Blood manganese and cognitive and motor skills at age 6-7 in Canadian cohort GESTE

Background: Increasing evidence suggests that high exposure to manganese (Mn) may impair brain development. In Canada, Mn exposure comes from several environmental sources and the most common is drinking water. Our objective was to examine the relationship between blood Mn and psychomotor skills in school-aged children from the Eastern Townships, Qc, Canada. In addition, we examined the association of Mn with Attention-Decit Hyperactivity Disorder (ADHD) diagnosis in our study group. Methods: Children were recruited at birth (through their mother) and followed prospectively. At 6-7 years of age, 210 children provided a blood sample for manganese testing and underwent a battery of neuropsychological tests. This battery assessed several major cognitive domains including general intelligence, attention and others via subtests from the Wechsler Intelligence Scale for Children – IV (WISC-IV), the Developmental NEuroPSYchological Assessment-II (NEPSYII) and the Test of Everyday Attention for Children (TEACH). Parents were asked if their child had ever received a physician diagnosis of ADHD. Blood was analysed by inductively coupled plasma mass spectrometry. Multivariate statistic modelling was used to control for potential confounding factors, including blood lead. Results: Median blood Mn was 9.9 µg/L (range 4.7 – 21.4 µg/L). The Design Copying – Fine motor score of the NEPSYII was positively associated with blood Mn (linear model β: 0.17 [95% Condence Interval: 0.03 to 0.32]; adjusted model β: 0.16 [95% Condence Interval 0.01 to 0.30]). Blood Mn was not associated with diagnosis of ADHD. Sex-stratied analyses indicated potential effect modication by child sex such that manganese had a benecial association on the Score DT test (a measure of sustained attention), but only among boys (β: 0.29, [95% condence interval: 0.098 to 0.49]) Conclusions: In agreement with studies from areas with similar environmental Mn levels, our study suggests that blood Mn level does not have wide-ranging associations with cognitive functions, psychomotor skills, or a diagnosis/suspicion of ADHD in school-aged children. To resolve the controversy about toxicity of environmental Mn on the developing brain, further studies should simultaneously focus on several biomarkers of Mn exposure, potential lifestyle protective factors, as well as brain imaging.


Introduction
Manganese (Mn) naturally constitutes approximately 0.1% of the earth's crust, and low levels of Mn in water, food, and air are ubiquitous. Certain geologic regions in Quebec have enriched Mn bedrock, thus the surrounding groundwater in these regions have higher Mn levels (1,2).
Health-based guidelines for the maximum level of manganese in drinking water are set at 300 µg/L by the U.S. Environmental Protection Agency (EPA) (2004) (3) and at 400 µg/L by the World Health Organization (WHO) (2008) (4). However, as of 2019 Health Canada (5) has a much stricter maximum allowable manganese concentration of 120 µg/L (with an acceptable taste level of 20 µg/L). Current WHO (4) and Health Canada (2) guidelines for water are not based on toxicological evidence in children (6). Thus, some Canadian clinicians and researchers argue that even these regulations do not go far enough to protect children, they consider the enforcement not strict enough and that the regulation should apply to private wells (7). In the Eastern Townships, a southeast region of Quebec, private or city wells often have a Mn concentration exceeding the maximum acceptable level for drinking water as stated by the health guidelines (8,9).
Mn is considered to be an essential microelement, however, in high concentrations Mn can have neurotoxic effects on the brain (10). In adults, workplace Mn exposure through inhalation has been associated with parkinsonism (manganism), which manifests in motor symptoms such as bradykinesia and cognitive symptoms such as decreased memory and attention (11). Mn is also toxic for children with long term parenteral nutrition, because of the way they are getting manganese and the concentrations too high in what they are being given -because the optimal and safe concentration for children is not known (11,12). Moreover, in individuals exposed to high amount of Mn, MRI studies show the accumulation of Mn in the brain, and mostly in the globus pallidus which regulates voluntary movement (13).
Disagreement among optimal daily intake for healthy children is common as Mn is also an essential element that plays a fundamental role in child growth and development (14). Fifteen studies have focused on associations between exposure to Mn and neurodevelopmental problems in children; 4 of them found no association (Table 1) while other studies showed a toxic effect of Mn on Children. Blood Mn is the best proxy for the pallidal index -a measure of Mn accumulation in the brain (15,16). Thus, blood Mn is thought to be a better indicator of the Mn body burden (15,16) than other biomarkers of exposure such as hairs or teeth. In Bangladesh, a country well known for high Mn and arsenic (As) levels in drinking water from the bedrock, blood Mn showed no correlation with psychomotor outcomes after adjusting for As levels. These non-signi cant ndings may be explained by the high concentration of As in Bangladeshi's water (i.e., the harmful effects of manganese may be undetectable compared to the effects of As). Mn levels found in well water in some Quebec communities -including the Eastern Townships -are similar to those reported in Bangladesh, while concentrations of arsenic are nowhere near as high here as in Bangladesh (17,18). Bouchard et al (2010) studied the relations between drinking water Mn, hair Mn and the cognition of children aged 6-13 from the Eastern Townships (19). They found a negative correlation between drinking water Mn level and cognition in children aged 6-13, but not with hair Mn.
The objective of this study was to examine the association between blood Mn and psychomotor skills, in a sample of children aged 6-7 years. We also explored the risk of ADHD and other neurodevelopmental comorbidities potentially related to Mn exposure. In addition, we included lead exposure as a covariate, given the known neurotoxic effect of lead in children and the strong correlation between lead and Mn (6).
Insert Table 1 Material And Methods

