Links between blood sugar and fat measures and overall and specific causes of death in people with metabolic syndrome

The datasets were derived from ten cycles of interviews (1999 to 2018 years) in the National Health and Nutrition Examination Survey (NHANES) database. The NHANES database systematically gathered nationally representative health-related data on the repetitive non-institutionalized U.S. population, utilizing a stratified, multistage probability sampling design [27]. The specific descriptions of the NHANES database can be found on the website (https://www.cdc.gov/nchs/nhanes/↗).

To evaluate the association between the TyG-related indices with the all-cause and cause-specific mortality of the MetS population, only participants with the diagnosis of MetS were included. Therefore, during the ten cycles of interviews between 1999 and 2018 years, there were 101,316 participants were reviewed. After excluding the participants aged below 18 years old or non-MetS, there were 21,333 participants were initially included. Additionally, participants without records of triglyceride (TG) and fasting blood glucose (FBG), body measurements, or loss to follow-up were further excluded. Ultimately, there were 10,734 participants were included in this study (Fig. 1).

The definition of MetS was based on the previous guideline [National Cholesterol Education Program (NCEP): National Cholesterol Education Program Adult Treatment Panel III] [28]. Participants who met three or more of the following criteria: [1] the FBG > 100 mg/dL or drug treatment for diabetes mellitus; [2] high-density lipoprotein cholesterol (HDL-C) < 50 mg/dL in females, < 40 mg/dL in males or drug treatment for reduced HDL-C; [3] Plasma TG > 150 mg/dL or drug treatment for raised TG; [4] Waist circumference > 88 cm in women or > 102 cm in men; [5] Blood pressure > 130/85 mmHg or drug treatment for raised blood pressure.

There were three TyG-related indices of interest measured in the present study. The TyG index was measured by the peripheral blood test (TG, and FBG) of the participants at baseline according to the previous literature [29, 30]. Two variables, including WC and WHtR, were selected as the obesity indicators in the MetS population [31]. Specifically, the indices of TyG, TyG-WC, and TyG-WHtR were calculated as the following equations [32–34]:

1 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm{TyG}}\,{\rm{ = }}\,{\rm{ln}}\,\left[ {{\rm{TG}}\,\left( {{\rm{mg/dL}}} \right)\,{\rm{ \times }}\,{\rm{FBG}}\,\left( {{\rm{mg/dL}}} \right){\rm{/2}}} \right]$$\end{document}

2 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\eqalign{{\rm{TyG - WC}}\,{\rm{ = }}\, & {\rm{ln}}\,\left[ {{\rm{TG}}\,\left( {{\rm{mg/dL}}} \right)\,{\rm{ \times }}\,{\rm{FBG}}\,\left( {{\rm{mg/dL}}} \right){\rm{/2}}} \right]\, \cr & {\rm{ \times }}\,{\rm{waist}}\,{\rm{circumference}}\,\left( {{\rm{cm}}} \right) \cr}$$\end{document}

3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\eqalign{{\rm{TyG - WHtR}}\,{\rm{ = }}\, & {\rm{ln}}\,\left[ {{\rm{TG}}\,\left( {{\rm{mg/dL}}} \right)\,{\rm{ \times }}\,{\rm{FBG}}\,\left( {{\rm{mg/dL}}} \right){\rm{/2}}} \right]{\rm{ }} \cr & {\rm{ \times waist}}\,{\rm{circumference}}\,\left( {{\rm{cm}}} \right){\rm{/height}}\,\left( {{\rm{cm}}} \right) \cr}$$\end{document}

The MetS population was classified into four groups by the quartiles of the TyG index, TyG-WC index, and TyG-WHtR index, respectively, and the group at the 1st quartile was set as the reference.

