This cross-sectional study focuses on children and adolescents aged 8–17 years. The data was obtained from U.S. NHANES cycles spanning from 2007 to 2016 (five consecutive NHANES cycles). The NHANES conducts a cross-sectional survey using a complex multistage probability design to collect data from the noninstitutionalized U.S. population. This involves conducting household interviews and physical examinations. Prior to participation, all individuals provided written informed consent, and the study protocol received approval from the NCHS Research Ethics Review Board. Out of the total of 9,386 participants, 7,162 had incomplete data, resulting in a final unweighted sample size of 2,224 individuals. The process of data acquisition is depicted in Supplementary Figure S1.

In our study, participants who underwent BP measurements at the Mobile Examination Centers (MECs) were instructed to sit in a seated position with their feet flat on the floor and rest for a duration of 5 min. Trained researchers then performed three consecutive BP measurements using a mercury manometer and an appropriately sized cuff, which were subsequently averaged to obtain the final result (21).

The diagnostic criteria for hypertension were based on the 2017 clinical practice guidelines and previous studies (17, 22). The specific criteria for diagnosis include the following: (1) For children aged 8–12 years, hypertension is diagnosed when their systolic and/or diastolic blood pressure exceeds the 95th percentile compared to children of the same age, gender, and height. Alternatively, a diagnosis of hypertension can also be made if their SBP exceeds 130 mmHg and/or their DBP exceeds 80 mmHg. Conversely, for adolescents aged 13–17 years, hypertension is diagnosed if their SBP exceeds 130 mmHg and/or their DBP exceeds 80 mmHg (22); or (2) Irrespective of the BP level, patients aged ≥16 or the parents/guardians of patients aged <16 report the patient’s diagnosis of hypertension or the use of antihypertensive drugs (17, 23).

In the NHANES, most metals were detected in urine samples rather than blood samples. The non-invasive, sensitive, and prompt detection capabilities of urine have increasingly positioned it as an alternative method for metal detection over blood samples. Therefore, all metal detection data in this study was based on urine samples. Upon arriving at the MECs, study participants were instructed by the coordinator to provide urine samples. Subsequently, urine samples undergo processing and analysis using inductively coupled plasma-mass spectrometry (ICP-MS) to determine the concentrations of 12 elements, including Ba, Cd, Co, Cs, Mo, Pb, Sb, TI, W, U, Hg, and As.¹ Comprehensive instructions for laboratory methods utilized to measure the urinary metal concentrations can be found on the NHANES website.² In our statistical analysis, the concentrations of urinary metals were used as a variable after being transformed with natural logarithm (Ln). Approximately 70% of the cleared As from the blood is excreted through urine. Thus, we opted to assess As exposure by measuring the total concentration of As in urine, rather than focusing on speciated As. Moreover, these data pertain to the overall healthy population, and restricting the analysis to speciated As may lead to an underestimation of long-term exposure levels. Nonetheless, this study conducted further analysis on speciated As, namely arsenobetaine (AsB), arsenic acid (As(V)), arsenocholine (AsC), arsenous acid (As(III)), monomethylarsinic acid (MMA), and dimethylarsinic acid (DMA). Exclusion of trimethylarsine oxide from the analysis was due to the unavailability of subject data for NHANES 2013–16 cycle (24).

Based on previous literature, several covariates were extracted as potential confounding factors in this study (17, 25, 26). The selected covariates included age, sex, race/ethnicity (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, and other race), family poverty income ratio (PIR, categorical variables: <1.3, 1.3–3.5, >3.5 denote low, middle and high income, respectively), serum creatinine (an indicator of renal function), urinary creatinine (Detection method: Enzymatic Roche Cobas 6,000 Analyzer) and serum cotinine levels (25) (considered to reflect exposure to environmental cigarette smoke). Physical activity was fell into three groups (never, moderate or vigorous, and no record) according to self-reported activity intensity. Consistent with a previous study, the population was divided into three age groups: 8–10 years old, 11–13 years old, and 14–18 years old, representing primary school, junior high school, and senior high school, respectively (27). Additionally, the consumption of fish in the past 30 days (Yes or No) was included as a covariate in the model to account for the impact of dietary factors on urinary heavy metal concentrations. Body mass index (BMI) was calculated using the formula weight (in kilograms) divided by the square of height (in meters). Underweight was an age-and gender-specific BMI below the 5th percentile on the 2000 Centers for Disease Control and Prevention (CDC) age-and gender-specific growth charts, normal weight was a BMI below the 85th percentile but at or above the 5th percentile, overweight was a BMI falling between the 85th and 95th percentiles, and obesity was a BMI at or above the 95th percentile (28).

Additionally, all participants in the NHANES are eligible for two separate 24-h dietary recall interviews. Nonetheless, fewer participants had completed two 24-h dietary recalls. Consequently, the research evaluated the daily intake of total energy, calcium, sodium, and potassium using data from the first recall (23), which were then incorporated as covariates in our analysis.