Study population
Participant's mothers were recruited between 2007 and 2009 during pregnancy (n = 761) in a prospective cohort (GESTE: GESTation and Environment) from the Eastern Townships region, Québec, Canada (20). Eligible participants were women from the Eastern Townships, Quebec region, aged > 18 y, and able to give informed consent. For this analysis, we excluded women illicit substance users, severely preterm births (< 33 weeks), resuscitated infants, and congenital malformations and families which were not certain to stay in the area for the follow up. The majority of GESTE families are caucasian French-Canadians. At 6-7 years of age, a total of 358 children completed a series of neuropsychological tests. The current analyses were conducted on a subsample 210 children who also provided a blood sample at age 6-7. Children using medication (including psychotropic drugs) were tested without any changes in posology. Test administration was done by two quali ed neuropsychologists (ASD, SG) and one trained graduate student (YSG) following a standardized procedure. Five mock administrations were lmed and evaluated to increase interrater reliability. Children with known neurobehavioral disorders were evaluated only by certi ed neuropsychologists. Final results were validated by one senior neuropsychologist (ASD).

Mn and lead (Pb) analysis
Blood was sampled after the psychomotor testing, collected in a BD Vacutainer®, K2EDTA certi ed metalfree tube (Becton-Dickinson, San Jose, California), and kept at − 20 °C until analysis which were done by the CTQ (Centre de Toxicologie  Outcomes:

1) Psychomotor Testing
Children completed subtests from the WISC-IV (Wechsler Intelligence Scale for Children -IV) (21) including; Block Design, Coding, Digit Span (forward and reverse), Information, and Vocabulary. Agespeci c normalized scores for each subtest were used in our analysis [mean (SD)]. In addition to the WISC-IV, participants completed NEPSY II (Design Copying and Visuomotor Precision subtests), a tool used by clinicians to asses 6 domains of child functioning (22). The current analyses focused on two domains, sensorimotor functioning and visuospatial processing, measured by tests such as design copying and visuomotor precision. Three subtests from the TEA-Ch (Test of Everyday Attention for Children) (23) were implemented: Sky search, Score! and Score DT. Sky search is used to measure selective attention (B: number of target correctly circle, C: time per target, G: Attentional score), this subtest also includes one motor score (F: time per target of the motor control). Score! and Score DT are used to measure sustained attention (24). Children unable to count to 15 were excluded from Score! and Score DT subtests (n = 10).

2) ADHD
Information on ADHD diagnosis and medication were obtained from caregivers during the visit. Given that some children were currently being evaluated for suspected ADHD diagnosis, we divided our study group into two categories: ADHD (including both suspected -reported by the caregiver -n = 10 and diagnosed ADHD n = 8 ; 7 children with ADHD have a comorbid developmental disorder, i.e. language disorder) and no ADHD [all other children: neurotypical (n = 169) and children with neurodevelopmental disorder other than ADHD (n = 23)].

3) Developmental Coordination Disorder
The Developmental Coordination Disorder Questionnaire -French Canadian (DCDQ-FC) (25) was also completed by the caregiver, the questionnaire is subdivided in 3 sections: control during movement, ne motor skills/ writing and global coordination. Children with DCDQ-FC score greater than 45 points are considered to be at risk of having a Developmental Coordination Disorder (DCD). In addition to the cutpoint, the continuous scores were also considered in this study.
Covariate Data: To estimate parental cognitive function, caregivers were evaluated by self-administered Raven matrices (26). If both caregivers were evaluated, we used maternal Raven score in our analyses. We found a strong correlation (r = 0.6) between mothers and fathers (n = 82), who were present and evaluated at the same time. A socio-demographic questionnaire was administered to obtain the information about parental education level, marital status, family income and other relevant information (like life stress events, industrial neighbourhood, …). Information about pregnancy was obtained from questionnaires and medical records at enrolment (20). The information on other neurodevelopmental disorders such as autistic spectrum disorder, language delay, motor coordination de cit or any pervasive developmental disorder was obtained from the caregiver.

