The association between different insulin resistance surrogates and all-cause mortality and cardiovascular mortality in patients with metabolic dysfunction-associated steatotic liver disease

Abstract

Background

Metabolic dysfunction-associated steatotic liver disease (MASLD) is closely associated with insulin resistance (IR). However, the prognostic value of different alternative IR surrogates in patients with MASLD remains unclear. This study aimed to evaluate the association between various IR indices and all-cause mortality and cardiovascular mortality in MASLD patients.

Methods

A total of 8,753 adults aged ≥ 20 years with MASLD from the National Health and Nutrition Examination Survey (NHANES, 2003–2018) were included, and their mortality data were obtained from the National Death Index (NDI). Insulin resistance surrogates [including the triglyceride-glucose (TyG) index, TyG-body mass index (TyG-BMI), TyG-waist circumference index, TyG-waist-to-height ratio index, and Homeostatic Model Assessment for IR] were stratified into quartiles. Cox proportional hazards models, receiver operating characteristic (ROC) curve analysis, restricted cubic spline (RCS), mediation analyses, and subgroup analyses were used to explore the associations between these indices and all-cause mortality as well as cardiovascular mortality in MASLD patients.

Results

During a median follow-up of 98 months, 1,234 deaths were observed, including 409 cardiovascular disease (CVD)-related deaths. In the fully adjusted model, higher quartiles of TyG-related indices were significantly associated with an increased risk of all-cause mortality in MASLD patients. Furthermore, the TyG-BMI index was associated with both all-cause mortality and CVD mortality [all-cause mortality: HR (95% CI) 2.84 (1.73–4.67), P < 0.001; CVD mortality: HR (95% CI) 5.32 (2.26–12.49), P < 0.001]. The RCS analyses indicated a U-shaped relationship between TyG-BMI and mortality, with a threshold value of 270.49. Subgroup analyses demonstrated that TyG-related indices had stronger associations with mortality in elderly MASLD patients.

Conclusions

Our findings highlight the prognostic value of IR indices, particularly TyG-BMI index, in predicting all-cause mortality and CVD mortality in MASLD patients.

Graphical abstract

This study highlights the prognostic value of IR indices, particularly TyG-BMI index, in predicting all-cause mortality and CVD mortality in MASLD patients.

Research insights

What is currently known about this topic?

• MASLD is linked to insulin resistance and metabolic dysfunction. Insulin resistance drives MASLD progression and related comorbidities. Cardiovascular risk is high in MASLD patients.

What is the key research question?

• What is the best insulin resistance indices which impact mortality in MASLD patients?

What is new?

• TyG-BMI index predicts all-cause and cardiovascular mortality in MASLD. Insulin resistance indices improve MASLD patient risk stratification.

How might this study influence clinical practice?

• Findings may guide risk assessment and treatment in MASLD patients.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), previously referred to as non-alcoholic fatty liver disease (NAFLD), has been recognized globally as the most prevalent chronic liver disease. Its global prevalence has increased dramatically, from 25.5% before 2005 to 37.8% after 2016 [1]. In 2023, the term MASLD was adopted through a Delphi consensus process, emphasizing that hepatic fat accumulation is primarily attributed to metabolic dysregulation rather than other secondary etiologies [2]. Because MASLD is associated with liver-related and systemic complications—including progression to cirrhosis, hepatocellular carcinoma, and increased cardiovascular risk—it has become a significant public health burden [3].

A hallmark of MASLD is insulin resistance (IR) [4]. IR not only promotes hepatic steatosis but also exacerbates systemic metabolic dysfunction, increasing the risk of cardiovascular disease (CVD) and type 2 diabetes, and ultimately leads to increased mortality [5, 6]. Although the hyperinsulinemic euglycemic clamp technique is the gold standard for IR assessment, its complexity and associated risks make it unsuitable for large-scale clinical studies [7]. Consequently, surrogate markers for IR, including the homeostatic model assessment of insulin resistance (HOMA-IR) and the triglyceride-glucose (TyG) index, have been developed [8, 9]. HOMA-IR and the TyG index have been found to be associated with prognosis in various populations, including patients with diabetes, coronary heart disease, and hypertension [10, 11].

Furthermore, obesity is another key risk factor for MASLD [12]. Evidence suggests that combining the TyG index with obesity-related indices, such as TyG-waist circumference (TyG-WC) index, TyG-body mass index (TyG-BMI), and TyG-waist-to-height ratio (TyG-WHtR) index, improves predictive accuracy compared to the TyG index alone [13, 14]. Despite the well-established link between IR and MASLD, the comparative prognostic value of different IR surrogates in predicting all-cause mortality and cardiovascular mortality remains underexplored.

