Association of cholesterol, high-density lipoprotein and glucose (CHG) index with mortality risk in metabolic dysfunction-associated steatotic liver disease (MASLD) adults: results from two prospective cohorts

Abstract

Background

Metabolic dysfunction-associated steatotic liver disease (MASLD) is a health issue of increasing concern worldwide. The cholesterol-high density lipoprotein-glucose (CHG) index integrates key metabolic pathways, but its relationship with the prognosis of MASLD remains unclear.

Methods

A total of 13,286 MASLD adults were included in the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2018. The mortality results were determined by linking to the National Death Index (NDI) records. A weighted Cox model, restricted cubic spline (RCS) and threshold analysis were used to examine the nonlinear relationship between baseline CHG and all-cause (ACM) or cardiovascular mortality (CVM). Subgroup analysis examined consistency among different populations. Mediating analysis explored the mediating role of weight-adjusted waist index (WWI), neutrophil (NE) and estimated glucose disposal rate (eGDR). External validation was conducted using 2,914 individuals from the Health and Retirement Study (HRS) 2016 cohort.

Results

In a follow-up with a median time of 112 months, 1,688 cases of ACM and 551 cases of CVM occurred. Weighted Cox regression showed that an increase in the CHG index was significantly associated with an increased mortality risk in MASLD. The RCS curve and threshold effect show a significant U-shaped nonlinear relationship of the CHG index with ACM and CVM. The results of the subgroup analysis showed that the association between the CHG index and mortality risk was more significant in the subgroups under 60 years old and lean. Mediating analysis indicates that WWI, NE and eGDR may partially mediate the effects of the CHG index on ACM and CVM. In the HRS cohort, the CHG index was significantly correlated with ACM.

Conclusions

Our study confirmed a robust association between the CHG index and ACM and CVM in patients with MASLD based on the two prospective cohorts of NHANES and HRS.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-026-03079-2.

Research insights

What is currently known about this topic?

• MASLD is characterized by metabolic disorders, which in turn can drive theprogression of MASLD and increase the mortality risk.

• The CHG index integrates multiple metabolic pathways and is an emergingmarker of metabolic disorders.

What is the key research question?

• What is the relationship between the CHG index and the mortality risk of MASLD?

What is new?

• A significant U-shaped association of the CHG index with the ACM and CVM in MASLD.

How might this study influence clinical practice?

• Maintaining the CHG index near the observed threshold may be associated with lower mortality risk for MASLD.

Introduction

Non-alcoholic fatty liver disease (NAFLD) is regarded as the most common chronic liver disease, with a global prevalence of approximately 30% and is expected to continue to rise [1]. In recent years, the terms “steatotic liver disease (SLD)” and “metabolic dysfunction-associated steatotic liver disease (MASLD)” have been introduced to gradually replace the old terms “fatty liver disease (FLD)” and “NAFLD” based on the Delphi consensus [2]. The diagnosis of MASLD requires that individuals with hepatic steatosis have at least one of the five cardiometabolic risk factors simultaneously. Therefore, compared with NAFLD, MASLD can more accurately identify SLD patients related to metabolism, emphasizing the role of metabolic factors in disease progression [3]. With lifestyle changes such as high-calorie diets and lack of exercise, MASLD has become one of the significant challenges in the field of public health [4]. The rising prevalence of MASLD has also brought an enormous socioeconomic burden to the world [5]. The pathogenesis of MASLD is complex and is mainly closely related to metabolic disorder factors such as insulin resistance (IR), obesity and dyslipidemia [6–8]. Beyond hepatic complications such as liver fibrosis, cirrhosis, and liver cancer, MASLD patients frequently experience a high incidence of cardiovascular diseases (CVD), and cardiovascular events represent one of the leading causes of mortality among these patients [9–12]. This suggests that all-cause mortality (ACM) and cardiovascular mortality (CVM) have significant clinical significance for patients with MASLD. In-depth research on the related risk factors is of great significance for improving the prognosis of MASLD patients.

Recently, the cholesterol, high-density lipoprotein and glucose (CHG) index, as an emerging biomarker for evaluating metabolic disorders, has attracted increasing attention from researchers. Comprising total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and fasting blood glucose (FBG), the CHG index reflects both lipid and glucose metabolism. Initially, the CHG index was proposed for its superior diagnostic efficacy in type 2 diabetes mellitus (T2MD) [13]. Several composite markers that combine lipid and/or glucose parameters, such as the triglyceride-glucose (TyG) index and lipid accumulation product (LAP), have been developed to capture metabolic dysfunction and predict cardiovascular events or diabetes [14, 15]. However, they all rely on a single lipid component and ignore total cholesterol or HDL-C. In contrast, the CHG index may more comprehensively reflect the balance among atherosclerotic cholesterol, protective HDL-C and glucose metabolism stress. Several comparative studies have revealed that the predictive ability of the CHG index is comparable to or even slightly superior to that of the TyG index in terms of short-term mortality in stroke patients and long-term CVD risk in middle-aged and elderly people [16, 17], and it is significantly better than the TyG index and LAP in terms of diabetic retinopathy [18]. These findings indicate that the CHG index potentially represents an integrative biomarker capable of assessing systemic metabolic dysregulation. To date, some studies have reported the potential link between the CHG index and the prognosis of cardiovascular and metabolic diseases such as stroke, aortic stenosis and metabolic syndrome [19–21]. However, there are relatively few studies on the CHG index in patients with liver diseases, especially in the specific group of MASLD patients, and its association with ACM and CVM has not been fully explored. Therefore, it is of great significance to conduct in-depth research on the role of the CHG index in the prognosis of patients with MASLD.

Beyond simple correlations, we attempt to dissect the pathways by which the CHG index may influence mortality risk, such as obesity, inflammation and IR. The weight-adjusted waist index (WWI), which is not only independent of overall weight but also captures central obesity, has been repeatedly associated with liver fat accumulation and poor prognosis of MASLD [22]. The neutrophil (NE) count is an easily accessible systemic immune indicator that reflects low-grade inflammation that drives the progression of MASLD disease [23]. Elevated NE is associated with more severe steatosis, fibrosis and adverse clinical outcomes [24]. The estimated glucose disposal rate (eGDR) is an effective alternative indicator of systemic IR caused by blood glucose levels, blood pressure and waist [25]. eGDR is closely related to CVD risk and long-term prognosis of MASLD [26]. Therefore, WWI, NE and eGDR are biologically reasonable mediators that link the CHG index to mortality risk.

