Associations of heavy metal mixtures with blood pressure among U.S. adults in NHANES 2017–2018 by four statistical models

To the Editor: Hypertension is one of the leading risk factors for cardiovascular disease and the global goal of the World Health Organization for the prevention of non-communicable diseases.[1] Hypertension, as a primary modifiable cause of premature death, affects approximately 30% of the adult population.[2] Environmental heavy metal pollution is an important public health concern. Toxic metals may increase the risk of hypertension or other cardiovascular diseases, even beyond the influence of conventional behavioral risk factors.[3] Meanwhile certain essential trace elements, such as selenium, zinc, and iron, are involved in various enzyme reactions directly related to blood pressure regulation and are beneficial to the human body within a specific range.[4] The analysis of health outcomes specific to metal mixtures is challenging by nonlinear relationships and interactions between toxic metals and essential trace elements.[5] The development of innovative statistical methods for analyzing mixtures has greatly facilitated the exploration of the health effects of multiple pollutant exposures, such as Bayesian kernel machine regression (BKMR) and quantile g-computation (QG-C). BKMR can be used to estimate the exposure-response function, including nonlinear exposure-response curves, linear functions with main effects and their interaction, as well as nonlinear functions incorporating a synergistic interaction between them. The QG-C model considers the nonlinearity and nonadditivity of individual exposures effects and the overall mixture, enabling valid inference on the individual contributions to that metal mixture by calculating the weight of each element in each direction. In this study, we adopted four novel statistical models, multivariable linear regression, elastic-net regression (ENET), BKMR, and QG-C to evaluate the associations and interactions between individual metals and metal mixtures with blood pressure and to identify important elements that may have the largest contribution to blood pressure alteration in the U.S. general population from the National Health and Nutrition Examination Survey (NHANES) data. We collected data from participants aged ≥20 years from the 2017–2018 survey cycle. We excluded participants with incomplete data on blood pressure, blood and urinary metal concentrations, and covariate information and eventually included 1348 individuals [Supplementary Figure 1, https://links.lww.com/CM9/B831]. A comprehensive set of 5 metals in the blood, namely, lead (Pb), cadmium (Cd), manganese (Mn), mercury (Hg), and selenium (Se), and 15 metals in the urine including arsenic (As), chromium (Cr), Hg, nickel (Ni), barium (Ba), Cd, cobalt (Co), cesium (Cs), molybdenum (Mo), Mn, Pb, antimony (Sb), stannum (Sn), thallium (Tl), and tungsten (W), were measured using inductively coupled plasma mass spectrometry (ICP-MS) (Agilent Technologies, Santa Clara, CA, USA). All determinations of blood pressure were taken in the mobile examination center (MEC) and were performed by trained inspectors using the Omron HEM-907XL digital electronic upper arm measuring device (Omron Healthcare, Kyoto, Japan). Descriptive analyses were performed to examine the baseline characteristics of all the participants. Continuous variables were calculated by the mean and standard deviation (SD), and the frequency and percentage were used to describe categorical variables. Blood and urinary levels of metals were naturally log-transformed to reduce skewness in analyses. In the preliminary analysis, the multivariable linear regression model and ENET model were used to examine the associations between blood pressure (systolic blood pressure [SBP] and diastolic blood pressure [DBP]) and metal concentrations while adjusting for all covariates, namely, age, sex, body mass index (BMI), smoking, alcohol use, race, and education. Smokers were defined as those smoking at least 100 cigarettes during their lifetime. Drinkers were defined as those drinking at least once a month for more than six months. In further analysis, we constructed BKMR and QG-C models to estimate the effects of the metal mixtures on blood pressure for comprehensive and accurate statistical analyses. All statistical analyses were performed in R (version 4.0.2, R Development Core Team, https://www.r-project.org/). P-values <0.05 were considered statistically significant. The general characteristics of the study population are shown in Supplementary Table 1, https://links.lww.com/CM9/B831. Among the participants included in this study, the mean age was 51.7 (17.1) years and 51.2% (690/1348) were male. The mean BMI was 29.9 (7.2) kg/m2. In addition, 47.0% (633/1348) of subjects were smokers, and approximately 55% (740/1348) were drinkers. Non-Hispanic White individuals accounted for the largest proportion (36.6%, 493/1348). A majority of the subjects had some college or AA degree (33.2%, 448/1348). Supplementary Table 2 (https://links.lww.com/CM9/B831) presents the distribution of the concentrations of metals in the blood and urine. Blood Se showed a