Intellectual Impairment in School-Age Children Exposed to Manganese from Drinking Water

We conducted this cross-sectional study in southern Quebec (Canada) between June 2007 and June 2009. Municipalities were considered as potential study sites if their aqueduct was supplied by groundwater. We selected eight municipalities to achieve a gradient of MnW. We explained the study to the principals and teachers of the elementary schools, who agreed to distribute recruitment letters to the children’s families. The response rate of the families was 60%, and the participation rate was 52%. Recruitment was restricted to children who had lived in the same house for > 3 months, to ensure continuous exposure to the same source of water for this minimum period of time; 362 children (age 6–13 years) participated in the study. The Human Research Ethics Board of the Université du Québec à Montréal approved the study protocol. We obtained signed informed consent from the parents and verbal assent from the child to participate in the study.

We collected a hair sample from the occiput of each child, cutting as close as possible to the root with surgical stainless steel scissors. We used the 2 cm closest to the scalp and washed samples to minimize external contamination, using the method described by Wright et al. (2006). In a test phase, we photographed hair strands with an electronic microscope and observed that the washing procedure effectively removed all particulates from the surface of the hair strand without compromising its structural integrity. We measured metals [manganese, lead (Pb), iron (Fe), arsenic (As), zinc (Zn), and copper (Cu)] by inductively coupled plasma-mass spectrometry (ICP-MS). Details of the analyses are in the Supplemental Material (doi:10.1289/ehp.1002321 via http://dx.doi.org/).

When manganese concentrations for certified hair material were outside of the designated concentrations, we excluded the measures from the analyses. Nine children who reported using hair dye in the preceding 5 months were also excluded, because hair dye could influence manganese hair content (Sky-Peck 1990). Children who reported use of hair dye had higher MnH compared with the others [geometric mean (GM), 1.1 μg/g, and 0.7 μg/g, respectively]. A total of 302 children were included in the analyses of MnH.

During the home visit, a parent responded to an interview-administered questionnaire about the source of the domestic tap water (private well/public well), residential history, and changes to domestic water treatments. We collected a water sample from the kitchen tap and also collected a second sample when there was a point-of-use filter (filter attached to the tap). We used the following procedure to standardize tap water sampling (van den Hoven and Slaats 2006): a) open the tap for 5 min, b) close and leave untouched for 30 min, and c) collect first draw. We added 0.15 mL nitric acid (50%) to the 50-mL water sample and stored samples at 4°C. We measured metals (manganese, Pb, Fe, As, Zn, and Cu) by ICP-MS. Calibration curves were run every 30 samples, along with field and laboratory blanks and quality controls (CRM TM-26.3; Environment Canada, Laboratory for Environmental Testing, Burlington, ON, Canada) every 15 samples.

For a subsample of participating families (n = 20), we sampled tap water on three occasions over a 1-year period to examine time-dependent variability.

During the home visit, we orally administered a semiquantitative food frequency questionnaire to the parent and the child to assess manganese intake from the diet and water consumption. We used 3-dimensional models of portion size to obtain more precise estimates for all sites except the first site; thus, data for these participants (n = 16) were not included in the analyses on dietary intake. We estimated manganese intake from water consumption for direct water ingestion and for water incorporated in food preparations. We estimated water consumption from different sources—bottled, tap, tap filtered with a pitcher, and tap with an attached filter. For each water source, the amount consumed was multiplied by the measured or estimated concentration of manganese, yielding a total intake in micrograms per month. Further methodological details can be found in the Supplemental Material (doi:10.1289/ehp.1002321 via http://dx.doi.org/).

We used the Wechsler Abbreviated Scale of Intelligence (WASI) to assess general cognitive abilities (Wechsler 1999). This standardized test yields a Verbal IQ score (based on the subtests Vocabulary and Similarities), a Performance IQ score (Block Design and Matrix Reasoning), and a Full Scale IQ score. Throughout the study, three psychometricians administered the WASI, but all scoring was performed by the same person. We administered the WASI within 1 week of tap water sampling.

