Abstract
Background
The prognostic utility of the triglyceride-glucose (TyG) index and its anthropometry-enhanced variants (TyG-WC, TyG-WHtR, TyG-BMI) for mortality risk in advanced cardiovascular–liver–kidney–metabolic syndrome (CLKM), a multisystem condition involving heart, liver, kidney, and metabolic health, remains unknown.
Methods
This nationwide prospective cohort study included 1384 adults with advanced CLKM syndrome from NHANES 1999–2018. The associations between the TyG index, its modified variants, and all-cause mortality were assessed using weighted multivariable Cox proportional hazards models. Restricted cubic splines (RCS) were used to identify nonlinear associations. To compare predictive performance, C-index, net reclassification index (NRI) and integrated discrimination improvement (IDI) were calculated.
Results
Over a mean 56-month follow-up, 360 deaths were recorded. RCS revealed U-shaped associations (i.e., lower risk at intermediate levels and higher risk at both low and high levels) between TyG indices and mortality (P for nonlinear< 0.05), with inflection points at TyG = 9.56, TyG-WC = 1,039.11, TyG-WHtR = 5.17, and TyG-BMI = 215.85. At values below the inflection points, higher indices were associated with reduced mortality risk. Comparison based on the C-index, NRI, and IDI showed that the modified TyG indices did not outperform the original TyG in mortality prediction. Subgroup analyses confirmed consistency (P for interaction >0.05).
Conclusion
In advanced CLKM syndrome, TyG indices exhibit U-shaped mortality association, revealing dual metabolic roles. The original TyG index performs comparably to anthropometry-enhanced variants, supporting its use as a parsimonious risk-stratification tool. Identified inflection points offer actionable thresholds for personalized management.
Supplementary Information
The online version contains supplementary material available at 10.1186/s41043-025-01187-5.
Keywords: Cardiovascular–liver–kidney–metabolic syndrome, All-cause mortality, Triglyceride glucose index, Modified triglyceride glucose indices
Introduction
Cardiovascular-kidney-metabolic (CKM) syndrome stems from pathophysiological interactions among subclinical or overt cardiovascular disease (CVD), chronic kidney disease (CKD), and metabolic dysfunction. This multisystem disorder shares common pathophysiological pathways—including oxidative stress, chronic inflammation, and immune dysregulation [1, 2]. However, the CKM framework notably underrepresents the liver’s role. Metabolic dysfunction-associated steatotic liver disease (MASLD)—the hepatic manifestation of metabolic dysregulation—is mechanistically tied to insulin resistance and obesity. It both initiates and results from metabolic disturbances, establishing bidirectional pathways that exacerbate atherosclerosis, heart failure, and CKD progression [3]. Updated 2024 EASL-EASD-EASO guidelines incorporate cardiometabolic parameters into MASLD diagnosis, replacing the former Non-alcoholic fatty liver disease (NAFLD) terminology [4]. Mounting evidence positions MASLD as both a frequent comorbidity and critical accelerator of CKM progression, warranting syndromic redefinition. Consequently, Zhou et al. [5] proposed cardiovascular–liver–kidney–metabolic (CLKM) syndrome, recognizing the liver’s integral role in interorgan crosstalk and refining clinical management approaches.
Insulin resistance (IR) serves as a core pathological driver across both CKM and NAFLD spectra. The Triglyceride-Glucose (TyG) index has gained recognition as a practical, economical surrogate marker for IR, demonstrating strong concordance with gold-standard hyperinsulinemic-euglycemic clamp measurements [6, 7]. Its ability to predict incident type 2 diabetes, cardiovascular events, and mortality across general and disease-specific cohorts is well-established [8, 9]. Acknowledging the interplay between IR and adiposity, modified TyG indices integrating anthropometric parameters have been developed to potentially improve risk stratification by concurrently reflecting metabolic dysfunction and body composition [10, 11]. Initial studies suggest these composite indices may surpass the original TyG index in predicting outcomes such as arterial stiffness and new-onset diabetes [10, 12].
While studies such as those by Hong et al. [13] and Li et al. [14] have demonstrated the value of TyG indices in predicting cardiovascular disease in early CKM stages (0–3), their prognostic utility for all-cause mortality in patients with advanced CLKM syndrome remains completely unexamined. It is unknown whether a similar linear pattern holds, or if a more complex, nonlinear association emerges in this state of advanced multisystem dysfunction. Consequently, the comparative prognostic value of the original versus modified TyG indices, as well as the potential existence and location of risk thresholds, are entirely undefined in this high-risk population. Therefore, this study utilizes data from a large, prospective, nationally representative cohort to investigate the association between TyG and modified TyG indices with all-cause mortality in advanced CLKM syndrome patients, aiming to facilitate early risk detection and intervention.
Methods
Study population and design
Data were obtained from the National Health and Nutrition Examination Survey (NHANES), an ongoing, complex, multistage probability survey designed to assess the health and nutritional status of the non-institutionalized civilian population in the United States. The survey employs a stratified, multistage probability cluster sampling design to produce nationally representative estimates [15, 16]. We analyzed pooled cycles from 1999 to 2018. Publicly available survey documentation and mortality linkage files are accessible at www.cdc.gov/nchs/nhanes and www.cdc.gov/nchs/ndi/, respectively.
The inclusion criteria for this study were: (1) age ≥ 20 years (as the NHANES protocol designates participants aged 20 and above as adults, and several key variables necessary for defining CLKM syndrome components were specifically targeted and available for this adult population), (2) availability of complete data for the classification of advanced CLKM syndrome (i.e., data for CKM staging and USFLI calculation), and (3) availability of mortality follow-up data. Conversely, exclusion criteria comprised: (1) age < 20 years, (2) incomplete data for CLKM syndrome classification, and (3) missing mortality follow-up. After exclusions, 1384 participants were retained (Fig. 1). Survival status was tracked through December 31, 2019. The National Center for Health Statistics Research Ethics Review Board approved all protocols (#2021-05), with participants providing written informed consent. Study procedures adhered to the Declaration of Helsinki.
Fig. 1.