We collected the demographic information of participants from the NHANES database. Detailly, the socioeconomic characteristics included gender (male or female), age, race (Mexican, Hispanic, non-Hispanic White, non-Hispanic Black, or other races), marital status (not married, married or living with a partner), educational level (≤ high school, college, or > college), and family income-poverty ratio (< 1.3, 1.3–3.5, or > 3.5) were collected. Besides, the living habits and history of comorbidities include smoking status (never, ever, or current), alcohol use (never, ever, or current), history of cancer, chronic kidney disease (CKD), and cardiovascular disease (CVD) (yes or no) were further collected. In addition, physical and laboratory examinations such as body mass index (BMI), energy intake (average kilocalorie derived from two 24-hour dietary recall interviews), serum levels of estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), uric acid (UA), glutamic-pyruvic transaminase (ALT), aspartate transaminase (AST), albumin (ALB), and total bilirubin (TBil) were also selected as the potential confounders.

The main outcome of the current study was the all-cause mortality of the MetS population. The second outcome was the cause-specific mortality (cardiovascular and diabetes mortality) of the MetS population. The mortality data for the follow-up population were obtained from the NHANES public-use linked mortality file as of December 31, 2019, which was correlated with NCHS with the National Death Index (NDI) through a probability matching algorithm. The ICD-10 (International Statistical Classification of Diseases, 10th revision) was used to check the causes of mortality. The period of follow-up was calculated from the date when the interview was initially taken to either the date of the patient’s death or December 31, 2019 [35].

Kolmogorov-Smirnov test was used to check the normality assumption distribution of each variable. The normally distributed variables were presented as mean (standard deviation). The non-normally distributed variables were presented as median (interquartile range). The categorical variables were presented as numbers (percentage, %). Continuous variables with normal distribution were evaluated using Student’s t-test, while continuous variables with non-normal distribution were tested using the Kruskal-Wallis test. Categorical variables were compared by using the Chi-Squared test. Univariate and multivariable Cox proportional hazard models were used to estimate the association of the TyG index with all-cause, cardiovascular as well as diabetic mortality of the MetS population.

The selections of controlled covariates in the present study were based on previous literature evaluating the survival of MetS [3, 8, 36]. As the WC, WHtR, and BMI were highly correlated, BMI was not adjusted during the Cox regression analysis for evaluating the association between TyG-WC and TyG-WHtR indices with the survival of the MetS population. Detailly, model 1 served as the unadjusted analysis. Besides, brief adjustments for age, gender, and race were made in Model 2. In the fully adjusted model, we accounted for age, gender, race, marital status, educational level, family income-poverty ratio, smoking status, alcohol use, cancer, CKD, CVD, BMI (only for TyG index), energy intake, serum levels of TC, HDL-C, LDL-C, BUN, UA, eGFR, ALT, AST, ALB, and TBil.

Kaplan-Meier (KM) curves were used to show the censored data and different survival patterns among the MetS population with different quartiles of the TyG index. Missing data (< 5%) was imputed by using the multiple imputation method. To evaluate the dose-effect correlations between TyG-related indices with all-cause and cause-specific mortality of the MetS population, restricted cubic splines (RCS) were employed. The selection of knots for the RCS curves was guided by the minimization of Akaike’s Information Criterion (AIC). The RCS curves were fitted using different numbers and positions of knots, and the corresponding AIC values were calculated for each model. The final knot configuration was chosen based on the model with the lowest AIC value, indicating the optimal balance between model fit and complexity. To check the robust connections between TyG-related indices and mortality of the MetS population, four sets of sensitive analyses were conducted to validate the main findings. First, we excluded participants who reported a history of diabetes mellitus at baseline. Second, we excluded participants who died within two years after the interview. Third, participants with a history of comorbidities were excluded. Last, participants with low or normal BMI were excluded.

All statistical analyses of the present study were conducted by using the R software (version 4.2.3, https://www.r-project.org/↗). To determine a more rigorous interpretation of the associations between TyG, TyG-WC, and TyG-WHtR with the survival of the MetS population, the Bonferroni-corrected P-value was used in this study [37]. Thus, the significance level P = 0.05 was divided by three, which provided a significance level corrected for multiple testing: P = 0.017.

