Abstract

INTRODUCTION

Cardiovascular diseases (CVD) are the leading cause of mortality worldwide, contributing to nearly a third of all deaths in 2019 [1], and are projected to remain a major burden of disease over the next decades [2]. Risk factors for the onset of CVD in adulthood include high blood pressure and dyslipidemia [3]. Although overt CVD does not typically emerge until adulthood, growing evidence indicates that the development of CVD is a lifelong process that begins in childhood [4-6]. Identifying modifiable risk factors in childhood is therefore important for informing early risk-reduction and prevention programs [7, 8].

In addition to behavioral factors, such as poor diet [9] and physical inactivity [10, 11], exposures to environmental contaminants [12], such as metals and metalloids (collectively referred to as “metals” hereafter) [13-15], contribute to an increased incidence of CVD and related mortality; in fact, the American Heart Association recently published a scientific statement highlighting the important role that nonessential metals play in the development of CVD [16]. Essential metals may also contribute to cardiovascular risk, with potential benefits at low levels of exposure and possible adverse effects at higher concentrations [17-22].

Human exposure to metals is widespread [13, 23], with common sources of exposure including contaminated drinking water, food, and air pollution [24, 25]. Although the mechanisms by which metal exposures affect CVD risk are not yet clear, increases in oxidative stress, inflammation, and endocrine disruption are hypothesized to play a role [14, 26, 27]. Emerging evidence indicates that metal exposures also contribute to dyslipidemia [28-39]. However, prior studies examining these relationships have largely focused on single metal exposures in adult populations [40-43].

Given that atherosclerosis begins in early life, improving understanding of metal impacts on CVD risk factors during childhood may identify early opportunities for public health interventions that benefit health across the life course. Despite accumulating literature supporting impacts of both toxic and essential metals on blood pressure in childhood [44-49], to our knowledge only one prior study has investigated metal mixture impacts on lipid levels in children [49]. However, this study focused on metal exposures during the prenatal period. In contrast, impacts of metal mixture exposures during the postnatal period, which is also a critical window for child development, remain largely unknown [50].

To address this knowledge gap, we assessed both the joint and individual impacts of multiple metal exposures on lipid concentrations in early childhood, including total cholesterol, triglycerides, high-density lipoprotein (HDL) cholesterol, and low-density lipoprotein (LDL) cholesterol. To account for possible non-linear exposure-response relationships and interactions between mixture components, we used a flexible environmental mixture modeling approach. We hypothesized that toxic metals are associated with higher levels of total cholesterol, triglycerides, and LDL cholesterol, but with lower levels of HDL cholesterol. We additionally hypothesized that essential elements have a U-shaped relationship with total cholesterol, triglycerides, and LDL cholesterol and an inverse U-shaped relationship with HDL cholesterol. The present study was conducted in the Rhea mother-child cohort study, which is based in Greece, a country that has a particularly high burden of cardiovascular mortality compared with other European countries [51, 52].

METHODS

RESULTS

Characteristics of the 291 mother-child pairs included in the analytic sample are presented in Table 1. At delivery, the mean age of participating mothers was 30.1 years (SD = 5.1), with approximately one-third of the mothers being college graduates. Approximately one third of mothers reported smoking during their pregnancy, with 11.2% smoking in early pregnancy only and 20.3% smoking throughout the pregnancy. Additionally, approximately half of the children (47.4%) were exposed to ETS at 4 years of age, and 17.5% reported consuming fish or seafood at least twice per week. On average, participants consumed red meat, eggs, and dairy 2.5, 2.4, and 27.6 times per week, respectively. Although participants in the analytic sample were more likely to be highly educated and to have consumed more fish/seafood compared with participants excluded from the analysis, other characteristics were comparable (Table S2).

Table 1.

Descriptive characteristics of the 291 mother-child pairs from the Rhea cohort.