Our study aimed to evaluate the potential prognostic role of various IR surrogates (HOMA-IR, TyG, TyG-WC, TyG-BMI, TyG-WHtR) in MASLD patients. These findings may provide novel insights to refine risk stratification and inform future clinical guidelines, facilitating more personalized management strategies for MASLD.

Methods

Data source

The National Health and Nutrition Examination Survey (NHANES) is administered by the National Center for Health Statistics (NCHS) under the U.S. Centers for Disease Control and Prevention (CDC). NHANES aims to evaluate the health and nutritional status of U.S. adults and children through interviews, physical examinations, and laboratory tests. NHANES has been approved by the NCHS Research Ethics Review Board, and all participants provided written informed consent. Therefore, no additional informed consent or independent ethical review was required for this study. The NHANES datasets used in this study are publicly available at https://www.cdc.gov/nchs/nhanes.

Study population

Data for this study were from 2003–2018 NHANES database. Initially, 80,312 participants were recruited. The following participants were excluded: (1) those under 20 years old. (2) those lacking data for IR surrogate calculations, such as fasting triglycerides (FTG), fasting plasma glucose (FPG), and fasting insulin (FINS). (3) those with ultrasound fatty liver index (USFLI) < 30. (4) those without cardiometabolic risk factors or with a history of moderate-to-severe alcohol intake. (5) those lacking follow-up outcome data. The final study population included 8,753 participants. The selection process is shown in Fig. 1.

Fig. 1

The flow diagram of the study participants from NHANE 2003–2018. Abbreviations: NHANES, National Health and Nutrition Examination Survey; IR, insulin resistance; USFLI, ultrasound fatty liver index; MASLD, metabolic dysfunction-associated steatotic liver disease;

Definitions of MASLD

In accordance with the Delphi method consensus criteria, MASLD was defined as steatotic liver disease (SLD) with at least one cardiometabolic risk factor, while excluding viral hepatitis, autoimmune liver disease, genetic liver disorders, drug-induced liver disease, or alcohol-related liver disease (alcohol intake ≥ 30 g/day for men or ≥ 20 g/day for women). A USFLI ≥ 30 has demonstrated greater validity for SLD diagnosis than the FLI, with an AUC (95% CI) of 0.80 (0.77–0.83)[15]. The USFLI was calculated using the following formula:

$$ \begin{aligned} {\text{USFLI }} & = \left( {\text{e}} \right.^{{ - 0.{\text{8}}0{\text{73}}*{\text{non-Hispanic black }} + {\text{ }}0.{\text{3458}}*{\text{Mexican American}}}} \\ & ^{{ + {\text{ }}0.00{\text{93}}*{\text{age }} + {\text{ }}0.{\text{6151}}*{\text{loge}}\left( {{\text{GGT}}} \right){\text{ }} + {\text{ }}0.0{\text{249}}*{\text{waist circumference}}}} \\ & \left. { ^{{ + {\text{ 1}}.{\text{1792}}*{\text{loge}}\left( {{\text{insulin}}} \right) + 0.{\text{8242}}*{\text{loge}}\left( {{\text{glucose}}} \right) - {\text{ 14}}.{\text{7812}}}} } \right) \\ & /\left( {{\text{1 }} + {\text{ e}}^{{ - 0.{\text{8}}0{\text{73}}*{\text{non-Hispanic black }} + 0.{\text{3458}}*{\text{Mexican American }}}} } \right. \\ & ^{{ + 0.00{\text{93}}*{\text{age }} + 0.{\text{6151}}*{\text{loge}}\left( {{\text{GGT}}} \right) + 0.0{\text{249}}*{\text{waist circumference}}}} \\ & \left. { ^{{ + {\text{ 1}}.{\text{1792}}*{\text{loge}}\left( {{\text{insulin}}} \right) + 0.{\text{8242}}*{\text{loge}}({\text{glucose}}){-}{\text{14}}.{\text{7812}}}} } \right)*{\text{ 1}}00 \\ \end{aligned} $$

The USFLI ranges from 0 to 100. Participants are assigned a value of 1 if they belong to the "Non-Hispanic Black" or "Mexican American" racial groups; otherwise, the value is 0.

