Associations of the Zhejiang University (ZJU) index with cardiovascular diseases and mortality among US adults: a national cohort study Authors: Zhenzhen Mo, Bilian Chen, Minyi Wu Published: 2025 ABSTRACT AbstractBackgroundThe Zhejiang University (ZJU) Index, a composite metabolic indicator incorporating body mass index (BMI), fasting plasma glucose (FPG), triglycerides (TG), and alanine aminotransferase/aspartate aminotransferase (ALT/AST) ratio, is associated with abnormalities like dyslipidemia and glucose intolerance. Yet its clinical significance in cardiovascular disease (CVD) is underinvestigated, and this study aims to clarify its associations with prevalent CVD, all-cause mortality, and CVD mortality.MethodsThe study included 18,609 adults from the National Health and Nutrition Examination Survey (NHANES) spanning 1999–2018. Associations of the ZJU Index with prevalent CVD were evaluated using multivariable Logistic regression. Associations with all-cause mortality and CVD mortality were assessed using multivariable Cox proportional hazards regression. Restricted cubic spline (RCS) regression and ROC curve analyses were further used to explore non-linear relationships and predictive performance, respectively. Sensitivity analyses, which included excluding participants who died within 2 years and those with baseline CVD, confirmed the robustness of the findings.ResultsAmong 18,609 participants, 2,121 had CVD (9.0% weighted prevalence). Logistic regression showed a positive association of the ZJU Index with prevalent CVD risk: after full adjustment, a 1-unit increase was linked to a 2% higher risk (OR = 1.02, 95% CI 1.01–1.03, P < 0.001), and participants in the highest quartile (Q4) had a 50% higher risk than those in the lowest quartile (Q1) (OR = 1.50, 95% CI 1.22–1.84, P < 0.001). Significant non-linearity between the ZJU Index and prevalent CVD was confirmed (P < 0.001). Over a median 71-month follow-up (36–121 months), 2,752 all-cause deaths (11.1% weighted rate) and 934 CVD deaths (3.4% weighted rate) occurred. After full adjustment, the association persisted for CVD mortality (Q4 vs. Q1: HR = 1.30, 95% CI 1.03–1.63, P = 0.027). Restricted cubic spline (RCS) analysis revealed that the ZJU Index had a U-shaped relationship with all-cause mortality and a J-shaped relationship with CVD mortality (both P < 0.001). Sensitivity analyses supported the robustness of these findings.ConclusionsIn conclusion, whether analyzed as a continuous or categorical variable, a higher ZJU Index is significantly associated with higher risks of CVD and CVD mortality, while it shows a U-shaped relationship with all-cause mortality. This indicates that the ZJU Index holds potential as a CVD risk stratification tool to identify high-risk individuals and guide targeted interventions. However, its utility as a CVD screening tool requires further validation to confirm optimal cut-offs and compatibility with existing protocols.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12944-025-02760-x.Keywords: ZJU index, Cardiovascular disease, Cardiovascular mortality, All-cause mortality, Nonlinear relationship Clinical perspective What is new? This is the first large, well - designed prospective study to assess the associations between ZJU Index and prevalent CVD, as well as all - cause and CVD mortality among U.S. adults. This is the first large, well - designed prospective study to assess the associations between ZJU Index and prevalent CVD, as well as all - cause and CVD mortality among U.S. adults. A higher ZJU Index, whether as a continuous or categorical variable, shows significant associations with increased risks of CVD and CVD mortality. For all-cause mortality, it has a U-shaped relationship. A higher ZJU Index, whether as a continuous or categorical variable, shows significant associations with increased risks of CVD and CVD mortality. For all-cause mortality, it has a U-shaped relationship. What are the clinical implications? The ZJU Index can serve as a potential indicator for CVD risk and CVD mortality assessment. Clinicians may consider incorporating it into clinical evaluations to identify individuals at higher risks, aiding in early intervention and personalized cardiovascular care. The ZJU Index can serve as a potential indicator for CVD risk and CVD mortality assessment. Clinicians may consider incorporating it into clinical evaluations to identify individuals at higher risks, aiding in early intervention and personalized cardiovascular care. Introduction In recent decades, while global population health and life expectancy have exhibited a steady upward trajectory, cardiovascular disease (CVD) remains the preeminent cause of adult morbidity and mortality [1]. Recognized as a critical public health challenge, CVD is primarily characterized by its high global prevalence, incidence, and mortality [2, 3]. With population aging acting as a key driver, the global burden of CVD is projected to escalate further [4, 5]. Thus, prioritizing the early identification and screening of individuals at high risk of CVD, especially those with metabolic risk factors, is pivotal. Numerous individual factors are known to contribute to CVD risk, including diabetes, hyperlipidemia, hypertension, and obesity (often characterized by elevated BMI), all of which are linked to increased CVD incidence and/or mortality [6–10]. Additionally, sex differences in CVD progression and outcomes [11, 12] as well as the alanine aminotransferase to aspartate aminotransferase ratio (AST/ALT) as a predictive biomarker for CVD mortality [13], have been highlighted. However, despite robust evidence for these isolated factors, no integrated metric currently quantifies the collective