Study flowchart. Flowchart showing the process of participant selection: of 101,316 participants from the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2018, 1384 (weighted sample sizes = 4964918) remained in the final analysis CKM, cardiovascular-kidney-metabolic syndrome; NAFLD, nonalcoholic fatty liver disease; CLKM, cardiovascular–liver–kidney–metabolic syndrome
Defining advanced CLKM syndrome
The definition of advanced CLKM syndrome is illustrated in Fig. 2. Briefly, advanced CLKM syndrome is defined by the presence of stage 3 or 4 CKM syndrome with concurrent NAFLD. CKM syndrome requires concurrent subclinical or overt CVD, CKD, and metabolic dysfunction. Clinical CVD was defined as a self-reported physician diagnosis of myocardial infarction, heart failure, stroke, or coronary artery disease. This was ascertained based on an affirmative response to the following question from the medical conditions questionnaire: “Has a doctor or other health professional ever told you that you had [coronary heart disease (MCQ160C)/a heart attack (MCQ160B)/a stroke (MCQ160F)/congestive heart failure (MCQ160D)]?“. Subclinical CVD was identified as 10-year CVD risk ≥ 20% or high-risk CKD, defined as an estimated glomerular filtration rate (eGFR) of < 60 mL/min/1.73 m² or a urinary albumin-to-creatinine ratio (UACR) ≥ 300 mg/g [17]. eGFR was calculated using the CKD-EPI 2021 creatinine equation [18]. The 10-year CVD risk was computed using a validated algorithm incorporating age, sex, smoking status, blood pressure, lipid levels, diabetes status, kidney function, and relevant medication use. Metabolic disorders encompassed overweight or obesity, abdominal obesity, dysglycemia, hypertension, dyslipidemia, and metabolic syndrome [19, 20]. Metabolic syndrome is defined by the presence of ≥ 3 of the following: waist circumference ≥ 102 cm for men or ≥ 88 cm for women; HDL-C < 40 mg/dL for men, < 50 mg/dL for women; TG ≥ 150 mg/dL; elevated blood pressure (SBP ≥ 130 mmHg, DBP ≥ 80 mmHg, a medical diagnosis, or taking antihypertensive medication); FBG ≥ 100 mg/dL. For each participant, the CKM syndrome stage was determined [21]: stage 0 (no CKM risk factors), 1 (excess or dysfunctional adiposity), 2 (additional metabolic risk factors or moderate-or high-risk chronic kidney disease), 3 (very high-risk chronic kidney disease or high predicted 10-year CVD risk) [22], or 4 (established CVD). For this study, we focused on stages 3 and 4. The operationalization of these stages using NHANES variables is described in detail in the Supplemental Methods. In brief, CKM Stage 3 was identified based of the presence of very-high-risk KDIGO CKD stages or a high-predicted 10-year CVD risk, and CKM Stage 4 was identified based on self-reported established cardiovascular disease. NAFLD, now commonly referred to as MASLD, was defined in this study according to US Fatty Liver Index (USFLI) ≥ 30, which demonstrates 0.80 AUC (95% CI: 0.77–0.83) for ultrasound-diagnosed NAFLD [23]. The USFLI was calculated using the published formula [23]:
 |
Fig. 2.
Definition of advanced cardiovascular-liver-kidney-metabolic (CLKM) syndrome. Advanced CLKM syndrome is defined as the presence of stage 3 or 4 cardiovascular-kidney-metabolic (CKM) syndrome with concurrent nonalcoholic fatty liver disease (NAFLD). CKM syndrome stages are defined according to the American Heart Association presidential advisory. NAFLD is defined by US Fatty Liver Index (USFLI) ≥ 30
Defnitions of TyG, TyG-BMI, TyG-WC, and TyG-WHtR
Baseline fasting blood glucose (FBG) and triglycerides (TG) were measured to calculate the TyG index: TyG = ln [TG (mg/dL) × FBG (mg/dL)/2] [24]. Furthermore, the body mass index along with the waist-to-height ratio were calculated. The participants were classified into four groups (Q1, Q2, Q3, Q4) by the quartiles of the TyG index, TyG-WC, TyG-WHtR, and TyG-BMI, respectively, and the Q1 group was used as the reference group.
TyG-WC, TyG-WHtR, and TyG-BMI were calculated according to the following formulas: (1) WHtR = waist circumference/height; (2) TyG-WC = TyG×waist circumference; (3) TyG-WHtR = TyG×WHtR; (4) TyG-BMI = TyG×BMI [24].
Ascertainment of mortality
As of December 31, 2019, the mortality rate data came from the NHANES public related mortality documents. The documents used a probabilistic matching algorithm to integrate data from survey participants with death certificate records from the National Death Index. All-cause mortality refers to the total number of deaths from all specific causes.
Data collection
The selection of covariates in this study considered a range of demographic characteristics and health-related information, including age (RIDAGEYR), gender (RIAGENDR), race (RIDRETH1), education (DMDEDUC2), marriage (DMDMARTL), poverty-to-income ratio (PIR) (INDFMPIR), BMI, waist circumference, smoking (SMQ020 and SMQ040), low-density lipoprotein cholesterol (LDL-C) (LBDLDL), glycohemoglobin (LBXGH), and uric acid (LBXURIC). Body weight, height, and waist circumference were obtained when people participated in the physical examinations at MECs. The NHANES questionnaire items were used to define current smokers as those who have smoked at least 100 cigarettes in their lifetime and currently smoke every day or some days. Demographic and socioeconomic factors (age, gender, race, education, marriage, PIR) are fundamental determinants of health and access to care, and were included to adjust for baseline population differences. Smoking status is a major risk factor for cardiovascular, metabolic, and all-cause mortality. uric acid is intricately linked to metabolic syndrome, insulin resistance, and cardiovascular risk. Conventional risk factors such as LDL-C and glycohemoglobin were adjusted for as they represent standard clinical risk assessment parameters. All biochemical assays were performed in accordance with the standard laboratory procedures detailed in the NHANES Laboratory/Medical Technologists Procedures Manual. Similarly, all body measurements were collected by trained health technicians following the protocols outlined in the NHANES Anthropometry Procedures Manual. The detailed acquisition process and measuring method of each variable are available at www.cdc.gov/nchs/nhanes.