In addition, cardiometabolic risk factors include overweight/obesity/central obesity, hyperglycemia or diabetes, hypertension, hypertriglyceridemia, and reduced high-density lipoprotein cholesterol (HDL-C), defined as follows: (a) BMI ≥ 25 kg/m² or waist circumference (WC) > 94 cm in men or > 80 cm in women; (b) fasting plasma glucose ≥ 5.6 mmol/L (100 mg/dL), 2-h post-load glucose ≥ 7.8 mmol/L (≥ 140 mg/dL), glycated hemoglobin (HbA1c) ≥ 5.7% (39 mmol/mol), or a diagnosis of diabetes; (c) blood pressure ≥ 130/85 mmHg or treatment with antihypertensive medications; (d) plasma triglycerides ≥ 1.70 mmol/L (150 mg/dL) or treatment with lipid-lowering medications; and (e) plasma HDL-C ≤ 1.0 mmol/L (40 mg/dL) in men or ≤ 1.3 mmol/L [50 mg/dL] in women or treatment with lipid-lowering medications.

Definitions of IR surrogates

The IR surrogates in this study included HOMA-IR index, TyG index, TyG-BMI index, TyG-WC index, and TyG-WHtR index, calculated using the following formulas:

$$ \begin{aligned} {\text{HOMA-IR index}} & = {\text{Fasting blood glucose }}\left( {{\text{mmol}}/{\text{L}}} \right) \\ & \times {\text{Fasting insulin}}(\mu {\text{U}}/{\text{mL}})/{22}.{5} \\ \end{aligned} $$

$$\begin{aligned} {\text{TyG index}} &= {\text{ln}}[{\text{Fasting triglycerides}}\left( {{\text{mg}}/{\text{dL}}} \right)\\& \times {\text{Fasting glucose}}\left( {{\text{mg}}/{\text{dL}}} \right)/{2}] \end{aligned} $$

$${\text{TyG-BMI index}} = {\text{TyG index}} \times {\text{BMI}}({\text{kg}}/{\text{m}^2}) $$

$$\begin{aligned} {\text{TyG-WC index}}& = {\text{TyG index}} \\&\times {\text{Waist circumference}}\left( {{\text{cm}}} \right) \end{aligned}$$

$$\begin{aligned} {\text{TyG-WHtR index}} &= {\text{TyG index}} \\& \times {\text{Waist circumference }}\left( {{\text{cm}}} \right) /{\text{Height }}\left( {{\text{cm}}} \right) \end{aligned}$$

Participants were stratified into four groups (Q1, Q2, Q3, Q4) based on the quartiles of each index, with the group Q1 as the reference.

Quartile stratification

We calculated the 25th, 50th, and 75th percentiles of each index in the full study cohort. Participants were then stratified into four groups based on these cut‑points: Q1: < 25th percentile; Q2: >  = 25th to < 50th percentile; Q3: >  = 50th to < 75th percentile; Q4: >  = 75th percentile. The cut off points for each index are provided in the supplementary files (Table S1).

Ascertainment of mortality

The primary outcome of this study was all-cause mortality, and the secondary outcome was cardiovascular mortality. All-cause mortality was defined as death from any cause. According to the 10th Revision of the International Classification of Diseases (ICD-10), cardiovascular mortality was defined as deaths caused by heart disease and cerebrovascular disease. Mortality data for the follow-up population were obtained from the NHANES Public Use Linked Mortality File through December 31, 2019. This file was linked to the National Center for Health Statistics (NCHS) and the National Death Index (NDI) through a probabilistic matching algorithm. The follow-up period was from the date of the interview to the date of death or the end of follow-up (December 31, 2019). The dataset used for mortality analysis in this study is available at https://www.cdc.gov/nchs/data-linkage/mortality-public.htm.

Assessment of covariates

Demographic characteristics, dietary data, physical examination, laboratory tests, and questionnaire data were collected for all participants. (1) Demographic characteristics included gender (male or female), age, race (Mexican, Hispanic, non-Hispanic white, non-Hispanic black or other race), marital status (married/living with a partner, widowed/divorced/separated, or never married), education level (< high ​​school, completed high school, or > high school), and Income-to-Poverty Ratio (PIR) (< 1.3, 1.3–3.5, or > 3.5). (2) Dietary data included energy intake (average kilocalories derived from two 24-h dietary recall interviews) and alcohol intake. (3) Physical examination included waist circumference, height, body mass index (BMI), systolic blood pressure (SBP), and diastolic blood pressure (DBP). (4) Laboratory test results included fasting blood glucose, fasting insulin, glycated hemoglobin (HbA1c), fasting triglycerides (TG), total cholesterol (TC), high-density lipoprotein (HDL), low-density lipoprotein (LDL), alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), lactate dehydrogenase (LDH), and total bilirubin (TBil). (5) Questionnaire data included smoking status (never, ever or current), alcohol intake (never, ever or current), self-reported history of diabetes, hypertension, CVD, and cancer. Diabetes was defined as a self-reported diagnosis, use of insulin or oral hypoglycemic agents, fasting blood glucose ≥ 7 mmol/L, 2-h plasma glucose ≥ 200 mg/dL (11.1 mmol/L), or HbA1c ≥ 6.5%. Hypertension was defined as self-reported diagnosis, mean SBP ≥ 140 mmHg, mean DBP ≥ 90 mmHg, or the use of antihypertensive medication. CVD was defined as self-reported diagnosis of heart failure, coronary heart disease, angina, heart attack, or stroke.