We collected information from the mother on factors that might confound the association between manganese exposure and cognitive abilities of the child, such as socioeconomic status indicators (i.e., maternal education, family income and structure), parity, and alcohol and tobacco consumption during pregnancy. We assessed maternal nonverbal intelligence with the Raven’s Progressive Matrices Test (Raven et al. 2003), home cognitive stimulation with a modified version of the Short-Form HOME (Home Observation for Measurement of the Environment) interview (Bradley et al. 2001), and maternal symptoms of depression with the Beck Depression Inventory-II (Beck et al. 1996). Data on family income were missing for four families, and the Raven score was missing for one mother; for missing data, we assigned the mean value of individuals with data.

The distributions of manganese concentrations in hair and water, as well as manganese intakes, were considerably skewed. We thus employed log₁₀ transformation to normalize residuals. Likewise, we log-transformed the concentrations of other elements measured in water. Manganese intakes from consumption of water and from the diet were divided by the weight of the child for use in the analyses (micrograms per kilogram per month). We used generalized estimating equations (GEE) to examine relationships between exposure to manganese and children’s IQ scores. GEE is an extension of generalized linear models for nonindependent data (Zeger and Liang 1986). These analyses were used to account for the community- and family-clustered data in our study. Some of the advantages of using GEE, instead of the more common approach of mixed models with random intercepts, include more efficient estimators of regression parameters and reasonably accurate standard errors [i.e., confidence intervals (CIs) with the correct coverage rates]. With GEE, the computational complexity is a function of the size of the largest cluster rather than of the number of clusters—an advantage and a source of reliable estimates when there are many small clusters (Hanley et al. 2003), such as in the present study (i.e., the 251 families). CIs were calculated with Wald statistics. An exchangeable working covariance matrix was used, with a robust estimator providing a consistent estimate of the covariance even when the working correlation matrix is misspecified.

Changes in IQ were examined in relation to four manganese exposure metrics reflecting different assumptions for exposure pathways and toxicokinetics: MnW, MnH, manganese intake from water consumption, and dietary manganese intake. The change in IQ (β) associated with a 10-fold increase in manganese exposure indicators was examined with adjustments for two sets of covariates. The first set of covariates (model A) was chosen based on the examination of directed acyclic graphs (Greenland and Brumback 2002) and included several socioeconomic indicators. Because manganese causes aesthetic problems, families who have the means might treat water domestically to remove it. The second set of covariates (model B) included the same variables as model A, as well as variables significantly associated with IQ or MnW, to reduce the unexplained variance, thus diminishing type 2 error (Bellinger 2007). We conducted sensitivity analyses on inclusion of additional covariates in the models. We used 0.05 as the threshold for statistical significance (two-sided tests). We examined residuals for normality and homoscedasticity and detected no problem.

This study included 362 children from 251 families. Most children (85%) had resided for > 12 months in their present home. The mean (± SD) age of the children was 9.3 ± 1.8 years; range, 6.2–13.4 years), and 99% of children were white. Seventy-eight percent of mothers had at least some college education (Table 1). Tap MnW ranged from 1 to 2,700 μg/L (Table 2), with an arithmetic mean of 98 μg/L and a GM of 20 μg/L. MnW from repeated sampling in the same residence over a period of 1 year had an intraclass correlation coefficient of 0.91.

MnW was not associated with socioeconomic or other family characteristics such as family income, family structure, home stimulation score, nonverbal maternal intelligence, or maternal education (Table 1). MnW was lower in houses with a private well (GM, 8 μg/L) than in those on a public well (GM, 55 μg/L). The concentration distribution for elements other than manganese in residential tap water is in the Supplemental Material (doi:10.1289/ehp.1002321 via http://dx.doi.org/). The Pearson correlation of MnW with other elements was 0.68 (Fe), 0.26 (Zn), 0.11 (Cu), 0.06 (As), and −0.02 (Pb).

The median of estimated manganese intakes from direct consumption of water (1.6 μg/kg/month) was similar to the median of intakes from water incorporated into food preparations (1.9 μg/kg/month) (Table 2). The estimated dietary manganese intakes were much higher than the intakes from water consumption, with a median of 2,335 μg/kg/month (Table 2).