contribution of these factors to CVD risk. This gap limits the accuracy of early risk stratification, making the development of a comprehensive risk score for improved risk assessment clinically imperative. As an innovative metabolic composite index, the ZJU Index incorporates various biomarkers, such as triglycerides (TG), liver enzymes (AST, ALT), body mass index (BMI), and fasting blood glucose (FBG) [14]. Unlike isolated biomarkers, this index enables comprehensive metabolic health profiling by identifying systemic metabolic dysfunction patterns and related health risks. The ZJU Index was initially applied in research on non-alcoholic fatty liver disease (NAFLD) [14]. Beyond its association with NAFLD, accumulating evidence has expanded its range of associations; subsequent studies have revealed that the index is not only associated with insulin resistance(IR) and the risk of developing diabetes, but also closely linked to abnormal lipid metabolism [15–17]. Additionally, its significant correlation with an elevated risk of sarcopenia in the U.S. population suggests its potential value as a predictive marker for sarcopenia [18]. Other studies have confirmed that a higher ZJU Index is significantly associated with an increased risk of obstructive sleep apnea syndrome (OSAS) [19], and it is also linked to periodontitis [20] and shows a positive linear relationship with kidney stone recurrence [21]. Across this wide range of applications, the ZJU Index consistently proves its efficacy in capturing the intricate metabolic shifts associated with various diseases. Specifically, it can reflect complex metabolic changes linked to diverse pathophysiological conditions, including abnormal lipid and glucose metabolism, as well as the subsequent inflammatory and oxidative stress responses. Notably, its potential to track early metabolic disorders, coupled with its integration of key CVD-related parameters (glucose, lipids, BMI), which are known to underpin metabolic-CVD crosstalk, positions it as a uniquely promising candidate for CVD risk assessment. However, its association with CVD incidence and mortality has not been systematically explored. We hypothesized that higher ZJU Index values would be associated with increased CVD risk and mortality in a general population, with clinical application facilitating early detection of CVD and improved outcomes. Materials and methods Data availability All data are publicly available, on research approval access, from the National Health and Nutrition Examination Survey (NHANES), the website (https://www.cdc.gov/nchs/nhanes/). Study participants This retrospective cohort study included 18,609 adult participants from NHANES 1999–2018, with mortality follow-up. Since 1999, NHANES has been conducted in biennial cycles, generating nationally representative data through a cross-sectional design. Our analysis included 10 biennial NHANES cycles (1999–2018), representing 20 years of nationally representative data. Consistent methodology was applied across cycles to ensure comparability. We included nonpregnant adults aged 20 years or older, and the sample was weighted to be representative of the noninstitutionalized US civilian population. Figure 1 illustrates the participant selection procedure. Exclusions were made based on the following criteria: participants lacking ZJU Index data (n = 71,199), those without mortality follow-up data (n = 5,297), pregnant participants (n = 716), and those with missing covariates (n = 5,495). After these exclusions, 18,609 participants were included in the final analysis. All participants in NHANES provided written informed consent, following approval by the NCHS Ethics Review Board. Flow diagram of the screening and enrollment of study participants Data collection Participant data on demographics, physical exams, blood tests, and medical history were collected as previously described [22]. Demographic characteristics, comprising age, sex, racial/ethnic background, educational attainment, marital status, and poverty-income ratio (PIR), were obtained through a computer-assisted personal interviewing system. The racial/ethnic distribution included five categories: (1) non-Hispanic White; (2) non-Hispanic Black; (3) Mexican American; (4) Other Race - Including Multi-Racial; and (5) Other Hispanic. Three education categories were used: below high school, high school level, and post-secondary education. Marital status included coupled and single/separated. PIR was divided into ≤ 1.3, 1.31–3.5, or ≥ 3.5. Health-related information included self-reported smoking and alcohol consumption behaviors. Smoking status was defined as never smoker, former smoker and current smoker. Drinking status (yes/no) was determined by annual alcohol consumption, with ≥ 12 drinks defining “yes” and < 12 drinks defining “no”. Assessment of ZJU index The ZJU Index was calculated using the formula: Body Mass Index (BMI, kg/m²) + Fasting Plasma Glucose (FPG, mmol/L) + Triglycerides (TG, mmol/L) + 3 * (Alanine Aminotransferase/Aspartate Aminotransferase [ALT/AST] ratio), with an additional + 2 for female participants [14]. ZJU Index = BMI (kg/m²) + FPG (mmol/L) + TG (mmol/L) + 3 * [ALT (U/L)/AST (U/L)] (+ 2 if females). This gender-specific adjustment was consistently applied in all subsequent analyses, including all regression models, subgroup analyses, and sensitivity analyses. Based on ZJU Index quartiles, the cohort was divided into Q1 (reference), Q2, Q3, and Q4 groups for comparative assessment. Definitions of outcome variables CVD case status was determined using a validated self-report methodology with a structured medical history questionnaire. Interviewers asked participants to confirm any previous