This study aims to utilize the National Health and Nutrition Examination Survey (NHANES) database and, through a large-scale cohort study design, deeply explore the association of the CHG index with ACM and CVM in MASLD, as well as the potential mediating effects. The Health and Retirement Study (HRS) database was further analyzed to verify the robustness of our research results. We uniformly selected the noninvasive hepatic steatosis index (HSI) as the tool for defining hepatic steatosis in both cohorts. The findings of this study will further enrich the understanding of the intrinsic connection between metabolic disorders and the mortality risk of MASLD, provide a theoretical basis and direction for future related basic research and clinical trials, and promote the improvement of clinical diagnosis and treatment levels of MASLD.

Method

Study design and population

NHANES is a nationally representative research project managed by the National Center for Health Statistics (NCHS) under the Centers for Disease Control and Prevention (CDC) of the United States. Designed to evaluate the health and dietary conditions of Americans, this initiative serves as a critical data source for informing evidence-based public health interventions and policy development. By collecting comprehensive health metrics, NHANES supports epidemiological research and guides strategies for disease prevention and health promotion across the U.S. population. NHANES is an annual continuous survey that includes approximately 5,000 participants each year, covering all age, racial, and gender groups to ensure that the data are representative [27]. The subjective reports, including demographic characteristics, health history and lifestyle behaviors, and objective measures, including physiological measurements and laboratory tests, of the participants are collected through home interviews and mobile examination center (MEC) physical examinations. The study protocol received ethical clearance from the National Center for Health Statistics’ Ethics Review Committee prior to data collection. All enrolled participants provided documented consent through written forms. As the current investigation utilizes deidentified NHANES records that are openly accessible without restriction through the official repository (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx), this secondary analysis was exempt from further institutional review board evaluation per standard research guidelines.

The analysis incorporated data from ten consecutive NHANES cycles spanning 1999 to 2018, initially comprising 101,316 participants. Through systematic application of predefined exclusion criteria in Fig. 1, we sequentially removed ineligible participants: (1) participants under 20 years old (n = 46,235), (2) those who were not diagnosed with MASLD (n = 24,607), (3) participants with missing CHG index values (n = 15,777), (4) individuals lacking mortality follow-up information (n = 18), pregnant women (n = 459) and participants without available weight data (n = 934). After these exclusions, our final analytical cohort consisted of 13,286 MASLD adults with complete follow-up records and all necessary clinical data.

Fig. 1.

The flow chart of inclusion and exclusion criteria in the study

Definition of MASLD

The current diagnostic framework in the latest consensus for MASLD requires the presence of hepatic steatosis in conjunction with at least one cardiometabolic abnormality, and excludes other causes [2]. These metabolic risk factors include: (1) body mass index (BMI) ≥ 25 kg/m² or waist circumference (WC) exceeding sex-specific thresholds (≥ 94 cm for males, ≥ 80 cm for females). (2) elevated FBG (≥ 100 mg/dL), increased hemoglobin A1c (HbA1c ≥ 5.7%), or current use of glucose-lowering medications. (3) systolic/diastolic blood pressure (SBP/DBP) ≥ 130/85 mmHg or ongoing antihypertensive therapy. (4) fasting triglyceride (TG) levels ≥ 150 mg/dL or receipt of lipid-modifying treatment. (5) reduced HDL-C (< 40 mg/dL in men, < 50 mg/dL in women) or current lipid-lowering medication use.

To ensure consistency and comparability across both the NHANES and HRS cohorts, we defined hepatic steatosis as an HSI > 36. The reasons for choosing this critical value are as follows: (1) HSI > 36 was initially verified through liver ultrasound in a large-scale cohort, with an area under the curve (AUC) of 0.812 [28]. (2) HSI only requires alanine aminotransferase (ALT), aspartate aminotransferase (AST), BMI, gender and diabetes, all of which are available in the NHANES and HRS cohorts, without the need for additional biomarkers [29].

The fatty liver index (FLI) is another common screening tool for hepatic steatosis, with an AUC of 0.84 verified by ultrasound [30]. Since the dry blood spot detection menu in the HRS cohort does not include γ-glutamyl transferase (GGT), which is an important component of FLI, we decided to abandon using FLI as a tool to define hepatic steatosis. Instead, we applied FLI in the NHANES cohort in the sensitivity analysis to test whether the conclusion was affected by the selection of indicators. The calculation formulas of HSI and FLI are as follows:

  

  

ALT: alanine aminotransferase (U/L); AST: aspartate aminotransferase (U/L); BMI: body mass index (kg/m²); TG: triglyceride (mg/dL); GGT: γ-glutamyl transferase (IU/L); WC: waist circumference (cm).

Definition of CHG index

The CHG index is mathematically expressed as follows: CHG index = Ln [TC (mg/dL) * FBG (mg/dL)/ 2 * HDL-C (mg/dL)] [13].

Assessment of mortality

The study examined mortality outcomes in MASLD patients, focusing on two principal measures: (1) ACM as the primary outcome and (2) CVM as the secondary outcome. CVM was classified according to ICD-10 criteria, encompassing cardiac conditions (codes I00-I09, I11, I13, and I20-I51) and cerebrovascular events (codes I60-I69). Survival data were acquired by cross-referencing NHANES participant records with the National Death Index, with observation periods spanning from enrollment through December 31, 2019, or until death. Follow-up intervals were quantified in months to facilitate accurate survival analysis. To maintain participant confidentiality while providing research access, NHANES employs stringent data protection measures in its publicly available linked mortality file (LMF). These include limited variable disclosure, implementation of statistical disclosure control techniques, and application of data perturbation methods to prevent reidentification.

Potential covariates

This study considered comprehensive covariates covering demographics, physical measurements, lifestyle, medical history and laboratory tests.

Demography includes gender (female/male), age (years), race, educational levels, marital status, and economic status. The economic situation is evaluated through the poverty income ratio (PIR), and a PIR value below 1.0 indicates poverty.

Body measurement indicators included height (cm), weight (kg), WC (cm), SBP (mmHg) and DBP (mmHg). BMI (kg/m²) = weight (kg)/ square of height (m). SBP and DBP were calculated as the average of three valid blood pressure measurements. The calculation formula of WWI is as follows: WWI = WC (cm)/ square root (sqrt) of weight (kg).