significant positive association with DBP in both the model 1(β: 25.12, 95% confidence interval [CI]: 13.88–36.35) and model 2 (β: 24.11, 95% CI: 12.72–35.50). Both blood Cd (β: 2.46, 95% CI: 0.39–4.53) and urinary Cd (β: 3.03, 95% CI: 0.76–5.30) were positively associated with DBP in the model 2 [Supplementary Table 3, https://links.lww.com/CM9/B831]. Interestingly, blood Mn is negatively associated with SBP, but positively associated with DBP in the multi-metal ENET model [Supplementary Figures 2 and 3, https://links.lww.com/CM9/B831]. Blood Se, Cd, and Pb showed a linear association with increased SBP. Blood Se and Pb showed a non-linear exposure-response relationship with DBP [Supplementary Figure 3A,B, https://links.lww.com/CM9/B831]. When the other four metals were set to the 25th, 50th, or 75th percentile, blood Se was significantly and positively associated with DBP levels [Supplementary Figure 3D, https://links.lww.com/CM9/B831]. The blood metal mixture was significantly associated with increased DBP overall in the BKMR model [Supplementary Figure 4, https://links.lww.com/CM9/B831]. There were similar results that the metal mixture was significantly and positively associated with SBP (β = 3.47, 95% CI: 1.39–5.54) and DBP (β = 2.66, 95% CI: 1.61–3.70) levels in QG-C model [Supplementary Figure 5, https://links.lww.com/CM9/B831]. Besides, blood Pb contributed 78.7% positive weight with SBP, and blood Se was assigned 57.7% positive weight with DBP. [Supplementary Figure 5C, D, https://links.lww.com/CM9/B831]. More and more evidence shows that high level of Se is a risk factor for hypertension. Cross-sectional studies have shown that chronic overexposure to environmental Se increases blood pressure and the prevalence of hypertension.[6] However, a cohort study found that high Se intake may be beneficial for hypertension prevention.[7] These inconsistent results may be explained by the different concentration ranges of Se, the race of the subjects, and their genetic backgrounds. Cd, a toxic element that is widely distributed in the environment, has been extensively studied for its health hazard to multiple systems in the human body. Based on animal and in vitro studies, Cd may increase blood pressure through affecting the expansion and contraction of blood vessels, such as inhibiting the endothelial nitric oxide synthase in blood vessels,[8] besides oxidative stress may also involve in the mechanism of Cd-induced hypertension.[9] Pb is a typical toxic heavy metal. Pb poisoning contributes to cardiac and vascular damage and induces hypertension through oxidative stress-mediated limitation of NO availability.[10] However, the underlying biomechanism still needs to be further studied. Mn widely exists in nature and is an essential trace element required for human health. Deficiency of Mn may increase oxidative stress and accelerate vascular cell proliferation and vasoconstriction.[11] Mn may involve in the mitochondrial damage and the interaction with calcium channels in the cardiovascular system.[12] Evidence also showed a protective effect of Mn on blood pressure. A cohort study in Bangladesh found that certain levels of Mn have a blood pressure lowering effect. There is also evidence of a protective effect of Mn on blood pressure. A cohort study in Bangladesh found that certain levels of Mn may have a blood pressure-lowering effect.[13] However, the causal relationships of Mn with SBP and DBP still need to be further investigated. Although many studies have evaluated the relationship of individual metal exposure with blood pressure, few studies have examined the combination or interaction effects of multiple elements associated with blood pressure in adults. Our findings showed that metal mixtures have a positive joint effect on blood pressure, which was consistent with the effects of multiple metals on hypertension in epidemiological studies of Chinese populations.[14,15] The mechanisms underlying the development of environmental metal-induced hypertension remain unclear. Oxidative stress, impaired nitric oxide (NO) signaling, a modified vascular response to neurotransmitters, disturbed vascular smooth muscle Ca2+ signaling, renal damage, and interference with the renin-angiotensin system might be the mechanisms underlying hypertension induced by metals associated with some metals.[9] In summary, the metal mixture may increase blood pressure, Se and Pb contribute the largest weight with DBP and SBP. Interestingly, blood Mn may have a protective effect on SBP and an adverse effect on DBP. However, further epidemiological studies and biological mechanism studies are warranted to confirm our findings. Acknowledgment We thank the investigators, the staff, and the participants of NHANES for their valuable contribution. Funding This study was supported by grants from the R&D Program of Beijing Municipal Education Commission (No. KM202210025026) and Young Elite Scientist Sponsorship Program by BAST (No. BYESS2023385). Conflicts of interest None. Availability of data and materials A full list of data sets supporting the results in this research article can be found at: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2017