physician diagnoses of the following conditions: (1) congestive heart failure; (2) coronary heart disease; (3) angina pectoris; (4) myocardial infarction; and (5) stroke. A positive response to ≥ 1 condition constituted the CVD case definition, with each component disease adhering to standard clinical diagnostic criteria. Diabetes mellitus (DM) was defined by the following criteria: self-reported physician-diagnosed diabetes, current use of hypoglycemic agents, glycated hemoglobin A1c (HbA1c) ≥ 6.5%, or random blood glucose >11.1 mmol/L. Hypertension was defined as an average of the three most recent blood pressure measurements with a systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg, or current use of antihypertensive medications. Hyperlipidemia was defined using specific thresholds: total cholesterol ≥ 200 mg/dL, triglycerides ≥ 150 mg/dL, high-density lipoprotein (HDL) ≤ 40 mg/dL in men and ≤ 50 mg/dL in women, or low-density lipoprotein (LDL) ≥ 130 mg/dL [23]. In addition, individuals who were told by a doctor that they have a high cholesterol level, or those who report taking cholesterol-lowering medications, are also classified as having hyperlipidemia. Mortality outcomes (all-cause and CVD mortality) were assessed using death certificate records from the National Death Index (NDI; https://www.cdc.gov/nchs/data-linkage/mortalityrecords_public.htm). Mortality data for each participant were obtained through linkage with the NDI, with follow-up through December 31, 2019. Using the ICD-10 classification system, we identified cause-specific mortality. CVD mortality was defined as deaths from cardiac diseases, cerebrovascular events, or hypertensive disorders, corresponding to ICD-10 codes I00-I09, I11, I13, I20-I51, and I60-I69. Statistical analysis The participants were divided into four groups according to their ZJU Index quartiles: Q1 (≤ 25th percentile), Q2 (> 25th to ≤ 50th percentile), Q3 (> 50th to ≤ 75th percentile), and Q4 (> 75th percentile). All statistical analyses were conducted in R (version 4.4.3; R Foundation for Statistical Computing) with packages including survey for complex survey design adjustment, rms for restricted cubic spline (RCS) modeling, and survival for Cox regression analyses, incorporating sampling weights, stratification variables, and cluster identifiers to adjust for the complex survey design features of NHANES. Continuous variables were presented as mean ± SD if normally distributed, otherwise as median (interquartile range [IQR]) based on the Shapiro-Wilk test. Continuous variables were compared using weighted Kruskal-Wallis tests, and categorical variables were analyzed using weighted Pearson’s chi-square (χ²) tests, accounting for the complex survey design. We employed multivariable Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% confidence intervals for time-to-event outcomes, specifically CVD mortality and all-cause mortality. For binary baseline outcomes, namely prevalent CVD, we used multivariable logistic regression to estimate odds ratios (ORs) and 95% confidence intervals. These two models assessed associations between the ZJU Index, analyzed as a continuous variable and by quartiles, and risks of CVD, CVD mortality, and all-cause mortality. Model 1 was unadjusted. Model 2 included adjustments for demographic and socioeconomic factors: age, sex, race, educational attainment, marital status, and poverty-income ratio. Model 3 additionally adjusted for behavioral and clinical covariates, including smoking status, alcohol consumption, hyperlipidemia, hypertension, and diabetes mellitus. RCS is a reliable method for analyzing nonlinear associations. In this study, RCS with four knots was fitted using the Akaike information criterion (AIC) to minimize model complexity. Sensitivity analyses were performed by varying the number of knots (3–5) in RCS models to test the robustness of nonlinear associations. We assessed multicollinearity in both logistic and Cox regression models using the Variance Inflation Factor (VIF), with a pre-specified threshold of VIF < 5 to indicate no significant multicollinearity. The VIF values of all variables in both models ranged from 1.10 to 4.05, all of which fell below this threshold, indicating that there was no significant multicollinearity in the models. In the Cox regression analysis, the proportional hazards assumption was evaluated via Schoenfeld residual tests, with the criterion that a p-value > 0.05 for both individual covariate tests and the global test indicated no violation of the assumption. Across all specifications of the Cox regression models, the p-values of all individual covariates were > 0.05, and the p-values of the global tests were also > 0.05, confirming that the proportional hazards assumption was not violated. Weighted Kaplan-Meier (KM) curves and log-rank tests were used to perform survival analyses between the ZJU Index and all-cause mortality and cardiovascular mortality. Interaction terms were tested to evaluate effect modification by key variables (e.g., sex, age groups), with statistical significance defined as P < 0.05 for interaction effects. All analyses were performed using R software (version 4.4.3). A two-tailed P < 0.05 was set as the threshold for statistical significance. Results Baseline characteristics of the participants A total of 18,609 participants were included in the study, with 9,455 (50.0% weighted) being male and 8,834 (71.3% weighted) being non-Hispanic White, and a mean age of 47.5 ± 16.7 years. These proportional data were obtained by weighting. The subjects were stratified into four groups according to ZJU Index quartiles, with significant differences in age, sex, race, marital