Heavy Metal(loid) Body Burden in Environmentally Exposed Children With and Without Internalizing Behavior Problems

Abstract

The prevalence of internalizing behavior disorders in children is increasing. Reasons for increasing anxiety and depression include several factors with a less studied consideration being the potential neurotoxic effects of environmental exposures. One group at risk for environmental exposures is children living near coal-burning power plants with coal ash storage facilities. Multivariate logistic regression was used to assess the relationship between metal(loid) exposures and internalizing behaviors in children aged 6–14 years. Metal(loid)s in nail samples were determined by Proton-Induced X-ray Emission and internalizing behavior problems were obtained from the parent ratings on the Child Behavior Checklist. Results indicated that concentrations of metal(loid)s in nails differ between children with internalizing behaviors and without internalizing behaviors. Logistic regression models suggested that exposure to zinc and imputed zirconium were associated with internalizing behaviors in children. However, when a sex-metal(loid) interaction term was included, none of the metal(loid)s were associated with internalizing behaviors indicating a role of sex differences in neurotoxicity with zinc and copper showing effects only for males. In all models, greater exposure to traffic was associated with internalizing behaviors. Zinc has previously been shown to increase risk for mental health problems, while zirconium has received less attention. Out findings indicate that environmental exposures of zinc and zirconium deserve further attention in studies of childhood internalizing disorders.

Introduction

Mental health problems in children and adolescents are often categorized as either externalizing or internalizing (Achenbach and Edelbrock 1981). Externalizing behaviors, such as aggressiveness and conduct disorder are often directed outward, whereas internalizing behaviors, such as affective problems depression, anxiety, and somatic complaints are often directed inward and thus can be more difficult to detect (Forns et al. 2011). Thackery and Harris (2003) defined internalizing disorders as “mental disorders with primary symptoms that involve inner emotions as opposed to outward behavior.” Research has shown that males are more likely to display externalizing behaviors (Campos et al. 2014; Ghandour et al. 2019), whereas females are more likely to display internalizing behaviors (Chaplin and Aldao 2013; Schäfer et al. 2017).

Risk factors for internalizing behaviors in children and adolescents include genetics, medical illness, societal stressors, and environmental factors (Ghandour et al. 2019). Research on the role of environmental factors in the development of internalizing behaviors in children is limited (Brook et al. 2008; Liu et al. 2014; Yolton et al. 2008; Chastang et al. 2015; Yang et al. 2018; Park et al. 2020) Some research on the impact of heavy metal(loid)s in children indicates concern for overexposure to arsenic, antimony, cadmium, copper, lead, manganese, and mercury leading to ADHD, conduct disorder, anxiety, internalizing behaviors, externalizing behaviors, and depression. (Menezes-Filho et al. 2014; Rahman et al. 2017; Winter and Sampson 2017; Lee et al. 2018; Soetrisno and Delgado-Saborit 2020; Lamoureux-Tremblay et al. 2020; Li et al. 2020; Munoz et al. 2020). Likewise, studies in adults have shown that exposure to these are associated with a variety of mental health conditions such as depression, and bipolar disorder (Orisakwe 2014; Scinicariello and Buser 2015; Mengmei et al. 2018).

Several biologic mechanisms have been suggested to increase risk of depression and anxiety from heavy metal(loid) exposure. Most neurobiological studies of internalizing disorders have addressed effects of lead in children which may apply to other heavy metal(loid)s. Studies of lead have shown changes in catecholaminergic and serotonergic neurotransmitter systems that have been related to depressive behavior in an animal model (Sprowles et al. 2018) Lead can also promote oxidative stress and neuroinflammation, based on both human and animal models (reviewed in Metryka et al. 2018) which has been linked to depression and anxiety (Hurley and Tazabi 2013). A third mechanism by which heavy metal(loid) overexposure may alter brain function with neurobehavioral consequences is through epigenetic mechanisms impacting DNA function based on findings from animal models (reviewed in Ijomone et al. 2020). The various mechanisms that heavy metal(loid) exposure may impact brain function indicate a need to further research mental health in children at risk of environmental exposures.

