS), raising levels of angiotensin II and aldosterone.This triggers vasoconstriction, sodium retention, and elevated blood pressure, all of which place additional strain on the kidneys [28, 29].IR also contributes to metabolic disturbances, causing abnormalities in blood glucose, lipid levels, and blood pressure, which collectively drive the onset and progression of CKM [30].

Therefore, assessing IR is crucial for understanding the disease’s progression and prognosis in patients with CKM Stage 4.However, the gold standard for evaluating IR—the hyperinsulinemic-euglycemic clamp technique—remains impractical for routine clinical use due to its complexity, time demands, and high cost. As a result, researchers have been working to identify simpler, more cost-effective biomarkers. The TyG index, a reliable surrogate biomarker for IR [31, 32], can be easily calculated using fasting glucose and triglyceride levels, offering the advantages of low cost and simplicity. It has been closely associated with increased all-cause mortality in critically ill patients [33]. Moreover, numerous studies have linked the TyG index to the incidence and mortality of metabolic syndrome (MS), cardiovascular disease (CVD), or chronic kidney disease (CKD).

A longitudinal prospective cohort study [34] with a follow-up period of 16.8 years demonstrated that the TyG index independently predicted all-cause mortality (HR = 1.39, 95% CI 1.07–1.79) and cardiovascular mortality in type 2 diabetes patients, and was also associated with major adverse cardiovascular events, diabetic nephropathy, and neuropathy. Qin Zhang [35] et al. reported that the TyG index is a strong predictor of cardiovascular and all-cause mortality in patients with diabetes or prediabetes and CVD, with higher baseline TyG levels correlating with greater mortality risk. Additionally, a meta-analysis by Sandeep Samethadka Nayak [36] et al. showed that an elevated TyG index was significantly associated with the incidence of chronic kidney disease (RR = 1.46, 95% CI 1.32–1.63). Research by Enmin Xie [37] et al. also found that the TyG index could be a valuable predictor of cardiovascular adverse events in patients with end-stage renal disease and coronary artery disease (CAD). Similarly, a prospective study by Setor K. Kunutsor [18] et al. demonstrated that a higher TyG index was associated with an increased risk of CKD and improved CKD prediction and classification, surpassing traditional risk factors. However, there is currently insufficient data to confirm the relationship between the TyG index and CKM stage 4, particularly in critically ill populations, and its predictive value for future mortality risk. Most studies have focused on CKM stages 0–3, with relatively few studies examining CKM stage 4 patients [38, 39]. A recent retrospective cohort study [40] demonstrated that the TyG-BMI index at admission was significantly negatively correlated with 1-year mortality risk in critically ill CKM stage 4 patients, showing an “L-shaped” association. Although TyG-derived indices (e.g., TyG-BMI, TyG-WC) may enhance predictive power, the TyG index itself holds independent clinical significance [41–43]. Additionally, in CKM stage 4 patients, due to fluid retention and muscle atrophy caused by combined heart-kidney failure, body measurement parameters such as waist circumference (WC) and BMI may not accurately reflect actual fat content and distribution, leading to inaccurate assessments of IR status and associated disease risks [44, 45].

It is important to note that previous studies have primarily relied on the “immediate admission” TyG index, which reflects the patient’s baseline metabolic status upon arrival. However, critically ill patients often receive initial treatments before being transferred to the ICU, which may introduce confounding factors across different clinical scenarios. In contrast, the first TyG measurement taken within 24 h of ICU admission provides a more accurate reflection of the patient’s IR status in the critical care environment. Research has shown that higher TyG levels measured within this timeframe are significantly associated with 90-day all-cause mortality in critically ill patients. Therefore, this study focuses on the TyG index measured within the first 24 h of ICU admission, as it better represents the patient’s immediate metabolic state upon entering the ICU and offers a more timely and targeted prognostic assessment for clinical application.

In the subgroup analysis, we found that the association between the TyG index and mortality risk varied across different patient groups. Notably, the association was stronger in non-diabetic and hypertensive patients, which is consistent with previous research [46–48]. In individuals without diabetes, elevation of the TyG index may reflect reduced metabolic flexibility and a heightened glycemic response to acute physiological stress, which could confer greater prognostic impact than in patients with established diabetes. Similar patterns have been observed in the KorAHF cohort, where glycemic variability predicted mortality only in non-diabetic patients, suggesting that this population may be more susceptible to stress-related metabolic instability [49].

Among hypertensive patients, long-standing endothelial dysfunction, impaired vasomotor regulation, arterial stiffness, and chronic activation of the renin–angiotensin–aldosterone system may amplify the adverse cardiovascular effects of insulin resistance [50]. These pathophysiological features could enhance vulnerability to metabolic disturbances and may help explain why the prognostic association between the TyG index and mortality was more pronounced in hypertensive individuals.