We utilized the NHANES weighting guidelines to weigh the analysis results. Our study incorporated sample weights, sampling units, and strata provided by NHANES. We combined two groups of 5 years to conform two periods for the calculation of a new multiyear sample, ensuring that the results represent the nationwide non-institutionalized civilian population aged 8–17 years. In descriptive analysis, means ± standard deviation (SD) and counts (percentage) are applied to, respectively, describe quantitative and qualitative data. Spearman’s rank correlation analysis was performed to examine the correlations of urinary toxicant concentrations. Subsequently, due to the presence of values below the detection limit for certain individuals, urinary toxicant concentrations were categorized into four quartiles (Q1, Q2, Q3, Q4) as categorical variables, with the quartile containing the lowest metal concentrations serving as the reference group. Survey-weighted logistic regression and survey-weighted multiple linear regression models were employed to calculate odds ratios (ORs), β, and 95% confidence intervals (CIs). The false-discovery rate (FDR) correction was applied to adjusted for errors resulting from multiple testing in regression models.

The WQS approach has been widely utilized to investigate the cumulative effect of environmental mixtures on health outcomes and to assess the contribution of individual metals (29, 30). The urinary metals composing a weighted index were divided into quartiles and then applied in the estimation of empirically deduced weights and a final WQS index by bootstrap sampling (31, 32). This WQS index denotes the cumulative effect of all urinary toxicants on BP. The weights sum to 1 and range from 0 to 1, and they can be applied to identify important urinary metals (the average weight surpasses the threshold of 1 divided by the total number of independent variables) in the mixture (33). The WQS index was constructed from the quartiles of urinary metals, with 40% of participants in this study divided into the test set and 60% into the validation set. Initially, we examined whether the relationship between heavy metal concentration and BP is influenced by body weight by adjusting the BMI category in the WQS model. Subsequently, we divided the samples into two groups depending on fish consumption in the past 30 days to investigate the impact of fish consumption on the association between heavy metals and BP. Lastly, to investigate the impact of heavy metals on BP in children and adolescents across different age groups, we categorized the age of pediatric patients into seven time intervals based on previous research (34). The first group comprised individuals aged 8–14, the second group included those aged 9–15, and so on, in order to analyze the trend of the effect of heavy metal concentration on BP with changing age.

In our sensitivity analysis, we initially considered the potential non-linear and non-additive relationships among urine metals. Bayesian kernel machine regression (BKMR) was employed to assess the combined effects of all metals and the dose–response relationship between individual metals and BP when fixing other metal concentrations. Secondly, participants were excluded if their urine samples were categorized as either diluted (urine creatinine <30 mg/dL, n = 139) or concentrated (urine creatinine concentration > 300 mg/dL, n = 63), according to previous study (26, 35). Subsequently, WQS analysis was conducted.

All statistical analyses were performed with R statistical software (V.4.4.0),³ and a two-sided p value <0.05 was considered statistically significant. The R packages gWQS and nhanesR were applied to construct the WQS model, weighted logistic regression and multiple linear regression, respectively.

Supplementary Figure S1 illustrates the process of acquiring the study population. Table 1 and Supplementary Table S1 present the weighted general characteristics and urinary metal concentrations of 2,224 children and adolescents aged 8–17 years from NHANES 2007–2008 to 2015–2016. The weighted average age of the participants was 12.81 ± 0.08 years, with male and female proportions of 50.63 and 49.37%, respectively. The overall prevalence of hypertension was 10.48% (n = 233). Most of the participants were Mexican American and had attained higher education levels. Among hypertension patients, there was a higher proportion of individuals who were overweight or obese, had increased daily energy and sodium intake, and a lower proportion of those with unavailable serum creatinine values. Additionally, both systolic and diastolic blood pressures were elevated, while no significant differences were observed in other general characteristics. In comparison to children and adolescents without hypertension, hypertensive patients exhibited lower concentrations of Co (0.56 μg/L, 95%CI, 0.49, 0.63), Cs (4.37 μg/L, 95%CI, 3.97, 4.78), TI (0.18 μg/L, 95%CI, 0.16, 0.20), and As (8.74 μg/L, 95%CI,6.62, 10.86) in their urine samples, whereas there were no significant differences in the concentrations of Ba (2.38 μg/L, 95%CI, 1.82, 2.94), Cd (0.08 μg/L, 95%CI, 0.07, 0.10), Mo (75.31 μg/L, 95%CI, 65.28, 85.33), Pb (0.38 μg/L, 95%CI, 0.32, 0.43), Sb (0.08 μg/L, 95%CI, 0.07, 0.09), W (0.19 μg/L, 95%CI, 0.15, 0.22), U (0.02 μg/L, 95%CI, 0.00, 0.03), and Hg (0.38 μg/L, 95%CI, 0.29, 0.46). The proportions of urinary Cd and Hg concentrations exceeding the limits of detection (LOD) were the lowest, at only 65.47 and 76.35%, respectively, whereas the proportions of other urinary metals exceeding the LOD were all above 80% (Supplementary Table S1).