Lifestyle includes daily energy intake, smoking status, drinking status and physical activity. The daily energy intake is obtained based on the 24-hour dietary information recorded in the dietary interview section. If the participants had two complete 24-hour dietary records, the energy intake was calculated as the average of the total energy intake over the two days; otherwise, it was equal to the energy intake in the first 24 h. The smoking status was classified based on the lifetime smoking quantity and current smoking situation, and nonsmokers (< 100 cigarettes), current smokers and former smokers were distinguished. Drinking was considered to be greater than or equal to 12 drinks in the past 12 months. Physical activity levels are classified into inactive, moderate (activities that cause a mild increase in heart rate/respiratory rate for ≥ 10 consecutive minutes per week), and vigorous (activities that cause a significant increase in heart rate/respiratory rate for ≥ 10 consecutive minutes per week).

The medical history focused on major chronic conditions, including hypertension, diabetes, CVD and cancer. For hypertension classification, participants met diagnostic criteria if they either (1) reported a previous physician diagnosis, (2) were currently prescribed antihypertensive medications, or (3) demonstrated persistently high blood pressure (defined as the mean systolic pressure ≥ 140 mmHg or diastolic ≥ 90 mmHg on repeated assessments). Diabetes status was determined by meeting at least one of four conditions: self-reported diagnosis, active use of glucose-lowering medications or insulin injection, HbA1c levels ≥ 6.5%, or FBG concentrations ≥ 126 mg/dL [31]. The CVD category included physician-confirmed cases of heart failure, ischemic heart conditions (such as myocardial infarction, angina pectoris, or coronary artery disease), or stroke. Cancer is diagnosed if a participant has been informed by a doctor of having cancer or any type of malignant tumor.

Laboratory tests included triglyceride (TG, mg/dL), low-density lipoprotein cholesterol (LDL-C, mg/dL), ALT (U/L), AST (U/L), serum creatinine (SCR, mg/dL), serum uric acid (SUA, mg/dL), total bilirubin (TBIL, mg/dL), HbA1c (%), neutrophil count (NE, 10⁹/L), platelet count (PLT, 10⁹/L), and lymphocyte count (LYC, 10⁹/L). The Fibrosis 4 score (FIB-4) is a noninvasive method for evaluating liver fibrosis with chronic liver diseases. The calculation formula is FIB-4 = Age (years) * AST (U/L)/ [PLT (10⁹/L) * sqrt(ALT) (U/L)]. The systemic immune inflammation index (SII) is a novel inflammatory marker that reflects the inflammatory state and immune system function, and it is calculated as SII = NE (10⁹/L) * PLT (10⁹/L)/ LYC (10⁹/L). The estimated glomerular filtration rate (eGFR) valuation was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. The calculation formula of eGDR is as follows: eGDR = 21.158 − [0.09 * WC (cm)] − (3.407 * hypertension status) - [0.551 * HbA1c (%)]. The hypertension status is assigned a value of 1 if it exists and 0 if it does not.

Statistical analysis

This study followed the NCHS guidelines and conducted statistical analysis using appropriate sampling weights. The FBG included in the CHG index is derived from fasting tests, so fasting sampling weights (WTSAF4YR or WTSAF2YR) are strictly adopted. For the participants in the 1999–2002 cycle, their weights were calculated as WTSAF4YR * 2/10, and for those in the 2003–2018 cycle, their weights were calculated as WTSAF2YR* 1/10. To reduce selection bias and handle missing data to varying degrees (Table S1), the “mice” R package was used for interpolation and iteration of the covariates. The missing values are filled in via the predictive mean matching method. The baseline characteristics of these MASLD subjects were described based on the quartiles of the CHG index. Continuous variables, presented as medians (interquartile ranges), were analyzed using nonparametric Mann Whitney U tests. Categorical variables, expressed as frequencies (percentages), were compared through chi-square tests. Survival differences in ACM and CVM of MASLD across CHG index quartiles were assessed using Kaplan-Meier survival curves with log-rank tests.

Three weighted multivariate Cox regression models were constructed to examine associations between the CHG index and mortality outcomes in MASLD. Model 1 has not adjusted any variables. Model 2 adjusts for gender, age, race, educational level, marital status, and economic conditions. Model 3 further adjusted for daily energy intake, smoking status, drinking status, physical activity, CVD, cancer, BMI, SBP, SUA, eGFR, LDL-C, TBIL, FIB-4 and SII on the basis of Model 2. The multicollinearity test calculates the variance inflation factor (VIF) of all covariates in the model (Table S2). The VIF of all variables is less than 5, indicating that there is no significant collinearity in the model [32]. The results of Cox regression are reported as hazard ratios (HRs) with corresponding 95% confidence intervals (95% CIs).

The restricted cubic splines (RCS) in the fully adjusted Model 3 were used to evaluate the nonlinear relationship of the CHG index with mortality risk in MASLD. The Akaike Information Criterion (AIC) is used to evaluate the model fit degree, and the nodes corresponding to the minimum value of AIC are selected as the nodes of the RCS curve. The threshold effect analysis is further carried out in the RCS curve of the nonlinear relationship. Segmented Cox proportional hazards regression analyses were performed on both sides of the cutoff values, and the likelihood ratio test was used to compare the segmented differences.

Subgroup analysis was conducted in the strata of gender (female or male), age (< 60 or ≥ 60 years), lean (BMI < 25 kg/m²) or nonlean (BMI ≥ 25 kg/m²), diabetes (yes or no), hypertension (yes or no), CVD (yes or no) and cancer (yes or no). These analyses aimed to identify factors potentially influencing the relationship between the CHG index and mortality outcomes in MASLD. Interaction effects were examined by incorporating multiplicative terms into weighted Cox regression models. According to the guidelines for subgroup analysis, conducting independent tests on multiple subgroups simultaneously increases the risk of Type I errors [33]. In this study, Bonferroni correction was adopted to reduce the risk of false positivity, and the p for interaction was compared with the corrected significance threshold α = 0.05/7 = 0.0071 [34].

A causal mediation model was adopted in this study to examine whether the effect of the CHG index on mortality risk is mediated by potential mediators, including WWI, NE and eGDR [25]. First, a mediator model is constructed with potential mediators as the continuous dependent variable and CHG as the independent variable to evaluate the path a coefficient of CHG with respect to potential mediators. Subsequently, an outcome model was constructed with the follow-up duration as the time scale, ACM or CVM as the endpoint events, and CHG and potential mediators as the exposure factors, obtaining the direct coefficient of CHG to the mortality risk (path c’ ) and the coefficient of potential mediators to the mortality risk (path b). The mediator model and the outcome model simultaneously controlled the confounding factors of Model 3. Through 1000 non-parametric Bootstrap samples with the “mediation” R package, the direct effect (DE), indirect effect (IE), and total effect (TE) were calculated, and the mediating proportion (IE / TE * 100%) was obtained.