Children living near power plants and coal ash storage facilities may be at risk for greater exposure to metal(loid)s. Coal ash is a waste product, generated when coal is burned for energy. In the United States, over 78 million tons of coal ash was produced in 2019, making it one of the largest industrial waste streams in the country (ACAA 2019). Coal ash is predominately composed of iron, aluminum, silicon, and oxygen; however, it may contain trace concentrations of neurotoxic metal(loid)s such as manganese, arsenic, lead, and chromium. Both stack emissions and fugitive dust from landfills (EPA 2010, 2015; Mueller et al. 2013) can impact residential communities. Long-term exposure may affect child behavior.

Toenails and fingernails are effective as biomarkers of metal(loid)s and other elements. Concentrations of trace elements in nails reflect long-term exposure (3–12 months) and are easy to collect and store, especially from children. Nails are unaffected by other metabolic activities in the body and have some of the highest concentrations of trace elements due to the keratin. (Barbosa et al. 2005; Button et al. 2009; He 2011; Clarkson et al. 2012; Gutierrez-Gonzalez et al. 2019).

The main objective of this study was to assess differences in heavy metal(loid) concentration in nail samples, between children with internalizing behaviors and those without internalizing behaviors, who live near power plants with coal ash storage facilities. Furthermore, we assessed the association between heavy metal(loid)s and internalizing behaviors in children. We hypothesized that children with higher levels of metal(loid) exposure would display higher scores on the internalizing behavior disorders scale as measured by the Child Behavior Checklist (Achenbach and Rescorla 2001).

Methods

Study Area

This study took place in Jefferson County, Kentucky which is home to two power plants that are approximately 10 miles apart. Both power plants, which sit on the Ohio River, stored coal ash in ponds and landfills on the plant’s property, within close proximity to neighborhoods (See Fig. 1). The Cane Run Generating Station began operation in November 1954 and burned coal until 2015, when it converted to natural gas. This plant had one main ash pond that was opened in 1972 and stored fly ash, bottom ash, and other materials (Louisville Gas and Electric (LG&E) 2010). The pond had a surface area of approximately 50 acres with a capacity of 2 million cubic yards (AECOM Tech Services 2020). In August 2017, the main ash pond was closed and capped. In addition to the capped pond, Cane Run had a large coal ash landfill that was opened in the early 1980s (LG&E 2010). This landfill, which stored fly ash and stabilized scrubber sludge is now capped.

Fig. 1.

Location of the two power plants in Jefferson County, Kentucky, and the participants enrolled in study

The Mill Creek Generating Station began operation in the early 1970s and continues to burn coal today. It is the largest coal-fired power plant owned by Louisville Gas and Electric (LG&E 2021). The plant’s main coal ash pond sits on over 40 acres and stores and estimated 6.4 million cubic yards of ash materials (Lockheed Martin 2009; LG&E 2016). The pond has been given a high hazard rating by the Environmental Protection Agency, indicating that collapse of the pond can lead to loss of life or major damage to dwellings, buildings, railroads, or important utilities (United States Environmental Protection Agency (EPA) 2009). In the 1980s, Mill Creek opened a coal ash landfill. As of 2019, it was estimated to have a volume of 13.495 million cubic yards and an elevation of 598 feet (LG&E 2019).

Participants of the Study

The participants of this study included 261 children aged 6–14 years old who lived within 10 miles of the centroid between the two power plants. Geographical Information Systems was used to successfully recruit participants from five buffer zones surrounding the plants (Odoh et al. 2019). The parents/guardians and their children provided written assent/consent before the study began. Exclusion criteria included children who did not live in the study area for a minimum of two years and children who had a genetic disorder known to cause neurobehavioral problems.

