You are on page 1of 10

Qualitative assessment of visuospatial errors in mercury-exposed

Amazonian children
Ce cile Chevrier
a,
*, Kimberly Sullivan
b
, Roberta F. White
b,c
, Callie Comtois
b
,
Sylvaine Cordier
a
, Philippe Grandjean
d
a
Inserm, U625, University of Rennes I, GERHM, IFR140, F-35000 Rennes, France
b
Department of Environmental Health, Boston University School of Public Health, Boston, MA 02118, USA
c
Department of Neurology, Boston University School of Medicine, Boston, MA 02118, USA
d
Department of Environmental Medicine, Institute of Public Health, University of Southern Denmark, DK-5000 Odense, Denmark
1. Introduction
The vulnerability of the developing brain to methylmercury
(MeHg) toxicity has been recognized since the 1950s, following
large prenatal exposures from maternal ingestion of contaminated
marine food in Japan (Harada, 1995) and a subsequent incident
caused by ingestion of contaminated our in Iraq (Amin-Zaki et al.,
1979). Both incidents resulted in severe neurodevelopmen-
tal abnormalities in children exposed prenatally, with mental
retardation, cerebral palsy, blindness, and deafness. More subtle
neurotoxicity such as developmental delays may be a concern due
to lower MeHg exposures in children who depend on marine-food
or freshwater sh as their main food staple. Two large-scale
longitudinal studies conducted in the Faroes and in the Seychelles
have examined possible adverse neurological effects that may
occur at lower levels of dietary MeHg exposure, but seemed to
differ on the effects of methylmercury on the visuospatial and
other cognitive functions (Grandjean et al., 1997; Myers et al.,
2003). Two cross-sectional studies were conducted among
residents of remote gold-mining areas in the Amazon region,
where freshwater sh is a primary food source (Pfeiffer and
Lacerda, 1988; Akagi et al., 1993; Grandjean et al., 1999; Frery
et al., 2001; Cordier et al., 2002). They both reported decrements in
motor speed, attention and visuospatial function associated with
MeHg exposure among Amazonian children aged 712 years old
(Grandjean et al., 1999; Cordier et al., 2002).
While large exposures to MeHg during brain development can
result in severe brain damage that is readily identiable during
infancy, lower levels of MeHg exposure can result in decrements in
NeuroToxicology 30 (2009) 3746
A R T I C L E I N F O
Article history:
Received 19 December 2007
Accepted 26 September 2008
Available online 17 October 2008
Keywords:
Methylmercury exposure
Fish consumption
Neuropsychological tests
StanfordBinet Copying test
A B S T R A C T
In order to better dene the effects of methylmercury (MeHg) exposure on neurodevelopment,
qualitative error types observed in the responses of exposed children to the StanfordBinet Copying Test
were categorized and quantied using raw data from two studies of 395 Amazonian children aged 712
years (fromBrazil and French Guiana). These outcomes were related to hair-mercury concentration as the
biomarker of MeHg exposure (range = 0.563.8 mg/g). The combined analysis of data from two separate
countries had two major goals: (1) to gain clues concerning the underlying neuropathological
mechanisms of observed effects based on convergent evidence of MeHg-related qualitative outcomes in
the two studies and (2) to explore possible cultural determinants of test response based on divergent
outcomes in the two countries. Multiple linear and logistic regression analyses were performed with
adjustment for confounders. In the combined data set, mercury exposure was negatively associated with
scores on the drawing task: a score reduction of 1.2 (s.e., 0.3) points was observed in the children with a
hair-mercury concentration above 10 mg/g compared to those with a hair level below 1 mg/g; this effect
appeared to be stronger in the younger children. Risk of committing one or more errors of rotation,
simplication or perseveration in the drawings increased with hair-mercury concentration in both
cultural settings, providing convergent evidence of specic types of MeHg-related neurocognitive
outcomes. However, relationships between mercury exposure and scores on the Block organization
component of the test varied according to the study site, indicating that other factors must be considered
in evaluating responses to the demands of this cognitive task.
2008 Elsevier Inc. All rights reserved.
* Corresponding author at: Avenue du Ge ne ral Leclerc, Unit 625 INSERM, Campus
de Beaulieu, F-35000 Rennes, France. Tel.: +33 2 2323 6126; fax: +33 2 2323 6126.
E-mail address: cecile.chevrier@rennes.inserm.fr (C. Chevrier).
Contents lists available at ScienceDirect
NeuroToxicology
0161-813X/$ see front matter 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuro.2008.09.012
brain function that may not be as obvious at birth or during infancy
(Grandjean et al., 1998). These associated functional disabilities or
developmental delays may not be detectable until the child
becomes old enough to perform testing that assesses specic
cognitive domains (Bernstein and Waber, 1990; White et al., 1994).
Neuropsychological instruments assessing cognition are crucial for
detecting subclinical decrements associated with low-level
exposure to MeHg, but test results may vary by culture. Therefore,
proper test selection is crucial in order to identify tests that are
sensitive to the particular neurotoxicant being assessed but
culturally sensitive. The StanfordBinet (SB) Copying test
(Thorndike et al., 1986), which assesses visuospatial skills, is a
sensitive measure of cognitive change related to MeHg (Grandjean
et al., 1999; Cordier et al., 2002; Debes et al., 2006) and pesticide
exposures (Grandjean et al., 2006). A major advantage of the task is
its practical convenience for testing populations living in remote
areas of the Amazon: it involves simple paper-and-pencil and
block design stimuli and responses. A noted disadvantage to using
the standard scoring of the SB Copying test is that it has been
criticized as lacking adequate objective scoring methods often
resulting in poor inter-rater reliability (Mason, 1992). Therefore, a
more consistent relaxed scoring system was developed by one of
the authors (RW) to improve the objectivity and sensitivity of the
task in these specic cultural groups of Amazonian children. The
same relaxed criteria were used in the study conducted in French
Guiana, where lower scores were associated with higher hair-
mercury concentrations in children (Cordier et al., 2002). A specic
error type (rotation on the drawings) in the copying task was
observed more frequently among the French Guiana children with
higher levels of MeHg exposure, especially among boys and among
younger children. Although drawing errors in copying tasks are
part of normal development, it was hypothesized that specic
patterns of error types would correlate with higher MeHg exposure
and relate to specic areas of brain vulnerability to MeHg
exposure.
The present study provides a joint analysis of the association
between the SB Copying scores and MeHg exposure levels from
both studies in the Amazonian area in order to answer two
questions: (1) can similar qualitative patterns of error types
associated with MeHg exposure level be documented in two
Amazonian communities (Brazilian, French Guiana) and (2) are
there heterogeneities in the exposureoutcome relationships for
the two study sites that may suggest cultural differences in the SB-
Copying task performances.