Statistical analyses
The analysis followed NHANES guidelines, utilizing sampling weights to accurately represent the U.S. population. All analyses were performed using R statistical software (Project R, version 4.4.1). Key R packages included rms (version 6.7-1.7.7.7) for restricted cubic splines and model validation, and segmented (version 2.0–3) for identifying inflection points in the threshold effect analysis. Participants’ baseline characteristics were described according to the quartiles of TyG (Q1-Q4). Continuous variables were reported as mean ± standard deviation (SD) or median (interquartile range). Categorical variables were expressed as percentages. Differences in baseline characteristics were compared using analysis of variance (ANOVA) for continuous variables and a weighted chi-square test for categorical variables.
Kaplan-Meier curves were used to illustrate the cumulative incidence of all-cause mortality. We established three models to control for confounders and used a weighted multivariable Cox proportional hazards model to estimate the association between TyG index and modified indices and all-cause mortality. Model 1 did not include any covariate adjustments. Model 2 was adjusted for age, gender, race, education, marriage and PIR. Model 3 further adjusted smoking, eGFR, GGT, insulin, LDL-C, glycohemoglobin, and uric acid.
To investigate the dose-response association between TyG index and modified indices and all-cause mortality, restricted cubic splines (RCS) analyses were performed at the 5, 50, and 95th percentiles of the distributions of TyG, TyG-WC, TyG-WHtR, and TyG-BMI. This knot placement is a commonly recommended and robust approach for RCS analyses, as it places more flexibility in the tails of the distribution where the relationship is often less stable, while adequately capturing the central trend. The number and location of knots were chosen a priori to balance model flexibility against overfitting. If a significant nonlinear association was detected (P for nonlinear < 0.05), the inflection point was estimated. This was achieved using a segmented Cox proportional hazards model. The solid line represented the risk ratio, and the shaded area represented the 95% confidence interval. In order to compare the predictive ability of TyG and modified indices for all-cause mortality, we calculated the timedependent Harrell’s C-index. Net reclassification index (NRI) and integrated discrimination improvement (IDI) index were also used to further evaluate the incremental predictive value. Subsequently, we performed subgroup analyses by CKM stages (stage 3 or 4), gender (male or female), and age (< 60 years or ≥ 60 years, the cutoff of 60 years was chosen a priori based on its common use in cardiometabolic literature to distinguish between middle-aged and older adult populations, where pathophysiology and risk factors may differ). Statistical significance was defined as a two-tailed P value < 0.05.
This study involved multiple comparisons in the evaluation of four related TyG indices and their associations with mortality. We acknowledge the potential for increased type I error due to these multiple comparisons. However, we deliberately chose not to apply formal statistical corrections for the following reasons: (1) Our primary aim was not to test each index in isolation but to compare their predictive performance directly within the same cohort, which is an inherently exploratory and comparative analysis; (2) The indices are highly correlated, deriving from the same core components (TG and FBG), making standard corrections overly conservative and inappropriate; and (3) Our interpretation focused on the consistency of findings and the magnitude and precision of effect estimates rather than relying solely on statistical significance.
Results
Characteristics of the study participants
In this study, 1384 eligible subjects, with a mean age of 69.35 years and a male proportion of 63.1%, were included and stratified by the level of TyG. The general characteristics were detailed in Table 1 according to TyG quartiles.
Table 1.
Baseline characteristics according to TyG quartiles
| TyG index | Quartile 1 (< 8.69) |
Quartile 2 (8.69–9.09) | Quartile 3 (9.09–9.45) | Quartile 4 (> 9.45) | P value |
|---|---|---|---|---|---|
| N | 342 | 347 | 347 | 348 | |
| Age (years) | 68.12 ± 12.67 | 68.31 ± 10.77 | 67.74 ± 11.34 | 66.51 ± 11.46 | 0.473 |
| Gender (n, %) | 0.986 | ||||
| Male | 232 (61.5) | 219 (60.9) | 206 (61.6) | 216 (62.5) | |
| Female | 110 (38.5) | 128 (39.1) | 141 (38.4) | 132 (37.5) | |
| Race (n, %) | 0.326 | ||||
| Mexican American | 46 (5.1) | 55 (5.4) | 57 (5.8) | 85 (8.8) | |
| Other Hispanic | 19 (2.5) | 27 (3.7) | 27 (3.8) | 25 (3.1) | |
| Non-Hispanic White | 205 (79.5) | 208 (79.4) | 209 (81.3) | 188 (78.2) | |
| Non-Hispanic Black | 53 (7.7) | 38 (5.5) | 42 (6.7) | 32 (5.8) | |
| Other Race - Including Multi-Racial | 19 (5.2) | 19 (6.0) | 12 (2.4) | 18 (4.2) | |
| Education (n, %) | 0.191 | ||||
| Less than high school | 113 (24.6) | 129 (23.9) | 124 (26.8) | 154 (31.7) | |
| High school | 86 (26.4) | 68 (25.2) | 95 (31.8) | 85 (28.2) | |