All statistical computations were performed using R software (version 4.3.3). We employed several specialized R packages to conduct our analyses: the “survey” package accounted for the complex sampling design, while survival analyses were implemented using the “survival” package for Cox proportional hazards modeling and interaction testing. Data visualization was achieved through the “survminer” package for KM curves, along with the “ggplot2” and “rms” packages for RCS plots. Threshold effect analysis was conducted using the “segmented” package, and subgroup analyses were processed with the “broom” package. For all statistical tests, significance was determined at p-value = 0.05 (two-tailed).

To assess the stability of the research results, sensitivity analyses were conducted from various aspects. (1) Considering that drug use can affect the level of the CHG index, the hypoglycemic and lipid-lowering drugs were further adjusted. (2) To reduce reverse causality, we excluded participants who died within two years after follow-up. (3) We selected FLI ≥ 60 and ZJU > 38 as the new diagnostic criteria for SLD to evaluate the impact of the CHG index on the risk of death in different MASLD populations. (4) To reduce the lever-driven effect of a few extreme CHG indices on the main association, the standardized Z-score method was used to eliminate outliers. A CHG index with an absolute value of Z-score greater than 3 is defined as an outlier. (5) SLD was defined using HSI > 36 for American adults in the HRS wave 2016 database. The HRS database is a national longitudinal survey database that focuses on adults aged 50 and above. It was initiated in 1992 and a survey is conducted every two years. The survival time of the participants was calculated based on the death dates obtained from the interviews in 2020. Similarly, three multivariate Cox regression models were used to verify the association between the CHG index and the mortality risk of MASLD, and there was no collinearity in these models (Table S3). Model 1 does not adjust for covariates. In Model 2, adjustments were made for age and gender. In Model 3, we further adjusted for smoking status, drinking status, physical activity, cancer, CVD, BMI, SBP, LDL-C, SCR, high-sensitivity C-reactive protein (hsCRP), TBIL and FIB4.

Results

Baseline characteristics of the participants

After strict inclusion and exclusion criteria, a total of 13,286 adult MASLD patients were included in this study. According to the weighting calculation, these study participants represented 130,399,038 MASLD adults across the United States (Table S4). The baseline characteristics of the study participants based on the quartiles of the CHG index are summarized in Table 1. Among these MASLD participants, the median age was 50 (36–63) years old. The majority are male (53.0%), the largest race group is white (42.6%), the majority have received higher education and 48.7% have a college degree or above, over 20% are economically poor, and over 60% are married or cohabiting. Nearly one-fifth of people suffer from diabetes and nearly half have hypertension. A small number of people suffer from CVD (11.2%) and cancer (8.7%). Compared with the lowest quartile group (Q1) of CHG, MASLD patients in the highest quartile group (Q4) were often male, older, white, less educated, poorer economically, married, more likely to suffer from diabetes, hypertension and CVD, and less active in exercise. They had higher levels of energy intake, BMI, WWI, SBP, DBP, TG, LDL, SUA, FIB-4, and SII.

Table 1.