2. Methods
2.1. Population
2.1.1. Brazil
The Brazil study was conducted in three villages, where the
Copying test was administered to 263 children aged 712 years
(80% of 322 eligible children). The villages included two located on
the Tapajo s River tributary (Sai Cinza and Brasilia Legal) and one
upstream on the main Amazon River (Santana do Ituqu). The
choice of the three villages considered their resemblance to each
other in community size and social environment, but with
differences in regard to proximity to gold mining activities and
levels of MeHg exposure (Malm, 1998). The Sai Cinza children
spoke Munducuru and were examined with an interpreter, while
the children fromthe two other villages were Portuguese-speaking
and culturally similar. School education is mandatory in all three
villages until age 12. Illiteracy among adults was common in the
Tapajo s river basin. Occupational activities generally involved
subsistence commercial farming and shing. Other sources of
pollution, such as trafc, deforestation, and industrialization, are
limited in these sites.
2.1.2. French Guiana
Eligible children for the French Guiana study were all children
aged 512 years belonging to the Wayana community living in the
remote Upper Maroni areas (97 included of 110 eligible), and a
random group of children from the Galipi community living in
Awala on the Atlantic coast with similar age and sex distributions
(82 included of 94 invited). The choice of these communities was
based on a previous survey, which showed that the Wayana
community on the Upper Maroni River was the most exposed to
MeHg, mainly due to their freshwater sh consumption (Cordier
et al., 1998). The Galipi from Awala had lower levels of exposure.
This latter area is accessible by road and is more developed than
other remote areas in the Amazonian region. Its social organization
slightly differs from that in Wayana: for instance, most of the
mothers living in Awala had completed primary school.
2.2. MeHg exposure assessment
Hair-mercury analysis was chosen to assess MeHg exposure
levels in these studies because of its acceptable validity for
population studies (Cernichiari et al., 1995; Grandjean et al., 1994,
2002) and the convenience of collecting hair samples rather than
blood in remote locations. Hair samples were taken from both the
mothers and the children as close as possible to the scalp in order
to estimate current mercury exposure. The rst two centimeters of
the hair samples were analyzed for total mercury by ow-injection
cold-vapor atomic absorption spectrometry. Each study included a
quality assurance program for the hair-mercury analyses.
2.3. StanfordBinet Copying test administration and scoring criteria
The administration of the Copying test was carried out by a
single examiner in Brazil (in Sai Cinza with an interpreter) and by
three examiners in French Guiana; all examiners were blind to the
mercury exposures of the participants.
The SB Copying test includes two components. The rst
component (Blocks) requires reproduction of the examiners three-
dimensional block designs (items 112). The second part, Copying,
requires the examinee to draw 16 two-dimensional geometric
designs with pencil and paper (items 1328). Using the standard
scoring system, several criteria must be met in order to receive
credit for correctly drawing a particular itemto copy. For example,
one of the stimulus items is a cube. Six criteria must be met to
obtain credit for the drawing: partial credit is not given and the
score for each drawing using these criteria is therefore either 1 or 0.
As previously discussed, the standard scoring criteria have been
criticized as being somewhat inconsistent and requiring subjective
judgment (Mason, 1992). For example, design rotations beyond108
lose credit in some designs but not others. Therefore, the relaxed
scoring system was designed to assess the overall gestalt (or
congural awareness) of the childrens drawings. Errors in minor
details were weighed less than clear visuospatial distortions; for
instance, drawings with rotation errors of less than 908 received
full credit if all other criteria for the drawing had been met. Full
credit was thus given for drawings with only minor details missing
to reect whether the childrens drawings captured the gestalt of
each stimulus item. Finally, a total Block score (012) and a total
design Copying relaxed score (016) were obtained.
The additional scoring technique included a tally of kinds of
error violations given that a single drawing might include several
types of errors. The specic qualitative error types that were scored
included rotations, distortions, simplications, perseverations,
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 38
overdrawing, micrographia, macrographia, and tremor. A rotation
error is dened as a shifting of the whole design more than 908
fromthe horizontal of the page or shifting a part of the design more
than 908. A distortion is a drawing that is modied fromits original
design in a way that would not earn credit for the design. Examples
include designs being made into letters, angles being rounded or
lines left out of drawings. Simplication involves a design that is
changed into a less complex one. Examples include diamond-
shaped designs being changed into triangles, or cubes being
changed into squares. Perseverations include repeating a previous
design, or part of a design, in lieu of the correct design.
Overdrawing includes heavy writing that is formed from drawing
over a design several times. Micrographia is dened as very small
drawing and macrographia is dened as an exaggeratedly large
drawing. Tremor is dened here as the appearance of shaky lines in
the drawings. The total qualitative error score for each examinee is,
in principle, unlimited.
Scoring of the SB Copying test with relaxed criteria was
performed by the same rater (RW) in both studies. For the Brazil
study, qualitative error types were also scored by the same rater
who then trained a second rater for the qualitative scoring in the
French Guiana study. In the original Brazilian study, two raters
scored the data with regard to both standard and relaxed scoring
criteria showing a good inter-rater correlation (rs = 0.91 and 0.86
respectively). In the original French Guiana study, the rotation
errors were scored by a psychologist, trained by the main rater, and
also showed a good reliability when compared to the scoring by the
second rater (r = 0.94).
2.4. Statistical analysis
To obtain comparable data for the two populations, 64 children
from French Guiana aged 56 years were excluded. The present
analysis therefore included 395 children (205 boys, 190 girls) aged
712 years. The hair-mercury concentration was available for 95%
of these children fromthe childs own sample and for 68% fromthe
mothers sample.
Performance on the Copying test depends on factors such as age
and sex. These two confounding variables were therefore included
in the analysis as mandatory covariates. Age was assessed in full
years. Results additionally adjusted on village were provided. In a
different additional adjustment strategy, potential confounding
factors were selected according to their availability and their
relevance in both studies. Married/cohabitating mothers were
distinguished from other mothers using a marital status variable.
The impact of maternal intelligence on childrens test performance
was large in the French Guiana study when the mothers Raven
Progressive Matrices score was used to predict Copying test
performance in their children (Cordier et al., 2002). The Raven test
was not measured in the Brazil study. Therefore, maternal
education was used as a proxy estimate of maternal intelligence.
In the French Guiana data, this variable discriminated between low
and intermediate/high Raven scores, with a mean Raven score of
13.2 (range: 425) among mothers who did not complete primary
school and a mean of 20.5 (range: 1135) among mothers with
more education. Also, maternal alcohol consumption during
pregnancy was included as a dichotomous variable (yes/no). To
obtain a nal model, a backward selection procedure (p < 0.20)
was used to determine whether to adjust for maternal marital
status, maternal education and maternal alcohol consumption.
Since the hair-mercury concentrations showed a skewed
distribution, a logarithmical transformation (base 10) was used
to normalize the distribution and avoid disproportionate inuence
by results froma fewhighly exposed children. Pearson correlations
(r) between hair-mercury concentrations from the child and the
mother were computed. Multiple linear regression analysis was
performed to assess the relationship between mercury exposure
levels and scores on the SB Copying test. The coefcient estimate
and its degree of signicance (p-value) are reported.