| More than high school | 143 (49.0) | 150 (50.9) | 128 (41.4) | 109 (40.1) | |
| Marriage (n, %) | 0.018 | ||||
| Married/living with partner | 215 (67.0) | 222 (68.8) | 206 (63.6) | 197 (60.5) | |
| Widowed/divorced/separated | 115 (30.5) | 116 (29.9) | 120 (31.7) | 128 (31.9) | |
| Never married | 12 (2.4) | 9 (1.3) | 21 (4.8) | 23 (7.6) | |
| PIR | 2.54 ± 1.51 | 2.80 ± 1.56 | 2.68 ± 1.45 | 2.67 ± 1.57 | 0.395 |
| BMI (kg/m²) | 32.60 ± 7.19 | 32.64 ± 5.95 | 32.41 ± 5.13 | 33.26 ± 6.08 | 0.497 |
| Waist circumference (cm) | 112.35 ± 13.81 | 112.77 ± 13.98 | 113.54 ± 12.11 | 114.29 ± 14.14 | 0.423 |
| Current smoker (n, %) | 56 (18.4) | 46 (12.1) | 57 (16.4) | 57 (18.2) | 0.286 |
| CKM stages (n, %) | 0.622 | ||||
| 3 | 115 (30.0) | 144 (34.6) | 137 (32.1) | 148 (35.6) | |
| 4 | 227 (70.0) | 203 (65.4) | 210 (67.9) | 200 (64.4) | |
| eGFR (ml/min/1.73m2) | 71.16 ± 22.00 | 70.18 ± 21.61 | 71.38 ± 21.28 | 72.92 ± 22.97 | 0.626 |
| GGT (U/L) | 43.68 ± 57.06 | 36.72 ± 50.86 | 35.05 ± 46.09 | 43.17 ± 42.46 | 0.059 |
| Insulin (pmol/L) | 121.48 ± 173.57 | 127.72 ± 100.81 | 134.55 ± 116.75 | 145.80 ± 117.24 | 0.158 |
| LDL-C (mg/dL) | 94.79 ± 30.82 | 102.15 ± 33.37 | 108.74 ± 40.99 | 97.63 ± 37.75 | 0.002 |
| Glycohemoglobin (%) | 5.86 ± 0.71 | 6.09 ± 0.95 | 6.27 ± 0.93 | 7.24 ± 1.69 | < 0.001 |
| Uric acid (mg/dL) | 6.13 ± 1.44 | 6.49 ± 1.61 | 6.42 ± 1.42 | 6.33 ± 1.56 | 0.082 |
Data are displayed as the weighted mean ± standard deviation or unweighted frequency counts (weighted percentage) as appropriate. P values for categorical variables were derived from survey-weighted Chi-square test. TyG, triglyceride glucose; PIR, Poverty income ratio; BMI, Body mass index; CKM, cardiovascular-kidney-metabolic syndrome; eGFR, estimated Glomerular Filtration Rate; GGT, gamma glutamyltransferase; LDL, low density lipoprotein cholesterol
Associations between TyG and modified TyG indices and all-cause mortality
During a mean follow-up period of 56 months, a total of 360 participant deaths were recorded. TyG and modified TyG indices (TyG-WC, TyG-WHtR, TyG-BMI) were divided into four groups based on their quartiles (Q1-Q4). Kaplan-Meier curves of all-cause mortality are shown in Fig. 3. The relative risk of all-cause mortality in relation to TyG index and modified TyG indices are shown in Table 2, using the weighted multivariable Cox proportional hazards regression model. Using the Q1 as a reference, the HRs and 95% CIs for all-cause mortality of the remaining three groups, namely TyG (Q2-Q4), were 0.88 [0.65–1.17], 0.77 [0.57–1.05] and 0.71 [0.52–0.99], respectively, after adjustment for potential confounders (model 3). According to the trend analysis, a stepwise decreasing association was observed between increasing TyG and risk of all-cause mortality (P for trend = 0.030). After adjustment for potential confounders, no significant linear associations were observed between increasing TyG-BMI, TyG-WC and TyG-WHtR and risk of all-cause mortality (P for trend > 0.05).
Fig. 3.
Kaplan-Meier plot of all-cause mortality by TyG and modified TyG indices. (A) TyG index; (B) TyG-WC index; (C) TyG-WHtR index; (D) TyG-BMI index. TyG, triglyceride glucose; WC, waist circumference; WHtR, waist circumference/height; BMI, body mass index
Table 2.
Associations of TyG index and modified indices with all-cause mortality onset
| Model 1 | Model 2 | Model 3 | ||||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | |
| TyG (P for trend) | 0.92 [0.84–1.01] | 0.093 | 0.97 [0.88–1.06] | 0.475 | 0.89 [0.80–0.99] | 0.030 |
| TyG quartile | ||||||
| Q1 | ref | ref | ref | ref | ref | ref |
| Q2 | 0.94 [0.70–1.25] | 0.651 | 0.92 [0.69–1.23] | 0.569 | 0.88 [0.65–1.17] | 0.376 |
| Q3 | 0.80 [0.60–1.07] | 0.133 | 0.86 [0.64–1.16] | 0.328 | 0.77 [0.57–1.05] | 0.095 |
| Q4 | 0.81 [0.60–1.09] | 0.156 | 0.91 [0.67–1.23] | 0.542 | 0.71 [0.52–0.99] | 0.042 |
| TyG-WC (P for trend) | 0.95 [0.86–1.04] | 0.281 | 0.96 [0.87–1.06] | 0.443 | 0.93 [0.83–1.03] | 0.162 |
| TyG-WC quartile | ||||||
| Q1 | ref | ref | ref | ref | ref | ref |
| Q2 | 0.95 [0.73–1.25] | 0.724 | 0.84 [0.64–1.11] | 0.213 | 0.84 [0.63–1.11] | 0.209 |
| Q3 | 0.72 [0.53–0.98] | 0.034 | 0.66 [0.49–0.90] | 0.008 | 0.64 [0.47–0.88] | 0.006 |
| Q4 | 0.94 [0.70–1.26] | 0.666 | 0.99 [0.73–1.35] | 0.968 | 0.89 [0.64–1.24] | 0.491 |
| TyG-WHtR (P for trend) | 0.97 [0.88–1.06] | 0.494 | 1.00 [0.90–1.11] | 0.981 | 0.97 [0.88–1.09] | 0.638 |
| TyG-WHtR quartile | ||||||
| Q1 | ref | ref | ref | ref | ref | ref |
| Q2 | 0.93 [0.71–1.23] | 0.622 | 0.85 [0.64–1.13] | 0.274 | 0.88 [0.66–1.18] | 0.387 |
| Q3 | 0.92 [0.69–1.23] | 0.574 | 0.95 [0.71–1.27] | 0.703 | 0.94 [0.70–1.27] | 0.695 |
| Q4 | 0.90 [0.66–1.22] | 0.499 | 0.97 [0.71–1.34] | 0.868 | 0.90 [0.64–1.26] | 0.541 |
| TyG-BMI (P for trend) | 0.88 [0.80–0.97] | 0.007 | 0.97 [0.87–1.07] | 0.503 | 0.95 [0.85–1.05] | 0.298 |
| TyG-BMI quartile | ||||||
| Q1 | ref | ref | ref | ref | ref | ref |
| Q2 | 0.82 [0.63–1.08] | 0.160 | 0.78 [0.59–1.02] | 0.074 | 0.80 [0.61–1.06] | 0.120 |
| Q3 | 0.68 [0.51–0.90] | 0.007 | 0.76 [0.57–1.01] | 0.061 | 0.76 [0.56–1.03] | 0.073 |