Statistical analysis

Because the NHANES uses a complex, multistage stratified probability sampling design, it was necessary to incorporate sample weighting, clustering, and stratification into the statistical analyses. In accordance with NHANES analysis guidelines, all participant data were weighted using the recommended two-day dietary sample weights (WTDR2D). Detailed information on the weighting calculation method is available at https://wwwn.cdc.gov/nchs/nhanes/tutorials/Weighting.aspx.

The study population was stratified into four groups based on the quartiles (Q1–Q4) of HOMA-IR index, TyG index, TyG-WC index, TyG-BMI index, and TyG-WHtR index. The Kolmogorov–Smirnov test was used to examine the normality assumption for each variable. As presented in Fig.S1, all continuous variables were normally distributed. Continuous variables were expressed as mean ± standard deviation (SD), and categorical variables were reported as unweighted frequency counts and weighted percentages. To compare baseline characteristics among the four groups, weighted one-way analysis of variance (ANOVA) was used for continuous variables, and a weighted chi-square test was used for categorical variables. Kaplan–Meier survival curves were constructed to visually assess the associations between these indices and survival outcomes, including all-cause mortality and cardiovascular mortality. A log-rank test was used to statistically compare survival curves across the different quartile groups. We constructed three regression models by adjusting for different covariates to control for confounding bias. Covariates were selected identified from previous studies as related to survival outcomes. Specifically, Model 1 was unadjusted; Model 2 was adjusted for age, sex, and race; and Model 3 was adjusted for age, sex, race, education level, BMI, energy intake, smoking status, alcohol intake, CVD, diabetes, hypertension, AST, ALT, and PIR.

A weighted multivariable Cox proportional hazards model was subsequently used to estimate the associations between these indices and all-cause as well as cardiovascular mortality in MASLD patients. To further evaluate the prognostic performance of each model, we conducted receiver operating characteristic (ROC) curve analyses to determine the sensitivity and specificity of each model in predicting all-cause and cardiovascular mortality in MASLD patients. The area under the curve (AUC) was calculated for each model as a measure of its discriminative ability. To further explore the potential nonlinear association between these indices and all-cause mortality and cardiovascular mortality in MASLD, a restricted cubic spline (RCS) function was incorporated into the Cox proportional hazards model. Four knots were placed at the 5%, 35%, 65%, and 95% percentiles of each index’s distribution. A likelihood ratio test was used to assess the significance of nonlinearity. Furthermore, in Model 3, a mediation analysis was performed to determine whether the associations between the various indices and all-cause and cardiovascular mortality could be mediated by HbA1c and insulin. To evaluate whether the effects of these indices on mortality differed across subgroups, stratified analyses and interaction tests were performed according to potential covariates. For each subgroup, an adjusted Cox proportional hazards model was fitted, and the significance of interactions was assessed using likelihood ratio tests. We applied the Benjamini Hochberg method to correct the false discovery rate (FDR), with statistical significance defined as an FDR-adjusted p < 0.05.

All statistical analyses were performed using (R version 4.2.2 R; Foundation for Statistical Computing, Vienna, Austria). Two-sided p-values < 0.05 were considered statistically significant.

Result

Baseline characteristics of the study population

A total of 8,753 participants were included in this study, with a mean age of 53.41 ± 17.46 years, 51.9% of whom were female and 48.1% were male. The baseline mean values of the HOMA-IR index, TyG index, TyG-BMI index, TyG-WC index, and TyG-WHtR index were 5.26 ± 7.71, 8.86 ± 0.63, 281.19 ± 62.43, 945.38 ± 151.55, and 5.67 ± 0.90, respectively. During a median follow-up of 98 months, 1234 all-cause deaths and 409 cardiovascular deaths were observed.