For each error type (except distortion error), the majority of the
children scored 0, thus preventing statistical analysis using
the error type score as linear. Logistic regression was used to
assess the association between mercury exposure levels and the
presence of at least one error (and at least two distortion errors). An
odds-ratio (OR) and a standardcondence interval (CI) are reported.
To take into account potential study-based heterogeneities,
interaction between hair-mercury concentration and the study site
(country) variable was assessed. In addition, the effects of mercury
exposure on the childs performance allowing for interaction with
age and sex were examined in stratied regression analyses. Using
bar charts, we illustrated the potential heterogeneities of the
copying scores and the frequencies of error types according to both
the childs hair-mercury concentration and the childs age. We
used 3 classes of the childs age (78, 910 and 1112 years) and 2
classes of the hair-mercury concentration (10 mg/g, >10 mg/g).
Similar graphs using the mothers hair-mercury concentration (not
shown) were also realized and showed very similar patterns.
Finally, we carried out additional analyses excluding the children
from the Sai Cinza village which had required the presence of an
interpreter.
3. Results
3.1. Population description
Among the 395 children included in the analyses, 64% were
from Brazil and 36% from French Guiana (Table 1). More girls were
examined in Sai Cinza and Brasilia Legal whereas a predominance
of boys was observed in the other villages, leading to a slight
majority of boys (52%) among the children examined. The children
had a mean age of 9.5 years. Missing data occurred for 24% of the
sample for the marital status of the mother, 31% for maternal
education and 30% for alcohol consumption of the mother; in
particular, the highest proportions of missing values involved the
Brasilia Legal (64% missing at least one variable), Sai Cinza (40%)
and Upper Maroni (38%) villages. Information available on the
mothers indicated that at least 85% were married/cohabitating,
except in Sai Cinza where only half of the mothers reported being
married/cohabitating. Except in Brasilia Legal and Awala, most
mothers had not completed primary school. Maternal consump-
tion of alcohol during pregnancy was more frequent in French
Guiana than in Brazil but was minimal: an occasional drink of
Cachiri, which has very low alcohol content.
The childrens hair-mercury concentrations ranged from
0.5 mg/g to 63.8 mg/g (mean, 9.8 mg/g). As expected in Brazil,
the children of Sai Cinza, the closest village to the gold-mining
areas of the three villages, had the highest average hair-mercury
concentrations (mean, 19.4 mg/g; range, 9.663.8 mg/g). The
children of Santana do Ituqu, located far from pollution sources,
showed lower average levels of MeHg exposure than the two other
Brazilian villages (mean, 4.4 mg/g; range, 0.512.4 mg/g), while
Brasilia Legal showed intermediate exposure levels (mean,
13.8 mg/g; range, 0.735.8 mg/g). In French Guiana, the children
from the Wayana community living near the Upper Maroni River
had higher hair-mercury concentrations (mean, 10.5 mg/g; range,
3.520.8 mg/g) than the Galipi children living on the Atlantic coast
(mean, 2.1 mg/g; range, 0.55.7 mg/g). Maternal hair-mercury
concentrations were slightly higher than those of the childs hair
(mean, 10.3 mg/g among all mothers fromboth sites for whomhair
samples were available) and ranged from 0.6 mg/g to 41.7 mg/g.
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 39
Table 1
Background data on the Amazonian children examined with the StanfordBinet Copying task, according to residence.
Sai-Cinza (Brazil) Brazilia Legal
(Brazil)
Santana do Ituqu
(Brazil)
Upper Maroni
(French Guiana)
Awala (French
Guiana)
All villages Scores of the StanfordBinet
Copying test
n % n % n % n % n % n % Relaxed design
Copying score;
mean (s.e.)
Block score;
mean (s.e.)
Children examined 65 94 94 71 71 395 13.8 (0.1) 8.3 (0.1)
Boys 25 38 42 45 51 54 42 59 45 63 205 52 14.0 (0.1) 8.8 (0.2)
Girls 40 62 52 55 43 46 29 41 26 37 190 48 13.6 (0.2) 7.7 (0.2)
Age in year; mean (s.e.) 9.7 (0.2) 9.5 (0.2) 9.7 (0.2) 9.2 (0.2) 9.3 (0.2) 9.5 (0.1)
78 years old 128 32 12.8 (0.3) 7.5 (0.3)
910 years old 141 36 14.0 (0.2) 8.4 (0.2)
1112 years old 126 32 14.6 (0.1) 9.0 (0.2)
Marital status of the mother (g)
Married/cohabitating 20 49 39 93 79 94 58 88 57 85 253 84 13.9 (0.1) 8.9 (0.2)
Other status 21 51 3 7 5 6 8 12 10 15 47 16 13.2 (0.4) 7.4 (0.5)
Missing 24 52 10 5 4 95 13.8 (0.3) 7.1 (0.3)
Educational level of the mother
Not completed primary school 24 62 7 21 43 51 49 86 8 14 131 48 1.2 (0.2) 8.5 (0.3)
Primary or higher school 15 38 27 79 41 49 8 14 50 86 141 52 14.2 (0.2) 8.9 (0.2)
Missing 26 60 10 14 13 123 13.9 (0.2) 7.4 (0.3)
Alcohol consumption of the mother
Yes 3 7 9 21 23 27 50 93 27 50 112 41 13.6 (0.2) 9.5 (0.2)
No 38 93 33 79 61 73 4 7 27 50 163 59 14.1 (0.2) 7.9 (0.2)
Missing 24 52 10 17 17 120 13.6 (0.2) 7.6 (0.3)
Hair-mercury concentration (mg/g)
Childs hair; mean (s.e.) n = 64; 19.4 (1.2) n = 89; 13.8 (0.7) n = 93; 4.4 (0.2) n = 65; 10.5 (0.4) n = 65; 2.1 (0.1) n = 374; 9.8 (0.4)
Maternal hair; mean (s.e.) n = 46; 18.8 (1.1) n = 42; 15.1 (1.3) n = 63; 4.8 (0.3) n = 58; 12.9 (0.5) n = 59; 3.5 (0.2) n = 269; 10.3 (0.5)
n: number; %: percent; s.e.: standard error. Range of the design Copying score: 016; range of the Block score: 012.
C
.
C
h
e
v
r
i
e
r
e
t
a
l
.
/
N
e
u
r
o
T
o
x
i
c
o
l
o
g
y
3
0
(
2
0
0
9
)
3
7

4
6
4
0
Correlations between the hair-mercury concentrations of the child
and the mother varied according to village (r ranged from 0.50 to
0.57 in Brazil; r = 0.28 in Awala; r = 0.09 in Upper Maroni).
3.2. The SB Copying test
The relaxed Copying score and the Block score of the SB
Copying test varied between communities; the lowest scores
occurred in Sai Cinza. Girls had slightly lower scores on average
than boys, especially for the Block task (Table 1). Increasing mean
scores were seen with the childs age.
