The data utilized in this study has been drawn from the NHANES database. The latter is a cross-sectional study conducted on a nationwide scale, overseen by the National Center for Health Statistics. The survey enables researchers to obtain information related to demographics, socioeconomics, diet, health, and physical examination from a sample of the US population that is representative of the whole population. The survey data can then be utilized to appraise the health and nutritional status of the population. In summary, from 2003 to 2016, relevant indicators were collected from approximately 70,000 U.S. residents. All data and indicator sources for the study can be referred to on public access.¹

Among the 71,058 subjects available for the study, the following conditions were set for data cleaning: (1) lack of data on metal exposure and data required for calculation of the ePWV index; (2) subjects under the age of 20 or pregnant; (3) missing one or more covariates. Eventually, 8,800 samples were included in the analysis to verify the association between metal exposure and ePWV. The detailed flow chart of the included subjects is shown in Figure 1.

The formula for ePWV is referenced in Greve et al. and is derived from the collaborative reference value for arterial stiffness (8). ePWV = 9.587–0.402 × age + 4.560 × 10⁻³ × age²–2.621 × 10⁻⁵ × age² × MBP + 3.176 × 10⁻³ × age × MBP - 1.832 × 10⁻² × MBP. MBP was calculated by the following formula: (DBP) + 0.4 (SBP − DBP). And ePWV was trisected as T1 (<6.88 m/s), T2 (6.88–9.20 m/s), T3 (>9.20 m/s).

Urine metal levels are favored as a noninvasive assay for large follow-up cohorts compared to serum metal levels. According to the guiding manual, the urine samples were pretreated, stored on appropriate medium and transported to the laboratory for analysis. Urine levels of 12 selected metals [including Barium (Ba), Cobalt (Co), Cesium (Cs), Molybdenum (Mo), Antimony (Sb), Tungsten (W), Uranium (U), Thallium (Tl), Lead (Pb), Cadmium (Cd), Mercury (Hg), Arsenic (As)] were measured by inductively coupled plasma mass spectrometry (ICP-MS) analysis (23). Based on these results, we extracted data of urine metal levels in the period 2003 to 2016, and all 12 metals were detected at a rate of more than 60%, and were corrected by creatinine (Supplementary Table S1). The distribution of these metals among the study participants can be seen in Supplementary Table S2.

In accordance with the previous literature, the core covariates selected for this study are outlined below: sociodemographic characteristics, including sex (male/female), age (under 40 years old/40–60 years old/over 60 years old), race (Mexican American/other Hispanic/non-Hispanic White/non-Hispanic Black/other), education level (less than 11th grade/ high school grade/ some college or above), poverty income ratio (PIR); life style and body measurement indicators, including physical activity, body mass index (BMI), TC, HDL, white blood cell counts; alcohol consumption, smoking status, total energy intake and urine creatinine. PIR (%) was divided into categorical variable according to the quartile [Q1 (<1.18), Q2 (1.18–2.20), Q3 (2.21–4.10), Q4 (>4.10)]. Physical activity was divided into three categories based on the questionnaire (never/moderate/vigorous). Smoking status was (never/former/current) based on cigarettes consumption (more than 100) and whether smoking currently. Drinking status was based on alcoholic beverage consumption. The conversion of alcohol consumption was in accordance with the previous method. The results were “No” and “Yes.” BMI was coded as continuous variables (24). Considering the correlation between arterial stiffness and related disease, we additionally regarded diabetes, hypertension, and CVD as covariates in our analysis (25).

The mean ± standard deviation (SD) was utilized to depict continuous variables conforming to a normal distribution, in order to provide descriptive information on the sociodemographic characteristics. Other variables were defined as categorical variables and are expressed as n (%). In this study, metal concentrations (μg/mg) were adjusted by urine creatinine to analyze the correlation between urine metal and ePWV. In order to circumvent the potential bias engendered by a skewed distribution, the concentrations of urine metals were transformed through log10 function in order to achieve approximate normal distribution (continuous variable), or categorized into quartiles (Q1, Q2, Q3, and Q4) as categorical variables. The internal relevance between the mixed urine metals was evaluated by Spearman’s correlation coefficient.