For the HOMA-IR index and TyG-BMI index, compared with participants in the lowest quartile, those in the higher quartiles were more likely to be male, younger, non-Hispanic Black, unmarried, have a lower PIR, be obese, and have a history of hypertension, diabetes, CVD, and cancer (P < 0.05). In contrast, participants in the higher quartile of TyG index, TyG-WC index, and TyG-WHtR index were more likely to be male, older, divorced or widowed, less educated, have a lower PIR, obesity, and above medical history (P < 0.05). The laboratory characteristics of the enrolled participants exhibited a similar trend across all indices. Those in the highest quartile showed significantly elevated levels of ALT, ALP, GGT, LDH, TG, and HbA1c, accompanied by significantly lower HDL levels. The baseline characteristics of all MASLD patients, stratified by quartiles of the different IR surrogates (HOMA-IR index, TyG index, TyG-BMI index, TyG-WC index, and TyG-WHtR index) are summarized in Table 1 and in the supplementary files (Tables S2–S5).

Table 1 Baseline characteristics of patients with MASLD based on HOMA-IR index quartiles

The relationship between IR surrogates and mortality in patients with MASLD

The TyG index, TyG-WHtR index, and TyG-WC index were significantly associated with both all-cause mortality and CVD mortality in MASLD patients (P < 0.05). Patients in the higher quartiles of these indices experienced poorer survival outcomes, whereas those in the lower quartiles showed better overall survival in Fig. 2 and Fig.S2.

Fig. 2

Kaplan–Meier curves show the overall survival probabilities of the MASLD patients with different quartiles of various IR surrogates. A HOMA-IR index; B TyG index; C TyG-BMI index; D TyG-WC index; E TyG-WHtR index. Abbreviations: MASLD, metabolic dysfunction-associated steatotic liver disease; IR, insulin resistance; HOMA-IR, homeostatic model assessment for IR; TyG, triglyceride-glucose; TyG-WC, TyG-waist circumference; TyG-WHtR, TyG-waist-to-height ratio

Three Cox regression models were constructed to explore the independent associations between the IR indices and the risk of all-cause mortality and CVD mortality in Fig. 3 and in the supplementary files (Fig.S3-S4). In the fully adjusted models(Fig. 3A), the highest quartiles (Q4) of the TyG index, TyG-WC index, and TyG-WHtR index were significantly and positively associated with all-cause mortality in MASLD patients (TyG: HR = 1.36, 95% CI 1.04–1.78, P = 0.03; TyG-WC: HR = 1.83, 95% CI 1.12–2.99, P = 0.02; TyG-WHtR: HR = 1.70, 95% CI 1.12–2.59, P = 0.01). Notably, the highest quartile (Q4) of the TyG-BMI index was significantly associated with both all-cause mortality (HR = 2.84, 95% CI 1.73–4.67, P < 0.001) and CVD mortality (HR = 5.32, 95% CI 2.26–12.49, P < 0.001) in MASLD patients.

Fig. 3

The forest plots show the associations between the different IR surrogates and mortality in MASLD patients in Model 3. A all-cause mortality; B cardiovascular mortality. Note: Model 3 was adjusted for age, gender, race, education level, BMI, energy intake, smoking, alcohol, CVD, diabetes, hypertension, AST, ALT, and PIR. Abbreviations: IR, insulin resistance; MASLD, metabolic dysfunction-associated steatotic liver disease; HOMA-IR, homeostatic model assessment for IR; TyG, triglyceride-glucose; TyG-WC, TyG-waist circumference; TyG-WHtR, TyG-waist-to-height ratio; CVD, cardiovascular disease; AST, aspartate aminotransferase; ALT, alanine aminotransferase; PIR, income-to-poverty ratio

Then, ROC curve analyses and AUC values were used to assess the predictive performance of the three models for all-cause and cardiovascular mortality in MASLD patients. The results showed that, for all-cause mortality (Fig. 4A), the AUC was 0.730 (95% CI [0.703, 0.747]) in the unadjusted model (Model 1), while the fully adjusted model (Model 3) showed a notable improvement (AUC = 0.866, 95% CI [0.844, 0.875], P < 0.001). Similarly, regarding cardiovascular mortality (Fig. 4B), the unadjusted model (Model 1) yielded an AUC of 0.737 (95% CI [0.690, 0.756]), with the fully adjusted model (Model 3) achieving the highest predictive performance (AUC = 0.902, 95% CI [0.883, 0.914], P < 0.001).

