INTERNATIONAL PROGRAMME ON CHEMICAL SAFETY
WORLD HEALTH ORGANIZATION
SAFETY EVALUATION OF CERTAIN FOOD
ADDITIVES AND CONTAMINANTS
WHO FOOD ADDITIVES SERIES: 44
Prepared by the Fifty-third meeting of the Joint FAO/WHO
Expert Committee on Food Additives (JECFA)
World Health Organization, Geneva, 2000
IPCS - International Programme on Chemical Safety
METHYLMERCURY
First draft prepared by Dr D.C. Bellinger1, Dr M. Bolger2, Dr C.
Carrington2, Dr E. Dewailly3, Dr L.P.A. Magos4 and Dr B.
Petersen5
1Harvard Medical School, Boston, Massachusetts, USA;, 2US Food and
Drug Administration, Washington DC, USA; 3Centre de Sante Publique du
Québec, Québec, Canada, 4TNO BIBRA International Ltd, Carshalton,
Surrey, United Kingdom; and 5Novigen Sciences Inc., Washington DC,
USA
Explanation
Biological data
Pharmacokinetics
Absorption
Distribution
Transfer from mother to offspring
Placental transfer
Lactation
Clearance
Biochemical aspects
Cleavage of carbon-mercury bond
Complexes with thiol radicals
Interaction with selenium
Toxicological studies
Acute toxicity
Renal and hepatic toxicity
Anorexia
Neurotoxicity
Small rodents
Non-human primates
Domestic animals
Reproductive and developmental toxicity (other than
neurotoxicity)
Developmental neurotoxicity
Exposure in utero
Exposure in utero and postnatally
Exposure after parturition
Carcinogenicity
Immunomodulation
Extrapolation between species
Observations in humans
Case series
Childhood development
Neurological status
Developmental milestones
Early development
Development later in childhood
Sensory, neurophysiological and other end-points
Adult neurological, neurophysiological, and sensory function
Bias: Covariates, confounders and effect modifiers
Study in the Faroe Islands
Study in the Seychelles
Study in the Amazon Basin
Study in New Zealand
Study in Peru
Reanalysis of the study in Iraq
Estimates of dietary intake
Environmental mercury
Biomarkers of exposure
Intake assessment
Residues
National intake estimates
Estimates based on WHO GEMS/Food diets
Estimates of intake by fish consumers at the 95th percentile
Comments
Evaluation
References
1. EXPLANATION
The Committee first evaluated methylmercury at its sixteenth
meeting (Annex 1, reference 30), when it established a provisional
tolerable weekly intake (PTWI) of 300 µg of total mercury per person,
of which no more than 200 µg should be present as methylmercury. At
its twenty-second and thirty-third meetings (Annex 1, references 47
and 83), the Committee confirmed the PTWI of 200 µg of methylmercury
(3.3 µg/kg bw) for the general population. At its thirty-third
meeting, the Committee noted that pregnant women and nursing mothers
may be at greater risk than the general population from the adverse
effects of methylmercury. The Committee considered the available data
insufficient to recommend a specific intake for this population group,
and it recommended that more detailed studies be undertaken.
At its present meeting, the Committee reviewed information that
had become available since the previous evaluation. The PTWI was not
reconsidered and was maintained at its present value. Two other WHO
publications have dealt with the effects of mercury and methylmercury
on human health (WHO, 1976, 1990). Relevant information from those
documents and the studies published since the report of the
thirty-third meeting are summarized and discussed in this monograph,
and the data were used to estimate the risks associated with exposure
to methylmercury. It should be noted that the doses given refer to the
mercury constituent of the organic mercury compound.