Multiple linear regression model was utilized to investigate the association between each urine metal and ePWV. Mediation analysis was performed to ascertain the mediating effect of serum lipid levels and white blood cell counts on the associations between urine metals and ePWV. Considering the complexity of real environmental exposure, WQS analysis was employed to investigate the joint effects of mixed metal exposures on ePWV. And the contribution of every single metal to the holistic indices effect can then be assessed through the weight assigned to each variable by the model. Additionally, we introduced the Qgcomp model to confirm the effects of each urine metal on ePWV. This was achieved by allocating positive or negative weights to each factor in the model. Restricted cubic spline (RCS) regression was employed to estimate the potential nonlinear relationship between each urine metal and ePWV. Spearman’s correlation coefficient was employed to examine relationships between urine metals.

All statistical methods have been adjusted for the aforementioned confounding variables. These analyses were executed utilizing the R software platform (R 4.1.1). In the statistical model, significance was affirmed by a two-sided p-value of under 0.05.

The basic characteristics of the participants which were involved in study are shown in Table 1. A total of 71,058 participants were collected for this study and the final number included in the analysis was 8,880, including 4,585 male and 4,295 female. Higher level of ePWV was prefer to exist in the subgroup which participants over 60 years old, male, of Non-Hispanic White, have lower level of education, with less exercise, had a history of smoking. Additionally, we observed the higher levels of ePWV were accompanied by a higher prevalence of hypertension, CVD, diabetes.

In order to investigate mutual correlations among the environmental metal exposures, we calculated Spearman’s correlation coefficients to check the correlation between pairwise urine metals. Positive but weak correlations were observed between almost all of the 12 metals in urine (Supplementary Figure S1). The highest levels of correlation were observed between U and Ba (r = 0.21, p < 0.05), Tl and Cs (r = 0.2, p < 0.05).

According to the results exhibited in Figure 2, Supplementary Table S3, we found that Cd (β = 0.012, 95% CI: 0.006, 0.017), Co (β = 0.012, 95% CI: 0.006, 0.017), Cs (β = 0.008, 95% CI: 0.001, 0.015), Pb (β = 0.024, 95% CI: 0.019, 0.029), U (β = 0.031, 95% CI: 0.012, 0.051) levels were positively correlated with the ePWV. And an inverse correlation has been observed between Sb (β = −0.009, 95% CI: −0.018, −0.001), Tl (β = −0.015, 95% CI: −0.024, −0.005), W (β = −0.008, 95% CI: −0.015, −0.002), Hg (β = −0.005, 95% CI: −0.009, −0.002) and the levels of ePWV (p < 0.05).

We conducted RCS analysis to visualize the potential nonlinear relationships between urine metals and ePWV. As exhibited in Figure 3, nine of all urine metals were significantly associated with ePWV (p-overall < 0.05), while U, Pb, Co, Cd showed positive associations with ePWV. The analysis results also revealed that Cs, Mo were positively correlated with the ePWV and in a nonlinear mode (p-non-linear < 0.05).

Our findings indicated that the WQS indices exhibited a substantial correlation with ePWV. As shown in Figure 4A, the upper quartile of the WQS indices was significantly related to elevated ePWV (β = 0.022, 95% CI: 0.017, 0.028), in which U (β = 0.15), Pb (β = 0.48), Co (β = 0.15), and Cd (β = 0.1) were the dominating contributors to the positive correlation (p < 0.01).

Unlike other mixed exposure models, Qgcomp analysis does not require all weight indices to be aligned in the same direction. Our findings showed mixed metals exposure were significantly and positively associated with ePWV (β = 0.022, 95% CI: 0.016, 0.027). For the levels of ePWV, the urine level of Pb (β = 0.4) showed the strongest positive correlation, followed by Cd (β = 0.18), Co (β = 0.16) and Cs (β = 0.10) (Figure 4B). When it comes to negative relationship between urine metals and ePWV, TI, Hg, W, Sb exhibited a negative tendency with ePWV.

To explore potential underlying mechanisms between metal exposure and ePWV, a parallel mediation analysis was conducted to evaluate the mediating effect of serum lipid levels and white blood cell counts on the aforementioned association. Parallel mediation analysis demonstrated that HDL and TC exerted a weak but significant mediation effect on the association between mixed metals WQS indices and ePWV, and the proportion of mediation was 2.14 and 3.70% (Table 2). The mediated proportion of HDL on the associations between Co, Cs, Pb, U, As and ePWV was 1.50, 5.73, 2.29, 5.29 and 18.82% (p < 0.05). TC similarly mediated the association between Ba, Cd, Co, Cs, Pb, U and ePWV, and the proportion of mediation was 11.88, 6.44, −6.04%, 5.55, 4.91 and 6.40% (Supplementary Table S4).