The study cohort 1 consisted of adult participants who participated in NHANES from 1999 to 2004, after excluding all individuals under 40 years old (n=4,064), BMI < 18.5 (n=157), a history of CVD (defined as self-reported coronary heart disease, heart attack, angina, stroke, or heart failure, n=10,371), missing data on a cardiac biomarker (NT-proBNP, n=9,920), missing data on TYG index and other covariates (n=382), substantial renal impairment (eGFR ≤15 ml/min/1.73 m², n=16). The final analytical population consisted of 6,216 participants.

The study cohort 2 included middle-aged and elderly patients, aged 40–79 years, who were hospitalized at the Department of Cardiology, JSHTCM, between December 2024 and June 2025. All subjects had no previous history of CVD and underwent routine electrocardiogram, echocardiography, and coronary CTA or coronary angiography during hospitalization. Exclusion criteria included: (1) any coronary artery stenosis >50% (n=18); (2) ejection fraction <40% (n=11); (3) rapid arrhythmias such as atrial fibrillation and atrial flutter (n=12). Ultimately, 302 middle-aged and elderly individuals were included in the analysis. The study design and exclusion details can be found in the flowchart ( Figure 1 ).

The TYG index was calculated as follows using these parameters as an exposure variable: Ln [triglycerides (mg/dl) * fasting glucose (mg/dl)/2] (22, 23). Fasting venous blood samples were collected from all participants after at least 8.5 hours of fasting. Serum triglycerides and glucose concentrations were measured with an automatic biochemical analyzer, and the specific levels were determined using coupling reaction enzyme methods and hexokinase reactions respectively.

The study cohort 1 aimed to analyze NT-proBNP in stored serum samples, and the data are currently applicable to 1999-2004. The Roche Cobase 601 automated analyzer measured NT-proBNP levels in serum. The lower limit of detection was 5 pg/ml, and the upper limit was 35000 pg/ml. The coefficients of variation were 3.1% (low, 46 pg/ml) and 2.7% (high, 32805 pg/ml), respectively (24). Study cohort 2 was based on the measurement of serum NT-proBNP levels by the Laboratory Department of JSHTCM, with a low detection limit of 20 pg/mL. In the main analysis, we take reference of other research findings and stratified NT-proBNP levels based on established clinical reference ranges for cardiovascular biomarkers, we defined elevated NT-proBNP as NT-proBNP ≥125 pg/mL (25–28).

In study cohort 1, from the start of the investigation (1999-2004), the follow-up period continued until December 31, 2019. Participant’s mortality data can be obtained by linking their personal identification code to the death certificate records in the National Death Index (NDI) based on the mortality files released by the National Center for Health Statistics (NCHS). Given the absence of endpoint event follow-up data within study cohort 2, the reporting of all-cause mortality outcomes is not applicable.

Participants in both study cohorts underwent detailed information collection to obtain sociodemographic data such as age, gender, Body Mass Index (BMI). Hypertension was defined by a mean systolic blood pressure of ≥140 mmHg, a diastolic blood pressure of ≥90 mmHg, self-reported physician-diagnosed hypertension, or the current use of antihypertensive medications. A detailed questionnaire was used to assess smoking status and categorize individuals as never, current, or former smokers. After fasting for 8.5 hours, participants provided a fasting venous blood sample, which was analyzed for total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), uric acid (UA), urea, creatinine, and glycated hemoglobin (HbA1c). Those with diabetes were defined as having a history of diabetes diagnosed by a physician or possessing an HbA1c of ≥6.5%. Renal function was evaluated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation, and subjects exhibiting substantial renal impairment (eGFR ≤15 ml/min/1.73 m²) were omitted from the analysis. Medication use was obtained from the prescription records of medications used by participants in the past month, mainly including the following two categories: hypoglycemic and lipid-lowering drugs.

According to elevated NT-proBNP status, we assessed sociodemographic and cardiovascular risk factor characteristics among middle-aged and elderly individuals without known CVD. Normally distributed quantitative data are expressed as mean ± standard deviation (SD), and intergroup variations are evaluated using analysis of variance (ANOVA). For non-normally distributed data, the median (M) and interquartile range (IQR) are presented, and intergroup comparisons are performed using the Kruskal-Wallis rank-sum test. Categorical data are described using rates or proportions, and the chi-square test or Fisher’s exact test is applied to component comparisons. In the study cohort 1, we evaluated the crude prevalence of elevated NT-proBNP, then age-standardized the prevalence based on the age distribution of the U.S. adult population in 2000 (29), and visualized the data by histograms ( Figures 2A, B ).