2. BIOLOGICAL DATA
2.1 Pharmacokinetics
2.1.1 Absorption
The dermal absorption of methylmercury is similar to that of
inorganic mercury salts (Friberg et al., 1961). Studies of
occupational exposure and studies on rats (Fang, 1980) and mice
(Ostlund, 1969) both indicate that pulmonary absorption accounts for
95% of the dose. Methylmercury in ligated segments of intestines was
absorbed 17-35 times faster than was an inorganic mercury salt (Sasser
et al., 1978). In volunteers (Aberg et al., 1969; Kershaw et al.,
1980), in squirrel monkeys (Berlin et al, 1975a), and in macaque
monkeys (Rice et al., 1989), the peak concentration in blood was
reached within 6 h of ingestion; 95% of an ingested dose was absorbed
by volunteers (Nittinen et al., 1973), squirrel monkeys (Berlin et
al., 1975a), and rats (Walsh, 1982).
2.1.2 Distribution
The distribution of methylmercury has three characteristics: (i)
a high concentration of mercury in blood and a high ratio of the
concentration in erythrocytes:plasma; (ii) greater ease of transfer
across the blood-brain and blood-placenta barriers than any other
mercury compound with the exception of elemental mercury vapour
(although the transfer of the latter is limited by its rapid
oxidation; Magos et al., 1989); and (iii) less renal deposition than
any other mercury compound.
Three days after a single intravenous dose of methylmercury to
rats at 0.5 µg/g, the blood concentration of mercury was 2.1 µg/ml and
that in the brain was 0.14 µg/g. In rats given the same dose as
mercuric acetate, the concentration of phenyl and methoxyethylmercury
was 0.042-0.068 µg/ml in blood and 0.018-036 µg/g in brain. The renal
concentration was 87% less in rats given methylmercury than in those
given the other mercury compounds (Swensson & Ulfvarson, 1967).
A high concentration of mercury in blood is associated with a
high concentration ratio in erythrocytes:plasma in every species
tested. The reported ratios are about 20 in humans (Miettinen, 1973;
Kershaw et al., 1980) and nearly 20 in squirrel monkeys (Berlin et
al., 1975a), guinea-pigs (Iverson et al., 1973), and sheep (Kostyniak,
1983). The ratio was 12 in hamsters (Omata et al., 1986), 9 in pigs
(Gyrd-Hansen, 1981), and 5-9 in mice (Ostlund, 1969; Sundberg et al.,
1998a) and rabbits (Berlin, 1963). In cats, the ratio was 42 (Hollins
et al., 1975), and in rats the reported ratios ranged from 128 to 288
(Ulfvarson, 1962; Norseth & Clarkson, 1970; Vostal & Clarkson, 1973).
The high ratio in rats has been attributed to the greater number of
thiol groups in rat haemoglobin (Naganuma & Imura, 1980), which
results in an eight times greater release of methylmercury from human
than from rat erythrocytes suspended in albumin (Doi & Tagawa, 1983).
Rat haemoglobin also has an increased capacity to bind alkyltin, which
has little affinity for thiol radicals (Rose & Aldridge, 1968). The
accumulation of methylmercury in rat blood is associated with
organ:blood concentration ratios of mercury that are lower than in any
other species.
Except in rats, the brain:blood ratios were greater than 1 in all
species tested (Table 1). The ratio was consistently high in monkeys,
and in all species it was higher after multiple dosing than after the
administration of a single dose. Differences in strain and sex
affected the concentration of mercury in blood of mice more than that
in brain. The concentration in the brain was higher in female than in
male mice. A similar sex difference in brain mercury concentrations,
but without a difference in the brain:blood ratio, was seen in rats.
Six to 12 days after four daily oral doses of methylmercury chloride
at 8 mg/kg bw, the concentration of mercury in brain was 8.8 µg/g in
female and 6.7 µg/g in male rats (Magos et al., 1981).
In heavily exposed squirrel monkeys, the brain stem had
approximately the same concentration as the cerebellum and most of the
cerebral regions, with the exception of the occipital lobe, which had
the highest concentration (Berlin et al., 1975c). The thalamus had
somewhat higher concentrations than the occipital pole (Vahter et al.