status, education, PIR, smoking, drinking, hyperlipidemia, hypertension, and diabetes (all P < 0.001). When comparing the quartiles, subjects in Quartile 4 had higher levels of BMI, ALT, AST, fasting plasma glucose, and triglycerides, as well as higher rates of hypertension, diabetes mellitus, and CVD, than those in the other groups, suggesting a gradient of risk factors with increasing ZJU Index. The overall prevalence of CVD was 9.0%. During a median follow-up period of 71 (36–121) months, 2,752 (11.1%) all-cause deaths were recorded, of which 934 (3.4%) were cardiovascular deaths. Further details are provided in Table 1. Table 1Relationship between ZJU Index and all-cause mortality and CVD mortality among participants in NHANES 1999–2018CharacteristicOverall N = 18,609(100%)1Q1 (< 34.68) N = 4653 (27%)1Q2 (34.68–39.19) N = 4652 (25%)1Q3 (39.19–44.73) N = 4652 (24%)1Q4 (≥ 44.73) N = 4652 (24%)1P Value2Age (years)47.5 ± 16.743.4 ± 17.548.7 ± 16.949.8 ± 16.148.6 ± 15.2< 0.001Sex< 0.001 Male9,455 (50.0%)2,546 (50.0%)2,566 (54.0%)2,453 (53.2%)1,890 (42.5%) Female9,154 (50.0%)2,107 (50.0%)2,086 (46.0%)2,199 (46.8%)2,762 (57.5%)Race< 0.001 Non-Hispanic White8,834 (71.3%)2,402 (73.2%)2,308 (73.1%)2,090 (69.9%)2,034 (68.6%) Non-Hispanic Black3,621 (9.9%)857 (9.0%)795 (8.3%)851 (9.5%)1,118 (12.9%) Mexican American3,172 (7.7%)486 (4.9%)777 (7.3%)985 (9.8%)924 (9.2%) Other Race - Including Multi-Racial1,519 (6.0%)617 (8.6%)391 (6.0%)301 (4.8%)210 (4.4%) Other Hispanic1,463 (5.1%)291 (4.3%)381 (5.2%)425 (6.0%)366 (5.0%)Marital status< 0.001 Coupled11,362 (65.0%)2,587 (60.2%)2,971 (67.1%)2,999 (68.2%)2,805 (64.8%) Single or separated7,247 (35.0%)2,066 (39.8%)1,681 (32.9%)1,653 (31.8%)1,847 (35.2%)Education attainment< 0.001 < High school2,046 (5.4%)373 (4.2%)494 (5.1%)632 (6.9%)547 (5.8%) Completed high school2,707 (10.9%)641 (10.2%)625 (9.8%)693 (11.7%)748 (12.0%) >High school13,856 (83.6%)3,639 (85.6%)3,533 (85.1%)3,327 (81.4%)3,357 (82.2%)PIR groups< 0.001 ≤ 1.35,509 (20.4%)1,341 (20.4%)1,238 (17.5%)1,368 (20.1%)1,562 (23.6%) 1.3–3.57,166 (36.2%)1,726 (34.6%)1,758 (34.5%)1,848 (38.0%)1,834 (38.0%) ≥ 3.55,934 (43.4%)1,586 (45.0%)1,656 (48.0%)1,436 (41.9%)1,256 (38.4%)Smoking, n (%)< 0.001 Never9,769 (52.2%)2,404 (51.7%)2,455 (52.7%)2,452 (52.0%)2,458 (52.5%) Former4,904 (26.3%)978 (20.9%)1,261 (26.8%)1,356 (28.8%)1,309 (29.3%) Current3,936 (21.5%)1,271 (27.4%)936 (20.5%)844 (19.2%)885 (18.2%)Drinking, n (%)12,297 (69.8%)3,291 (74.3%)3,221 (72.5%)3,052 (69.6%)2,733 (62.0%)< 0.001Hyperlipidemia14,057 (74.4%)2,583 (53.8%)3,541 (76.1%)3,886 (82.6%)4,047 (87.4%)< 0.001Hypertension8,050 (37.9%)1,237 (20.9%)1,851 (34.1%)2,225 (43.1%)2,737 (55.6%)< 0.001Diabetes< 0.001 No15,472 (87.6%)4,472 (97.6%)4,212 (93.5%)3,838 (87.8%)2,950 (69.9%) Yes or Borderline3,137 (12.4%)181 (2.4%)440 (6.5%)814 (12.2%)1,702 (30.1%)CVD2,121 (9.0%)361 (5.6%)520 (8.6%)580 (10.1%)660 (11.9%)< 0.001BMI (kg/m²)28.8 ± 6.722.0 ± 2.126.3 ± 1.730.1 ± 2.237.8 ± 6.2< 0.001ALT (U/L)25.8 ± 25.820.8 ± 21.924.0 ± 16.127.6 ± 17.431.3 ± 40.2< 0.001AST (U/L)25.2 ± 18.124.8 ± 22.625.0 ± 18.025.1 ± 13.625.9 ± 16.4< 0.001Fasting plasma glucose (mmol/L)5.9 ± 1.75.2 ± 0.65.5 ± 0.85.8 ± 1.16.9 ± 2.9< 0.001Triglyceride (mmol/L)1.5 ± 1.31.0 ± 0.51.3 ± 0.71.7 ± 1.02.1 ± 2.2< 0.001All-cause mortality2,913 (11.1%)717 (9.7%)759 (11.5%)758 (12.1%)679 (11.4%)0.003CVD mortality934 (3.4%)196 (2.4%)248 (3.5%)251 (3.7%)239 (4.0%)< 0.001Continuous variables are presented as means (standard deviations) or medians (interquartile ranges), and categorical variables are presented as n (%). PIR is calculated by dividing family income by family size, year, and geographic location, as measured by the Department of Health and Human ServicesCVD Cardiovascular disease, ALT Alanine Aminotransferase, AST Aspartate Aminotransferase, BMI Body Mass Index Relationship between ZJU Index and all-cause mortality and CVD mortality among participants in NHANES 1999–2018 Continuous variables are presented as means (standard deviations) or medians (interquartile ranges), and categorical variables are presented as n (%). PIR is calculated by dividing family income by family size, year, and geographic location, as measured by the Department of Health and Human Services CVD Cardiovascular disease, ALT Alanine Aminotransferase, AST Aspartate Aminotransferase, BMI Body Mass Index Association between ZJU index and CVD Table S1 illustrates the association between the ZJU Index and CVD. A total of 18,609 individuals were included in the analysis, among whom 2,121(9.0%) were diagnosed with CVD. Compared with the non-CVD group, participants in the CVD group were generally older, predominantly male, and mostly non-Hispanic White or non-Hispanic Black. They also had lower educational attainment, a higher proportion of former smokers or drinkers, a lower family PIR, and higher levels of BMI, AST, plasma glucose, and triglycerides (all P < 0.001). Additionally, the CVD group had an increased risk of hyperlipidemia, hypertension and diabetes mellitus (all P < 0.001). Furthermore, continuous ZJU levels were significantly higher in the CVD group than in the non-CVD group (P < 0.001). When the ZJU Index was stratified into quartiles (Q1 to Q4), a significant association with CVD was also observed: the proportion of CVD cases in each quartile increased progressively from Q1 (16.7%) to Q2 (24.4%), Q3 (27.1%), and Q4 (31.7%). The relationship of ZJU Index levels with the risk of CVD was evaluated using logistic regression. As shown in Table S2, the ZJU Index, as a continuous variable, was significantly associated with CVD in both the unadjusted and adjusted models. In the fully adjusted Model 3, each 1-unit increase in the ZJU Index was associated with a 2% higher risk of CVD (odds ratio [OR]: 1.02, 95% confidence interval [CI]: 1.01–1.03, P < 0.001). When the data were stratified into quartiles, Q2, Q3 and Q4 were positively associated with CVD in the unadjusted Model. Compared to Q1 as the reference level, the analysis revealed a significant increase in the risk of CVD with the ZJU Index at Q4 (OR = 2.28, 95% CI: 1.91–2.83, P < 0.001), as shown in Model 2. This association was slightly attenuated in the fully adjusted Model 3 (additionally adjusted for behavioral and clinical covariates). The risk of CVD progressively increased with rising ZJU Index, showing a dose-response pattern (Q2 compared to Q1, OR = 1.17, 95% CI: 0.98–1.40, P = 0.086; Q3 compared to Q1, OR = 1.24, 95% CI: 1.00–1.54.00.54, P = 0.047; Q4 compared to Q1, OR = 1.50, 95% CI: 1.22–1.84, P < 0.001). Moreover, the findings from the generalized additive model (GAM) and smooth curve fitting, conducted following propensity matching, revealed that the relationship between the ZJU Index and CVD is a nonlinear relationship (P for non-linear < 0.001) (Fig. 2). Dose-response density relationship between ZJU Index and CVD. The shaded area represents the 95% CI. Each point represents the value of the ZJU Index and is connected to form a continuous line. Adjusted for age, sex, race, education attainment, marital status, PIR factor, smoking, drinking, hyperlipidemia, hypertension, diabetes We assessed ROC curves for two models in the entire cohort: a baseline risk model incorporating traditional CVD risk factors (age, sex, smoking status, hyperlipidemia, diabetes mellitus, and hypertension) and an extended model further including the ZJU Index. The baseline risk model yielded an area under the curve (AUC) of 0.827, whereas the model with the integration of the ZJU Index demonstrated an AUC of 0.829. Statistical analysis confirmed that this difference in AUC values was statistically significant (P = 0.0099) (Fig. 3). Incremental effect of the ZJU Index for predicting CVD. The baseline risk model includes age, sex, smoking, hyperlipidemia, diabetes, and hypertension; CVD, cardiovascular disease; AUC, area under curve Clinical outcomes for all-cause and CVD mortality When stratified by ZJU Index quartiles, in the unadjusted Model 1, Q2 (HR = 1.21, 95% CI: 1.07–1.40, P = 0.002), Q3 (HR = 1.32, 95% CI: 1.17–1.49, P < 0.001), and Q4 (HR = 1.34, 95% CI: 1.18–1.51, P < 0.001) were positively associated with all-cause mortality, as shown in Table 2. In Model 2 (partially adjusted for demographic factors), the association pattern changed: Q2 showed a significantly negative association with all-cause mortality (HR = 0.88, 95% CI: 0.79–0.99, P = 0.030), while the associations for Q3 (HR = 0.94, 95% CI: 0.83–1.05, P = 0.273) and Q4 (HR = 1.11, 95% CI: 1.00–1.23.00.23, P = 0.056) were weakened and no longer statistically significant. In the fully adjusted Model 3, Q3 (HR = 0.88, 95% CI: 0.79–1.00.79.00, P = 0.043) exhibited a significantly negative association with all-cause mortality, whereas Q2 (HR = 0.90, 95% CI: 0.80–1.01, P = 0.070) showed a marginally non-significant negative trend, and Q4 (HR = 0.96, 95% CI: 0.85–1.08, P = 0.493) had no significant association. Table 2Relationship between ZJU index and all-cause mortality and CVD mortality among participants in NHANES 1999–2018CharacteristicModel 1Model 2Model 3HR195% CI1p-valueHR195% CI1p-valueHR195% CI1p-valueAll-cause Mortality Continuous ZJU1.011.01–1.02< 0.0011.011.00–1.02.00.02< 0.0011.001.00–1.01.00.010.527 Quartered ZJU Q1111 Q21.211.07–1.400.0020.880.79–0.990.0300.900.80–1.010.070 Q31.321.17–1.49< 0.0010.940.83–1.050.2730.880.79–1.00.79.000.043 Q41.341.18–1.51< 0.0011.111.00–1.23.00.230.0560.960.85–1.080.493 P for trend< 0.0010.0310.316CVD Mortality Continuous ZJU1.021.02–1.03< 0.0011.031.02–1.04< 0.0011.021.00–1.03.00.030.006 Quartered ZJU Q1111 Q21.481.19–1.85< 0.0011.090.87–1.350.4741.040.84–1.300.698 Q31.621.29–2.04< 0.0011.200.95–1.510.1291.040.82–1.310.769 Q41.851.50–2.29< 0.0011.771.39–2.11< 0.0011.301.03–1.630.027 P for trend< 0.001< 0.0010.0441HR Hazard Ratio, CI Confidence IntervalModel 1: no covariates were adjustedModel 2: adjusted for age, sex, race, education attainment, marital status, PIR factorModel 3: adjusted for age, sex, race, education attainment, marital status, PIR factor, smoking, drinking, hyperlipidemia, hypertension, diabetes Relationship between ZJU index and all-cause mortality and CVD mortality among participants in NHANES 1999–2018 1HR Hazard Ratio, CI Confidence Interval Model 1: no covariates were adjusted Model 2: adjusted for age, sex, race, education attainment, marital status, PIR factor Model 3: adjusted for age, sex, race, education attainment, marital status, PIR factor, smoking, drinking, hyperlipidemia, hypertension, diabetes As shown in Table 2, the ZJU Index was significantly associated with CVD mortality across all models. Compared with Q1 (reference), Q4 was associated with a higher HR for CVD mortality in the unadjusted model. In Model 1, Q2 had an HR of 1.48 (95% CI: 1.19–1.85, P < 0.001), Q3 had an HR of 1.62 (95% CI: 1.29–2.04, P < 0.001), and Q4 had an HR of 1.85 (95% CI: 1.50–2.29, P < 0.001), with a significant trend (P for trend < 0.001). After multivariable adjustment, in Model 2, the ZJU Index at Q4 was associated with a significantly elevated CVD mortality, showing an HR of 1.77 (95% CI: 1.39–2.11, P < 0.001), and there was also a significant trend (P for trend < 0.001). In Model 3, the association remained significant for Q4, with an HR of 1.30 (95% CI: 1.03–1.63, P = 0.027), and the P for trend was 0.044, indicating a tendency of increasing CVD mortality with the increasing ZJU quartiles in a relatively stable manner across models, though the specific HR