Interestingly, we observed that patients with BMI ≥ 30 had a slightly lower mortality risk compared to those with BMI < 30 (HR = 1.21, 95% CI 1.11–1.33 vs. HR = 1.28, 95% CI 1.16–1.42). This finding mirrors the “obesity paradox” identified by Pan [40] et al. using the TyG-BMI index. The association between higher BMI and lower mortality risk may be explained by mechanisms such as the buffering effect of free fatty acids from peripheral lipolysis during stress, which helps delay cachexia progression [51]. Additionally, anti-inflammatory factors secreted by adipose tissue (such as adiponectin) can suppress systemic inflammation [52, 53], while increased blood volume in obesity reduces the cardiac stress burden [54]. Lower sympathetic activity and resistance to excessive activation of the renin-angiotensin system may also reduce heart strain, providing a survival advantage [40, 55]. However, in our study, this difference was not statistically significant (interaction P = 0.432), suggesting that BMI did not significantly alter the relationship between the TyG index and mortality risk. Thus, we interpret these findings with caution. On one hand, the TyG-BMI index may be more sensitive to obesity, capturing the combined effects of obesity and metabolic disorders. On the other hand, BMI, as a relatively crude measure of obesity, does not account for fat distribution [13], which may prevent it from accurately reflecting the metabolic status of critically ill CKM stage 4 patients. Additionally, our study may be subject to survivor bias, as some patients who died early were excluded from the study, leading to potential discrepancies in the characteristics of survivors compared to the general population. Moreover, the timing of variable selection may influence the interpretation of study results. Therefore, further research is needed to confirm the presence of the obesity paradox in this specific population.

Limitations and future directions

This study has several limitations. First, as a retrospective analysis of the MIMIC-IV database, selection or information bias cannot be completely ruled out, despite strict inclusion criteria and multiple sensitivity analyses. Second, missing or incomplete laboratory records are inevitable in routine clinical practice; although multiple imputation was applied, residual measurement error may still influence TyG calculations and covariate adjustments.

Another important consideration is medication use. Agents such as insulin, statins, or corticosteroids can affect glucose and triglyceride concentrations and may therefore influence the TyG index. Although medication administration data are available in MIMIC-IV, the timing and completeness of these records vary substantially, particularly within the first 24 h after ICU admission when the TyG index was measured. Because many early medications reflect acute physiological responses rather than baseline metabolic status, adjusting for them may introduce misclassification. For these reasons, medication variables were not included in our multivariable models.

Third, the study design limits our ability to establish a causal relationship between the TyG index and all-cause mortality. While multivariable models and RCS analysis assess the association, unmeasured confounders could still influence the results. In addition, BMI, which was used to characterize metabolic dysregulation, does not fully capture adiposity patterns in critically ill CKM patients. Finally, all data originated from a single academic medical center, which may limit the generalizability of the findings.

Future prospective, multicenter studies with serial metabolic measurements, more complete and time-resolved medication data, and more precise body-composition assessments (such as waist circumference or body-fat percentage) are warranted to validate and extend these findings.

Conclusion

This study is the first to investigate the impact of the initial TyG index measurement within 24 h of ICU admission on 1-year mortality risk in critically ill patients with CKM stage 4. The results demonstrate a significant positive correlation between the TyG index at this time point and 1-year all-cause mortality, with a linear relationship. This correlation remained robust even after adjusting for confounding factors and conducting sensitivity analyses. In summary, the TyG index measured within 24 h of ICU admission was linearly associated with 1-year all-cause mortality in critically ill patients with CKM stage 4. The TyG index may provide a simple indicator of acute metabolic disturbances and could help clinicians identify patients at higher risk, but its role in guiding management needs to be confirmed in prospective studies.

Data availability

The dataset used in this study is available upon request from the corresponding author. However, due to the specific nature of the MIMIC-IV database, access must comply with the database’s terms and conditions. Users must submit a request through the official MIMIC-IV access channels. Once the necessary permissions are obtained, users can request the relevant data subsets from the corresponding author. The corresponding author will ensure that data sharing complies with all ethical and legal guidelines and will provide the necessary documentation and support for data retrieval.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviatons

AHA

American Heart Association

AG

Anion gap

APS III

Acute Physiology Score III

AF

Atrial fibrillation

BMI

Body mass index

BUN

Blood urea nitrogen

CAD

Coronary artery disease

Ca

Calcium

CI

Confidence interval

CKD

Chronic kidney disease

CKM

Cardiovascular–kidney–metabolic syndrome

Cl

Chloride

Cr

Creatinine

CVD

Cardiovascular disease

CVA

Cerebrovascular accident

DM

Diabetes mellitus

eGFR

Estimated glomerular filtration rate

HF

Heart failure

HR

Hazard ratio

HTN

Hypertension

ICD

International Classification of Diseases

ICU

Intensive care unit

IHD

Ischemic heart disease

IR

Insulin resistance

K

Potassium

MI

Myocardial infarction

MS

Metabolic syndrome

Na

Sodium

PVD

Peripheral vascular disease

RAAS

Renin–angiotensin–aldosterone system

RCS

Restricted cubic spline

RRT

Renal replacement therapy

SD

Standard deviation

SOFA

Sequential Organ Failure Assessment

TG

Triglyceride

T2DM

Type 2 diabetes mellitus

TyG

Triglyceride–glucose

UACR

Urine albumin-to-creatinine ratio

WBC

White blood cell count

Author contributions

X.Y. and S.H. designed the study. X.Y. and Z.Z. collected and analyzed the data. L.T. and J.L. interpreted the data. Y.Z. critically revised the manuscript. All authors drafted and reviewed the manuscript and approved the final version for publication.All authors contributed equally to this work.

Funding

This research received no specific funding from any public, commercial, or not-for-profit organization.

Data availability

The dataset used in this study is available upon request from the corresponding author. However, due to the specific nature of the MIMIC-IV database, access must comply with the database’s terms and conditions. Users must submit a request through the official MIMIC-IV access channels. Once the necessary permissions are obtained, users can request the relevant data subsets from the corresponding author. The corresponding author will ensure that data sharing complies with all ethical and legal guidelines and will provide the necessary documentation and support for data retrieval.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The MIMIC-IV database is a publicly available, de-identified dataset. Individual patient consent was not required because all data were anonymized. Access to the database for research purposes was approved by the Institutional Review Board of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center.