1994). In pigs, the concentrations in the cerebrum, cerebellum, and
optic nerve differed only slightly, and all had higher concentrations
than the spinal cord (Platonow, 1968). In guinea-pigs, the cerebellum
had the lowest concentration (Iverson et al., 1974). In rats, the
highest concentration was found in the spinal roots and ganglia,
closely followed by the cerebral cortex and the cerebellum (Somjen et
al., 1973a), but the concentrations in the cerebellum, medulla
oblongata and various areas of the cerebrum differed only slightly
(Magos et al., 1981).
Table 1. Organ:blood concentration ratios of mercury after treatment with methylmercury
Species Treatmenta Blood Organ:blood ratio Reference
(µg/g)
Brain Liver Kidney
Squirrel monkey Single dose, 0.63 3.1 (cc) 5.9 5.1 Berlin et al.
8 days (1975b)
Squirrel monkey < 2 months, 1.4 5.3 - - Berlin et al.
9-22 days (1975c)
Macaque monkey < 2 months, 2.4 2.7 13 21 Evans et al.
1 day (1977)
Macaque monkey < 2 months, 2.0 3.1 - - Stinson et al.
1 day (1989)
Macaque monkey < 2 months, 1.1 4.4 (o) - - Vahter et al.
1 day (1994)
Macaque and < 2 months, 0.45 3.1 12 47 Kawasaki et
rhesus monkeys 1 day al. (1986)
Pig Single dose, 0.39 3.3 12 17 Gyrd-Hansen
28 days (1981)
Pig 4-10 doses, 1.8 1.8 (c) 11 8.5 Platonow
1 day (1968)
Rabbit Single dose, 0.08 5.4 10 17 Petersson et
7 days al. (1989)
Cat < 2 months, 14 2.1 (cc) 5.2 2.6 Charbonneau
1 day et al. (1974)
Guinea-pig < 2 months, 3.4 1.8 (of) 8.2 21 Iverson et al.
1 day (1974)
Table 1. (cont'd)
Species Treatmenta Blood Organ:blood ratio Reference
(µg/g)
Brain Liver Kidney
Guinea-pig Single dose, 3.5 1.3 4.0 6.7 Iverson et al.
7 days (1973)
Rat, male Single dose, 3.5 0.08 - 1.2 Farris et al.
7 days (1977)
Rat, female Single dose, 36 0.08 0.4 1.1 Fang (1980)
4 days
Rat, male Single dose, 3.6 0.08 0.23 1.2 Thomas et al.
Rat, female 4-10 days 3.3 0.10 0.25 2.0 (1986)
Rat, male 4-10 doses, 40 0.07 (c) 0.3 1.7 Friberg (1959)
17 days
Rat, female < 2 months, 95 0.07 0.03 1.2 Magos &
1 day Butler (1976)
Hamster Single dose, 1.5 1.9 4.2 10 Omata et al.
16 days (1986)
Hamster 4-10 doses, 9.0 3.8 (cc) 5.5 9.8 Omata et al.
9 days (1986)
Table 1. (cont'd)
Species Treatmenta Blood Organ:blood ratio Reference
(µg/g)
Brain Liver Kidney
Hamster Single dose. 1.7 2.5 5.1 12 Petersson et
7 days al. (1989)
Mouse, CBA Single dose, 0.17 0.77 3.2 14 Kostyniak
Mouse, CWV 8 days 0.04 1.4 3.3 16 (1980)
Mouse, NMRI, male Single dose, 0.05 0.9 8.1 20 Nielsen at al.