Baseline characteristics of participants by quartiles of the CHG index

Variables	Total	Quartiles of CHG				p-value
		Q1 (< 5.09)	Q2 (5.09–5.34)	Q3 (5.34–5.60)	Q4 (≥ 5.60)
Count, n	13,286	3322	3322	3320	3322
Gender, n (%)						< 0.001
Female	7036(53.0)	2462(74.1)	1888(56.8)	1458(43.9)	1228(37.0)
Male	6250(47.0)	860(25.9)	1434(43.2)	1862(56.1)	2094(63.0)
Age, years	50.00(36.00,63.00)	47.00(33.00,63.00)	49.00(36.00,63.00)	50.00(37.00,63.00)	52.00(40.00,63.00)	< 0.001
Race, n (%)						< 0.001
Mexican American	2756(20.7)	545(16.4)	699(21.0)	718(21.6)	794(23.9)
Non-Hispanic White	5658(42.6)	1306(39.3)	1417(42.7)	1474(44.4)	1461(44.0)
Non-Hispanic Black	2755(20.7)	999(30.1)	700(21.1)	567(17.1)	489(14.7)
Other races	2117(15.9)	472(14.2)	506(15.2)	561(16.9)	578(17.4)
Education level, n (%)						< 0.001
Below high school	3688(27.8)	745(22.4)	864(26.0)	966(29.1)	1113(33.5)
High school	3126(23.5)	719(21.6)	806(24.3)	788(23.7)	813(24.5)
College or above	6472(48.7)	1858(55.9)	1652(49.7)	1566(47.2)	1396(42.0)
Economic condition, n (%)						0.011
PIR < 1	2697(20.3)	652(19.6)	645(19.4)	659(19.8)	741(22.3)
PIR ≥ 1	10,589(79.7)	2670(80.4)	2677(80.6)	2661(80.2)	2581(77.7)
Marital status, n (%)						< 0.001
Married/Living with a partner	8405(63.3)	1892(57.0)	2129(64.1)	2152(64.8)	2232(67.2)
Widowed/Divorced/Separated	2937(22.1)	825(24.8)	721(21.7)	686(20.7)	705(21.2)
Never married	1944(14.6)	605(18.2)	472(14.2)	482(14.5)	385(11.6)
Daily energy intake, kcal/d	1908.00 (1445.01,2477.50)	1807.00 (1367.50,2331.50)	1902.00 (1444.63,2431.25)	1973.00 (1504.38,2563.25)	1957.00 (1459.00,2575.50)	< 0.001
Smoking status, n (%)						< 0.001
Never smoker	7320(55.1)	2035(61.3)	1956(58.9)	1748(52.7)	1581(47.6)
Current smoker	2441(18.4)	494(14.9)	542(16.3)	650(19.6)	755(22.7)
Former smoker	3525(26.5)	793(23.9)	824(24.8)	922(27.8)	986(29.7)
Drinking status, n (%)	5253(39.5)	1413(42.5)	1345(40.5)	1283(38.6)	1212(36.5)	< 0.001
Physical activity, n (%)						< 0.001
Inactive	6813(51.3)	1555(46.8)	1671(50.3)	1707(51.4)	1880(56.6)
Moderate	3570(26.9)	930(28.0)	882(26.6)	906(27.3)	852(25.6)
Vigorous	2903(21.9)	837(25.2)	769(23.1)	707(21.3)	590(17.8)
Diabetes, n (%)	2627(19.8)	265(8.0)	384(11.6)	534(16.1)	1444(43.5)	< 0.001
Hypertension, n (%)	6324(47.6)	1436(43.2)	1527(46.0)	1543(46.5)	1818(54.7)	< 0.001
CVD, n (%)	1494(11.2)	348(10.5)	320(9.6)	368(11.1)	458(13.8)	< 0.001
Cancer, n (%)	1161(8.7)	283(8.5)	295(8.9)	301(9.1)	282(8.5)	0.803
ACM, n (%)	1688(12.7)	361(10.9)	378(11.4)	392(11.8)	557(16.8)	< 0.001
CVM, n (%)	551(4.1)	117(3.5)	126(3.8)	123(3.7)	185(5.6)	< 0.001
Follow-up time, month	112.00(61.00,167.00)	107.00(58.25,162.00)	112.00(61.00,166.00)	116.00(65.75,171.00)	113.00(57.00,168.00)	0.001
BMI, kg/m2	30.89(28.12,35.00)	29.90(27.41,33.70)	30.80(28.10,35.00)	31.09(28.30,35.33)	31.69(28.82,35.80)	< 0.001
WWI	11.22(10.72,11.75)	11.06(10.53,11.62)	11.20(10.70,11.75)	11.26(10.78,11.77)	11.34(10.85,11.84)	< 0.001
SBP, mmHg	122.67(113.33,135.33)	120.00(110.00,132.67)	122.00(112.67,133.33)	122.67(114.00,134.67)	126.00(116.00,138.67)	< 0.001
DBP, mmHg	72.00(64.00,78.67)	69.33(62.17,76.00)	71.33(64.00,78.00)	72.67(65.33,79.33)	74.00(66.67,81.33)	< 0.001
TG, mg/dL	119.00(83.00,173.00)	76.00(58.00,102.00)	105.00(80.00,141.00)	135.00(104.00,178.00)	191.00(138.00,269.00)	< 0.001
LDL-C, mg/dL	116.00(93.00,140.00)	95.00(78.00,113.00)	115.00(95.25,134.00)	127.00(106.00,148.00)	134.00(106.00,161.00)	< 0.001
TBIL, mg/dL	0.60(0.50,0.80)	0.60(0.50,0.80)	0.60(0.50,0.80)	0.70(0.50,0.80)	0.60(0.50,0.80)	< 0.001
eGFR, ml/min/1.73 m²	98.15(81.93,111.40)	98.18(81.07,113.21)	98.15(82.67,111.56)	97.54(82.83,110.86)	98.72(81.01,110.18)	0.215
SUA, mg/dL	5.60(4.70,6.60)	5.00(4.30,6.00)	5.60(4.60,6.40)	5.90(5.00,6.80)	6.10(5.10,7.00)	< 0.001
FIB-4	0.92(0.61,1.35)	0.91(0.59,1.41)	0.93(0.60,1.35)	0.92(0.61,1.33)	0.93(0.67,1.33)	0.027
SII	470.65(338.16,657.32)	458.35(325.06,643.45)	479.27(344.31,663.92)	480.42(340.77,668.12)	466.18(342.11,650.58)	0.002

The continuous variables are expressed as medians and quartiles, and the categorical variables are expressed as numbers and percentages. Abbreviations: PIR, poverty income ratio; CVD, cardiovascular diseases; ACM, all-cause mortality; CVM, CVD mortality; BMI, body mass index; WWI, weight-adjusted waist circumference index; SBP, systolic blood pressure; DBP, diastolic blood pressure; TG, triglyceride; LDL-C, low-density lipoprotein cholesterol; TBIL, total bilirubin; eGFR, estimated glomerular filtration rate; SUA, serum uric acid; FIB-4, fibrosis 4 score; SII, systemic immune inflammation index

Association of the CHG index with ACM and CVM in MASLD

A long-term follow-up with a median time of 112 (61–167) months was conducted on 13,286 patients with MASLD. A total of 1,688 (12.7%) deaths were observed, among which 551 (4.1%) were caused by CVD. We used the KM survival analysis method to conduct a comparative analysis of the survival data of participants in the quartiles of the CHG index. The results in Fig. 2 show that with increasing follow-up time, the cumulative mortality rate of all quartile groups showed an upward trend. Compared with the other quartile groups (Q1–Q3), the growth trend of ACM and CVM in the highest quartile (Q4) group was the fastest, indicating that its survival status was relatively poor. The log-rank test revealed significant differences in the survival curves of ACM and CVM in the quartile of CHG index .

Fig. 2.

KM curve analysis of CHG index with ACM (A) and CVM (B) in MASLD

Weighted Cox regression analysis revealed the association of the CHG index with the ACM and CVM of MASLD (Table 2). In Model 1, the CHG index was significantly correlated with the risk of ACM (HR = 1.74, 95% CI 1.53–1.98; p < 0.001). Taking the lowest quartile of the CHG index as the reference, the effect of Q4 on ACM is 1.56 (1.34–1.82). After adjusting for covariates, the correlation between the CHG index and ACM remained significant in Model 2 (HR = 1.46, 95% CI 1.27–1.67; p < 0.001) and Model 3 (HR = 1.50, 95% CI 1.30–1.74; p < 0.001). The effect of Q4 on ACM is 1.29 (1.09–1.52) in Model 2 and 1.30 (1.07–1.57) in Model 3. The effects of the CHG index and its quartiles on CVM are similar to their effects on ACM. In Model 1, the effects of the CHG index and Q4 are 1.82 (1.43–2.31) and 1.65 (1.28–2.13), respectively; in Model 2, they are 1.50 (1.14–1.97) and 1.33 (1.02–1.73); and in Model 3, 1.53 (1.16–2.02) and 1.32 (0.97–1.80).

Table 2.