Nail Samples and Analysis

Nail samples were collected as participants were enrolled in the study, from fall 2015 to spring 2020. All children that completed the study lived at the same address during their nail collection. Nail sample collection has been detailed (Zierold et al. 2021). In brief, participants were provided with stainless steel nail clippers and instructions on how to correctly clip nails. Children provided an initial sample during the first visit and continued until a minimum of 150 mg of combined toenail and fingernail samples were collected. Nail collection in participants varied between 3 weeks to 6 months. Once 150 mg of nails were collected, they were washed with one acetone and two deionized water washes and dried on a paper towel. Nail samples were stored in round, closed, plastic containers at room temperature until they were taken to the laboratory for analysis by Proton Induced X-ray Emission (PIXE).

PIXE is an analytical method that can determine the concentrations of 72 inorganic elements across the periodic table of elements, from sodium to uranium. It is a non-destructive technique that is useful when elements are unknown or at low concentrations (Lowe 1993; Ishii 2019). Although less utilized than other analytical methods for metal(loid) analysis in nail samples, it has been used for determining trace elements in nails (Biswas et al. 1984; Favaro 2013; Olabanji et al. 2005). Before PIXE analysis, the nail samples were cryogenically frozen and then ground and mixed with a binding agent called Somar-Mix Powder #210 which is a mixture of boric acid and water. The grounded nails were bound into a 3/8-inch pellet and analyzed by PIXE to ascertain the elements and concentration of the elements in the nails.

Behavioral Assessment

Assessment of children’s internalizing behaviors was completed using the parent form of the Child Behavior Checklist (CBCL), which assesses behaviors associated with anxiety, depression, and mood regulation (Achenbach and Rescorla 2001). The CBCL is one of the most widely used instruments to assess behavior problems in children. It has been translated into more than 90 languages and has shown to have good internal consistency (Chronbach’s alpha for internalizing behaviors ranges from 0.686 to 0.746) (Guttman-nova et al. 2008). The parent completed behavior survey consists of 113 questions that are scored of a three-point Likert scale (0 = absent, 1 = occurs sometimes, 2 = occurs often).

The CBCL’s questions are associated with problems that fall into eight syndrome scales: anxious/depressed, withdrawn/depressed, somatic complaints, social problems, thought problems, attention problems, rule-breaking behavior, and aggressive behavior. Anxious/depressed, withdrawn/depressed, and somatic complaints are broadly categorized as internalizing behaviors. Overall, the CBCL yields scores for internalizing and externalizing behaviors, total problems, and six Diagnostic and Statistical Manual of Mental Disorders (DSM)-oriented subscales. Based on age and sex, these scores are compared to clinical cut-off points for the particular comparison group. For this study, children who ranked in the “borderline” range (score of 60–63) or “clinical” range (score of 64 +) of the internalizing behavior scale were considered to have internalizing behavior problems.

Statistical Analysis

Analyses were performed with SAS software, version 9.4 (Cary, NC). To assess the characteristics of the population, Chi-Square tests were used. If fewer than 10 observations were in a category, Fisher’s Exact Test was used. Analysis was stratified by whether children had internalizing behavior problems or not. Since the prevalence of internalizing behavior problems frequently differ by sex, additional analysis was conducted by sex. Normality of the metal(loid)s was assessed using the Shapiro–Wilk test. When variables were normally distributed, unpaired t tests were used to assess differences in the means. When variables were not normally distributed, comparison between medians were done using the Mann–Whitney U Test. A p-value of 0.05 was considered statistically significant.