The joint analysis of the Brazil and the French Guiana studies
showed that the hair-mercury concentrations of both the child and
the mother correlated signicantly and negatively with both the
relaxed Copying score and the Block score of the SB test (Table 2).
For example, the drawing task score was 1.2 points lower
(s.e. = 0.3) for children with hair-mercury concentrations of
10 mg/g, compared to those with hair levels of 1 mg/g. For
comparison, the adjusted regression coefcient for age was 0.47
(s.e. = 0.07, p < 0.001). The association between the mercury
concentrations and the Block score was signicant only for the
Brazilian children. No signicant heterogeneity between studies
was observed in the relationship between the performance on the
design Copying task and hair-mercury concentration. Analyses
excluding the children of Sai Cinza reduced the strength of the
associations which however remained statistically signicant with
the hair-mercury concentrations of the child.
No interaction between sex and mercury exposure was
observed for performance on the SB Copying test. The association
between the relaxed score and hair-mercury concentration varied
with age of the child: stronger associations were observed among
younger children (Table 2 and Fig. 1a). In addition, the mean of the
relaxed score among the children aged 78 years and with hair-
mercury concentration of less than 10 mg/g (13.6, 95% condence
interval (CI): 13.2, 14.0) was similar to the score mean among the
more exposed children aged 910 years (14.0, CI: 13.6, 14.4) or
aged 1112 years (14.2, CI: 13.6, 14.8).
3.3. The qualitative error types of the SB copying task
The most common error type was the distortion error (among
89% of the children; Table 3), the overdrawing error (37%), the
rotation error (27%) and the simplication error (24%). Very few
children committed no errors (n = 12) and more than two types of
errors occurred for 58% of the children. Among the multiple
observedcombinations of error types, the most commonones were
distortionoverdrawing (8.3% of the children), distortionsimpli-
cation (4%), distortionrotation (3.8%) and distortionoverdraw-
ingrotation (3.2%).
Through multivariate analyses, sex was found to be signi-
cantly associated with the presence of perseveration errors, with
the presence of macrographic errors and with the presence of
tremor in the drawings; boys showed an increased risk of these
error types compared to girls. As expected, the younger children
had greater risk of committing drawing errors, including simpli-
cations, distortions, perseverations, and micrographic errors.
However, the risk of overdrawing errors increased with age. There
was no evidence for association between sex or age and presence of
rotation errors for these Amazonian children aged 712 years.
Table 4 shows that the risks of committing at least one rotation
error, one simplication error or one perseveration error on the 16
drawings increased signicantly with the levels of hair-mercury
fromboth the child and mother, whereas the risk of making at least
one tremor error decreased with the MeHg levels. For instance,
children with hair-mercury concentrations of 10 mg/g had a 5-fold
higher risk of committing a rotation error than children with an
exposure-level of 1 mg/g. No evidence of heterogeneity by study
site was observed for the association between mercury exposure
and error types, except for the hair-mercury level of the child and
the overdrawing error (Table 4). Analyses excluding the children of
Sai Cinza reduced the strength of the associations which however
remained statistically signicant for rotations, perseverations and
tremor errors.
Statistically signicant heterogeneity of the risk of making a
rotation error, micrographic error or simplication error asso-
ciated with the MeHg levels was observed with age (Table 4).
Among these errors, Fig. 1b showed that the decit (i.e. difference
between the frequencies of the error type of the most and the least
mercury-exposed groups) in the drawing task is smaller among the
older children for rotations and simplications, whereas the
heterogeneity observed with age for the micrographic error was
due to substantial risk of errors in children aged 1112 years
Table 2
Relationships between Copying scores and maternal and child hair-mercury
concentrations.
Child hair-mercury
concentration
(mg/g, log
10
scale)
Mother hair-mercury
concentration
(mg/g, log
10
scale)
n b
a
p-Value n b
a
p-Value
Relaxed copying score
376 1.17 <0.001 268 1.35 0.003
1.02
b
0.04 0.50
b
0.41
243 0.96
c,d
0.009 224 0.80
c,e
0.05
Stratied by country
Brazil 1.40 0.002 1.33 0.007
French Guiana 1.87 0.002 1.80 0.002
Stratied by sex
Boys 0.91 0.01 1.22 0.008
Girls 1.49 0.002 1.54 0.01
Stratied by age
78 years 2.32
*
0.002 2.84
*
0.001
910 years 0.53
*
0.23 1.11
*
0.06
1112 years 0.34
*
0.28 0.02
*
0.96
Block score
376 3.01 <0.001 268 2.63 <0.001
0.54
b
0.24 0.55
b
0.33
243 2.27
c,f
<0.001 224 2.24
c,f
<0.001
Stratied by country
Brazil 3.50
*
<0.001 3.25
*
<0.001
French Guiana 0.10
*
0.78 0.18
*
0.69
Stratied by sex
Boys 2.65 <0.001 2.06 <0.001
Girls 3.51 <0.001 3.43 <0.001
Stratied by age
78 years 2.81 <0.001 2.84 <0.001
910 years 3.01 <0.001 2.03 0.01
1112 years 3.20 <0.001 3.13 <0.001
n: Number; b: multiple regression coefcient; p-value: degree of signicance of
regression coefcient (two-sided).
a
Adjustment on sex and childs age (in year); adjustment only on age when
stratied by sex and only on sex when stratied by age.
b
Additional adjustment on village if p < 0.20.
c
Analyses restricted among subjects without missing data for the three
covariates (alcohol consumption, marital status and education of the mother).
d
Additional adjustment with backward selection (p < 0.20) on marital status of
the mother, maternal education and maternal alcohol consumption.
e
Additional adjustment with backward selection (p < 0.20) on marital status of
the mother and maternal education.
f
Additional adjustment with backward selection (p < 0.20) on marital status of
the mother and maternal alcohol consumption.
*
Statistically signicant heterogeneity between strata (p < 0.05).
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 41
(Fig. 1b). We observed that for most of the error types (except the
overdrawing error), the frequency of errors observed among the
youngest and the least exposed children group was similar or
higher than the frequency observed among the oldest but the most
exposed children group.
Among the French Guianan children aged 56 years, the
frequency of error types were generally consistent with the
suggested tendency with age among the children aged 712 years:
62%, 100%, 76%, 19%, 10%, 41%, 9% and 21% for respectively
rotations, distortions, simplications, perseverations, overdraw-
ing, micrographic, macrographic and tremor errors. Their risk of
making rotation errors increased and their risk of making
micrographic errors decreased with the hair-mercury concentra-
tion of both the child (respectively, odds-ratio (OR) = 6.1, CI: 1.6,
22.8; OR = 0.3, CI: 0.1, 1.0) and the mother (respectively, OR = 13.7,
CI: 2.4, 79; OR = 0.2, CI: 0.0, 0.8) but there was no association
between simplication errors and MeHg levels.