Fig. 4

The ROC curves display the sensitivity and specificity of each model in predicting MASLD mortality. The AUC represents the discrimination ability of each model. A all-cause mortality; B cardiovascular mortality. Note: Model 1 was unadjusted; Model 2 was adjusted for age, gender, and race; Model 3 was adjusted for age, gender, race, education level, BMI, energy intake, smoking, alcohol, CVD, diabetes, hypertension, AST, ALT, and PIR. Abbreviations: ROC, receiver operating characteristic; MASLD, metabolic dysfunction-associated steatotic liver disease; area under the curve (AUC); CVD, cardiovascular disease; AST, aspartate aminotransferase; ALT, alanine aminotransferase; PIR, income-to-poverty ratio

RCS analysis of IR surrogates and mortality in MASLD patients

In the multivariable Cox regression analyses, potential nonlinear relationships were observed between the IR surrogates and all-cause/cardiovascular mortality in MASLD patients. To further verify this idea, RCS analysis conducted. As shown in Fig. 5 and supplementary files (Fig. S5), the results indicated a linear trend between the HOMA-IR index and all-cause mortality in MASLD patients (P for overall < 0.001), while the TyG index and TyG-WHtR index showed significant nonlinear associations (P for nonlinearity < 0.001). In contrast, the TyG-BMI index and TyG-WC index exhibited U-shaped associations with both all-cause mortality and CVD mortality in MASLD patients.

Fig. 5

The nonlinear relationship between the different IR surrogates and all-cause mortality in MASLD patients. A HOMA-IR index; B TyG index; C TyG-BMI index; D TyG-WC index; E TyG-WHtR index. Abbreviations: MASLD, metabolic dysfunction-associated steatotic liver disease; IR, insulin resistance; HOMA-IR, homeostatic model assessment for IR; TyG, triglyceride-glucose; TyG-WC, TyG-waist circumference; TyG-WHtR, TyG-waist-to-height ratio

Moreover, two-piece linear regression models were used to assess the relationship between the IR indices and mortality outcomes (all-cause/CVD mortality) among MASLD patients (Table 2 and Tables S6). We found that the inflection points for all-cause mortality for the TyG index, TyG-BMI index, and TyG-WHtR index were approximately the same as those for CVD mortality, which were 8.8, 270.49, and 5.55, respectively. When the TyG index exceeded 8.61, each unit increase was associated with a 77.7% elevation in the adjusted hazard ratio (HR) for all-cause mortality (HR = 1.777; 95% CI 1.401–2.253; log-likelihood ratio P = 0.008), whereas no significant relationship with CVD mortality was observed. For the TyG-BMI index, when values exceeded 270.49, each unit increase was associated with a 0.3% increase in the adjusted HR for all-cause mortality (HR: 1.003; 95% CI: 1.002–1.005, log-likelihood ratio test P < 0.05) and a 0.4% increase in the adjusted HR for CVD mortality (HR: 1.004; 95% CI: 1.002–1.007, log-likelihood ratio test P < 0.05). When the TyG-WHtR index exceeded 5.55, each unit increase corresponded to a 28.7% increase in the adjusted HR for all-cause mortality (HR: 1.287; 95% CI 1.169–1.418; log-likelihood ratio test P < 0.001). However, no significant association was found between the TyG-WHtR index and CVD mortality. The two-stage Cox proportional hazards regression model for the HOMA-IR index indicated no significant associations with all-cause mortality or cardiovascular mortality (log-likelihood ratio test P > 0.05).

Table 2 Threshold effect analysis of the different IR surrogates on all-cause mortality in MASLD patients using the two-piecewise linear regression model

Mediation analyses between IR surrogates and mortality in MASLD patients

Mediation analyses showed that HbA1c and insulin indirectly mediated the associations between the IR surrogates and both all-cause mortality and CVD mortality in MASLD patients (Fig.S6). For the TyG index, the proportion of the indirect effect mediated by HbA1c was 81.4% for all-cause mortality and 94.1% for cardiovascular mortality. For the TyG-WC index, HbA1c mediated 37.7% of the indirect effect on all-cause mortality and 96.9% on CVD mortality. For the HOMA-IR index, HbA1c-mediated indirect effects on all-cause and CVD mortality were 18.4% and 23.3%, respectively.

For all-cause mortality, the insulin-mediated indirect effect proportions for the TyG index, TyG-WC index, and HOMA-IR index were 48.0%, 37.8%, and 13.0%, respectively. The proportion of the indirect effect mediated by insulin was 17.4% for the TyG index, 18.5% for the TyG-BMI index, 17.9% for the TyG-WC index, and 39.6% for the TyG-WHtR index. Interestingly, for the HOMA-IR index, the indirect effect mediated by insulin was − 45.3%, suggesting a potential suppression effect rather than mediation.

Subgroup analyses

The results of interaction tests from subgroup analyses stratified by age, gender, education level, race, marital status, PIR, diabetes, cancer, hypertension, CVD, smoking status, and alcohol intake are presented in Fig. 6 and the supplementary files (Fig. S7–S10). Only the age subgroup demonstrated significant interactions across all indices. The significant associations of the TyG-WC index, TyG-WHtR index, and TyG-BMI index with all-cause mortality were more likely in patients aged ≥ 60 years and females, while their associations with CVD mortality were primarily observed in those aged ≥ 60 years. However, among males aged < 60 years, the HOMA-IR index was significantly associated with both all-cause mortality and cardiovascular mortality.