Multivariate analysis was conducted using logistic regression with a CBCL cut-off score of 60 for borderline/clinical internalizing problems. Copper, zinc, and zirconium were considered for analysis based on the results from univariate analysis and based on the significant concentration differences in the nail samples. Due to sample size concerns, zirconium was modeled as a binary variable in which participants were grouped based on whether or not participants had zirconium detected in their nails. Zirconium was also modeled as a continuous variable by imputing $\frac{\mathrm{L}\mathrm{O}\mathrm{D}\left(\mathrm{Z}\mathrm{r}\right)}{\sqrt{2}}$ when zirconium concentrations were missing. This model was purely exploratory due to having to impute data for more than seventy percent of participants. Metal concentrations were transformed via natural log to rein in outliers. However, three participants were dropped from the zinc model because their zinc concentrations were more than four times higher (x > 350 ppm) than the median concentration (87 ppm) prior to transformation. These observations were noticeable outliers after log transformation. Models were adjusted for sex, age, dichotomized race (white and non-white), and dichotomized traffic exposure (above median exposure or less than or equal to the median).Traffic exposure data were collected from the EPA’s Environmental Justice database. Traffic proximity is measured as the average annual daily traffic at major roads within 500 m from each block group and is then divided by the distance to those roads (in meters) (EPA 2015). Dichotomized house age was considered as a parameter in the models as a proxy for lead exposure but was ultimately not included. There was not a significant association in any of the models between internalizing problems and house age, and removing the term did not noticeably affect the other parameters. Furthermore, there were concerns regarding overfitting the logistic models, as there were 62 participants with borderline/clinical internalizing problems. Models were fit with and without interaction terms between sex and metal(loid) concentrations. p-values were adjusted by the false discovery rate to control for the potential of multiple comparisons.

Results

There were 260 children aged 6–14 years old who provided demographic information and CBCL data. Of those 260, 62 (24%) had internalizing behavior problems. Table 1 reports the demographics of the children. There were no sex differences (p = 0.454) or race differences (p = 0.072) in children with internalizing behavior problems. Children who were older were more likely to experience internalizing behavior problems (p = 0.016), compared with younger children. Children with traffic exposure above the median were more likely to have internalizing behavior problems (p = 0.028). House age, dichotomized as pre-1978 and post-1978 as a measure of possible lead exposure, was not associated with internalizing behavior problems (p = 0.889).

Table 1.

Demographics of study participants (N = 260)

Characteristic	Children with borderline or clinical internalizing behaviorsa (n = 62)	Children with no internalizing behaviorsa (n = 198)	p-value
Sex			0.454b
Female	27 (43.6%)	97 (49.0%)
Male	35 (56.5%)	101 (51.0%)
Age, years (median, IQR)	12.0 (10, 14)	11.0 (8, 13)	0.016c
Race			0.072d
White	51 (82.3%)	138 (69.7%)
Other	11 (17.7%)	60 (30.3%)
Traffic			0.028b
Median or less	23 (37.7%)	105 (53.9%)
Greater than median	38 (62.3%)	90 (46.2%)
House age			0.889b
Post-1978	19 (31.2%)	61 (32.1%)
Pre-1978	42 (68.9%)	129 (67.9%)

^aMeasured by the Internalizing Scale of the Child Behavior Checklist

^bChi-square test

^cWilcoxon-Mann-Whitney test

^dFisher’s Exact test

Table 2 reports the heavy metal(loid) concentrations in nails by internalizing behaviors. Children with internalizing behaviors had higher concentrations of zirconium (p ≤ 0.01) and borderline higher concentrations of zinc (p = 0.07) and tin (p = 0.07), compared to children without internalizing behaviors.

Table 2.

Heavy metal concentrations in children with internalizing behavior problems compared with children without internalizing behavior problems (N = 260)

Element	Children with internalizing behavior (n = 62)			Children without internalizing behavior (n = 198)			p-value
	N	Median	25th, 75th percentile	N	Median	25th, 75th percentile
Aluminum	52	148.6	(117.8, 229.3)	177	146.2	(110.4, 211.1)	0.38
As	3	2.500	(1.53, 11.3)	14	1.66	(1.50, 1.99)	0.15
Copper	62	5.26	(4.05, 7.50)	198	4.84	(3.63, 6.43)	0.30
Cr	33	5.90	(5.00, 7.15)	112	6.36	(4.70, 9.20)	0.45
Iron	62	70.0	(50.0, 108.2)	198	69.3	(47.0, 107.0)	0.94
Mn	14	3.68	(3.24, 4.30)	40	3.95	(3.30, 6.05)	0.17
Ni	45	2.02	(1.34, 3.20)	133	2.10	(1.45, 3.10)	0.76
Se	7	1.80	(1.70, 2.06)	20	1.65	(1.48, 1.86)	0.12
Sr	5	4.61	(4.20, 5.20)	10	3.47	(2.90, 3.97)	0.10
Sn	4	507.4	(375.6, 658.4)	17	266.1	(233.5, 440.0)	0.07
Ti	29	29.0	(23.0, 35.8)	94	24.0	(17.0, 34.0)	0.10
Zinc	64	90.2	(80.8, 99.2)	198	85.2	(75.0, 97.0)	0.07
Zr	19	12.7	(8.90, 14.0)	50	8.35	(6.57, 10.2)	< 0.01