Mouse, NMRI, female 14 days 0.21 0.6 4.1 3 (1994)
c, cerebrum; cc, cerebral cortex; o, occipital pole; of, occipital and frontal lobes
a Type of dosing and number of days between last or single dose and sacrifice
Histochemical localization (by silver amplification) of mercury
showed a different distribution. The first deposits of mercury in rat
brain became apparent 10 days after exposure to 16 mg/L of
methylmercury chloride in drinking-water. The deposits were found
initially in the brain stem, then in the cerebral cortex and
supraoptic nucleus, and finally in the cerebellum and thalamus. After
20 days, the deposits in the cerebellar cortex were restricted to
Purkinje cells and Golgi epithelial cells and those in the spinal cord
to the anterior motor neurons; the granule cells of the cerebellar
cortex remained unstained (Moller-Madsen, 1990, 1991). Similar
staining was seen after daily intraperitoneal administration of
methylmercury chloride at 0.16-0.8 mg/kg bw (Moller-Madsen, 1990). As
the cerebellar granule cells are target cells for methylmercury
(Chang, 1977), the absence of staining indicates that only
demethylated mercury can be detected with the silver amplification
method. When the cortex of the calcarine sulcus of macaque monkeys was
stained by the same method, large deposits were seen in the astrocytes
and microglia after six months, whereas staining of neurons appeared
later and remained faint even after 18 months (Charleston et al.,
1995). In squirrel monkeys given weekly doses of [3H]methylmercury,
the amount in protein increased, and it was found in damaged but not
in undamaged neurons (Berlin et al., 1975a).
2.1.3 Transfer from mothers to offspring
2.1.3.1 Placental transfer
Methylmercury passes about 10 times more readily through the
placenta than other mercury compounds, like mercuric mercury and
phenylmercury (Suzuki et al., 1967). Consequently, when 2 mg of
methylmercury chloride were infused intravenously into female rats,
the whole-body retention 1 h later was higher in pregnant than in
non-pregnant rats, but the deposition in blood, kidney, liver, and
brain was lower, as the fetus acted as a sink for methylmercury
(Aschner & Clarkson, 1987). When pregnant rats were given a single
(King et al., 1976) or multiple doses of methylmercury (Magos et al.,
1980a), the fetal content of mercury increased daily, with no increase
in the fetal concentration.
The concentration ratio in fetal brain:maternal brain is > 1,
except in hamsters given a single dose of mercury at 0.32 mg /kg bw on
day 2 or 9 of gestation. One day before parturition, the maternal
brain concentration was higher than that of the fetus (Dock et al.,
1994a). In macaque monkeys (Stinson et al., 1989), rats (Satoh et al.,
1985a), and mice (Satoh & Suzuki, 1983) given longterm dietary
exposure to methylmercury, the concentration of mercury in fetal brain
consistently exceeded that of maternal brain by a factor of 1.5.
Similarly, when methylmercury was given to dams during gestation,
1.7-4.8 times higher concentrations were found in fetal than in
maternal brain in rats (Null et al., 1973; Aaseth, 1976; King et al.,
1976) and mice (Inouye et al., 1985), although the ratio to whole-body
concentration. was 1 in rats (Magos et al., 1980a) and < 1 in mice
(Childs, 1973).
The fetal:maternal concentration ratio was slightly > 1 in liver
and 0.3-0.5 in kidney (King et al., 1976; Wannag, 1976; Inouye et al.,
1986), and the ratio in blood was 1.1 (Burbacher et al., 1984) or 1.2
in macaque monkeys (Stinson et al., 1989) and 0.6 in rats (Wannag,
1976). In contrast to animals, the fetal: maternal blood ratio in
humans is high. Thus, the cord blood:maternal blood ratio in Inuits
with a high consumption of marine foods (Hansen et al., 1990) and in
Swedish women who ate large amounts of fish (Skerfving, 1988) was 3.2.
In the offspring of squirrel monkeys exposed to methylmercury,
the highest concentration in the brain was found in the pituitary
gland, followed by caudatus, striatum, thalamus, cerebrum, cerebellum,
medulla, and cervical spinal cord (Lögdberg et al., 1993). In rats,
the concentration was higher in the cerebellum than in the cerebrum
(Yamaguchi & Nunotani, 1974; King et al., 1976).