Fig. 6

Subgroup analyses of the association between the TyG-BMI index and mortality in MASLD patients. A all-cause mortality; B cardiovascular mortality; Abbreviations: TyG, triglyceride-glucose; TyG-BMI, TyG -body mass index; MASLD, metabolic dysfunction-associated steatotic liver disease; HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease

Discussion

This study explored the associations between various IR surrogate indices and both all-cause mortality and cardiovascular mortality among MASLD population in an American community. After adjusting for multiple covariates in Cox regression and RCS analyses, the TyG-BMI index was found to be significantly associated with both all-cause mortality and cardiovascular mortality, exhibiting a “U-shaped” relationship. Additionally, threshold effect analyses identified an inflection point of 270.49 for the TyG-BMI index regarding both mortality outcomes, suggesting that values higher or lower than this threshold may increase the risk of death in this population. In contrast, the TyG index, TyG-WC index, and TyG-WHtR index were only significantly associated with all-cause mortality, whereas HOMA-IR index was not significantly related to either mortality.

IR is a critical driver of hepatic steatosis, inflammation, and fibrosis, and plays an essential role in the pathogenesis of MASLD [16,17,18]. Through stimulating de novo lipogenesis, increasing the influx of free fatty acids into the liver, and impairing fatty acid oxidation, hepatic IR promotes hepatic lipid accumulation [19,20,21]. Such excessive lipid deposition induces oxidative stress and increases hepatic mitochondrial metabolism, leading to hepatocyte injury, apoptosis, and subsequent fibrosis [22, 23]. Moreover, IR has been reported to exert extrahepatic effects, notably exacerbating CVD risk, the primary cause of mortality among MASLD patients. Indeed, IR has been linked to atherogenic dyslipidemia and hypertension, and it initiates multiple proatherogenic, prothrombotic, and proinflammatory pathways that contribute to the progression of CVD [24].

In this study, we observed that compared with the robust associations of TyG-related indices, the HOMA-IR index exhibited relatively weak or insignificant relationships with both all-cause and cardiovascular mortality in MASLD patients. This may be related to the differences in variability, biological relevance, or measurement issues. First, the calculation of HOMA-IR depends on fasting insulin and blood glucose levels. Previous studies have shown that insulin detection is easily influenced by technique, antibody specificity and sample storage conditions, resulting in a high coefficient of variation between individuals and inter-laboratories [25]. The TyG index is based on triglyceride and glucose, which are more standardized and have stronger stability [26]. Secondly, HOMA-IR mainly captures insulin-glucose homeostasis, whereas the TyG index more comprehensively reflects systemic insulin resistance and abnormal lipid metabolism, which may make TyG index more biologically relevant [27].

Recently, the TyG index and its related indices have been found to be alternatives to the gold standard for assessing IR. These TyG-related indices reflect imbalances in both glucose and lipid metabolism—two key contributors to atherosclerosis. Hypertriglyceridemia increases the formation of highly atherogenic low-density lipoprotein (LDL) particles [28, 29], while hyperglycemia leads to the overproduction of advanced glycation end products (AGEs), resulting in endothelial dysfunction and increasing CVD risk [30]. This mechanistic interplay may explain why TyG-related indices serve as strong predictors of cardiovascular mortality in MASLD patients. Previous studies have similarly confirmed that these indices are associated with all-cause mortality and cardiovascular mortality across diverse populations. For example, Xie et al. [31] and Lu et al. [11] demonstrated that TyG-related indices were closely associated with both all-cause mortality and CVD mortality in patients with cardiometabolic syndrome and diabetes. Another study showed a U-shaped association between TyG index and both all-cause mortality and CVD mortality in CVD patients [32]. Our study extends upon these findings, indicating that combining the TyG index with obesity indicators can more effectively predict mortality in MASLD patients, especially TyG-BMI index and TyG-WC index. Our findings show that the TyG-BMI index is better than the TyG-WC index. Although WC is widely used to assess abdominal adiposity and central obesity, it cannot distinguish between visceral and subcutaneous fat [33]. Studies have demonstrated that visceral fat plays a more important significant in IR [34]. BMI is more comprehensive to assess overall metabolic burden and exhibits lower inter-observer variability compared with WC.