| Q4 | 0.71 [0.53–0.96] | 0.026 | 0.97 [0.71–1.32] | 0.828 | 0.89 [0.64–1.24] | 0.494 |
Survey weight-adjusted multivariable Cox proportional hazard models were performed for all-causes mortality
Model 1 was unadjusted
Model 2 was adjusted for age, gender, race, education, marriage, and PIR
Model 3 was further adjusted for smoking, eGFR, GGT, insulin, LDL-C, glycohemoglobin, and uric acid, based on model 2
Detection of nonlinear associations
In Fig. 4, we used RCS to flexibly Cox proportional hazards models and visualize the association between TyG and modified TyG indices and all-cause mortality in individuals at advanced CLKM syndrome for examining potential nonlinear trends further. Notably, we discovered a significant U-shaped association between TyG and modified TyG indices and all-cause mortality, with P-nonlinear < 0.05. Utilizing the “segmented” package, we identified that the inflection points for TyG, TyG-WC, TyG-WHtR, and TyG-BMI in relation risks of all-cause mortality were 9.56, 1039.11, 5.17 and 215.85, respectively. Furthermore, we employed a segmented Cox proportional hazards model and Table 3 depicts the cumulative hazard of all-cause mortality in the groups with different TyG and modified TyG indices. The segmented Cox model confirmed a significant U-shaped association. Below the inflection point (TyG = 9.56), every unit increase in the TyG index was associated with a 33% lower mortality risk (HR = 0.67, 95% CI: 0.50–0.89, P = 0.006). Above this point, the association was no longer significant (HR = 1.73, 95% CI: 0.74–4.05, P = 0.206). Similar threshold effects were observed for the modified indices.
Fig. 4.
RCS analysis of the association between TyG index and modified TyG indices with all-cause mortality. The dose-response associations were assessed using RCS with knots placed at the 5, 50, and 95th percentiles of each index’s distribution. The solid line represents the adjusted hazard ratio (HR), and the shaded area represents the 95% confidence interval (CI). The bar chart show the distribution of TyG and modified TyG indices in the population. The restricted cubic spline curves and confidence intervals were derived from weighted Cox proportional hazards models. HR, hazard ratios; TyG, triglyceride glucose; WC, waist circumference; WHtR, waist circumference/height; BMI, body mass index
Table 3.
Threshold effect analysis of TyG index and modified indices on all-cause mortality
| Adjusted HR (95%CI), P value | |
|---|---|
| TyG | |
| Total | 0.815 [0.660–1.006] 0.057 |
| Segmented cox proportional hazards mode | |
| Inflection point | 9.56 |
| TyG < 9.56 | 0.670 [0.502–0.894] 0.006 |
| TyG ≥ 9.56 | 1.730 [0.739–4.049] 0.206 |
| P for Log-likelihood ratio | 0.018 |
| TyG-WC | |
| Total | 0.999 [0.998–1.000] 0.220 |
| Segmented cox proportional hazards mode | |
| Inflection point | 1039.11 |
| TyG-WC < 1039.11 | 0.997 [0.995–0.999] 0.005 |
| TyG-WC ≥ 1039.11 | 1.002 [1.000–1.005] 0.045 |
| P for Log-likelihood ratio | 0.003 |
| TyG-WHtR | |
| Total | 0.950 [0.812–1.111] 0.518 |
| Segmented cox proportional hazards mode | |
| Inflection point | 5.17 |
| TyG-WHtR < 5.17 | 0.255 [0.087–0.748] 0.013 |
| TyG-WHtR ≥ 5.17 | 1.102 [0.914–1.330] 0.309 |
| P for Log-likelihood ratio | 0.016 |
| TyG-BMI | |
| Total | 0.998 [0.996–1.001] 0.150 |
| Segmented cox proportional hazards mode | |
| Inflection point | 215.85 |
| TyG-BMI < 215.85 | 0.931 [0.883–0.981] 0.008 |
| TyG-BMI ≥ 215.85 | 0.999 [0.997–1.002] 0.619 |
| P for Log-likelihood ratio | 0.027 |
The model was adjusted for age, gender, race, education, marriage, PIR, smoking, eGFR, GGT, Insulin, LDL-C, Glycohemoglobin, and Uric acid
Compare the predictive ability of TyG and modified indices
The time-dependent Harrell’s C-indices of TyG and modified indices are shown in Fig. 5. Using the adjusted Cox regression model 3, the overall C-index values were 0.701 for TyG-WC and TyG-BMI, followed by 0.700 for TyG and TyG-WHtR. Table 4 shows the NRI and IDI indices comparing the models. Compared to TyG alone, TyG-WC yielded a significant NRI (−0.36%; P = 0.011), while the NRI and IDI for TyG-WHtR and TyG-BMI were not significant.
Fig. 5.
Time-dependent predictive capacity of TyG and modified indices for all-cause mortality. The time-dependent C-indices were calculated from weighted Cox regression models. TyG, triglyceride glucose; WC, waist circumference; WHtR, waist circumference/height; BMI, body mass index
Table 4.
NRI and IDI index of TyG and modified indices
| Comparison | NRI (%), P value | IDI (%), P value |
|---|---|---|
| TyG-WC vs. TyG | −0.36, 0.011 | −0.34, 0.068 |
| TyG-WHtR vs. TyG | 0.22, 0.071 | 0.10, 0.464 |
| TyG-BMI vs. TyG | 0.07, 0.768 | 0.06, 0.707 |
Subgroup analyses
To further elucidate the association between TyG and modified TyG indices and the risks of all-cause mortality, we conducted a series of subgroup analyses. In the subgroup analysis (Fig. 6), stratified by CKM stages, gender, and age. Across all subgroups, the association between TyG and modified TyG indices and all-cause mortality coincided. Clearly, there was no significant interaction observed between the TyG and modified TyG indices and the stratifed variables.