Additional adjustments of the results have been shown within
Tables 2 and 4. Additional adjustment for village provided
substantially larger condence intervals. Except for tremor errors,
conclusions remained unchanged (Table 4). Analyses with
additional adjustment using a backward strategy included a
smaller number of participants due to missing information on
some covariates. The strength of the associations was reduced
(when compared with results fromthe same subgroup of children
with complete data but without the additional adjustment;
results not shown), but statistically signicant results were
maintained.
Fig. 1. Exploring the potential heterogeneities (a) of the scores of the StanfordBinet Copying test and (b) of the frequencies of error types in the drawing task between the
youngest but the least exposed children to mercury and the oldest but the highest exposed children. s.e.: Standard error
Table 3
Frequency (%) of error types of the StanfordBinet Copying task among Amazonian children, according to residence, sex and age.
Type of error Rotation
(at least one)
Distortion
(at least two)
Simplication
(at least one)
Perseveration
(at least one)
Overdrawing
(at least one)
Micrographia
(at least one)
Macrographia
(at least one)
Tremor
(at least one)
All children (n = 373) 27.1 63.0 23.6 5.6 37.3 15.5 5.1 8.6
Village-areas
Sai-Cinza (Brazil) 52.3 64.6 53.8 9.2 10.8 21.5 1.5 4.6
Brasilia Legal (Brazil) 23.8 51.3 8.8 6.3 53.8 3.8 3.8 2.5
Santana do Ituqu (Brazil) 33.0 36.2 18.1 3.2 51.1 14.9 0.0 4.3
Upper Maroni (French Guiana) 19.1 92.6 29.4 10.3 27.9 25.0 8.8 11.8
Awala (French Guiana) 6.1 83.3 13.6 0.0 33.3 15.2 13.6 22.7
Sex
Boys 26.3 60.8 20.1 7.2 40.2 15.5 7.2 11.9
Girls 27.9 65.4 27.4 3.9 34.1 15.6 2.8 5.0
Age
78 years old 29.4 73.1 37.8 10.1 27.7 25.2 2.5 10.9
910 years old 29.6 64.4 16.3 3.7 35.6 11.1 7.4 8.1
1112 years old 21.8 51.3 17.6 3.4 48.7 10.9 5.0 6.7
All the values are given in percentage.
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 42
4. Discussion
The present study provides results of a joint analysis of
Amazonian children fromBrazil and French Guiana, indicating that
the SB Copying task, including both relaxed and qualitative
scoring systems, is a sensitive measure for assessing mercury
neurotoxicity effects in multiple populations. Using an objective
relaxed scoring system, the study rened previously published
results that MeHg exposure had a strong negative association with
performance on copying tasks in children (Grandjean et al., 1999;
Cordier et al., 2002). Among children aged 712 years, the decit
on the SB Copying task of children with a 10 mg/g hair-mercury
exposure (compared to children with a 1 mg/g level) corresponds
to a developmental delay equivalent of at least 2 years. The decit
on the Copying task linked with the hair-mercury concentration
was characterized by different error types: there were increased
risks of making rotation or simplication errors on the drawings
among children with an increased level of MeHg exposure. The
statistically signicant results of an increased risk of perseveration
errors and a decreased risk of tremor errors on the drawings among
children associated with an increased level of MeHg exposure
should be interpreted with more caution since the frequency of
these errors was low. In addition, the tremor error was no longer
signicant when adjusted by village parameter. Also we should
note that this error type was not a very sensitive measure of tremor
of the childs hand but rather a crude measure of motor functioning
indicating whether the drawings appeared shaky or not. The score
Table 4
Odds-ratio of error presence in the StanfordBinet Copying task with the hair-
mercury concentration of both the child and the mother, according to different
types of error (stratied analyses are shownwhen there were more than 12 children
in each stratum).
Child hair-mercury
concentration
(mg/g, log
10
scale)
Mother hair-mercury
concentration
(mg/g, log
10
scale)
n OR
a
95% CI OR
a
95% CI
Rotation error (at least one) 355 5.1 2.6, 10.1 258 3.3 1.5, 7.6
7.6
b
2.0, 29.2 4.5
b
1.0, 20.4
235 3.4
c
1.4, 7.9 217 2.1
c
0.8, 5.3
Stratied by country
Brazil 233 3.5 1.6, 7.7 147 2.2 0.8, 5.7
French Guiana 122 4.8 1.0, 21.8 111 3.7 0.6, 23.2
Stratied by sex
Boys 185 5.0 2.0, 12.1 136 2.5 0.9, 7.2
Girls 170 5.4 1.9, 15.5 122 4.2 1.4, 19.2
Stratied by age
78 years 116 3.5
*
1.2, 10.2 83 5.7 1.3, 24.3
910 years 127 21.8
*
5.4, 88.2 91 4.5 1.1, 18.7
1112 years 112 1.9
*
0.6, 6.4 84 1.2 0.3, 5.6
Distortion error (at least two) 1.5 0.8, 2.6 1.4 0.7, 3.0
3.6
b
1.1, 11.3 1.2
b
0.3, 5.1
1.1
d
0.5, 2.3 1.4
e
0.6, 3.2
Stratied by country
Brazil 4.9 2.2, 10.9 3.5 1.3, 9.6
French Guiana 3.1 0.7, 13.6 2.0 0.4, 11.2
Stratied by sex
Boys 1.3 0.6, 2.6 1.3 0.5, 3.2
Girls 1.8 0.8, 4.3 1.8 0.6, 5.7
Stratied by age
78 years 1.6 0.6, 4.3 1.5 0.4, 5.5
910 years 1.2 0.5, 2.9 1.5 0.4, 5.2
1112 years 1.5 0.6, 4.2 1.2 0.4, 4.2
Simplication error (at least one) 3.4 1.7, 6.8 5.1 2.1, 12.4
2.2
b
0.5, 9.0 4.9
b
0.9, 26.0
4.5
f
1.8, 11.0 5.4
c
1.9, 15.5
Stratied by country
Brazil 0.7 1.5, 9.4 5.5 1.7, 18.4
French Guiana 3.9 1.2, 12.3 4.5 1.1, 17.6
Stratied by sex
Boys 3.5 1.3, 9.0 5.0 1.5, 17.0
Girls 3.7 1.4, 10.1 5.7 1.5, 21.3
Stratied by age
78 years 5.1 1.8, 14.6 15.8
*
3.5, 72.6
910 years 6.5 1.5, 27.3 14.8
*
2.1, 10.7
1112 years 1.0 0.3, 3.9 0.4
*
0.1, 2.4
Perseveration error (at least one) 5.6 1.5, 21.1 3.6 0.6, 20.9
2.9
g
0.3, 23.8 3.4
c
0.3, 36.7
Overdrawing error (at least one) 0.6 0.4, 1.1 0.6 0.3, 1.3
0.9
b
0.3, 2.7 1.5
b
0.4, 5.8
0.7
h
0.3, 1.6 0.5
h
0.2, 1.1
Stratied by country
Brazil 0.3
*
0.1, 0.6 0.5 0.2, 1.4
French Guiana 1.2
*
0.5, 3.4 0.7 0.2, 2.2
Stratied by sex
Boys 0.6 0.3, 1.2 0.5 0.2, 1.2
Girls 0.7 0.3, 1.6 0.8 0.2, 2.8
Stratied by age
78 years 1.1 0.4, 2.9 0.7 0.2, 2.6
910 years 0.5 0.2, 1.3 0.8 0.2, 2.8
1112 years 0.4 0.2, 1.2 0.5 0.1, 1.7
Micrographia error (at least one) 1.3 0.6, 2.6 1.4 0.6, 3.6
1.8
b
0.4, 8.1 1.8
b
0.3, 10.3
0.9
c
0.3, 2.2 1.1
c
0.4, 3.2
Table 4 (Continued )
Child hair-mercury
concentration
(mg/g, log
10
scale)
Mother hair-mercury
concentration
(mg/g, log
10
scale)
n OR
a
95% CI OR
a
95% CI
Stratied by country
Brazil 1.4 0.5, 4.1 1.3 0.4, 5.1
French Guiana 2.1 0.7, 6.5 2.3 0.6, 8.7
Stratied by sex
Boys 1.1 0.4, 2.9 1.4 0.4, 4.5
Girls 1.8 0.6, 5.4 1.7 0.4, 7.2
Stratied by age
78 years 0.6
*
0.2, 1.6 0.6
*
0.2, 2.3
910 years 1.0
*
0.3, 3.9 1.9
*
0.3, 12.3
1112 years 34.1
*
2.9, 394 14.1
*
0.9, 210
Macrographia error (at least one) 0.5 0.2, 1.6 1.7 0.4, 7.8
1.2
d
0.2, 6.9 1.0
g