Although in this study, there were no sex differences in children who had internalizing behaviors, Table 3 reports concentrations of heavy metal(loid)s by sex to assess if there were differences in the nail concentrations of children by internalizing behaviors. Table 3 reports these differences. In males with internalizing behaviors, concentrations of copper (p = 0.05), zinc (p = 0.03), and zirconium (p = 0.04) were increased compared to males without internalizing concerns. Among females, only zirconium differed between those with internalizing behavior problems and those without these behaviors (p = 0.02).

Table 3.

Heavy metal(loid) concentrations by sex in children with and without internalizing behaviors (N = 260)^a

	Children with internalizing behaviors			Children without internalizing behaviors			p-valueb
	n	Median	25th, 75th percentile	n	Median	25th, 75th percentile
Males (N = 136)	n = 35 (25.7%)			n = 101 (74.3%)
Aluminum	30	178	119, 233	87	139	103, 193	0.13
Arsenic	3	2.50	1.53, 11.3	5	1.50	1.20, 1.82	0.10
Copper	35	5.28	4.09, 10.1	101	4.49	3.50, 6.44	0.05
Chromium	21	6.70	5.02, 7.60	62	6.30	4.70, 8.60	0.83
Iron	35	81.3	52.6, 127	101	71.0	49.0, 99.5	0.25
Manganese	9	4.10	3.42, 4.60	20	3.83	3.20, 6.13	0.67
Nickel	25	1.7	1.34, 3.90	68	1.90	1.29, 2.88	0.75
Selenium	3	1.73	1.65, 1.97	12	1.72	1.50, 2.00	0.77
Strontium	3	4.61	3.20, 5.20	6	3.15	2.90, 3.54	0.30
Tin	3	587	323, 730	10	392	250, 521	0.24
Titanium	13	25.0	22.0, 29.0	33	21.1	15.0, 25.9	0.22
Zinc	35	89.0	79.0, 102	101	82.0	71.0, 92.3	0.03
Zirconium	10	12.8	8.90, 13.6	25	8.70	7.30, 9.60	0.04
Females (N = 124)	n = 27 (21.8%)			n = 97 (72.2%)
Aluminum	22	142	117, 226	90	149	113, 233	0.84
Arsenic	0	0	0	9	1.72	1.60, 1.99	-
Copper	27	5.07	3.8, 6.4	97	5.30	3.80, 6.21	0.57
Chromium	12	5.70	4.55, 6.93	50	7.14	4.65, 10.2	0.18
Iron	27	57.4	44.0, 83.9	97	67.0	44.0, 114	0.22
Manganese	5	3.50	2.50, 3.80	20	4.02	3.54, 6.05	0.09
Nickel	20	2.08	1.31, 3.15	65	2.12	1.47, 3.41	0.48
Selenium	4	1.93	1.75, 2.46	8	1.59	1.43, 1.75	0.06
Strontium	2	5.60	4.20, 7.00	4	3.94	3.30, 4.44	0.17
Tin	1	428	428, 428	7	254	211, 291	0.13
Titanium	16	35.4	25.4, 46.2	61	27.0	19.0, 36.8	0.08
Zinc	27	92.3	82.0, 97.8	97	87.5	79.4, 101	0.61
Zirconium	9	10.0	9.15, 14.0	25	8.10	6.20, 10.9	0.020

^aOne participant did not provide their sex

^bThe significance of the difference between participants within each sex was tested using the Kruskal-Wallis test

Table 4 reports the results from the logistic regression without the sex-metal(loid) interaction term. Copper (p = 0.21) and dichotomized zirconium in nails (p = 0.20) were not significantly associated with internalizing behaviors. Zinc (p = 0.02 and imputed zirconium (p = 0.02) were both significantly associated with borderline/clinical internalizing problems in the logistic model. A one-unit increase in the natural log of zinc increases the log odds of clinical internalizing problems by 2.08 after adjusting for covariates. A one-unit increase in the natural log of imputed zirconium increases the log odds of borderline/clinical internalizing problems by 0.47, controlling for other covariates. Both associations remain significant when adjusting for the false discovery rate (p < 0.05).