Characteristic	Mean ± SD or N (%)
Maternal characteristics
Maternal age at delivery (years)	30.1 ± 5.1
Maternal educational attainment
<9 years	39 (13.4)
9–12 years	158 (54.3)
>12 years	94 (32.3)
Parity
Nulliparous	123 (42.3)
Parous	168 (57.7)
Smoking during pregnancy
None	189 (68.5)
During early pregnancy only	31 (11.2)
Throughout pregnancy	56 (20.3)
Child characteristics
Sex
Male	160 (55.0)
Female	131 (45.0)
Gestational age at birth (weeks)	38.2 ± 1.6
Secondhand tobacco smoke exposure at 4 years
Yes	138 (47.4)
No	153 (52.6)
Regular fish/seafood intake at 4 years
Consumed fish/seafood ≥2 times per week	51 (17.5)
Consumed fish/seafood <2 times per week	240 (82.5)
Red meat consumption (times per week)	2.5 ± 1.2
Egg intake (times per week)	2.4 ± 1.8
Dairy intake (times per week)	27.6 ± 12.1
Supplement use	20 (7.1)
Total cholesterol (mg/dL)	157.0 ± 27.0
Low-density lipoprotein (mg/dL)	93.5 ± 23.3
High-density lipoprotein (mg/dL)	49.3 ± 11.4
Triglycerides (mg/dL)	71.1 ± 29.4

Open in a separate window

Blood metal concentrations at 4 years of age and the percentage of measurements below the LOD for each metal are shown in Table 2. Spearman rank correlation coefficients (P < 0.05) were calculated for each metal pair (Fig. S2). Correlations between most of the blood metal pairs were weak (<0.40) and not statistically significant. The strongest correlation was between Se and Hg (rho = 0.35, p < 0.01).

Correlations between each pair of lipid measures were generally positive and statistically significant, except for the correlation between HDL cholesterol and triglycerides (rho = −0.35, p < 0.01) and between HDL cholesterol and LDL cholesterol (rho = −0.02, P > 0.05; Fig. S3). The strongest correlation was between total cholesterol and LDL cholesterol (rho = 0.91, p < 0.01).

There was a positive association between the metal mixture and total cholesterol levels (Table 3 and Fig. S4). Setting all metals simultaneously to their 75th compared with 25th percentiles was associated with an 8.27 mg/dl (95% credible interval [CI] = 0.07, 16.47) higher total cholesterol level, accounting for maternal education, child sex, child ETS exposure at 4 years of age, and child fish/seafood intake at 4 years of age. Selenium contributed the most to this association (PIP = 0.91), followed by Cd (PIP = 0.27; Table 4). We observed potential evidence of an interaction between Se and Cd, such that the positive association between Se and total cholesterol was stronger at lower (10th percentile) compared with higher (90th percentile) concentrations of Cd (Fig. S4).

When examining the joint association of the metal mixture with other lipid measures (LDL cholesterol, HDL cholesterol, triglycerides), we observed a similar relationship for LDL cholesterol as that for total cholesterol in both magnitude and direction (7.13 mg/dl difference [95% CI = 0.52, 13.73] when setting all metals to their 75th versus 25th percentiles; Table 3 and Fig. 1). Consistent with findings for total cholesterol, the highest PIP for LDL cholesterol was estimated for Se (0.95), followed by Cd (0.71; Table 4). When evaluating associations for individual metals, accounting for the rest of the mixture, we observed a difference in LDL cholesterol of 8.22 mg/dl (95% CI = 1.85, 14.59) when setting Se to its 75th compared with 25th percentile. The potential antagonistic interaction identified visually between Se and Cd for total cholesterol was also observed for LDL cholesterol (Fig. 1). No associations were identified between the metal mixture and either HDL cholesterol (Table 3 and Fig. S5) or triglycerides (Table 3 and Fig. S6).

Open in a separate window

Fig. 1

Joint associations of the metals with low-density lipoprotein at 4 years of age, estimated by BKMR (n = 291).