Fig. 6.
Subgroup analyses of the association between TyG index and modified indices with all-cause mortality (per 1 SD). All subgroup analyses were derived from weighted models that accounted for the NHANES complex survey design. CKM, cardiovascular-kidney-metabolic syndrome; HR, hazard ratios; TyG, triglyceride glucose; WC, waist circumference; WHtR, waist circumference/height; BMI, body mass index
Discussion
This nationwide prospective cohort study yields three principal findings regarding advanced CLKM syndrome: (1) A U-shaped association exists between TyG indices and all-cause mortality, with inflection points identified (TyG = 9.56, TyG-WC = 1039.11, TyG-WHtR = 5.17, TyG-BMI = 215.85). The relationship was characterized by a threshold pattern, wherein indices below the inflection points showed a significant inverse association with mortality, whereas the trend above the thresholds, while often pointing towards increased risk, was not consistently statistically robust. It most likely stems from reduced statistical power in the upper extremes of the exposure distributions, where fewer participants and mortality events lead to wider confidence intervals and less precise hazard ratio estimates. Therefore, while our data robustly confirm the perils of very low TyG values, they are less definitive regarding the risks at the very highest values, though the point estimates are consistently > 1. This nuanced interpretation does not diminish the clinical utility of the identified inflection points but underscores that the risk escalates most predictably once values fall below a critical lower threshold; (2) Contrary to studies in early-stage cardiometabolic populations, the modified TyG indices provided no incremental predictive value, and in the case of TyG-WC, may actually impair prediction accuracy, as evidenced by its significantly negative net reclassification index; and (3) These associations remained consistent across subgroups stratified by CKM stage, sex, and age. These insights redefine risk stratification in advanced multisystem syndromes and highlight the duality of metabolic dysregulation in end-organ disease.
The U-shaped mortality curve reveals a critical dichotomy. Below the inflection points, higher TyG indices were paradoxically associated with a reduced mortality risk. It is highly probable that this association is not causal but is heavily confounded by the profound alterations in body composition and energy metabolism characteristic of advanced multisystem syndromes. Specifically, the lower TyG values may be a marker of a catabolic state, characterized by sarcopenia (muscle wasting), cachexia, and frailty—all of which are powerful independent predictors of mortality. While our models adjusted for BMI and waist circumference, these are poor proxies for muscle mass and do not capture the critical balance between fat and lean tissue. Therefore, a more plausible explanation for our findings is the “sarcopenia and frailty paradox”: in end-stage disease, patients with significant tissue wasting may present with deceptively “better” metabolic profiles, yet they are in fact at extreme risk due to their loss of metabolic and functional reserve [25–28]. Conversely, above thresholds, TyG indices trended toward increased mortality, consistent with glucolipotoxicity-driven mitochondrial dysfunction, oxidative stress, and accelerated organ damage [29, 30]. This biphasic relationship underscores the limitations of linear models in complex multisystem syndromes and emphasizes the need for threshold-based risk assessment.
Surprisingly, the anthropometry-enhanced variants (TyG-BMI, TyG-WC, TyG-WHtR) showed no incremental predictive value over the original TyG index. This finding contrasts sharply with reports from earlier disease stages. For instance, Hong et al. [13] demonstrated the superiority of TyG-WC for predicting cardiovascular disease in individuals with CKM stages 0–3. Similar advantages have been noted in general populations [10, 31]. We propose that conventional adiposity measures lose discriminatory power in advanced CLKM due to sarcopenic obesity and fluid overload altering body composition metrics [25, 32], pervasive metabolic dysregulation overshadowing subtle anthropometric differences, and the cohort’s severe adiposity (mean BMI 32.5 kg/m², waist circumference 112 cm) creating a ceiling effect.
The identified inflection points offer actionable clinical thresholds: TyG = 9.56 corresponds to established insulin resistance cutoffs [33] but newly serves as a mortality risk transition point. For patients with a TyG index below the inflection point, our data suggest a paradigm of cautious, supportive management. The observed inverse association with mortality risk indicates that aggressive risk-factor control may be ineffective or even harmful in this frail subpopulation. The protective effect at lower TyG levels cautions against aggressive hypoglycemic therapy in frail CLKM patients, echoing concerns about iatrogenic harm in advanced metabolic syndromes [34]. For patients above thresholds, interventions targeting glucolipotoxicity (e.g., SGLT2 inhibitors, GLP-1 RAs) should be prioritized, given their cardiorenal-hepatic benefits [35].
Mechanistically, the liver’s centrality in CLKM may explain TyG’s robustness. As the primary site for triglyceride and glucose metabolism, hepatic insulin resistance directly amplifies systemic inflammation and atherogenic dyslipidemia [36], rendering TyG—a hepatic IR surrogate—sufficient for risk stratification without anthropometric augmentation in advanced disease.
Our findings diverge from studies in earlier disease stages: Early CKM (Stages 0–3): Hong et al. [13] and Li et al. [14] demonstrated TyG-WC/TyG-WHtR’s superiority in CVD prediction, attributed to visceral adiposity’s role in driving insulin resistance. Advanced CLKM: Here, TyG alone performed comparably to modified indices, suggesting disease-stage-dependent utility of obesity metrics. The U-shaped association further distinguishes advanced CLKM from linear associations in early stages [13, 37]. While Hong et al. [13] used C-indices/NRI/IDI to validate TyG-WC’s predictive gains, our nonlinear models (RCS, segmented Cox) revealed thresholds unaddressed in prior work.
Our findings provide a potential framework for risk-stratified management in advanced CLKM syndrome. The TyG index inflection point of 9.56 offers a actionable clinical benchmark. Patients below this threshold warranting a conservative management strategy that avoids aggressive risk factor control. In contrast, for patients above this threshold, the elevated risk likely reflects predominant glucolipotoxicity, justifying intensive intervention with therapies like SGLT2 inhibitors and GLP-1 receptor agonists [35]. Validating this TyG-guided approach in prospective cohorts could pave the way for more personalized and effective management strategies.