Table 4.

Multivariate logistic regression without interactions

Parameter	Copper			Zinc			Zirconium in Nails			Zr (imputed)
	Estimate (SE)	95% CL	p-valueb	Estimate (SE)	95% CL	p-valueb	Estimate (SE)	95% CL	p-valueb	Estimate (SE)	95% CL	p-valueb
(Intercept)	− 3.65 (0.95)	(− 5.58, − 1.84)	<0.001	−12.98 (4.34)	(− 21.70, − 4.59)	0.003	− 3.08 (0.76)	(− 4.62, − 1.64)	<0.001	− 3.60 (0.82)	(− 5.27, − 2.06)	<0.001
Ln(metal)/Zr in nailsa	0.33 (0.26)	(− 0.20, 0.84)	0.21	2.08 (0.88)	(0.37, 3.82)	0.02	0.44 (0.34)	(− 0.24, 1.10)	0.20	0.47 (0.21)	(0.07, 0.88)	0.02
Sex (male)	0.19 (0.31)	(− 0.41, 0.80)	0.53	0.36 (0.32)	(− 0.26, 1.00)	0.26	0.20 (0.31)	(− 0.40, 0.81)	0.51	0.21 (0.31)	(− 0.39, 0.83)	0.49
Race (nonwhite)	− 0.76 (0.39)	(− 1.56, − 0.03)	0.05	− 0.62 (0.40)	(− 1.43,0.13)	0.12	− 0.88 (0.39)	(− 1–69, −0.14)	0.02	− 0.89 (0.39)	(− 1.70, −0.15)	0.02
Age	0.14 (0.06)	(0.02, 0.27)	0.02	0.19 (0.07)	(0.06, 0.32)	0.006	0.13 (0.06)	(0.01, 0.25)	0.04	0.13 (0.06)	(0.01, 0.25)	0.04
Traffic above median level	0.83 (0.32)	(0.22, 1.46)	0.009	0.78 (0.32)	(0.17, 1.42)	0.01	0.87 (0.32)	(0.25, 1.51)	0.007	0.88 (0.32)	(0.26, 1.53)	0.006

^aIn the multivariate logistic regression models, Zr was utilized as a dichotomized variable (above LOD, below LOD) and not as the imputed Zr

^bp-values were not adjusted for the false discovery rate (FDR). See manuscript for details

Adding interaction terms to the models did not significantly improve the fit of the models (Table 5). The interaction terms in the copper (p = 0.08) and zinc models (p = 0.08) were borderline significant after controlling for covariates. A one-unit increase in the natural log of copper results in the odds of internalizing problems decreasing by 32% among females and the odds doubling among males. The estimate for the interaction between zinc and male is rather large, but it offsets the effect of being male in the model. The copper and zinc interaction terms with male were not significant after adjusting for the false discovery rate (p > 0.10).

Table 5.