Mn manganese, Co cobalt, Se selenium, Mo molybdenum, Cd cadmium, Hg mercury, Pb lead. All models were adjusted for maternal education, child sex, child secondhand smoke exposure at 4 years of age, and child fish/seafood intake at 4 years of age. A Difference in the outcome when setting all metals in the mixture to different percentiles compared with the reference (50th percentile) with corresponding 95% credible intervals. B Exposure-response functions and corresponding 95% credible intervals for each metal while holding the other exposures at their 50th percentiles. C Exposure-response function for each metal exposure (column) with the second exposure (row) fixed at its 10th, 50th, and 90th percentiles while holding the other exposures at their 50th percentiles.

When using quantile g-computation, we found that a simultaneous one decile increase in all metals was associated with a 3.79 mg/dl higher total cholesterol level (95% CI = 1.45, 6.14; Table S3 and Fig. S7). A similar association was observed for LDL cholesterol ($\psi =3.19$; 95% CI = 1.13, 5.24). Se was assigned the largest positive weight for both total cholesterol (0.46) and LDL cholesterol (0.38; Table S4). Co was assigned the largest negative weight (0.53 for total cholesterol and 0.71 for LDL cholesterol), followed by Cd (0.48 for total cholesterol and 0.29 for LDL cholesterol). The metal mixture was not associated with either HDL cholesterol ($\psi =0.27$; 95% CI = −0.66, 1.19) nor triglyceride levels ($\psi =1.69$; 95% CI = −0.67, 4.05).

BKMR results were similar when using 200,000 MCMC iterations versus 10,000 MCMC iterations. We therefore used 10,000 MCMC iterations for secondary analyses. Effect estimates remained very similar after adjusting for red meat intake (8.44 mg/dl difference in total cholesterol [95% CI = 0.04, 16.84] when setting all metals to their 75th versus 25th percentiles; Tables S5, S6 and Figs. S8-S11). Similarly, results were robust after additional adjustment for egg intake (9.18 mg/dl difference in total cholesterol [95% CI = 1.23, 17.13] when setting all metals to their 75th versus 25th percentiles; Tables S7, S8 and Figs. S12-S15), dairy intake (9.60 mg/dl difference in total cholesterol [95% CI = 0.50, 18.70] when setting all metals to their 75th versus 25th percentiles; Tables S9 and S10 and Figs. S16-S19), and dietary supplement use (8.24 mg/dl difference in total cholesterol [95% CI = −0.03, 16.51] when setting all metals to their 75th versus 25th percentiles; Tables S11, S12 and Figs. S20-S23). Selenium remained the primary contributor (PIP > 0.90) to the joint impact of metal exposures on lipid levels after additionally accounting for these dietary variables. In sex-stratified analyses, the joint associations between the metals and both total cholesterol and LDL cholesterol appeared stronger, although not statistically significant, among male children (Table S13 and Figs. S24, S25). Among male children, selenium was identified as the most important contributor to the metal mixture’s association with total cholesterol (PIP = 0.91) and LDL cholesterol (PIP = 0.90; Table S14). Metal mixture associations with HDL cholesterol (Fig. S26) and triglycerides (Fig. S27) were null, consistent with the unstratified results. We did not observe strong evidence of a joint association between the metal mixture and lipid measures among female children (Tables S13, S14; Figs. S28-S31). When stratified by child’s ETS exposure, the combined association of the metals with lipid measures (especially LDL cholesterol) was stronger among children exposed to ETS at 4 years of age (Table S15; Figs. S32-S35) compared with children not exposed to ETS at 4 years of age (Table S15; Figs. S36-S39). Among the ETS-exposed children, selenium was consistently assigned the largest PIP for total cholesterol (0.91) and LDL cholesterol (0.78) (Table S16). While a potential antagonistic interaction between Se and Cd was identified among male children for both total and LDL cholesterol (Figs. S24, S25), we did not observe any evidence of this interaction when stratifying by child’s ETS exposure (Figs. S32-S35). The association between the metal mixture and child’s BMI at age 4 was null (difference in BMI when setting all metals to their 75th compared with 25th percentiles: 0.14 kg/m²; 95% CI: −0.19, 0.48 kg/m²).