In the study cohort 1, multivariable logistic regression and multivariable Cox proportional hazards models were used to evaluate the relationship between TYG index (exposure indicator) and elevated NT-proBNP levels and all-cause mortality risk, respectively. Leveraging prior research and accounting for potential confounders influencing exposure-outcome relationships (24–27), three different statistical inference models were used: Crude model: unadjusted; Model I: adjusted for age, gender, race, education level; Model II: further adjusted for BMI, hypertension, diabetes, smoking status, TC, HDL-C, urea, creatinine, UA, eGFR and use of hypoglycemic/lipid-lowering medications. Sensitivity analysis involved categorizing the TYG index into low (T1: 0-33.33%), medium (T2: 33.33%-66.67%), and high (T3:66.67%-100%) tertiles to test the robustness of the results, with the low TYG index (T1) set as a reference dummy variable to evaluate elevated circulating NT-proBNP and all-cause mortality risks in the tertiles, and the trend test was carried out. Moreover, we compared survival rates between the TYG index groups and subgroups using Kaplan-Meier curves, based on log-rank tests of survival rates.

Statistical inference model for the study cohort 2: Crude model: unadjusted; Model I: adjusted for age, gender; Model II: further adjusted for BMI, hypertension, diabetes, smoking, TC, HDL-C, urea, creatinine, UA, HbA1c, eGFR and use of hypoglycemic/lipid-lowering medications. The sensitivity analysis was consistent with study cohort 1.

In order to determine the potential nonlinear relationship between the TYG index and circulating levels of NT-proBNP and all-cause mortality risk, we employed a RCS model. This model facilitates the visual representation of the relationship between continuous variables and outcomes by segmenting the variable’s range into intervals through the placement of knots. To mitigate overfitting, we utilized the quantile method, establishing four knots at the 5th, 35th, 65th, and 95th percentiles of the data distribution, respectively. The model adjustment was consistent with the multivariable regression model.

In the subgroup analysis of the study cohort 1, data were stratified by gender (male/female), age (40-59y/≥ 60y), BMI (normal/overweight/obese), eGFR (< 60/60-< 90/≥ 90), hypertension (yes/no), diabetes (yes/no), and medications use (yes/no), with interaction tests conducted to assess consistency with the overall population results. These stratification factors are considered potential effect modifiers. Moreover, further analysis was conducted by stratifying elevated circulating NT-proBNP (yes/no) and age (40-59y/≥ 60y) to assess the association between TYG index and all-cause mortality in different subgroups.

In the study cohort 1, the recommended sampling survey weights were utilized in order to obtain unbiased results. R software (version 4.3.3; http://www.r-project.org) was used for conducting all statistical analyses. A two-tailed P-value < 0.05 was considered statistically significant.

We confirm that all analysis methods involving NHANES data in this study have been strictly adhered to in accordance with the NHANES survey methods and analytic guidelines recommended by the Centers for Disease Control and Prevention (CDC)/NCHS.

In the study cohort 1 (males vs females: 3,039 vs 3,177), the mean TYG index of the participants was 8.68 (0.67), with an average age of 55.2 years (54.8, 55.6). Examining the sociodemographic characteristics, it can be observed that individuals with a higher TYG index tend to be male, older in age, have less education, and are primarily non-Hispanic white. Regarding CVD risk factors, individuals with a higher TYG index frequently presented with comorbidities such as hypertension and diabetes, along with increased levels of BMI, smoking rates (former and current), serum urea, creatinine, uric acid, total cholesterol, triglycerides, fasting blood glucose, glycated hemoglobin, and mortality rate, while eGFR, HDL levels were relatively lower. Additionally, in the study cohort 2, apart from age and eGFR, the baseline characteristics of the participants remained largely consistent with those in the study cohort 1 ( Table 1 , Supplementary Table S1 ).