0.1, 7.9
Tremor error (at least one) 0.3 0.1, 0.8 0.3 0.1, 1.0
1.6
b
0.2, 11.6 1.1
b
0.1, 11.1
0.1
f
0.0, 0.5 0.3
f
0.0, 1.0
OR: Odds-ratio; CI: condence interval.
a
Adjustment on sex and childs age (in year); adjustment only on age when
stratied by sex and only on sex when stratied by age.
b
Additional adjustment on village if its p < 0.20.
c
Analyses restricted among subjects without missing data for the three
covariates (alcohol consumption, marital status and education of the mother),
and with additional adjustment with backward selection (p < 0.20) on maternal
education and maternal alcohol consumption.
d
Additional adjustment with backward selection (p < 0.20) on maternal alcohol
consumption.
e
Additional adjustment with backward selection (p < 0.20) on marital status of
the mother and maternal alcohol consumption.
f
Additional adjustment with backward selection (p < 0.20) on no additional
adjustment.
g
Additional adjustment with backward selection (p < 0.20) on maternal
education.
h
Additional adjustment with backward selection (p < 0.20) on marital status of
the mother.
*
Statistically signicant heterogeneity between strata (p < 0.05).
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 43
on the Block task showed heterogeneous association with mercury
exposure depending on study site; this result suggests that the
Block score may be sensitive to subtle factors in the administration
of the task. For example, the examination room was not the same
for all children although efforts were made to secure undisturbed
and uniform conditions.
As anticipated, the age of the child had a large effect on the
childrens performance on SB Copying task, with older children
generally having better scores than the younger children, except
for the overdrawing errors that were committed more commonly
by the older children. This nding suggests that items of the SB
Copying task possess the range of difculty necessary to detect
visuospatial developmental differences in these samples of
Amazonian children. In addition, the pattern of increased over-
drawing errors in older children was also noted in clock drawings
of American children (Cohen et al., 2000) where erasures and
second attempts to further correct their drawings were signi-
cantly more common in the 12-year-old children compared with
younger children. This performance likely reects improved self-
regulatory abilities of the maturing frontal lobes in the older
children and their attempts to self-correct their errors (Stuss,
1992). Moreover the high frequency of rotations errors among the
French Guianan children aged 56 years, coincides with the study
of clock drawing in which 33% of 6-year-olds committed number
reversal errors but only 4% of 7-year-olds made the same errors
(Cohen et al., 2000). Plausible associations between the sex of the
child and test performance were also observed, in particular lower
Block scores, fewer perseveration errors, macrographia and tremor
among girls compared to boys. These associations may reect
different developmental trajectories in visuospatial organizationof
the two sexes (Kimura, 1996) particularly with regard to
myelination rates and lateralization of the hemispheres (Kara-
petsas and Vlachos, 1997).
The statistical analyses of the present study included different
adjustment strategies, as applied in previous studies on develop-
mental neurotoxicity of mercury in various settings and different
study designs (Grandjean et al., 1997; Crump et al., 1998; Murata
et al., 1999; Weihe et al., 2002). The rst strategy included sex and
age of the child as established predictors and therefore mandatory
covariates. Additional adjustment for village that may control for
some potential unmeasured confounders may not be appropriate
because the levels of mercury exposure differed substantially
between the villages. However, the main drawback of this study
designof intercommunitycomparisonis the difcultyincompletely
controlling for confounders and there is no strict statistical rule
dealing with the inclusion of highly correlated variables in the same
regressionmodel. Wehavethereforeprovidedtheseresultsadjusted
on village that despite the larger condence intervals for the hair-
mercury coefcient showed mostly similar conclusions; incon-
sistent results were observed for tremor errors probably due to a
very low frequency of this type of errors in some villages. Maternal
intelligence should be considered as an established predictor. In the
Brazilian study, maternal intelligence was not measured. We
therefore dened a low maternal education as an optional
adjustment factor after having veried that it was associated with
the level of the maternal Raven score assessed only in the French
Guiana study. Since the most prudent approach is generally to
include all potential confounders in the adjustment, the choice of a
backward selection with a non-strict criterion (p < 0.20) for the
second strategy therefore seems to be appropriate. However these
adjusted-analyses were limited by a large proportion of missing
values from communities with the highest levels of exposure. They
provided consistent results, however.
This study relied on surrogate measures of developmental
exposure of the child, since exposure was assessed several years
after birth. The assumption was therefore made that the level of
mercury pollution fromsh contamination was fairly stable during
the previous decade and that dietary habits, in particular sh
consumption, did not change signicantly over time. Such
hypotheses seem to be reasonable, particularly in the French
Guiana area, where hair-mercury levels assessed in 1994 and in
1997 remained constant (Cordier et al., 1998, 2002). Despite
incomplete data in the Brazilian study on maternal exposures, the
correlation between the hair-mercury concentrations of child and
mother were high (r from 0.50 to 0.57 according to villages, all ps
<0.001): this suggests that the current hair-mercury concentration
of the childs hair may also reect past exposures, perhaps
including prenatal exposure. In French Guiana weaker correlations
were observed between the hair-mercury levels of the child and
the mother. Their lowvalues may suggest that the children did not
share meals with their mothers. Thus, these low correlations may
partly explain the differences in the strength of the MeHg exposure
effects for child and mother on the Copying scores. Further, it is
more likely to introduce measurement errors when assessing
prenatal and early developmental exposures among older children.