Multivariate logistic regressions with interaction between metals and sex

Parameter	Copper			Zinc			Zirconium in nails			Zr (Imputed)
	Estimate (SE)	95% CL	p-valueb	Estimate (SE)	95% CL	p-valueb	Estimate (SE)	95% CL	p-valueb	Estimate (SE)	95% CL	p-valueb
(Intercept)	− 2.40 (1.23)	(− 4.76, 0.07)	0.05	− 6.33 (5.90)	(− 17.78, 5.51)	0.28	− 3.10 (0.76)	(− 4.66, − 1.65)	<0.001	− 3.60 (0.88)	(− 5.39, − 1.94)	<0.001
Ln(metal)/Zr in metalsa	− 0.38 (0.54)	(− 1.54, 0.53)	0.48	0.57 (1.27)	(− 2.00, 3.00)	0.65	0.51 (0.49)	(− 0.48, 1.46)	0.3	0.47 (0.32)	(− 0.16, 0.11)	0.14
Sex (male)	− 1.66 (1.08)	(− 3.87, 0.34)	0.12	− 12.73 (7.63)	(− 28.13, 1.88)	0.10	0.24 (0.37)	(− 0.48, 0.98)	0.51	0.21 (0.67)	(− 1.09, 1.54)	0.75
Race (nonwhite)	− 0.83 (0.39)	(− 1–64, − 0.09)	0.04	− 0.62 (0.40)	(− 1.44, 0.13)	0.12	− 0.87 (0.39)	(− 1.68, − 0.13)	0.03	− 0.89 (0.39)	(− 1.70, − 0.15)	0.02
Age	0.14 (0.06)	(0.01, 0.27)	0.04	0.20 (0.07)	(0.07, 0.34)	0.003	0.13 (0.06)	(0.01, 0.25)	0.04	0.13 (0.06)	(0.01, 0.25)	0.04
Traffic above median level	0.79 (0.32)	(0.18, 1.43)	0.01	0.78 (0.32)	(0.16, 1.42)	0.02	0.87 (0.32)	(0.25, 1.51)	0.006	0.88 (0.32)	(0.26, 1.53)	0.006
Ln(metal)/Zr present * male	1.11 (0.64)	(− 0.03, 2.46)	0.08	2.92 (1.70)	(− 0.33, 6.36)	0.09	−0.13 (0.67)	(− 1.45, 1.19)	0.84	− 0.00 (0.41)	(− 0.82, 0.82)	0.99

^aIn the multivariate logistic regression models, Zr was utilized as a dichotomized variable (above LOD, below LOD) and not as the imputed Zr

^bp-values were not adjusted for the false discovery rate (FDR). See manuscript details

Discussion

Internalizing behaviors in children are often considered behaviors turned inward and include affective disturbance, anxiety, somatic complaints, and obsessive–compulsive problems. The results of our study showed that each of these problems were increased in children with internalizing behaviors compared to children without internalizing behaviors. In a recent study, it was reported that among children aged 3 to 17 years old, 7.1% had anxiety problems and 3.2% had depression. The values reported by Ghandour et al. (2019) were much less than the prevalence of anxiety (13.8%) and of affective problems (which includes depression, 16%) in our population. Our sample of children in this study had significant larger numbers of children with internalizing behavior problems compared to other exposure studies.

In our study, when describing nail concentrations by sex, males with internalizing behaviors had higher concentrations of zinc, copper, and zirconium than males without internalizing behaviors. However, females with internalizing behaviors only had a significantly different concentration of zirconium in their nails. The sex differences in males may be associated with different biologic mechanisms than those underlying the higher prevalence of internalizing disorder in females. However, when overall prevalence odds ratios were calculated, only zinc and imputed zirconium were statistically significant in the relationship between internalizing behaviors and exposure to these metal(loid)s. When we further explored the data by adding a sex-metal(loid) interaction term, these differences by sex were not statistically significant after adjusting for age, race, and traffic exposure. Although there were no differences by sex, there was a relationship between internalizing behaviors and zinc and imputed zirconium, which need further investigation.

Zinc is an essential trace element that is needed for the function of enzymes and DNA-binding transcription factors, brain growth and function, and cellular metabolism (Jurowski et al. 2014; Bitanihirwe and Cunningham 2009). There has been much research on the role of zinc as an important factor in several mental health outcomes, including depression, anxiety, and attention-deficient hyperactivity disorder (ADHD). Research has mostly focused on how low zinc status has been associated with symptoms of depression in adults, and in anxiety and ADHD in children (Maserejian et al. 2012; Toren et al. 1996; Arnold et al. 2005; Arnold and DiSilvestro 2005; McLoughlin and Hodge 1990; Hubbs-Tait et al. 2007; Wang et al 2018).