During the study years, there were 10,734 participants with a diagnosis of MetS included in the present study (Table 1). The median age of the MetS population was 59 years and females counted for a slightly higher proportion than males (5,570 cases vs. 5,164 cases). However, males with Mets showed a higher mortality risk than females (1,182 cases vs. 1,033 cases). Nearly half of the MetS population were non-Hispanic white (4,940 cases) and less than 40% of them had a degree of college or above. More than 30% of the MetS population was overweight (3,573 cases) and nearly 57% were obese (6,051 cases). At the time of the interview, over half of the study population were still drinking (5,982 cases) and nearly 20% of them were still smoking (2,015 cases). The prevalence of comorbidities was 12.19% (1,309 cases) for cancer, 20.48% (2,198 cases) for CKD, and 14.48% (1,554 cases) for CVD, respectively. The median level of the TyG index was 7.41 and the TyG-WC as well as TyG-WHtR indices were 787.28 and 4.72, respectively. With a median follow-up of 100 months, 2,215 deaths were observed, with 360 for diabetes mortality and 615 for cardiovascular mortality. Non-survivors were observed to have characteristics of males, Non-Hispanic White, unmarried status, older age, less daily energy intake, lower educational level, lower BMI, higher levels of FBG, TyG index, and having a history of comorbidities. The detailed information of each group of the MetS population with varied quartile levels of TyG-related indices was summarized in Table S1-S3.

The KM curves showed that the MetS population at the low quartile levels of the TyG index group showed significantly higher overall as well as diabetes-specific survival probabilities when compared with those in the high quartile groups (p = 0.02 and p < 0.0001, respectively) (Fig. 2A and C). Additionally, significant differences were only observed in the diabetes-specific survival of the MetS population with varied quartiles of the TyG-WC index (p < 0.0001) (Fig. 2D-F). The TyG-WHtR index was adversely associated with overall and cause-specific survival among the MetS population (p = 0.018 for overall survival, p = 0.049 for cardiovascular-specific survival, and p < 0.0001 for diabetes-specific survival, respectively) (Fig. 2G and I).

The Cox regression analyses showed the association between TyG, TyG-WC, and TyG-WHtR indices with the all-cause mortality of MetS population [TyG index at 4th quartile: Unadjusted hazard ratio (HR) = 1.17, 95% confidence interval (CI): 1.04–1.31, p = 0.011; Less adjusted HR = 1.45, 95%CI: 1.28–1.63; p < 0.001; Full adjusted HR = 1.36, 95%CI: 1.18–1.56, p < 0.001 (Table S4); TyG-WC index at 4th quartile: Unadjusted HR = 1.01, 95% CI: 0.90–1.14, p = 0.868; Less adjusted HR = 1.40, 95%CI: 1.24–1.58; p < 0.001; Full adjusted HR = 1.17, 95%CI: 1.03–1.34, p = 0.020 (Table S5); TyG-WHtR index: Unadjusted HR = 1.16, 95% CI: 1.03–1.30, p = 0.017; Less adjusted HR = 1.54, 95%CI: 1.37–1.74; p < 0.001; Full adjusted HR = 1.29, 95%CI: 1.13–1.47, p < 0.001 (Table S6)] (Fig. 3A).

In our analyses, both of the TyG-WC and TyG-WHtR indices were significantly associated with the cardiovascular mortality of the MetS population [TyG-WC index: Unadjusted HR = 1.21, 95% CI: 0.97–1.51, p = 0.090; Less adjusted HR = 1.73, 95%CI: 1.37–2.17; p < 0.001; Full adjusted HR = 1.45, 95%CI: 1.13–1.85, p = 0.004 (Table S5); TyG- WHtR index: Unadjusted HR = 1.29, 95% CI: 1.03–1.61, p = 0.026; Less adjusted HR = 1.83, 95%CI: 1.46–2.30 p < 0.001; Full adjusted HR = 1.50, 95%CI: 1.17–1.92, p = 0.002 (Table S6)] (Fig. 3B). However, a borderline positive correlation was observed between the TyG index and cardiovascular mortality with quartile classification [(Full adjusted: HR = 1.29, 95%CI: 0.99–1.69, p = 0.060 (Table S4)] (Fig. 3B).