Weighted multivariate Cox regression analysis of CHG index with mortality risk in MASLD

CHG	Model 1		Model 2		Model 3
	HR (95% CI)	p-value	HR (95% CI)	p-value	HR (95% CI)	p-value
ACM
Continuous	1.74(1.53,1.98)	< 0.001	1.46(1.27,1.67)	< 0.001	1.50(1.30,1.74)	< 0.001
Quartile
Q1	Ref		Ref		Ref
Q2	0.98(0.80,1.20)	0.843	0.89(0.73,1.08)	0.248	0.93(0.76,1.15)	0.527
Q3	0.99(0.82,1.19)	0.911	0.89(0.73,1.08)	0.244	0.88(0.70,1.10)	0.265
Q4	1.56(1.34,1.82)	< 0.001	1.29(1.09,1.52)	0.003	1.30(1.07,1.57)	0.008
p for trend		< 0.001		0.001		0.004
CVM
Continuous	1.82(1.43,2.31)	< 0.001	1.50(1.14,1.97)	0.004	1.53(1.16,2.02)	0.002
Quartile
Q1	Ref		Ref		Ref
Q2	0.92(0.69,1.24)	0.597	0.83(0.61,1.13)	0.233	0.88(0.62,1.25)	0.476
Q3	0.90(0.68,1.20)	0.469	0.79(0.57,1.09)	0.156	0.79(0.54,1.15)	0.213
Q4	1.65(1.28,2.13)	< 0.001	1.33(1.02,1.73)	0.038	1.32(0.97,1.80)	0.082
p for trend		< 0.001		0.026		0.055

Model 1 adjusted for no variables. Model 2 adjusted for gender, age, race, educational level, marital status, and economic conditions. Model 3 adjusted for gender, age, race, educational level, marital status, economic conditions, daily energy intake, smoking status, drinking status, physical activity, CVD, cancer, BMI, SBP, SUA, eGFR, LDL-C, TBIL, FIB-4 and SII

Threshold effect of the CHG index with ACM and CVM in MASLD

The RCS curves in Fig. 3 show the significant U-shaped nonlinear relationships of the CHG index with ACM (nonlinear p-value < 0.001) and CVM (nonlinear p-value < 0.001) in MASLD in Model 3. Through the piecewise linear Cox regression model in Table 3, we further explored the possible threshold effect. We found that the impact of the CHG index on the ACM risk of MASLD changed when it reached the threshold of 5.31. Below the threshold (< 5.31), the CHG index was negatively correlated with ACM (HR = 0.58, 95% CI 0.38–0.90; p = 0.015); above the threshold (≥ 5.31), the association was positively correlated (HR = 2.05, 95% CI 1.71–2.47; p < 0.001). Similarly, the impact of the CHG index on the CVM in MASLD changes when it reaches the threshold of 5.35. Below the threshold (< 5.35), the association is negatively correlated (HR = 0.39, 95% CI 0.19–0.84; p = 0.016); above the threshold (≥ 5.35), the association is positively correlated (HR = 2.60, 95% CI 1.81–3.74; p < 0.001). The results of the likelihood ratio test indicate that the influence of the CHG index on both sides of the threshold on the mortality risk is significantly different, whether it is ACM or CVM.

Fig. 3.

The nonlinear relationship of CHG index with ACM (A) and CVM (B) in MASLD through RCS analysis. The HR (red solid lines) and 95%CI (red shaded areas) was adjusted for gender, age, race, educational level, marital status, economic conditions, daily energy intake, smoking status, drinking status, physical activity, CVD, cancer, BMI, SBP, SUA, eGFR, LDL-C, TBIL, FIB-4 and SII

Table 3.

Threshold effect analysis of the CHG index with mortality risk in MASLD

Threshold effect analysis	Adjusted HR (95%CI)	p-value
ACM
Inflection point = 5.31
CHG index < 5.31	0.58 (0.38–0.90)	0.015
CHG index ≥ 5.31	2.05 (1.71–2.47)	< 0.001
Log-likelihood ratio test		< 0.001
CVM
Inflection point = 5.35
CHG index < 5.35	0.39 (0.19,0.84)	0.016
CHG index ≥ 5.35	2.60 (1.81,3.74)	< 0.001
Log-likelihood ratio test		< 0.001

Threshold analysis adjusted for gender, age, race, educational level, marital status, economic conditions, daily energy intake, smoking status, drinking status, physical activity, CVD, cancer, BMI, SBP, SUA, eGFR, LDL-C, TBIL, FIB-4 and SII

Subgroup analysis

Subgroup analysis was performed using Model 3 to examine potential effect modification by demographic and clinical characteristics on the CHG index-mortality associations in MASLD. The α = 0.0071 by Bonferroni correction is used as the threshold for significant difference in the interaction test. Notably, there is a significant interaction in age (p for interaction < 0.001) and BMI (p for interaction = 0.006) groups for ACM, and only age (p for interaction < 0.001) has a significant interaction for CVM (Fig. 4). The results show that among people under 60 years old, an increase in the CHG index is significantly associated with an increased risk of ACM (HR = 2.23, 95% CI 1.68–2.96; p < 0.001) and CVM (HR = 3.11, 95% CI 1.72–5.63; p < 0.001). Compared with the nonlean population, the CHG index in lean MASLD has a stronger association with ACM (HR = 5.22, 95% CI 1.83–14.94; p = 0.002). Although there was no significant interaction of BMI for CVM (p for interaction = 0.487), the association of CHG in lean MASLD (HR = 3.64, 95% CI 0.81–16.29; p = 0.091) was significantly stronger than that in nonlean MASLD (HR = 1.47, 95% CI 1.09–1.97; p = 0.01).

Fig. 4.

Each subgroup analysis was adjusted for gender, age, race, educational level, marital status, economic conditions, daily energy intake, smoking status, drinking status, physical activity, CVD, cancer, BMI, SBP, SUA, eGFR, LDL-C, TBIL, FIB-4 and SII, except for stratified variables. The threshold of p for interaction is the Bonferroni corrected α = 0.0071

Mediating analysis

The results of the mediating analysis showed that WWI, NE and eGDR mediated the association of the CHG index with the ACM and CVM of MASLD. The mediating proportions of WWI to ACM and CVM were 12.56% and 9.88%, respectively (Fig. 5A, B). The mediating proportions of NE were 13.91% and 22.75%, respectively (Fig. 5C, D). The mediating proportions of eGDR were 24.23% and 47.44%, respectively (Fig. 5E, F).

Fig. 5.