This study has several limitations. First, the definition of NAFLD. We used the USFLI as a non-invasive surrogate for NAFLD diagnosis, which, while practical for a large epidemiological cohort like NHANES, is inferior to imaging techniques or histology. This misclassification bias could have attenuated the observed associations between TyG indices and mortality, potentially leading to an underestimation of the true effect sizes. Second, despite extensive adjustment for confounders, residual confounding from unmeasured variables remains possible. Such as detailed inflammatory markers and medication adherence, could influence both metabolic dysregulation and mortality risk. Third, the generalizability of our findings. Our cohort was predominantly elderly, male, and non-Hispanic White, reflecting a specific high-risk demographic within the US population. Therefore, our results may not be directly generalizable to younger populations, women, or other racial and ethnic groups. Future validation in more diverse, prospective cohorts is essential. Fourth, the follow-up duration. The median follow-up of 56 months may be insufficient to capture the very long-term mortality outcomes and the full trajectory of CLKM syndrome progression. Fifth, our operational definition of advanced CLKM syndrome combined CKM stage 3 and stage 4. Although the non-significant interaction test is reassuring, we acknowledge that important pathophysiological differences may exist between these stages. A well-powered sensitivity analysis to rigorously test for effect modification by stage was not feasible in our cohort due to the limited sample size and number of events when stratifying, which would have resulted in underpowered and unstable estimates. Therefore, we cannot rule out the possibility that the overall associations are influenced by the cohort’s composition. Finally, an important methodological consideration is the presence of competing risks. In this elderly cohort with advanced multisystem disease, patients are at high risk of death from multiple causes. Our use of standard survival analysis for all-cause mortality assumes that these competing events are non-informative. Future work should validate TyG inflection points in diverse populations while exploring their therapeutic implications, investigate molecular mechanisms linking TyG thresholds to organ damage, and integrate these thresholds with novel biomarkers for precision risk stratification.
Conclusion
In this nationwide prospective cohort of patients with advanced CLKM syndrome, TyG indices exhibited a U-shaped association with mortality, revealing dual protective and harmful roles of metabolic dysregulation. The original TyG index performed comparably or even superiorly to anthropometry-enhanced variants, supporting its utility as a parsimonious risk stratification tool. The identified inflection points provide clinically actionable thresholds for personalized management, underscoring the need for nonlinear models in multisystem syndromes.
Supplementary Information
Acknowledgements
Not applicable.
Author contributions
Conceptualization: J.L., J.W., and D.W. Methodology: H.L. Data curation and validation: P.L. and Z.H. Writing: Y.Z. All authors reviewed the manuscript.
Funding
This work was supported by the Chinese Medicine Guangdong Laboratory (HQL2024PZ028), State Key Laboratory of Traditional Chinese Medicine Syndrome (09005651007), Guangzhou University of Chinese Medicine "Strengthening the Foundation" (GZY2025GB0109), National Natural Science Foundation (82405285), Project of Administration of Traditional Chinese Medicine of Guangdong Province of China (20261406) and National Natural Science Foundation of China (12302409).
Data availability
Details about the surveys and the corresponding death index are available at www.cdc.gov/nchs/nhanes and www.cdc.gov/nchs/ndi/, respectively.
Declarations
Ethics approval and consent to participate
The National Center for Health Statistics Research Ethics Review Board approved all protocols (#2021-05), with participants providing written informed consent. Study procedures adhered to the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Dawei Wang, Email: 13902236233@139.com.
Junmao Wen, Email: 13760723750@163.com.
References
- 1.Sebastian SA, Padda I, Johal G. Cardiovascular-Kidney-Metabolic (CKM) syndrome: a state-of-the-art review. Curr Probl Cardiol. 2024;49(2):102344. [DOI] [PubMed] [Google Scholar]
- 2.Claudel SE, Verma A. Cardiovascular-kidney-metabolic syndrome: a step toward multidisciplinary and inclusive care. Cell Metab. 2023;35:2104–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Chew NWS, Mehta A, Goh RSJ, et al. Cardiovascular-liver-metabolic health: recommendations in screening, diagnosis, and management of metabolic dysfunction-associated steatotic liver disease in cardiovascular disease via modified Delphi approach. Circulation. 2025;151(1):98–119. [DOI] [PubMed] [Google Scholar]
- 4.European Association for the Study of the Liver (EASL), European Association for the Study of Diabetes (EASD), European Association for the Study of Obesity (EASO). EASL-EASD-EASO clinical practice guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). Obes Facts. 2024;17(4):374–444. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Zhou XD, Zheng MH. Cardiovascular-kidney-metabolic syndrome and MASLD: integrating medical perspectives. Nat Rev Cardiol. 2025;22(11):843. [DOI] [PubMed] [Google Scholar]
- 6.Simental-Mendía LE, Rodríguez-Morán M, Guerrero-Romero F. The product of fasting glucose and triglycerides as surrogate for identifying insulin resistance in apparently healthy subjects. Metab Syndr Relat Disord. 2008;6(4):299–304. [DOI] [PubMed] [Google Scholar]
- 7.Vasques AC, Novaes FS, de Oliveira Mda S, et al. TyG index performs better than HOMA in a Brazilian population: a hyperglycemic clamp validated study. Diabetes Res Clin Pract. 2011;93(3):e98–100. [DOI] [PubMed] [Google Scholar]