Meanwhile, the three TyG-related indices were all significantly association with diabetes mortality [TyG index: Full adjusted HR = 4.06, 95%CI: 2.81–5.87, p < 0.001(Table S4); TyG-WC index: Full adjusted HR = 2.55, 95%CI: 1.82–3.58, p < 0.001 (Table S5); TyG-WHtR index: Full adjusted HR = 2.53, 95%CI: 1.81–3.54, p < 0.001 (Table S6)] (Fig. 3C).

As shown in Fig. 4, the multivariable-adjusted RCS analyses were conducted to visualize the relation of TyG-related indices with all-cause and cause-specific mortality in the MetS population. Specifically, a Non-linear trend was observed in the relationships between TyG as well as TyG-WC indices with the all-cause mortality (P for non-linearity = 0.004 and 0.001, respectively) (Fig. 4A and D). However, the TyG-WHtR index showed a linear association with the all-cause mortality of the MetS population (P for overall = 0.006) (Fig. 4G). The RCS curves showed broad line significant non-linear associations between the TyG index and cause-specific mortality (P for non-linearity = 0.050 and 0.057, respectively) (Fig. 4B and C). Besides, TyG-WC and TyG-WHtR indices presented consistent linear associations with cardiovascular and diabetes mortality (all P for overall < 0.05) (Fig. 4E, F, H and I).

The stratified analyses revealed TyG-related indices had stronger associations with all-cause mortality of the MetS population with the clinical characteristics of non-Hispanic Black race, normal or overweight BMI, and married or living with a partner (Figure S1-S3). Additionally, subgroup analyses revealed some significant interaction effects (p < 0.05), indicating that the relationship between TyG-related indices and all-cause mortality of the MetS population varied significantly across different subgroups. Specifically, the effect of TyG-related indices on all-cause mortality of the MetS population differed among subgroups defined by age, gender, race, marital status, history of cancer, and serum levels of ALB, ALT, AST, BUN, TC, and UA. (Figure S1-S3). To check the robustness of the main findings, several sensitive analyses were performed. First, after excluding the participants with a diagnosis of diabetes mellitus at baseline, the TyG-related indices remained significantly associated with diabetes mortality during the follow-up (Table S7-S9). After excluding the participants who died within 2 years after the survey, we determined consistent associations between TyG-related indices with the all-cause as well as cause-specific mortality risk among the modified MetS population (Table S10-S12). Similarly, the results were still robust in the study population without a history of comorbidities (Table S13-S15) or with a high BMI index (Table S16-S18).

Some noteworthy strengths of the present study should be highlighted. Initially, this is the first study focusing on comprehensively evaluation the associations between TyG-related indices with the all-cause as well as cause-specific mortality of the MetS population. Our results highlight the promising application prospect of TyG-related indices in the clinical management of MetS, the group suffering from systemic metabolic disorders. Second, the prospective, population-based study design helps us to determine the robust evidence between the TyG index and mortality of the MetS population. Last, the sensitive analysis shows consistent associations as we determine in the main findings.

Nevertheless, there are some limitations existing in the present study. First, while the present study was conducted based on a large-scale, population-based cohort, the data of blood indicators and anthropometric measurements were only collected at the baseline. Whether the trajectories of TyG-related indices would offer valuable insights into the clinical management of individuals with MetS is worth exploring. Second, the diagnosis of MetS was solely based on the NCEP III criteria, which might allow us to underestimate its incidence among different populations. Additionally, while we controlled a series of covariates including socioeconomic status, comorbidities, blood tests, and personal habits, some residual confounders (such as inherent genetic heterogeneity and changes of lifestyle during the lifetime) were observed to play a role in the MetS were unavailable in the present database. Besides, the smoking status was based on a categorical variable instead of pack-per year might also result in residual bias. Of note, the findings we determined need to be interpreted cautiously due to the identification of negative confounders. The observational study design and imbalanced baseline clinical characteristics might contribute to these confounding factors during the analyses. Future well-designed, longitudinal, propensity score-matched works are warranted to validate our findings and mitigate the impact of negative confounders. Last, although we observed the association between the TyG-related indices and mortality risk among the U.S. MetS population, its practical clinical application in different phenotypes of MetS population and varied races from other countries needs to be further validated.

Below is the link to the electronic supplementary material.

Supplementary Material 1