values and precision of confidence intervals varied by model adjustments. To assess the nonlinear relationships between the ZJU Index and mortality, a multivariable Cox proportional hazards regression model with RCS analysis was used. After covariate adjustment, the ZJU Index showed a significant U-shaped relationship. The U-curve reflects that the all-cause mortality hazard ratio first decreases then increases with the ZJU Index. Data points like (37.3, 1) and (40.1, 1) in Fig. 4 (A) mark ZJU Index ranges with a hazard ratio of 1 (reference risk), corresponding to relatively lowest all-cause mortality risk. Statistical tests (Pₒvₑrₐₗₗ ​< 0.001; Pₙₒₙ₋ₗiₙₑₐr < 0.001) confirmed significance. The ZJU Index was associated with CVD mortality in a nonlinear relationship. The curve in Fig. 4 (B) shows a J-shaped nonlinear trend: the hazard ratio for CVD mortality remains relatively stable at low ZJU Index levels, and starts to increase significantly once the ZJU Index exceeds approximately 39.6, with a more pronounced upward trend as the index further rises. The data points (34.7, 1) and (39.6, 1) indicate that at these ZJU Index values, the hazard ratio for CVD mortality is 1, representing the baseline risk level. Meanwhile, the results of statistical tests (Poverall < 0.001, indicating that the overall association is significant; Pnon−linear < 0.001, indicating that there is a significant nonlinear component in this association) also verify the statistical significance of the relationship between the two. Weighted restricted cubic spline regression of ZJU Index and all-cause mortality (A) and CVD mortality (B). The adjusted restricted cubic spline Mode l was adjusted for the covariates in Model 3. As shown in Kaplan–Meier survival curves (Fig. S1), there were significant differences in the occurrence of CVD mortality among the four groups (Log rank P = 0.025). Higher ZJU Index tended to be positively associated with lower survival probability for all-cause mortality, though no statistically significant differences in all-cause mortality were observed between groups (Log rank P > 0.05). Subgroup analysis This study stratified participants by age, sex, race, marital status, educational attainment, PIR groups, smoking behavior, drinking behavior, hyperlipidemia, hypertension, diabetes, CVD, and BMI in the subgroup analysis. Fig.S2 showed that the relationship between the ZJU Index and all-cause mortality (left plot) and CVD mortality (right plot) exhibited certain patterns across various subgroups. However, in subgroup analyses centered on the ZJU Index, significant interactions were identified between the ZJU Index and age (P for interaction < 0.001 for both all-cause and CVD mortality), hyperlipidemia (P for interaction = 0.030 for all-cause mortality and P = 0.031 for CVD mortality), and marital status (P for interaction = 0.011 for all-cause mortality and P = 0.024 for CVD mortality) with regard to all-cause mortality and CVD mortality. Sensitivity analysis To evaluate the robustness of these findings, we performed several sensitivity analyses using multivariate Cox regression. First, we conducted a sensitivity analysis excluding short-term deaths (deaths within 2 years), given that some participants might have been in the terminal stage of the disease at baseline—their biomarkers could reflect advanced disease status rather than predictive factors (Table S3). Second, considering the impact of a history of cardiovascular disease on all-cause and cardiovascular mortality outcomes, we excluded 2,121 participants with CVD (Table S4). Fig. S3 presents the results of weighted restricted cubic spline regression between the ZJU index and all-cause mortality as well as cardiovascular disease (CVD) mortality, which is used to compare the association patterns under the two conditions of “3 knots” and “5 knots”. All Poverall​<0.001 and Pnon−linear​<0.001, indicating that the relationships between ZJU and all-cause mortality as well as cardiovascular mortality are robust under different numbers of knots. Consistently, all analyses confirmed that the association of the ZJU Index with CVD mortality retained statistical significance, while the U-shaped relationship with all-cause mortality remained robust. Discussion Statistical analysis revealed a robust positive association between the ZJU Index and CVD. This association remained significant after adjusting for confounders, which indicates the index’s potential to reflect CVD risk independently of traditional factors. Notably, this association was non-linear with P < 0.001, which suggests distinct risk thresholds. These thresholds may guide clinical decision-making, such as targeted interventions for individuals with rapidly increasing ZJU levels. For mortality outcomes, participants in the highest ZJU quartile (Q4) had a 30% higher risk of CVD mortality compared to those in Q1. This result persisted even after multivariable adjustment, with an HR of 1.30(95% CI 1.03–1.63; P = 0.027). Meanwhile, Kaplan–Meier analysis confirmed that Q4 participants had worse CVD-specific survival, with a log-rank P = 0.025. Subgroup analyses showed broadly consistent relationships across most demographic and clinical subgroups, including those stratified by sex, race, and lifestyle factors. This consistency strengthens the index’s utility for predicting CVD mortality. In subgroup analyses, significant interactions with age, hyperlipidemia, and marital status were also observed. These interactions hint at potential nuances in association strength across specific populations. For example, the association is stronger in middle-aged adults or those with hyperlipidemia, which suggests the index has tailored utility. However, these