Although low concentrations of zinc are a needed element, at high concentrations, zinc is a neurotoxin. Few studies have reviewed the neurotoxin effects of zinc in children, but the two that have showed a relationship between zinc exposure and behavioral effects. Yousef et al. (2011) found that increased zinc blood concentrations were associated with ADHD. Horton et al (2018) reported that children from birth through 10–11 months of age with dentine concentrations of zinc, were more likely to report anxiety problems. The authors found that at 6 months of age, a one-unit increase in dentine zinc was associated with a 0.25 unit increase in anxiety scores. The authors did not find an association between dentine zinc and depression (Horton et al. 2018).

The health impacts of exposure to zirconium have not been well-researched or established. Ryu et al. (2014) provided a case report of a women presenting with Parkinson-like symptoms and skin discoloration from zirconium toxicity. In their report, the authors stated that there needed to be more research to “elucidate the nature of the presumed zirconium-induced neurological manifestation.” Zirconium is present in soil and water and is found in fly ash suggesting a possible source of exposure for children in this study (Hood et al. 2017; Sorokin et al. 2019).

In our study, traffic exposure was found to have an effect on internalizing behaviors in children in the models with and without the sex-metal(loid) interaction. This finding is supported by other research that shows that exposure to traffic-related air pollution (TRAP) is associated with depression and anxiety in children. Yolton et al (2019) utilized elemental carbon attributable to traffic (ECAT) as a proxy for diesel exposure. The researchers found that exposure to ECAT at birth was associated with depression and anxiety in childhood. Zinc and other heavy metal(loid)s are found in diesel exhaust (Wang et al. 2003).

Although this is the first study to assess elemental concentrations in the nails of children with internalizing behaviors, there are some limitations to this study. First, we used PIXE analysis for determining the concentrations of heavy metal(loid)s in the nail samples. Although PIXE is a good analytical method for initial use because it is non-destructive and can simultaneously determine concentrations of 72 elements at one time, some of the limits of detection might be higher than would be found in other analytical methods, such as ICP-MS, for example. In archeological work, it has been reported that PIXE detection limits are in the ppm range, whereas ICP-MS can be three times lower than PIXE in the ppt range (Pillay 2001). Concentrations of elements in water samples that could not be determined by PIXE, were found when analyzed by ICP-MS (Patnaik et al. 2011). Second, we measured body burden for metal(loid)s from nails reflecting potential exposure over 3–6 months, while likely exposures throughout development contribute to behavior. Additional associations may have been found if a lifetime exposure to metal(loid)s was measured. Another challenge is assessing exposure in the context of the multiple genetic and environment factors contributing to anxiety and depression. Possibly the exposures increase internalizing behavior concerns in children already at risk for these concerns and further study of genetics and environmental interactions will be useful in better understanding the role of environmental exposures. Additionally, although millions of children are exposed to coal ash in the United States, the concentrations of heavy metal(loid)s might be higher or lower in coal ash emissions from the stack and in emissions from storage facilities in different locations throughout the country. The varying concentrations might lead to higher or lower concentrations in the nail samples and thus might affect the behaviors of children. Furthermore, although we controlled for traffic exposure, there may be other sources of metal(loid) s in nails that contribute to the children’s body burden, that we did not account for. Exposures from coal ash are likely a small source of heavy metal(loid)s exposure in children. A variety of air, water, soil, and dietary sources can lead to metal(loid) overexposure (Briffa et al. 2020). Future research will benefit from analysis of multiple sources of exposure when considering coal ash-related neurotoxicity.

In conclusion, children living near coal-burning power plants and coal ash storage facilities have a higher prevalence of internalizing behavior disorders. Likely several factors are involved in this increased prevalence of internalizing behaviors, but exposures to zinc and potentially zirconium are of concern. Zinc has been shown to be related to internalizing disorders and are plausible risk factors for depression and anxiety. Less research has been done on zirconium. Further research of these elements on internalizing disorders is suggested considering the increasing prevalence of depression and anxiety in children.