Therefore, we cannot exclude the possibility that a stronger
association between MeHg exposure and scores in younger
children may be partly due to the fact that the hair-mercury
concentration is a slightly better predictor of developmental
exposures in the younger children. Finally, both sets of exposure
variables used in this study (as well as in the previous studies) are
prone to variability that cannot be measured. Moreover, because
the total imprecision is likely to be substantially greater than the
laboratory error of the mercury analysis and because of the
substantial difference of MeHg-exposure levels between villages,
additional adjustment for village should not be advisable in the
present study and would cause bias due to the imprecision of the
exposure parameter (Budtz-Jrgensen et al., 2003).
Because all of the examined children were chronically exposed
to MeHg during gestation and into childhood, the separate impacts
of prenatal and postnatal exposure could not be distinguished.
Results indicating that MeHg exposure was associated with poorer
overall visuospatial functions in a dose-related manner coincide
with previous ndings from the Faroe Islands cohort of 7-year-old
children (Grandjean et al., 1997, 1998). Additionally, individuals
from Japan with a history of Minamata disease showed signicant
difculty completing the Bender Gestalt drawings (Harada, 1989).
The sensitivity of the visuospatial domain to effects of MeHg may
reect preferential effect of mercury on posterior brain areas (Choi,
1989; Davis et al., 1994) including the parietal and occipital lobes.
Alternatively, subtle changes in white matter and other brain
structures including the frontal lobes attributable to the diffuse
neuropathological effects noted in prenatal and childhood MeHg
exposures may disrupt the integrity of information processing by
the brain, differentially affecting visuospatial function (Choi et al.,
1978; Lapham et al., 1995). The added qualitative error scoring in
the present study was carried out in order to parse out differential
performances of the children with regard to space perception,
visuomotor coordination or self-regulation, and therefore to
differentiate specic developmental patterns of sensitivity to
MeHg. Possible neurotoxicological mechanisms for the results of
the current study involve several specic brain areas. The parietal
lobes have long been thought to be involved in the completion of
tasks requiring copying to command (Kaplan, 1988) and the frontal
lobes are known to be involved in executive system functions such
as self-regulation and awareness of decit (Stuss, 1992). Severe
parietal and occipital lobe atrophy was noted in brain imaging
studies of two New Mexico adults exposed as children to pork
contaminated with MeHg (Davis et al., 1994) and mild insults to
these areas would be expected in children with chronic MeHg
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 44
exposure. Although the levels of exposure in this previous study
were signicantly higher than in the present study, it suggests
plausible mechanisms that may underlie the decits indicated by
the current results. It is therefore likely that early damage to
specic visuospatial processing areas (parietal and/or occipital
lobes) could be reected in the specic qualitative errors made by
the MeHg-exposed children including rotations and simplica-
tions of the Copying task designs. Further neurotoxicological
studies of specic error types in graphomotor tasks performed
longitudinally are required to corroborate these ndings.
No sex-specic association between performance on the SB
Copying task and MeHg exposure was suggested in the present
study. However, more effects were generally observed for the
younger children than for the older children, except for the
micrographic error. The greater association between mercury
exposure and risk of rotation errors among younger children aged
56 years could have been explored only in the French Guiana data
but supports this nding. The further exploration of these age-
heterogeneities proposed in Fig. 1 has provided additional but only
descriptive results. It is entirely possible that some error types
could reect developmental delays while other error types reect
more permanent deviance in development in the same individuals,
particularly since the MeHg exposures have continued since
gestation. The present study had a cross-sectional design. There-
fore, whether the effects measured in the present study,
particularly apparent among younger children, reect temporary
developmental delays or permanent decrements cannot be reliably
determined at this point in time.
Conversely, whether the better performances of the older
children actually reect functional recovery of visuospatial
processing vs. the employment of compensatory mechanisms to
circumvent the visuospatial decit is unknown. This latter
phenomena was observed in children with congenital focal brain
injuries resulting in decits to the right hemisphere where the
children appeared to have functionally recovered from their
graphomotor decits on a house drawing task as they aged;
however, when asked to draw something more novel where they
did not have a ready-made compensatory strategy or graphomotor
formula, these children performed much lower than aged-matched
control children (Stiles et al., 1997). At the 14-year follow-up of the
Faroese cohort, decrements in visuospatial function appeared to be
only slightly smaller than at age 7 (Debes et al., 2006). In addition,
studies of children with fetal alcohol syndrome have also reported
lasting decits in visuospatial functioning (Uecker and Nadel,
1998). The permanence of decrements in cognitive abilities
acquired following exposure to other neurotoxicants during
development has been studied in adults who had documented
histories of childhood lead intoxication. The results indicated
lasting effects, including poorer cognitive testing performances in
adulthood accompanied by lower employment status (White et al.,
1993). Chronic exposure to low levels of lead has also been
associated with lower educational attainment than that of
unexposed controls (Needleman et al., 1990). When taken together
and applied to MeHg, these results suggest that cognitive
development may take longer in the more exposed groups and
may not be as complete, resulting in potentially serious, long-term
consequences.
Acknowledgements
The French Guiana study was supported by a grant from the
French National Network of Public Health (RNSP). The Brazil study
was supported by grants from the European Commission DG-1
Tropical Rainforest Programme (B7-6201), Imperial College Con-
sultants, London, and from the Danish Medical Research Council.
We are grateful to David Cleary, Ph.D., Elizabeth Oliviera Dos
Santos and the eld teams who examined the children in Brazil and
in French Guiana.
Conict of interest
None.
References
Akagi H, Malm O, Kinjo Y, Harada M, Branches F, Pfeiffer WC, et al. Methylmercury
pollution in Amazon, Brazil. In: Proceedings of the international symposium on
assessment of environmental pollution and health effects from methylmercury.
World Health Organization, Japan; 1993.
Amin-Zaki L, Majeed MA, Elhassani SB, Clarkson TW, Greenwood MR, Doherty RA.
Prenatal methylmercury poisoning, clinical observations over ve years. Am J Dis
Child 1979;133:1727.
Bernstein JH, Waber DP. Developmental neuropsychological assessment: the systemic
approach.In: Boulton AA, Baker GB, Hiscock M, editors. Neuromethods. Neurop-
sychology, vol. 17. Clifton, NJ: Humana Press; 1990.
Budtz-Jrgensen E, Keiding N, Grandjean P, Weihe P, White RF. Consequences of
exposure measurement error for confounder identication in environmental
epidemiology. Stat Med 2003;22:3089100.
Cernichiari E, Brewer R, Myers GJ, Marsh DO, Lapham LW, Cox C, et al. Monitoring
methylmercury during pregnancy: maternal hair predicts fetal brain exposure.