DISCUSSION

In a well-characterized cohort in Greece, we examined the joint associations of seven metal concentrations in blood with serum lipid levels in early childhood. Two different mixture modeling approaches, BKMR and quantile g-computation, consistently showed positive associations between the metal mixture and both total cholesterol and LDL cholesterol. These associations were primarily driven by selenium. In secondary analyses, we observed that the positive associations between selenium and both total cholesterol and LDL cholesterol were stronger among male compared with female children and for children exposed to ETS during childhood compared with unexposed children. We did not observe evidence that any of the metals examined, either individually or as a mixture, were associated with HDL cholesterol or triglycerides.

Selenium is an essential element that supports redox homeostasis, immune system function, fertility, and thyroid hormone metabolism [93, 94]. The main route of selenium exposure in the general population is diet [95-97], with major dietary sources including breads, grains, cereals, Brazil nuts, red meat, poultry, fish, and eggs [62, 98-100]. Although protective effects were initially hypothesized for selenium intake and cardiometabolic diseases [74], recent findings from both observational studies and randomized controlled trials have indicated that high concentrations of selenium, measured in serum/plasma [21, 101-105] or estimated dietary intake [106, 107], may lead to excess risk of cardiometabolic diseases [21, 101-107], including hyperlipidemia [108], as has been reported for other outcomes such as cancer, type 2 diabetes, and neurodegenerative disorders [109, 110]. While the potential mechanisms linking high selenium intake to dysregulated lipid metabolism are not clear, insulin sensitivity [111], inflammation [112], oxidative stress [113, 114], liver damage [115, 116], and increased LDL-receptor mRNA expression and activity [117, 118] have been hypothesized as possible mediators.

In the current study, we observed a difference in total and LDL cholesterol of 8.17 mg/dl (95% CI = 0.58, 15.76) and 8.22 mg/dl (95% CI = 1.85, 14.59), respectively, when setting selenium to its 75th compared with 25th percentile, while holding the other metals constant. Prior studies have reported that abnormal cholesterol levels in childhood contribute to cardiometabolic diseases in adulthood [119, 120]. For example, a recent study conducted among Finnish participants, aged 9–24 years, concluded that the risk of cardiovascular events in mid-adulthood increased by 25% for each 1 SD increase in LDL cholesterol (equivalent to 53.46 mg/dL) [120]. As an essential element with a narrow therapeutic window, the optimal range of selenium intake, particularly for children, is not yet known and currently debated [121]. Notably, findings from our study suggests potential harmful effects of selenium on cholesterol levels in childhood at concentrations lower than what has been reported by previous studies conducted in China (mean ± standard deviation blood selenium concentration = 128 ± 178 μg/L) [122] and Mexico (mean ± standard deviation = 250 ± 45 μg/L) [49]. Our findings highlight the need for more research on the optimal range of selenium concentrations for young children across different populations.

To the best of our knowledge, this is the first study to investigate the joint associations of multiple metals exposures with lipid measures in early childhood. Our results are consistent with several prior studies that examined selenium individually in relation to lipid measures among children. For example, a previous study [123] including children aged 6–12 years in the U.S. National Health and Nutrition Examination Survey (NHANES) reported a positive association between serum selenium concentrations (mean = 183 μg/L; range = 109, 327 μg/L) and total cholesterol levels. Although prior studies have most frequently used serum selenium as a biomarker of selenium exposure, both whole blood and serum selenium concentrations are reasonable biomarkers of selenium status and correlate well with dietary selenium intake (rho = 0.78 and 0.74, respectively) [58, 124]. Consistent with our findings using whole blood selenium as a biomarker of exposure, a cross-sectional study in China [122], which included participants 4 to >60 years of age, also reported a positive association between selenium concentrations in whole blood and total cholesterol levels. Furthermore, they observed that elevated whole blood selenium concentrations were associated with an increased odds of abnormal total cholesterol levels. Notably, the whole blood selenium cutoff used in that study (≥133 μg/L) exceeded the maximum whole blood selenium concentration observed in our study (123 μg/L). Importantly, these prior studies did not consider metal co-exposures. We are aware of only one prior study [49], conducted in Mexico, which investigated the joint association of prenatal metal mixture exposures (antimony, As, Cd, cesium, chromium, Co, Cu, Mn, Pb, Se, and Zn), measured in whole blood and including selenium (mean ± standard deviation = 250 ± 45 μg/L), with HDL cholesterol and triglycerides in children of a similar age (4–6 years). Consistent with our findings, results were null for HDL cholesterol. However, in contrast with our null findings for triglycerides, higher prenatal selenium was associated with lower triglycerides when accounting for metal co-exposures. This inconsistency could be attributed to the different developmental window investigated, differences in metal exposure levels, the specific metal co-exposures considered, or other population differences.