The crude prevalence of elevated NT-proBNP was 21.91% in study cohort 1, with a decline as the TYG index increased. Compared to middle-aged and elderly males, females exhibit a higher prevalence of elevated NT-proBNP (female vs male: 33.34% vs 20.2%). After further age standardization, the age-standardized prevalence rates of elevated circulating NT-proBNP are as follows: female: 29.21%, male: 17.08% ( Supplementary Table S3 , Figures 2A, B ).

In the study cohort 1, when sociodemographic factors such as age, gender, race, and education level were adjusted, multivariate logistic regression models revealed that high TYG indexes were negatively associated with elevated NT-proBNP levels [OR (95% CI): 0.70 (0.61, 0.82), p < 0.001]. Upon further adjustment for model covariates (Model 2), the negative correlation between the two remained stable and still held significant statistical significance [OR (95% CI): 0.77 (0.64, 0.93), p=0.008]. Sensitivity analysis indicated that taking the low TYG index group (T1) as a reference, the risk of elevated NT-proBNP decreased by approximately 27% in the high TYG index group (T3) [OR (95% CI): 0.73 (0.55, 0.96), p for trend=0.027].

The results of study cohort 2 indicate that as the TYG index progressively increases, the serum NT-proBNP levels in middle-aged and elderly individuals show a declining trend, which is consistent with the multiple regression analysis of study cohort 1. Specifically, for each standard unit increase in the TYG index, the serum NT-proBNP level decreased by approximately 37.58 pg/mL [β (95% CI): -37.58 (-59.11, -16.06)]. Furthermore, the T3 group exhibited a more pronounced decrease in serum NT-proBNP levels than the T1 group (p for trend=0.002) Table 2 , Supplementary Table S2 shows detailed data.

In both study cohorts, the TYG index and elevated NT-proBNP were visualized on a continuous scale using RCS plots. The findings suggest that as the TYG index increases, the risk of elevated NT-proBNP gradually decreases, in line with the multivariable regression model ( Figures 3A-C ).

Subgroup analyses were conducted in study cohort 1 to further validate the relationship between TYG index and elevated NT-proBNP across diverse populations stratified by gender, age, BMI, eGFR, hypertension, diabetes, and medications use. Our results indicate that individuals with a high TYG index group (T3) in different subgroups exhibit a lower risk of elevated NT-proBNP, and this negative correlation persists when the TYG index is continuous. Furthermore, interaction tests did not reveal any significant influence of gender, age, BMI, eGFR, hypertension, diabetes, and medications use on the association between TYG index and elevated NT-proBNP ( Figure 4 ).

In the study cohort 1, a total of 2,165 middle-aged and elderly individuals had died during the follow-up period as of December 31, 2019. The cumulative mortality rate in the non-elevated NT-proBNP group was 17.8%, compared to 51.8% in the elevated NT-proBNP group [absolute risk differences (ARD): 34%]. Kaplan-Meier survival curves showed that a higher TYG index was positively associated with an increased risk of all-cause mortality regardless of whether individuals had elevated NT-proBNP levels [survival probabilities: non-elevated NT-proBNP vs elevated NT-proBNP=82.2% vs 48.2%, ( Figures 5A-C , all log-rank p value< 0.001)]. According to the multivariate Cox proportional hazards model, the adjusted all-cause mortality risk for the entire population increased by 25% for each unit increase in the TYG index [HR (95% CI): 1.25 (1.08, 1.44), p=0.003], with the risk ratio for those with elevated NT-proBNP at 1.12 (95% CI: 0.97, 1.29), while those with non-elevated NT-proBNP were 1.34 (95% CI: 1.13, 1.60). Sensitivity analysis suggests that compared to the lower TYG index group, the high TYG index group is associated with a 20% increased risk of adjusted all-cause mortality [HR (95% CI): 1.20 (1.02, 1.41), p for trend=0.031] ( Table 3 ).

In the study cohort 1, the RCS curves adjusted for multiple variables indicate that with the gradual increase in the TYG index, the overall population experiences a gradual rise in all-cause mortality risk. This trend is consistent in both the non-elevated circulating NT-proBNP group and the elevated circulating NT-proBNP group, mirroring the findings observed in the overall population ( Supplementary Figures S1A-C ). Furthermore, the subgroup analysis results also support these findings ( Supplementary Table S4 , Supplementary Figure S2-S5 ).