Supplementary Figure S2 displays weak to moderate correlations (−0.02 ≤ r ≤ 0.67) among all toxic metals, as calculated using Spearman’s rank correlation analysis. Cs and TI exhibit the strongest correlation (r = 0.67, p < 0.05), followed by Mo and W, while the correlation between Cd and W is the weakest. Consequently, it may be necessary to construct a multi-pollution model to detect the impact of toxic metals on BP.

Survey-weighted logistic regression and multiple linear regression models were utilized to investigate the association between Ln-transformed urinary metal concentrations and BP. These models were adjusted for selected potential confounding factors. The results presented in Supplementary Table S2 indicate a null significant correlation between urinary metal concentrations and hypertension (all p for trend >0.05). Supplementary Table S3 revealed a negative association between an increase in urinary concentrations of Pb (p for trend =0.036), Hg (p for trend =0.036) and SBP. Moreover, an increase in Ln-transformed Mo concentration was associated with a decreasing trend in DBP (p for trend <0.001). However, for the remaining urinary metal concentrations, no significant trend effects on BP were observed in the single-metal models.

The WQS regression model was employed to examine the impact of mixed metals on BP. After adjusting for all selected confounding factors, null association was observed between low concentrations of urinary metal mixtures and the risk of hypertension (OR_index: 0.08, 95%CI: −0.31, 0.47, p = 0.681) (Table 2; Figure 1A). Conversely, these mixtures were correlated with lower SBP (β_index: −0.67, 95%CI: −1.24, −0.10, p = 0.002) and DBP (β_index: −0.59, 95%CI: −1.06, −0.12, p = 0.036) (Table 2). Metal mixtures primarily affected SBP through Pb (23.62%), As (19.22%), Hg (18.62%), and Co (18.52%) (Figure 1B), while the effects on DBP were primarily attributed to Cs (24.48%), Mo (15.22%), U (14.74%), and Co (11.24%) (Table 2; Figure 1C). Furthermore, excluding the degree of obesity from the model yielded similar results to the obesity-corrected model (Figures 1D–F). Subgroup analyses were conducted to determine how fish consumption influenced the impact of heavy metals on BP. The analyses were based on participants’ fish consumption in the past 30 days. The results indicated that heavy metal exposure did not significantly affect BP values in the subgroup without fish consumption (Figures 1G–I). However, in the subgroup with fish consumption, heavy metals were associated with lower SBP (β_index: −3.30, 95%CI: −4.73, −1.87, p ≤ 0.001) (Table 2; Figures 1J–L). The main contributors were Hg (27.61%), Cd (27.49%), Cs (17.98%), and TI (8.49%) (Table 2; Figure 1K).

Following the consideration of potential confounding factors, the impact of urinary heavy metals on the BP of children and adolescents across various age groups was investigated using the WQS regression model. The findings are outlined as follows: within the 10–17 age group, a 25th percentile increase in the WQS index corresponded to a 1.48 mmHg (95% CI, −2.66, −0.30) reduction in SBP. Meanwhile, in the 11–18 age group, each 25th percentile rise in the WQS index led to reductions of 1.42 mmHg (95% CI, −2.44, −0.40) in SBP and 2.62 mmHg (95% CI, −4.00, −1.22) in DBP, contributing to a 0.42 times (95% CI, −0.81, −0.03) decrease in hypertension risk. Notably, no significant association between heavy metal exposure and BP was observed among the 8–15 and 9–16 age groups (Table 3).

To examine the association between speciated As and BP, we adopted the WQS to model these categories, comprising AsB, As(V), AsC, As(III), MMA, and DMA (Supplementary Table S4; Figures 2A–C). A 25th percentile increase in the WQS index was associated with a decrease in SBP by 0.88 mmHg (95% CI, −1.62, −0.13), predominantly influenced by As (III) (40.00%) and AsC (34.04%) (Supplementary Table S5).

The results of the BKMR model reveal that when utilizing the median concentration (50th percentile) of all metals as the reference exposure level, a concentration of the mixture at or above the 55th percentile is associated with a downward trend in both systolic and diastolic blood pressure (Figures 2D–F). Furthermore, regardless of the percentile level (25th, 50th, or 75th) of other metals, Ba exhibits a significant association with elevated SBP (Supplementary Figure S3). The analysis yielded consistent results with the preliminary analysis, not just in the subgroups analyzed based on fish consumption, but also among participants with a urine creatinine concentration ranging between 30 mg/dL and 300 mg/dL (Supplementary Table S6; Supplementary Figures S4, S5).