The TyG-BMI index, which integrates TG, fasting blood glucose, and BMI, represents a comprehensive IR assessment tool [14]. Obesity, especially visceral obesity, secretes proinflammatory factors such as leptin and cytokines [e.g., interleukin-6 (IL-6) and tumor necrosis factor α (TNF-α)], thereby inducing a chronic inflammatory state [35]. This chronic inflammation exacerbates IR, leading to systemic inflammation, endothelial dysfunction, and atherosclerosis [36]​. Studies have shown that the TyG-BMI index is closely associated with MASLD, metabolic syndrome, cardiovascular disease, and type 2 diabetes [37,38,39,40,41]. Different from previous studies, our research shows that a U-shaped relationship was observed between the TyG-BMI and both all-cause mortality and cardiovascular mortality in MASLD patients. This finding highlighted the synergistic effect of obesity and IR on adverse hepatic and cardiovascular outcomes. These results are consistent with those reported by Er et al. [13]. It is noteworthy that the U-shaped association between TyG-BMI index and MASLD mortality reflects, on the one hand, the negative effect of excessive adiposity on metabolic disorders, and on the other hand, the adverse influence of low TyG–BMI on metabolic and immune functions. We hypothesize this may be because sarcopenia impairs energy metabolism and glucose storage capacity, worsening insulin resistance and lipid metabolism disorders. Second, the extremely low TyG-BMI may reflect the potential malnutrition or energy deficiency, which is associated with impaired immune function, decreased metabolic reserve and increased susceptibility to infection in MALD patients [42]. Moreover, in advanced MASLD, systemic inflammation and catabolism driven by cachexia may lead to abnormal decline in BMI and adipose tissue mass [43]. This may amplify cardiovascular risk through cytokine-mediated pathway, and thereby contribute to increased mortality.

Our study also determined that the significant associations between TyG-related indices and both all-cause and CVD mortality were more obvious in MASLD patients aged ≥ 60 years. This is consistent with previous studies that older patients with IR face a heightened risk of cardiovascular events and mortality [44]. This may be attributed to the reduced ability of older patients to resist metabolic and vascular injury, amplifying the impact of IR on adverse outcomes. Consequently, exploring targeted interventions—such as lifestyle modifications or pharmacotherapies that enhance insulin sensitivity—among older MASLD patients could provide valuable insights for reducing mortality in this high-risk population.

The results of this study show that high-risk groups with different insulin resistance related indices have significant differences in socio-demographic characteristics, which may guide precision prevention strategies. Specifically, similar to the findings of VanWagner et al. and Zou et al., men are at high risk of MASLD [45, 46]. Males exhibit inherently differences of adipose tissue distribution, hormone regulation, and lifestyle behaviors—such as dietary habits and physical activity, which makes them more prone to insulin resistance compared with females [47]. Second, we believe that marital status (unmarried, divorced or widowed) and socioeconomic factors (low education level and low PIR) are also high-risk factors. This is consistent with a Swedish study [48]. The study has shown that unmarried, divorced, or widowed may feel lack of social support and high psychological stress, which may promote the development of unhealthy behaviors and metabolic diseases. Besides, socioeconomic factors are frequently associated with poorer health literacy, limited medical care, and adverse living conditions, which will increase the risk of obesity, diabetes and other metabolic diseases [49].

Advantages and limitations.

This study has several notable strengths and limitations. First, the study population was drawn from a nationally representative NHANES cohort selected through a complex, multistage probability sampling method. A large, well-characterized cohort with long-term follow-up enhances the validity and generalizability of these findings. Second, this study evaluated both the linear and the nonlinear relationships between TyG-related indices and all-cause as well as CVD mortality, identified the segmented effects between variables, and determined the inflection point.

However, several limitations should be considered. First, this study was focused on a U.S. population, and racial difference is an important consideration. Whether these results can be widely applied to other races has yet to be verified and further research is needed. Second, the inflection points are the threshold for this cohort rather than absolute clinical cutoff points. These values may fluctuate with population characteristics such as age, race, complications, and lifestyle factors. Moreover, as an observational and retrospective study, our results demonstrated the association between the insulin resistance indices and the mortality in MASLD patients, but the causal relationship could not be determined. Further prospective interventional research is required to verify these findings. Despite these limitations, the present study helps to address existing knowledge gaps, and more extensive data collection in future studies may further clarify these associations.

Conclusions

In conclusion, this study provides important insights into the associations between IR and mortality outcomes in MASLD patients. Among various IR surrogates, the TyG-BMI index emerged as the most valuable predictor of both all-cause mortality and cardiovascular mortality risk in this population. Notably, a significant U-shaped association was observed between the TyG-BMI index and both all-cause mortality and cardiovascular mortality.