2.1.3.2 Lactation
The passage of methylmercury from blood to milk is low, in
contrast to passage through the blood-brain and blood-placenta
barriers. The average concentration ratio for milk:maternal blood was
0.2 in hamsters (Nordenhäll et al., 1995a), 0.03 in guinea-pigs
(Yoshida et al., 1994), and 0.04 in mice (Sundberg et al, 1998a) and
rats (Sundberg et al., 1991). Inorganic mercury passes more readily
into milk than methylmercury: after injection of equivalent doses of
inorganic and methylmercury, the concentration of mercury in the milk
was five times higher with the inorganic form in lactating mice
(Sundberg et al., 1998b) and 2.5 times higher in guinea-pigs. In
guinea-pigs, the milk:maternal blood concentration ratio was 0.12 for
inorganic mercury and 0.023 for methylmercury, but the milk:plasma
concentration ratios were similar (Yoshida et al., 1994). One
consequence of the difference in the passage of inorganic and
methylmercury is that mothers poisoned with the organic form had 95%
less mercury in their milk than in their blood, and 40% of the mercury
in milk was inorganic (Bakir et al., 1973). This tendency was
confirmed in studies on women who ate large quantities of fish
(Skerfving, 1988; Oskarsson et al., 1996) and in experimental animals
(Nordenhäll et al., 1995b; Sundberg et al., 1998a,b).
In the milk of hamsters, the concentration of mercury decreased
with a half-time of four days, and 5% of the injected dose was
excreted in milk. The proportion of inorganic mercury in milk was 16%
on the first day and 22% after five to six days (Nordenhäll et al.,
1995a). An average of 1.7% of a maternal dose given on the day of
parturition was transferred to a litter (Nordenhäll et al., 1995b).
In suckling pups of mouse dams given a single intravenous
injection of methylmercury chloride at 0.4 mg/kg bw, the
concentrations of mercury in plasma and brain peaked after six to
seven days. While that in plasma then immediately decreased, the
concentration of mercury in brain showed no variation for a further
four days (Sundberg et al., 1998a). In pups of hamster dams given a
single dose of 3.2 mg/kg bw methylmercury chloride by gavage on the
day 1 post partum, the whole-body and tissue concentrations of
mercury increased for 10-15 days and then decreased. When the pups
were 21 days of age, 50% of the body burden was in the pelt
(Nordenhäll et al., 1995a).
The body burden of mouse pups was increased to a lesser extent by
exposure during lactation than by exposure in utero. In
cross-fostering studies, the body burden at 14 days of age was twofold
higher in pups exposed in utero than in those exposed by lactation
(Nielsen & Andersen, 1992). Similar differences were observed in
hamster pups (NordenhäIl et al., 1998).
Although loss of methylmercury during lactation is too small to
affect maternal clearance, the clearance half-time was significantly
lower in lactating women (Greenwood et al., 1978), rats (Magos et al.,
1981), and mice (Greenwood et al., 1978) than in non-lactating ones.
2.1.4 Clearance
In animals, mercury is cleared by three main routes: in urine,
faeces, and hair. Faecal excretion predominates over urinary
excretion. While in humans excretion in hair is important only for
biological monitoring, in furry animals this route of excretion can
strongly alter clearance from the whole body and from toxicologically
important soft tissues (Table 2).
In volunteers who ingested a single dose, faecal excretion
reached a peak faster than urinary excretion; faecal excretion peaked
at 3% and urinary excretion at 0.11% of the dose (Miettinen, 1973). In
another study, the cumulative faecal excretion was 31% and urinary
excretion was 4% of the dose (Smith et al., 1994). In pigs, the ratio
of cumulative faecal:urinary excretion during the first 15 days after
a single dose was 17 (Gyrd-Hansen, 1981); in rats, the ratio was 4.8
in the first three days (Swensson & Ulfvarson, 1967). The ratio
decreased in rats with increasing dose and prolonged exposure (Magos &
Butler, 1976). In hamsters given a low dose, the ratio was 2.1 (Dock
et al., 1994a), while urinary excretion was greater than faecal
excretion after a renally toxic dose (Petersson et al., 1989),
probably because of loss of mercury with desquamated tubular cells.