- 8.da Silva A, Caldas APS, Hermsdorff HHM, et al. Triglyceride-glucose index is associated with symptomatic coronary artery disease in patients in secondary care. Cardiovasc Diabetol. 2019;18(1):89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Luo E, Wang D, Yan G, et al. High triglyceride-glucose index is associated with poor prognosis in patients with acute ST-elevation myocardial infarction after percutaneous coronary intervention. Cardiovasc Diabetol. 2019;18(1):150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Er LK, Wu S, Chou HH, et al. Triglyceride glucose-body mass index is a simple and clinically useful surrogate marker for insulin resistance in nondiabetic individuals. PLoS One. 2016;11(3):e0149731. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Guerrero-Romero F, Villalobos-Molina R, Jiménez-Flores JR, et al. Fasting triglycerides and glucose index as a diagnostic test for insulin resistance in young adults. Arch Med Res. 2016;47(5):382–7. [DOI] [PubMed] [Google Scholar]
- 12.Lee SB, Ahn CW, Lee BK, et al. Association between triglyceride glucose index and arterial stiffness in Korean adults. Cardiovasc Diabetol. 2018;17(1):41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hong J, Zhang R, Tang H, Wu S, Chen Y, Tan X. Comparison of triglyceride glucose index and modified triglyceride glucose indices in predicting cardiovascular diseases incidence among populations with cardiovascular-kidney-metabolic syndrome stages 0–3: a nationwide prospective cohort study. Cardiovasc Diabetol. 2025;24(1):98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Li W, Shen C, Kong W, et al. Association between the triglyceride glucose-body mass index and future cardiovascular disease risk in a population with Cardiovascular-Kidney-Metabolic syndrome stage 0–3: a nationwide prospective cohort study. Cardiovasc Diabetol. 2024;23(1):292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Chen TC, Clark J, Riddles MK et al. National health and nutrition examination Survey, 2015–2018: sample design and Estimation procedures. Vital Health Stat 2. 2020;(184):1–35. [PubMed]
- 16.Patel CJ, Pho N, McDuffie M, et al. A database of human exposomes and phenomes from the US National health and nutrition examination survey. Sci Data. 2016;3:160096. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Li J, Lei L, Wang W, et al. Social risk profile and cardiovascular-kidney-metabolic syndrome in US adults. J Am Heart Assoc. 2024;13(16):e034996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Inker LA, Eneanya ND, Coresh J, et al. New creatinine- and cystatin C-based equations to estimate GFR without race. N Engl J Med. 2021;385(19):1737–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.April-Sanders AK. Integrating Social Determinants of Health in the Management of Cardiovascular-Kidney-Metabolic Syndrome. J Am Heart Assoc. 2024;13:e036518. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ding Y, Wu X, Cao Q, et al. Gender disparities in the association between educational attainment and Cardiovascular-Kidney-Metabolic syndrome: cross-sectional study. JMIR Public Health Surveill. 2024;10:e57920. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Ndumele CE, Rangaswami J, Chow SL et al. Cardiovascular-Kidney-Metabolic Health: A Presidential Advisory From the American Heart Association [published correction appears in Circulation. 2024;149(13):e1023. doi: 10.1161/CIR.0000000000001241.]. Circulation. 2023;148(20):1606–1635. [DOI] [PubMed]
- 22.Khan SS, Matsushita K, Sang Y, et al. Development and validation of the American Heart Association’s PREVENT equations. Circulation. 2024;149(6):430–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ruhl CE, Everhart JE. Fatty liver indices in the multiethnic United States National Health and Nutrition Examination Survey. Aliment Pharmacol Ther. 2015;41(1):65–76. [DOI] [PubMed] [Google Scholar]
- 24.Dang K, Wang X, Hu J, et al. The association between triglyceride-glucose index and its combination with obesity indicators and cardiovascular disease: NHANES 2003–2018. Cardiovasc Diabetol. 2024;23(1):8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Cruz-Jentoft AJ, Bahat G, Bauer J, et al. Sarcopenia: revised European consensus on definition and diagnosis. Age Ageing. 2019;48(4):601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.von Haehling S, Ebner N, Dos Santos MR, Springer J, Anker SD. Muscle wasting and cachexia in heart failure: mechanisms and therapies. Nat Rev Cardiol. 2017;14(6):323–41. [DOI] [PubMed] [Google Scholar]
- 27.Tandon P, Raman M, Mourtzakis M, Merli M. A practical approach to nutritional screening and assessment in cirrhosis. Hepatology. 2017;65(3):1044–57. [DOI] [PubMed] [Google Scholar]
- 28.Kalantar-Zadeh K, Horwich TB, Oreopoulos A, et al. Risk factor paradox in wasting diseases. Curr Opin Clin Nutr Metab Care. 2007;10(4):433–42. [DOI] [PubMed] [Google Scholar]
- 29.Samuel VT, Shulman GI. Nonalcoholic fatty liver disease as a nexus of metabolic and hepatic diseases. Cell Metab. 2018;27(1):22–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Perry RJ, Camporez JG, Kursawe R, et al. Hepatic acetyl coa links adipose tissue inflammation to hepatic insulin resistance and type 2 diabetes. Cell. 2015;160(4):745–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Xia X, Chen S, Tian X, et al. Association of triglyceride-glucose index and its related parameters with atherosclerotic cardiovascular disease: evidence from a 15-year follow-up of Kailuan cohort. Cardiovasc Diabetol. 2024;23(1):208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Jia X, Zhang L, Yang Z, et al. Impact of sarcopenic obesity on heart failure in people with type 2 diabetes and the role of metabolism and inflammation: a prospective cohort study. Diabetes Metab Syndr. 2024;18(5):103038. [DOI] [PubMed] [Google Scholar]
- 33.Sánchez-Íñigo L, Navarro-González D, Fernández-Montero A, et al. The TyG index may predict the development of cardiovascular events. Eur J Clin Invest. 2016;46(2):189–97. [DOI] [PubMed] [Google Scholar]
- 34.Lai W, Lin Y, Gao Z, Huang Z, Zhang T. Joint association of TyG index and LDL-C with all-cause and cardiovascular mortality among patients with cardio-renal-metabolic disease. Sci Rep. 2025;15(1):5854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Lopaschuk GD, et al. Cardiometabolic benefits of novel antidiabetics. Nat Rev Cardiol. 2024;21(1):37–55.37563454 [Google Scholar]
- 36.Tilg H, et al. Liver crosstalk with kidney and heart. Nat Rev Gastroenterol Hepatol. 2024;21(4):258–71. [Google Scholar]
- 37.Caussy C, Reeder SB, Sirlin CB, et al. Noninvasive, quantitative assessment of liver fat by MRI-PDFF as an endpoint in NASH trials. Hepatology. 2018;68(2):763–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Details about the surveys and the corresponding death index are available at www.cdc.gov/nchs/nhanes and www.cdc.gov/nchs/ndi/, respectively.