A growing body of evidence supports that the health effects of selenium may differ by sex. This has been reported not only for lipids [74, 75], but also for blood pressure [125], mortality [126], cognitive performance [127], sepsis [128], and microalbuminuria [129]. For example, an analysis using data from the 2007–2014 NHANES reported a U-shaped relationship between selenium intake and the total cholesterol/HDL cholesterol ratio among women only [75]. However, in a study of healthy, non-smoking blood donors in Italy, a stronger positive association between urinary selenium and LDL cholesterol levels was observed in male participants [74]. Although the potential mechanisms underlying sex differences in selenium’s health impacts are currently unknown, possible explanations include reduced susceptibility to oxidative stress among women [114, 130] and increased translational activity of selenoproteins in the male liver [131, 132].

To our knowledge, this is one of the first studies to identify a potential antagonistic interaction between selenium and cadmium in relation to lipid measures in childhood. Cadmium is a toxic metal [133], which ranks very highly in the Agency for Toxic Substances and Disease Registry Substance Priority List due to both its toxicity and the high potential for human exposure [134]. While prior literature has suggested that cadmium exposure may increase lipid levels [135, 136], we observed an unexpected inverse association between cadmium exposure and blood lipid concentrations during childhood. To our knowledge, no prior studies have investigated potential interactions between cadmium and selenium in relation to lipid profiles, although antagonistic interactions between selenium and cadmium have been reported for other outcomes in prior animal and in vitro studies [137-140]. Multiple potential mechanisms by which selenium and cadmium may act antagonistically have been proposed. For example, selenium can reduce cadmium-induced oxidative stress and can also form complexes with cadmium that reduce its toxicity [140-149]. The interaction that we identified is therefore biologically plausible. However, it is important to acknowledge that this interaction was no longer observed after stratifying by child’s ETS exposure, which is likely a source of cadmium exposure for this population. We, therefore, cannot rule out the possibility that this interaction was a chance finding or a finding driven by residual or unmeasured confounding.

Our results indicated a stronger, positive association between selenium and LDL cholesterol among children exposed to ETS compared with unexposed children. While the prevalence of smoking among adults in Greece has substantially declined since the early 2010s [150], the adult smoking rate is still very high compared with other European countries, with 23.6% of the adult population smoking daily. Secondhand smoke exposure is a particular concern in Greece where more than 87% of children are exposed to tobacco smoke in the home [151]. Prior studies have linked ETS exposure in childhood to detrimental health outcomes, including premature atherosclerosis and dyslipidemia [152]. A growing number of studies have also reported that relationships between selenium concentrations and health outcomes [153-156], including cancer [153, 154], differ by smoking status. One potential biological mechanism that may explain this is elevated oxidative stress [157, 158] due to both excess selenium [113, 114] and tobacco smoke exposure [79-84], which may induce liver damage [159] and thereby interrupt cholesterol homeostasis, leading to an accumulation of cholesterol [160]. Another possible explanation is increased arsenic toxicity due to tobacco smoke exposure, which can exacerbate the impact of selenium exposure [161]. This finding highlights that ETS-exposed children may be particularly vulnerable to metal-induced cardiovascular risk and should be prioritized in future public health interventions.