Neurotoxicology 1995;16:70510.
Choi BH. The effects of methylmercury on the developing brain. Prog Neurobiol
1989;32:44770.
Choi BH, Lapham LW, Amin-Zaki L, Saleem T. Abnormal cell migration, deranged
cerebral cortical organization and diffuse white matter astrocytosis of human
fetal brain: a major effect of methylmercury poisoning in utero. J Neuropathol Exp
Neurol 1978;37:71933.
Cohen MJ, Ricci CA, Kibby MY, Edmonds J. Developmental progression of clock face
drawing in children. Child Neuropsychol 2000;6:6476.
Cordier S, Grasmick C, Paquier-Passelaigue M, Mandereau L, Weber JP, Jouan M.
Mercury exposure in French Guiana: levels and determinants. Arch Env Health
1998;53:299303.
Cordier S, Garel M, Mandereau L, Morcel H, Doineau P, Gosme-Seguret S, et al.
Neurodevelopmental investigations among methylmercury-exposed children in
French Guiana. Env Res 2002;89:111.
Crump KS, Kjellstrom T, Shipp AM, Silvers A, Stewart A. Inuence of prenatal mercury
exposure upon scholastic and psychological test performance: benchmark analysis
of a New Zealand cohort. Risk Anal 1998;18:70113.
Davis LE, Kornfeld M, Mooney HS, Fiedler KJ, Haaland KY, Orrison WW, et al.
Methylmercury poisoning: long-termclinical radiological toxicological and patho-
logical studies of an affected family. Ann Neurol 1994;35:6808.
Debes F, Budtz-Jorgensen E, Weihe P, White RF, Grandjean P. Impact of prenatal
methylmercury exposure on neurobehavioral function at age 14 years. Neurotox-
icol Teratol 2006;28:36375.
Frery N, Maury-Brachet R, Maillot E, Deheeger M, de Merona B, Boudou A. Gold-mining
activities and mercury contamination of native amerindian communities in French
Guiana: key role of sh in dietary uptake. Env Health Perspect 2001;109:44956.
Grandjean P, Weihe P, Nielsen JB. Methylmercury: signicance of intrauterine and
postnatal exposures. Clin Chem 1994;40:1395400.
Grandjean P, Weihe P, White RF, Debes F, Araki S, Yokoyama K, et al. Cognitive decit in
7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Ter-
atol 1997;19:41728.
Grandjean P, Weihe P, White RF, Debes F. Cognitive Performance of children prenatally
exposed to safe levels of methylmercury. Env Res Sect A 1998;77:16572.
Grandjean P, White RF, Nielsen A, Cleary D, Santos EO. Mercury neurotoxicity in
Amazonian children downstream from gold mining. Env Health Perspect
1999;107:58791.
Grandjean P, Jrgensen PJ, Weihe P. Validity of mercury exposure biomarkers. In:
Wilson SH, Suk WA, editors. Biomarkers of environmentally associated disease.
Boca Raton, FL: CRC Press/Lewis Publishers; 200223547.
Grandjean P, Harari R, Barr DB, Debes F. Pesticide exposure and stunting as indepen-
dent predictors of neurobehavioral decits in Ecuadorian school children. Pedia-
trics 2006;117:54656.
Harada M. The intrauterine methylmercury poisoning known as congenital Minamata
diseasea 20 year serial investigation and its recent problems. In: Tsuru S, editor.
Proceedings of international forum on Minamata diseaseKesiou Syobo, Tokyo,
1988; 1989.
Harada M. Minamata disease: methylmercury poisoning in Japan caused by environ-
mental pollution. Crit Rev Toxicol 1995;25:124.
Kaplan E. A process approach to neuropsychological assessment. In: Boll T, Bryant BK,
editors. Clinical neuropsychology and brain function: research, measurement and
practice. Washington, DC: American Psychological Association; 1988.
Karapetsas AB, Vlachos FM. Sex and handedness in development of visuospatial skills.
Perceptual Motor Skills 1997;85:13140.
Kimura D. Sex, sexual orientation and sex hormones inuence human cognitive
function. Curr Opin Neurobiol 1996;6:25963.
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 45
Lapham LW, Cernichiari E, Cox C, Myers GJ, Baggs RB, Brewer R, et al. An analysis of
autopsy brain tissue from infants prenatally exposed to methymercury. Neuro-
toxicology 1995;16:689704.
MalmO. Gold mining as a source of mercury exposure in the Brazilian Amazon. Env Res
Sect A 1998;77:738.
Mason EM. Percent of agreement among raters and rater-reliability of the copying
subtest of the StanfordBinet Intelligence Scale: fourth edition. Perceptual Motor
Skills 1992;74:34753.
Murata K, Weihe P, Renzoni A, Debes F, Vasconcelos R, Zino F, et al. Delayed evoked
potentials in children exposed to methylmercury from seafood. Neurotoxicol
Teratol 1999;21:3438.
Myers GJ, Davidson PW, Cox C, Shamlaye CF, Palumbo D, Cernichiari E, et al. Prenatal
methylmercury exposure from ocean sh consumption in the Seychelles child
development study. Lancet 2003;361:168692.
Needleman HL, Schell A, Bellinger D, Leviton A, Allred EN. The long-term effects of
exposure to low doses of lead in childhood, an 11-year follow-up report. N Engl J
Med 1990;322:838.
Pfeiffer WC, Lacerda LD. Mercury inputs into the Amazon region Brazil. Env Technol
Lett 1988;9:32530.
Stiles JL, Trauner D, Engel M, Nass R. The development of drawing in children with
congenital focal brain injury: evidence for limited functional recovery. Neurop-
sychologia 1997;35:299312.
Stuss D. Biological and psychological development of executive functions. Brain Cognit
1992;20:823.
Thorndike RL, Hagen EP, Sattler JM. StanfordBinet intelligence scale. 4th ed. Chicago:
Riverside Publishing Company; 1986.
Uecker A, Nadel L. Spatial but not object memory impairments in children with fetal
alcohol syndrome. Am J Ment Retard 1998;103:128.
Weihe P, Hansen JC, Murata K, Debes F, Jrgensen PJ, Steuerwald U, et al. Neurobe-
havioral performance of inuit children with increased prenatal exposure to
methylmercury. Int J Circumpolar Health 2002;61:419.
White RF, Diamond R, Proctor S, Morey C, Hu H. Residual cognitive decits 50 years
after childhood lead poisoning during childhood. Br J Ind Med 1993;50:61322.
White RF, Debes F, Dahl R, Grandjean P. Development and eld testing of a neurop-
sychological test battery to assess the effects of methylmercury exposure in the
Faroe Islands. In: Proceedings of the international symposium on assessment of
environmental pollution and health effects of methylmercury. Kumamoto;
1994;11240.
C. Chevrier et al. / NeuroToxicology 30 (2009) 3746 46

You might also like