Statistical analysis:
Statistical analyses were conducted using SAS Software version 9.4 (SAS Institute Inc., Cary, NC, USA).
To assess the risk of potential selection bias, we compared data from children included in this analysis with those from children whose parents did not consent to blood sampling (n = 148). Continuous variables were assessed using Wilcoxon-Mann-Whitney test because of a non-normal distribution and categorical data were compared using Chi² test.
Blood Mn and Pb and outcome variables distributions were close to normality, except Visuomotor Precision -Total Error score. Visuomotor Precision -Total Error score was Log10 transformed to achieve normal distribution. In addition, given that for Visuomotor Precision -Total Error score and Sky Search scores there is no age-normalized scale, we corrected it for children age using a regression to the mean of residues, a method previously described (27).
There was one missing value for the WISC-IV -Vocabulary because the child was not able to complete the test. We choose not to replace this value.
Linear regression models (SAS PROC GLM), both simple and multivariate, were used to test the association between blood Mn and psychomotor outcomes. WISC-IV subtests, NEPSY II, TEA-Ch and DCDQ scores -treated as continuous variables -were entered as dependent variables, while blood Mn was used as independent variable. Child age, sex, family income, caregiver intelligence, blood Pb, alcohol consumption (yes/no) and cigarette consumption of the mother during pregnancy (yes/no) were tested as potential confounders. Blood Mn was correlated with sex using a T-Test. Dependent variables were correlated with children sex, family income, age, caregiver intelligence and current use of tobacco in pregnancy. Only sex was correlated with both blood Mn and most outcome measures. Based on previous literature, we also considered blood Pb as a potential confounder in our models. Finally, all multivariate models included children sex, family income, age, caregiver intelligence, current use of tobacco in pregnancy and blood Pb as covariates. A statistical interaction with sex was tested for each outcome separately. In addition, a quadratic Mn term was included in the model to test for a potential nonlinear, Ushape, dose-response relationship.
We conducted three sensitivity analyses to evaluate the robustness of our results. First, we re-ran models excluding children with any reported neurodevelopmental disorder (i.e.: Autism spectrum disorder, pervasive development disorder and language delay) (n = 30). Second, we conducted an analysis to evaluate if there were any differences in the psychomotor outcomes between children non medicated for ADHD (i.e. suspected for ADHD) (n = 10) and medicated for ADHD (i.e. diagnosed for ADHD) (n = 8). Third, we investigated potential differences between children with medicated ADHD (n = 8) and those without any reported neurodevelopmental disorder (n = 169).
To examine the risk of diagnosis of ADHD and other neurodevelopmental disorders, we used logistic regression (SAS PROC LOGISTIC) to estimate the odds ratio (OR) and 95% con dence intervals (CI) of blood Mn using an unadjusted and adjusted model. To examine the risk of being suspected of Developmental Coordination Disorder (DCD), we used a logistic regression (SAS PROC LOGISTIC). The dependent variable was the DCDQ-FC > 45 (n = 13). Potential confounders were children age, sex, blood Pb, alcohol and cigarette consumption of the mother during pregnancy. Dependent variables were not correlated with any of the potential confounders. We kept children sex, blood Pb, alcohol and tobacco use during pregnancy in our models to be consistent with literature data (28,29).

Results
Cohort data Table 2 shows participants demographics and distribution of model covariates and outcomes data.
There were no differences between participants included in present analyses and those who declined to give a blood sample (n = 148) for any of the characteristics. A total of 13 included participants were using medication for other disease (like asthma) and 5 in the excluded group. There was no difference between results obtained with the whole study group and those from the analysis restricted to neurotypical children only (for more details see additional le 1). For this reason, we show results from the whole study group. The second sensitivity analysis found no difference in psychomotor outcomes between ADHD children with or without medication. The third sensitivity analysis found no difference between children with medicated ADHD and those without any reported neurodevelopmental disorder.
Insert Table 2 Blood Mn and psychomotor scores Table 3 shows simple and multivariate model parameters for psychomotor scores in relation blood Mn.
Globally, only NEPSY II-Design Copying ne motor score was slightly positively correlated with blood Mn in both adjusted and non-adjusted models.  Block design 0.00 (-0.14 to 0.14) 0.00 (-0.14 to 0.14) The sex-strati ed analyses yielded one statistically signi cant association. Boys performed better on the Score DT test (β: 0.29, CI: 0.098 to 0.49) when exposed to higher Mn concentration, while no association was found among girls (β: -0.09, 95% CI: -0.33 to 0.14).
The writing score of the DCD-FQ questionnaire was signi cantly and positively correlated with blood Mn in the non-adjusted models, but not in the adjusted model.
Insert Table 3 Blood Mn and at risk of ADHD/DCD