findings do not undermine the overall robustness of the identified links. RCS analysis identified distinct non-linear patterns. The ZJU Index showed a U-shaped relationship with all-cause mortality, which suggests complex interactions with other health factors. It also exhibited a J-shaped relationship with CVD mortality; this relationship is characterized by a slight initial decrease followed by a continuous increase, which indicates a more direct link to cardiovascular-specific risks. Sensitivity analyses were conducted to validate the stability of these findings. The analyses included three approaches: excluding short-term deaths within 2 years, excluding participants with baseline CVD, and adjusting the number of RCS knots. The results of these sensitivity analyses confirmed the stability of the findings, further supporting the ZJU Index’s reliability as a CVD risk assessment tool. The ZJU Index was developed to reflect systemic metabolic dysregulation, incorporating adiposity (BMI), FPG, TG, and hepatic injury (ALT/AST ratio), all of which are closely linked to CVD risk. Specifically, elevated BMI is associated with obesity, a chronic condition that increases risks of CVD, comorbidities, and premature death [24, 25]. Increased FPG levels reflect impaired glucose metabolism, with type 2 diabetes mellitus (T2DM), undiagnosed T2DM, and other glucose disorders being established CVD risk factors [26, 27]. IR, a core feature linked to elevated triglycerides and impaired glucose metabolism, which are key components of the ZJU Index, contributes to cardiovascular risks via inflammation, oxidative stress, and lipid disorders [28–30]. Notably, IR amplifies CVD risk independently of obesity, while glucose intolerance further strengthens this association [31]. Additionally, the AST/ALT ratio is significantly linked to elevated CVD incidence in men [32]. While existing studies have confirmed the association between individual components of the ZJU Index and CVD, research on the ZJU Index itself as a comprehensive indicator for CVD risk assessment remains scarce. Prior studies have linked the ZJU Index to NAFLD [14, 17] (a condition associated with elevated cardiovascular risk [33]), and incident diabetes [16], supporting its role as a marker of metabolic dysfunction. Considering its associations with a range of metabolism-related diseases, our findings further underscore its potential as an integrated indicator of metabolic health. Compared with other metabolic indices, the ZJU Index offers distinct advantages. For instance, metabolic syndrome (MetS) criteria focus on waist circumference, blood pressure, glucose, and lipids but lack hepatic biomarkers [34], and the Fatty Liver Index (FLI) does not comprehensively integrate obesity or glucose metabolism markers [35], whereas the ZJU Index includes the ALT/AST ratio, which captures liver-metabolic crosstalk, and balances multiple metabolic dimensions. This suggests it may be a more holistic tool for assessing both metabolic and cardiovascular risk. It is widely recognized that obesity and dyslipidemia are critical risk factors for both all-cause mortality and CVD mortality [36, 37]. Notably, the non-linear associations of the ZJU Index with mortality align with the risk profiles of its individual components. BMI shows a U-shaped relationship with all-cause mortality, with the lowest risk observed at 22.5–25.0 kg/m² [38–40], and a J-shaped relationship with CVD mortality [41]; fasting plasma glucose has a J-shaped association with all-cause mortality [42]; triglycerides exhibit a U-shaped relationship with all-cause mortality [43]; and finally, prediabetes (including impaired fasting glucose) has a non-linear association with CVD mortality, with risks accelerating beyond the diabetes threshold [44]. This consistency confirms the biological plausibility of the ZJU Index’s associations with CVD and mortality. Mechanistically, the ZJU Index’s non-linear associations likely stem from synergistic interactions between its components. For example, its U-shaped relationship with all-cause mortality may reflect overlapping U-shaped risks of two components: BMI and triglycerides. Specifically, low BMI combined with impaired glucose metabolism contributes to such risks, while for triglycerides, both extremely low and high levels can drive mortality [38, 43]. The non-linear upward trend with CVD mortality may arise from the combined effects of BMI’s J-shaped risk and glucose intolerance-driven cardiovascular damage [41, 44]. Additionally, the ZJU Index’s integration of glucose metabolism and liver function markers aligns with prior findings that impaired fasting glucose and the AST/ALT ratio predict mortality risk [13, 45], further supporting its value as a comprehensive metabolic health indicator. In summary, by integrating multi-dimensional biomarkers, the ZJU Index overcomes the limitation of single indicators in capturing systemic metabolic imbalances. Its associations with CVD incidence, all-cause mortality, and CVD mortality provide a more precise means of early identification of high-risk groups for metabolic disorders. Future studies should evaluate its performance in different age groups and populations with comorbidities, and explore its links to specific metabolic processes. For instance, oxidative stress is one such process, and investigating this link could help clarify the underlying biological mechanisms. These findings hold significant clinical and public health implications. Clinically, the ZJU Index can complement existing tools for routine CVD risk assessment, and it is particularly useful for individuals with subclinical metabolic dysfunction such as prediabetes and NAFLD, supporting the early