The mediating effects of WWI (A and B), NE (C and D) and eGDR (E and F) on the relationship of CHG index with ACM and CVM in MASLD. The mediating analysis adjusted for gender, age, race, educational level, marital status, economic conditions, daily energy intake, smoking status, drinking status, physical activity, CVD, cancer, BMI, SBP, SUA, eGFR, LDL-C, TBIL, FIB-4 and SII. Abbreviations: TE: total effect; DE: direct effect; IE: indirect effect

Sensitivity analysis

To comprehensively examine the robustness of the association between the CHG index and the risk of death from MASLD, this study conducted a sensitivity analysis from five dimensions. First, we further adjusted hypoglycemic treatment and lipid-lowering drug use to the weighted multivariate Cox model. The results show that the association of the CHG index with ACM (HR = 1.41, 95% CI 1.21–1.64; p < 0.001) and CVM (HR = 1.38, 95% CI 1.02–1.86; p = 0.037) remains significant (Table S5). Second, we excluded participants who died within the first two years after the start of follow-up (177 cases, accounting for approximately 10.5% of the total death events). The results show that the effect of the CHG index on ACM (HR = 1.48, 95% CI 1.26–1.73; p < 0.001) and CVM (HR = 1.50, 95% CI 1.09–2.07; p = 0.014) remains basically unchanged (Table S6). Third, we adopted FLI ≥ 60 and ZJU index > 38 as the new cut-off points for steatosis. The results showed that the effect of the CHG index was not affected by the selection of diagnostic criteria for fatty liver (Tables S7, S8). Fourth, we adopted the standardization method to identify and eliminate outliers of the CHG index, and the effect values remained basically consistent (Table S9). Finally, external validation was conducted using independent samples from wave 2016 of the HRS cohort. Specifically, wave 2016 had a total of 20,912 participants. After excluding those under 50 years old, without MASLD, and missing the CHG index, 2,914 participants with MASLD were ultimately included. The baseline characteristics of the participants in the HRS cohort are shown in Table S10. After a follow-up period of up to 4 years, 204 (7%) deaths occurred. The Cox regression results (Table S11) showed that although there was no significant difference in single-factor Model 1 (HR = 1.25, 95% CI 0.95–1.66; p = 0.116), there was a significant positive correlation between the CHG index and ACM in Model 2 (HR = 1.55, 95% CI 1.16–2.08; p = 0.003) and Model 3 (HR = 1.49, 95% CI 1.11–2.01; p = 0.009), and the effect was consistent with that in the NHANES cohort.

Discussions

This is the first prospective study to explore the impact of the CHG index on ACM and CVM in MASLD based on a large-scale population. The KM survival curve indicates that the MASLD population with the highest quartile of the CHG index is accompanied by a higher risk of ACM and CVM. The RCS curve shows the significant U-shaped characteristics of the CHG index in relation to the ACM and CVM risk. Further threshold effects determined that the inflection point of the CHG index for ACM was 5.31 and for CVM was 5.35. This might imply that both excessively high and low CHG index levels could increase the mortality risk for patients with MASLD. The results of the subgroup analysis showed that the association between the CHG index and mortality risk was more significant in the subgroups under 60 years old and lean. Mediating analysis indicates that WWI, NE and eGDR may partially mediate the effects of the CHG index on ACM and CVM. This association was further verified in the MASLD population through the HRS cohort. Other sensitivity analyses further demonstrated the robustness of our results.

At present, a large number of studies have explored the role of the TyG index in patients with MASLD [35, 36]. The TyG index provides a simple and effective indicator for assessing IR by combining FBG and TG levels [37]. In contrast, as an emerging metabolic indicator, the relationship between the CHG index and MASLD has not been fully studied, highlighting the novelty and significance of our research results. The CHG index was first proposed by Mansoori and his team. By comparing the diagnostic effects of the CHG index and the TyG index on T2DM, they found that the CHG index (AUC = 0.864) has a stronger diagnostic ability than the TyG index (AUC = 0.825) [13]. Similarly, a cohort study from China further confirmed the advantage of the CHG index in assessing CVD risk, demonstrating a predictive ability comparable to or even higher than that of the TyG index [17]. These research results consistently point to the potential of the CHG index as a more comprehensive indicator of IR. Compared with the TyG index, a significant advantage of the CHG index may be attributed to its inclusion of a broader range of lipid parameters, especially TC and HDL-C. The TyG index mainly focuses on FBG and TG levels while neglecting other important lipid metabolism indicators. In contrast, the CHG index integrates TC, FBG and HDL-C. By incorporating these additional lipid parameters, the CHG index can more accurately reflect an individual’s metabolic health status, thus having potential advantages in assessing IR and metabolic disorders.

This study revealed a significant nonlinear, U-shaped association between the CHG index and ACM and CVM in MASLD. This result indicates that both excessively high and low CHG indices may increase the mortality risk for MASLD patients, and the best survival rate occurs within the middle range. The essence of the U-shaped association is the dual risk of metabolic imbalance represented by different CHG index levels. A high CHG index indicates a combined state of elevated TC and FBG and decreased HDL-C, reflecting the superimposed load of atherosclerosis and glucose metabolism stress. Increased TC deposition can accelerate vascular endothelial damage, raise the risk of atherosclerosis and aggravate lipid toxicity in the liver [38]. HDL-C also has antioxidant and anti-inflammatory effects and improves vascular endothelial function [39]. Low HDL-C levels may weaken its protective effect on the cardiovascular system [40]. High FBG levels may further exacerbate cardiovascular system damage by triggering oxidative stress and inflammatory responses [41]. High blood sugar can also lead to IR, further aggravating metabolic disorders and affecting the metabolic function of the liver [42]. The increased risk observed at a very low CHG index should be interpreted with caution. Extremely low CHG may reflect hypocholesterolemia, low normal fasting blood glucose or significantly elevated HDL-C, each of which may be caused by nonpathological characteristics (such as genetic variations, strict exercise), chronic diseases or malnutrition. Low cholesterol may weaken the integrity of cell membranes and reduce the synthesis of steroid hormones and vitamin D, while persistently low glucose levels may induce sympathetic nerve excitation and cardiovascular events [43, 44]. Meanwhile, extremely high HDL-C levels are partly due to genetic variations, and their antioxidant and anti-inflammatory functions may have been impaired [45]. Our observations cannot clarify these possibilities or establish causal relationships. Therefore, these mechanisms of the U-shaped curve remain hypothetical. The inflection points we observed in the threshold effect were 5.31 for ACM and 5.35 for CVM. In the management of MASLD patients, the CHG index should be consciously shifted to the observed “optimal” range to avoid being too high or too low. In principle, lifestyle plans that maintain or increase HDL-C while reducing fasting blood glucose and atherosclerotic cholesterol, such as combining physical activity, dietary quality and moderate weight loss, are expected to move CHG values in the desired direction [46]. Similarly, the hypoglycemic and lipid-lowering drug treatments already recommended in the MASLD guidelines may also incidentally achieve this metabolic consistency [47, 48]. However, it remains unknown whether these exact thresholds are still valid in non-US MASLD patients. Prospective validations need to be conducted in Asian, European and other cohorts before these critical values are considered universally applicable.