Discussion
Our study shows no correlation between blood Mn and psychomotor skills in school-age children from the general population, contrary to our hypothesis. This nding aligns with previous reports by Lucchini (28) and Wasserman (18). In addition, we observe no association with the diagnosis or suspicion of ADHD, as well as non-signi cant associations found in the attention scores from TEA-Ch. We also found no association between high blood Mn and the risk of DCD, con rmed by our results from the motor scores in NEPSY II.
One potential explanation for these non-signi cant ndings are the low absolute levels of Mn in the blood that may not reach a toxic level in our population, thus not affecting children psychomotor skills. Also, at 6-7 years of age the physiological regulation of Mn in the body -reduction of absorption in the intestines and increase excretion by the liver-may be su cient to reduce blood Mn levels and its potential toxic effect (30). Another potential explanation is that our population is not exposed to any Mn-releasing industry as seen in other studies (31,32). The only Mn-releasing industry in the Eastern Townships is a small paper mill plant, that was closed approximately 2 years before our study. Thus, we may speculate that for the majority of children in our study group the major source of Mn would be water and/or food.  Table 1).
A recent publication discusses the presence and the role of Type III error in environmental health science.
Type III error is de ned as correctly rejecting the null hypothesis (in our case Mn in drinking water is not toxic for school age children) for the wrong reasons (37,38). This error occurs most often when a causal factor, such as socio-economic status or other environmental factors, etc., is homogeneously distributed in the population. Type III error mostly comes from a dichotomic reasoning. We may inconsistently conclude that only one factor is the cause rather than the interaction between factors (e.g. genetic and environment, socio-economic status and environment, etc.) (38).
It is possible that some studies did not consider the interactions that may lead to a toxic effect of Mn.
Studies based on socio-economically disadvantaged populations often show a toxic effect of Mn. These are predominantly people living in poor rural areas in Mexico, Brazil (31,32), and Bangladesh (39). Low socio-economic status can lead to an increasing of malnutrition in the population. The exposure to Mn is higher among malnourished children, they absorb Mn more e ciently and eliminate it less e ciently (40,41). Socio-economically disadvantaged communities also face higher concentrations of pollutants in their environment (42), thus increasing the risk of toxic interaction with Mn.
Studies focused mainly on communities exposed to Mn through food and water remain inconsistent.  (7) found no association between Mn levels in blood, hair, or water and decreased psychomotor performances.
In adults, the route of exposure plays a crucial role in total intake and Mn accumulation. When exposed to Mn through air, it will bypass physiological control and leads to a rate of absorption of almost 70%. This inhaled Mn will be directly transported to the brain through the olfactive nerve (14). The ingestion of signi cant quantities of Mn is tightly controlled by the liver, which allows only about 2% of ingested Mn to reach the general circulation, the rest being eliminated through bile into the GI tract (14). Elimination of Mn by hepatobiliary mechanisms is decreased in children (44) (45), when they estimated that at the same exposure level a child's Mn inhalation is at a greater proportion of the maximum recommended levels compared to adults. Inhalation provides more rapid uptake of Mn into the blood from lungs, as shown in welders and mice (45). In a review of the inhalation dosimetry methods applied to children's risk assessment, the authors (46) recommended to use a higher uncertainty factor due to increased toxicity of most inhaled chemicals in children.
Mn is suspected to have gender-dependent effect (6). In our study, only one attentional score is positively affected in boys, and tends to be negatively affected in girls. To date, mechanisms underlying this relationship remain unclear and there is no published data on the gender-dependent toxic effect. As only one score was affected, we presume that this result may be due to chance.
Our study has several strengths. Children included in the analyses participated in a prospective population-based birth cohort decreasing the risk of selection bias. We also adjusted for lead levels and most of confounders and effect modi ers we had access to. Our statistical analyses included non-linear modelling to account for a potential U shape association. Our study population is relatively homogenous socially, economically and racially, given that our root population is stable, composed of historical descendants from French and Irish families. This homogeneity is advantageous when we are looking for small effects of environmental pollutants because it reduces a background noise from potential confounders (i.e. "quasi-experimental design"). Blood Mn, our biomarker of exposure, has been shown to better approximate pallidal index, an indicator of Mn accumulation in the brain (relevant in the context of studying its neurodevelopmental effects), than other potential markers including hair or water Mn.