Children’s MnH increased with MnW and estimated manganese intake from water consumption (Figure 1A), but not from the estimated dietary manganese intake (Figure 1B). In a multivariate model, MnH was significantly associated with manganese intake from water consumption (p < 0.001) but not with dietary intake (p = 0.76). In this multivariate model, there was no difference in MnH between boys and girls (p = 0.46; GM for both, 0.7 μg/g), and age was not associated with MnH (p = 0.88).

Estimated dietary manganese intake was not significantly associated with IQ scores in unadjusted or adjusted analyses (results not shown). Table 3 presents unadjusted and adjusted changes in IQ scores for a 10-fold increase in exposure level for three exposure indicators: MnW, estimated manganese intake from water consumption, and MnH. In unadjusted analyses, the three indicators were significantly associated with lower Full Scale IQ scores. Adjustment for covariates, with either model A or model B, did not considerably change the point estimates. Higher MnW was significantly associated with lower Full Scale IQ scores in model A [change in scores for a 10-fold increase in concentration (β) = −1.9 (95% CI, −3.1 to −0.7)] and model B [β = −2.4 (95% CI, −3.9 to −0.9)]. Higher MnW was also significantly associated with lower Performance IQ scores in model A [β = −2.3 (95% CI, −3.7 to −0.8)] and model B [β = −3.1 (95% CI, −4.9 to −1.3)]. Higher MnW was associated with lower Verbal IQ scores, significantly for model A but not for model B. (The point estimates were similar for both models but the 95% CI were larger for model B.) Higher estimated manganese intake from water consumption was significantly associated with lower Full Scale and Performance IQ, in both model A and model B. Sex-stratified analyses on MnW and Full Scale IQ resulted in higher point estimate for girls [using model B, β = −3.2 (95% CI, −5.0 to −1.5)] than for boys [β = −2.3 (95% CI, −4.8 to 0.2)], but the term for interaction with sex was not significant (p = 0.14).

When MnH was examined as the predictor of IQ in unadjusted analyses, it was significantly associated with lower Full Scale IQ scores but not Performance or Verbal IQ scores. In adjusted analyses, higher MnH was associated with lower Full Scale IQ scores, both in model A [β = −3.7 (95% CI, −6.5 to −0.8)] and in model B [β = −3.3 (95% CI, −6.1 to −0.5)]. MnH was also associated with lower Performance and Verbal IQ scores, although these relations did not reach statistical significance except for Verbal IQ in model A [β = −3.1 (95% CI, −5.9 to −0.3)]. Sex-stratified analyses resulted in higher point estimates for the association between MnH and Full Scale IQ for girls [β = −4.8 (95% CI, −8.1 to −1.6)] than for boys [β = −3.5 (95% CI, −8.7 to 1.6)], but the term for interaction with sex was not significant (p = 0.55).

The shape of dose–response relations are shown in Figure 2. IQ scores decrease steadily with increasing MnW (Figure 2A). Children in the highest MnW quintile (median, 216 μg/L) scored 6.2 points below those in the lowest quintile (median, 1 μg/L). For estimated manganese intake from water ingestion (Figure 2B), children in the lowest quintile had the highest IQ scores, and those in the highest quintile had the lowest scores, but point estimates in the middle quintiles did not show a consistent pattern of increasing or decreasing trend. For MnH (Figure 2C), IQ scores decreased only slightly between children in the lowest quintile and the middles quintiles, and there was a steeper decrease for children in the highest quintile. A similar plot for dietary manganese intake showed no association, even in the higher range of intakes (data not shown).

We conducted sensitivity analyses on the inclusion of additional covariates in the models: birth weight of the child, rank of the child in the family, maternal smoking or alcohol consumption during pregnancy, maternal depressive symptoms, psychometrician, and concentration of Pb, As, Cu, and Zn in tap water. We did not retain these covariates, because they did not change point estimates by > 10% and were not significantly associated with IQ scores (p > 0.2).