The association between the CHG index and mortality risk in MASLD is more significant in people under 60 years old and those who are lean. Similarly, a study showed that several metabolic parameters were positively correlated with ACM or CVM in young people with metabolic syndrome, but no correlation was found in the elderly group [49]. This indicates that our discovery is not accidental: young people may be more vulnerable to metabolic disorders. This conclusion may be related to the metabolic characteristics and disease progression process of young patients. Among people under the age of 60, metabolic activities are more active, and changes in the CHG index may have a more significant impact on the metabolic system. In addition, the accumulation of chronic diseases among young people is relatively low, so even minor changes in the CHG index may more easily affect their health conditions. In contrast, people over 60 years old often have multiple chronic diseases, and the interaction among these diseases may mask the impact of the CHG index on mortality. Moreover, we observed that in lean MASLD, the increase in ACM (HR 5.22 vs. 1.50) and CVM (HR 3.64 vs. 1.47) per unit increase in the CHG index was numerically stronger than in nonlean MASLD. This finding is consistent with recent reports that describe lean MASLD as a unique metabolic phenotype characterized by more adverse consequences [50]. Recent multiple prospective studies from the West and Asia have shown that compared with nonlean individuals, lean MASLD individuals have a higher risk of liver-related events, ACM and CVM [51]. The distinct prognoses between lean and nonlean MASLD populations may be explained by reduced muscle mass, intestinal bioimbalance and genetic variations [52–54].

By evaluating the mediating effects of WWI, NE and eGDR, we quantified the proportions of the association between the CHG index and mortality risk explained by obesity, inflammation and IR, respectively. Metabolic disorders can promote visceral fat accumulation and trigger lipid toxicity. The accumulation of free fatty acids in skeletal muscle cells can inhibit the IRS-1/PI3K signaling pathway, accelerate muscle protein breakdown, and thereby lead to an increase in WWI [55]. Elevated WWI can cause a series of reactions to increase the mortality risk of MASLD [56]. Abnormal glycolipid metabolism leads to excessive accumulation of lipids in the liver, triggering inflammation and activating NE infiltration [57]. NE further exacerbates liver damage and poor prognosis by releasing proinflammatory mediators [24]. IR is a key link connecting disorders of glycolipid metabolism and the progression of MASLD [58]. IR not only has a bidirectional promoting relationship with MASLD but can also activate chronic inflammation [42, 59]. In addition, a lower eGDR may increase the risk of cardiovascular events, further affecting the survival prognosis of MASLD [60]. The cross-sectional nature of biomarkers, in which the exposure factors and mediating factors were measured at the same time point, and the observation design excluded the establishment of a time or causal sequence. Therefore, the above explanation should be regarded as a biologically reasonable hypothesis rather than an explicit mechanism inference.

Our study, with its prospective design and large sample size, provides reliable data support for the association between the CHG index and mortality risk in MASLD. The CHG index can more comprehensively reflect the metabolic status than a single indicator. The external validation of the HRS cohort further verified the robustness of the research results. Despite these advantages, the research still has some undeniable limitations. First, although this study adopted a prospective design and was able to observe the association between the CHG index and mortality risk, it could still not fully determine the causal relationship. Second, all components of the CHG index were measured only once at baseline. Therefore, we are unable to capture the long-term trajectory and dynamic changes in the CHG index. Repeating the CHG index over time may capture metabolic trajectories better, and future surveys will repeat these measurements in consecutive interviews. Third, the HRS cohort successfully replicated the main results of the NHANES cohort, but its baseline recruitment began at the age of 50 years. Therefore, the external validation of the relationship between the CHG index and mortality risk is limited to middle-aged and elderly people, and there is uncertainty in younger MASLD individuals under 50 years. A large cohort recruiting adults across the entire age range is needed to confirm the universality of our results in the younger population. Fourth, despite the comprehensive adjustment available in the NHANES and HRS cohorts, unmeasured or residual confounding cannot be excluded. Variables such as dietary quality changes, medication adherence, or genetic polymorphisms affecting lipid and glucose metabolism were not available, and their absence may inflate or attenuate the observed results. Fifth, hepatic steatosis was inferred from the HSI rather than imaging or histology in our study. Although HSI has acceptable sensitivity and specificity in large epidemiological settings, potential misclassification remains possible and could bias the associations toward the null. Finally, all the study participants were from the United States and may not reflect the actual situation of MASLD patients in other countries or regions.

Conclusion

Our study for the first time confirmed a robust association between the CHG index and ACM and CVM in patients with MASLD based on the two prospective cohorts of NHANES and HRS. Clinically, the CHG index may be used as a simple and economical risk stratification tool, thereby achieving the goal of improving the long-term prognosis of MASLD.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

HZ, LL and SY: Designed the research plan, collected and analyzed the data, explained the research results, and wrote the initial and final drafts. YF and YP: Participated in research design, data analysis, and manuscript revision. QF: Participated in data collection and analysis. XZ and FD: Supervised the research project, guided the research direction and content, and reviewed the final manuscript. All authors have approved to the final version of the paper.

Funding

This study was supported by grants from the National Natural Science Foundation of China (81760524, 82260599) and the Double Thousand Talents Plan of Jiangxi (jxsq20232010310).

Data availability

The datasets supporting the conclusions of this article are available in https://wwwn.cdc.gov/nchs/nhanes/Default.aspx and https://hrs.isr.umich.edu/.

Declarations

Ethics approval and consent to participate

The research design and implementation of NHANES follow the ethical principles outlined in the Helsinki Declaration. This study utilized data from the NHANES project that were publicly available and approved by the National Center for Health Statistics (NCHS) Ethics Review Board (ERB). NHANES requires informed consent from participants and ensures their privacy and information security. The HRS is approved by the institutional review board (IRB) of the University of Michigan, and all participants provided written informed consent. All of the data used in this study are deidentified, and most of the data are publicly available.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.