identification of high-risk groups. From a public health perspective, targeted strategies can be developed to reduce the ZJU Index through weight management, glycemic control, and liver function improvement; such strategies may help alleviate the global burden of CVD. Strengths and limitations This study has several merits. First, it drew on comprehensive and trustworthy medical data from NHANES, a nationwide prospective cohort with a long follow-up duration. Second, we adjusted for potential confounding factors in our analyses when possible and carried out subgroup and sensitivity analyses to minimize bias, thus guaranteeing the stability of our findings. Third, our study provides novel evidence regarding the prognostic value of the ZJU Index in the general population, demonstrating that maintaining the ZJU Index at an optimal level can significantly reduce the risk of CVD, all-cause mortality, and CVD mortality in the general population. Despite these strengths, our study has several limitations. First, this study relies on self-reported diagnoses from NHANES, which inherently carry limitations: such data are susceptible to nonresponse bias, measurement errors, and subjectivity. Additionally, potential inaccuracies in the diagnostic tests used in NHANES may introduce uncertainties into our findings. Second, as an observational study, it cannot establish a causal relationship between the ZJU Index and CVD or mortality. Reverse causality cannot be excluded; for example, progressive CVD damage may affect components of the ZJU Index such as hepatic enzyme levels.Third, unmeasured confounding factors such as detailed lifestyle habits including dietary patterns and physical activity intensity, genetic predispositions to CVD, and medication use including lipid-lowering or glucose-lowering drugs may have influenced the observed associations. Fourth, the NHANES dataset primarily includes U.S. populations, and the applicability of our findings to other ethnic or regional groups requires further validation. Finally, the optimal cut-off value of the ZJU Index for clinical risk stratification remains undefined, which limits its direct use in routine clinical practice—even though our findings confirm its potential as a CVD risk stratification tool. Future studies should prioritize validating optimal cut-offs across diverse populations to bridge this gap and translate its prognostic value into clinical utility. Conclusion In this nationwide prospective cohort study, we found a significant association between the ZJU Index and CVD, all-cause mortality, and CVD mortality in the general population. Maintaining the ZJU Index within an optimal range may significantly reduce the incidence of CVD, along with all-cause mortality and CVD mortality. The ZJU Index can provide clinicians with a practical tool for screening high-risk populations for CVD. Physicians should advise patients with a higher ZJU Index to adopt a healthy lifestyle, including a balanced diet and regular exercise, to manage weight, blood glucose, and blood lipids, thereby reducing their risk of CVD, as well as all-cause and CVD mortality. Continuous variables are presented as means (standard deviations) or medians (interquartile ranges), and categorical variables are presented as n (%). PIR is calculated by dividing family income by family size, year, and geographic location, as measured by the Department of Health and Human Services. CVD cardiovascular disease, ALT Alanine Aminotransferase, AST Aspartate Aminotransferase, BMI Body Mass Index. Supplementary Information Supplementary Material 1 Supplementary Material 2 Acknowledgements We would like to thank the data collection team and the NHANES administrators for granting access to their data. We also sincerely appreciate Figdraw (www.figdraw.com) for providing the platform which enabled the creation of our graphical abstract and thus enhanced the effective communication of the paper's central theme. Abbreviations Zhejiang University Cardiovascular disease Body mass index Fasting plasma glucose Triglycerides Alanine aminotransferase/aspartate aminotransferase ratio Nonalcoholic fatty liver disease Insulin resistance Obstructive sleep apnea syndrome National Health and Nutrition Examination Survey National Center for Health Statistics National death index Restricted cubic spline regression Hazard ratio Confidence interval Standard deviation Akaike information criterion Kaplan-Meier Type 2 diabetes mellitus Authors’ contributions ZZM: Responsible for investigation, data extraction, formal analysis, and writing and editing of the original manuscript; MYW and BLC: Responsible for study conceptualization, manuscript writing, review, and revision. All authors have reviewed the manuscript and approved its final version. Funding There is no funding to report. Data availability All data are publicly available, on research approval access, from the National Health and Nutrition Examination Survey (NHANES), the website (https://www.cdc.gov/nchs/nhanes/). Declarations Ethics approval and consent to participate The study protocol was approved by the National Center for Health Statistics Ethics Review Committee. All participants provided written informed consent before participating in the study. 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. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement All data are publicly available, on research approval access, from the National Health and Nutrition Examination Survey (NHANES), the website (https://www.cdc.gov/nchs/nhanes/). All data are publicly available, on research approval access, from the National Health and Nutrition Examination Survey (NHANES), the website (https://www.cdc.gov/nchs/nhanes/).