In Greece, the main sources of metal exposure include contaminated drinking water (lead) [162], dietary sources such as fish and seafood (mercury) [163], and secondhand tobacco smoke exposure (cadmium) [164, 165]. After accounting for metal co-exposures, we did not observe strong evidence that cobalt, manganese, molybdenum, cadmium, mercury, or lead were associated with lipid levels in early childhood, despite previous reports of their cardiometabolic effects in children and adolescents [49, 123, 166-170]. For example, higher concentrations of mercury, lead, and cadmium have each been associated with higher levels of cholesterol or other cardiometabolic risk factors in participants spanning 6–19 years of age, although most of these studies assessed these metals individually [123, 166-170]. The discrepant findings for these metals could also be due to differences in the age range of participants, the statistical analysis methods employed, the metal biomarkers selected, the distributions of metal concentrations, and the covariates that were considered.

The current study had several limitations that should be acknowledged. First, due to the cross-sectional design, the temporality of the relationship between the metal mixture and lipid measures cannot be established. It will therefore be important to confirm these findings in a future prospective study. Another important limitation is that serum lipids were measured using non-fasting blood samples due to the young age of the children. Nevertheless, while fasting blood samples are generally used for lipid assessment, cholesterol screening results from a non-fasting group of children were found to be similar to those from a fasting group, suggesting that non-fasting blood is reasonable for lipid assessment in children [171]. Another limitation is that total selenium concentrations were measured in whole blood; we therefore do not have measures of specific selenium species which might have differing effects on lipid levels [172, 173]. Assessing potential differential effects of selenium species on child lipid levels is therefore an important future direction. The generalizability of this study’s results may also be limited to pediatric populations with similar blood metal concentrations. Finally, given the observational nature of the study, we cannot rule out the possibility of residual confounding. For example, even though we accounted for fish and seafood consumption in our models, which is a known source of many metals [174, 175] that is known to influence cholesterol levels [176], these measures were based on dietary information rather than biomarkers of fish and seafood intake, such as urinary arsenobetaine concentrations [177, 178].

There are also several notable strengths of the study. To our knowledge, this is the first study to examine the impacts of metal mixture exposures during the postnatal period on lipid levels in childhood. Using BKMR enabled us to estimate potential non-linear, non-additive joint relationships between multiple metals and lipid levels in childhood, and results were robust when using another mixture modeling method (quantile g-computation). Additionally, by studying young children (4 years of age) we have identified a potential target group for very early cardiovascular health interventions. The high prevalence of ETS exposure in the current study also allowed us to investigate whether relationships between metal mixture exposures and lipid levels in childhood differ by ETS exposure status. To determine whether child BMI may mediate the relationship between metal co-exposures and lipid levels in early childhood, we conducted a secondary analysis examining the relationship between the metal mixture and child BMI. This association was null, which suggests that the association observed between selenium and lipid levels in childhood is not likely due to metal-related increases in child BMI.

In conclusion, our findings indicate that blood selenium concentrations are associated with higher LDL cholesterol levels in early childhood, with potentially stronger associations among male children and among children exposed to ETS. This has important public health implications, as higher LDL cholesterol in early life may predispose children to adverse cardiometabolic outcomes later in life [55]. Given that CVD accounts for ~45% of all deaths in Greece, our findings support the development of early prevention programs – including dietary and environmental health interventions—that target modifiable risk factors for CVD, such as hypercholesterolemia [51]. Importantly, while selenium is an essential element, its optimal range in children is currently unknown. This work, as well as future studies, will therefore be critical for establishing that optimal range. In particular, large-scale longitudinal studies which span different populations and selenium exposure ranges and account for the potential modifying effects of sex and ETS exposure, will be essential for designing the most effective public health interventions. Further studies are also warranted to elucidate the biological mechanisms underlying selenium’s impacts on lipid levels and other cardiometabolic risk factors in childhood.

Supplementary Material

ACKNOWLEDGEMENTS

Footnotes

REFERENCES