This study has some limitations. The homogeneity of our population can decrease the external validity of our results, which should be carefully extrapolated to a community with different genetic or socioeconomic characteristics. In the same way, we noticed that more educated families tend to stay within the follow up (data not shown), as it was recently observed for other prospective cohorts (47).
Another potential limitation is that we used blood Mn as the only biomarker of exposure, and blood Mn is weakly corelated with ingested Mn (18,48). However, blood Mn has been reported to have longitudinal stability (33). Furthermore, data on iron de ciency in the cohort was not collected, but there is low prevalence -3.5% of the children population -of iron de ciency in Canada (49). Children with iron de ciency tend to have high blood Mn (50). Yet increased blood Mn in patients with iron de ciency do not typically increase Mn accumulation in the brain (51,52). However, even if iron de ciency increases blood Mn, we observed no effect of blood Mn on psychomotor scores thus we expect the direction of results to be the same. In addition, we have no estimation of Mn intake. Most of the Mn in our region comes from water (19) and there is no association between dietary Mn intake and cognitive scores in children (19). Moreover it has been reported that blood Mn is poorly correlated with intake in children and is not affected by subtle intake variation (18). Thus, the lack of estimation of Mn intake should not alter our results.
For this study we used ve subscales of the WISC-IV -which did not allow us to estimate the children's full IQ -, two of the NEPSY II and three of the TEACh. These subscales were chosen in order to cover a variety of non-verbal skills within a limit of about 1.5 hour of evaluation, which is feasible in children aged 6-7.
ADHD group selection was based on whether the child took prescription ADHD medication and if the caregiver reported a diagnosis. We may have misclassi ed ADHD cases as controls if their caregiver did not report physician's diagnosis. Children with ADHD were not asked to stop taking their medicines during the visit, meaning they may appear more neurotypical on continuous assessments of their behaviors, which could weaken any true association. TEA-Ch is a test commonly used to diagnose ADHD and is known to differentiate children medicated for ADHD and those non-medicated (24). However, because children diagnosed with ADHD who were medicated had similar scores on the psychomotor evaluations compared to those who were unmedicated (for more details see additional le 2), any effect of medication is likely minimal. Although we could not con rm DCD diagnosis in medical records, our nding of no association between Mn and DCD as assessed by the parentally-administered DCDQ is supported by our null ndings with the NEPSY-II, which also measures dexterity but is more objective because it is administered by study staff.
In conclusion, our study in school-aged children from the general population we did not show evidence of Mn toxicity related to psychomotor and attention skills. Our study does not rule out prenatal Mn toxicity given that fetuses/neonates have immature hepatobiliary mechanisms of Mn elimination (44). Pre and postnatal exposure to Mn has toxic effects in toddlers (48). Those effects may be persistent during the childhood and alter the developmental trajectories of the child (53). There is a need for further comprehensive studies using different matrices and MRI to con rm these results. New biomarker that can better re ect ingested Mn (like meconium or stools) should also be considered in consort with other established biomarkers. Additionally, MRIs can be used to con rm Mn accumulation in children's brains and provide more information on the related anatomical or functional modi cations.

Declarations
Ethics approval and consent to participate The study protocol was approved by the ethical review board of the Centre Hospitalier Universitaire de Sherbrooke. All participants provided written informed consent before their inclusion.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This research was supported by the National Institute of Environmental Health Sciences (R21ES024841 and R01ES027845) and the Canadian Institutes of Health Research (MOP-84551). The funding sources had no involvement in study design, collection, analysis, data interpretation, manuscript preparation, or the decision to submit the article for publication.
Authors' contributions LS analyzed and interpreted the data regarding blood Mn and psychomotor outcomes. NA enrolled families. ASD conducted psychomotor testing. VG was in charge of the biobank and sample analysis. AB was the coordinator of the project. EW, CB, KB and HEL read and critically reviewed this article. MF assisted in enrolment of participants. AAB and LT are the principal investigator of the GESTE study and LT created the GESTE cohort. All authors read and approved the nal manuscript.    Block design 0.00 (-0.14 to 0.14) 0.00 (-0.14 to 0.14)