Reports of concentrations in excreta also suggest the importance
of faecal over urinary excretion, if it is assumed that the difference
in volume (or mass) is not great. The faecal:urinary concentration
ratio was 100 for squirrel monkeys (Berlin et al., 1975a) and 50 for
cats (Hollins et al., 1975). Although faecal and urinary excretion
could be used to calculate the clearance half-time for the whole body,
this method (like chemical or radiochemical estimation of whole-body
burden) results in an underestimate of clearance from the
toxicologically important soft tissues in furry animals. In rats given
repeated oral doses of methylmercury, the contribution of blood to the
body burden declined to 28% and the contribution of the pelt increased
to 38% (Magos & Butler, 1976); 98 days after a single dose, the body
burden represented 12% of the dose and nearly 90% of the body burden
was in the fur (Farris et al., 1993). In hamster pups exposed to
methylmercury either in utero or by lactation, 80% of the body
burden was in the pelt by 28 days of age (Nordenhäll et al., 1998).
Correction for mercury in fur (Table 2) decreased the clearance
half-time by 33% in cats and by 44% in rats. As the only toxicological
significance of methylmercury in fur is as a source of intake during
grooming (Farris et al., 1993), the clearance half-time in blood is
more meaningful than that in the whole body; however, even values
based on total mercury result in underestimates of clearance since
decomposition is ignored. Clearance half-times show wide interspecies
variation, depending on the body mass of the species: the larger the
body mass, the longer the clearance half-time.
In adult mice exposed to unlabelled methylmercury before and
after a single dose of [203Hg]methylmercury, the clearance of
radiolabelled compound was not affected (Nielsen & Andersen, 1996).
Similarly, in rats, the biliary excretion of labelled mercury was not
influenced by treatment with unlabelled compound (Cikrt et al., 1984),
indicating complete distribution of label between the first and second
injections.
Clearance from offspring is slow during the first weeks of life.
Thus, the body burden of rats declined by 5% of the dose during the
first 10 days and by a further 25% during the next 12 days (Thomas et
al., 1988). Suckling hamsters (Nordenhäll et al., 1998) and mice
(Rowland et al., 1983; Sundberg et al., 1998a,b) show similar changes
in distribution and clearance.
Table 2. Clearance half-times (days) of methylmercury
Species Compartment Reference
Whole body Brain Blood
Human 72 - - Aberg et al. (1969)
- - 52 Kershaw et al. (1980)
74 - 50 WHO (1990)
44a - 44a Smith et al. (1994)
Squirrel monkey 134 - 49 Berlin et al. (1975b)
Macaque monkey - 47 14 Rice et al. (1989)
- - 24 Stinson et al. (1989)
- - 26 Vahter et al. (1994)
Pig - - 27 Gyrd-Hansen (1981)
Sheep - - 14 Kostyniak (1983)
Cat - - 39 Charbonneau et al.
(1974)
117 (76b) - - Hollins et al. (1975)
Guinea-pig - - 13 Iverson et al. (1973)
Rat, female 34 26 14 Magos & Butler (1976)
Rat, male 24 - - Farris et al. (1993)
Rat, male 13a - 11 Farris et al. (1993)
Rat, male 11a,b - - Farris et al. (1993)
Hamster 7.7 - - Dock et al. (1994a)
Mouse, female 7.3 - - Clarkson et al. (1973)
Mouse, CR1:CD, - - 7 - Sundberg et al. (1998a)
female
Mouse, CBA, male approx. 1 7.4 - Kostyniak (1980)
Mouse, CFW, male 3.0 - - Kostyniak (1980)
Mouse, BALB/c, male - 15 5.0 Doi & Kobayashi (1982)
Mouse, C57Bl, male - 16 7.8 Doi & Kobayashi (1982)
Mouse, NMRI, male 6.3 - - Nielsen & Andersen (1991)
Mouse, NMRI, female 14 - - Nielsen & Andersen (1991)
a Clearance includes decomposition to inorganic form
b Half-time for whole body without hair
2.1.5 Biochemical aspects
2.1.5.1 Cleavage of carbon-mercury bond
Methylmercury is the most stable organic mercury compound. The
addition of even one carbon to the alkyl radical can accelerate
decomposition in brain and other tissues (Magos et al., 1985a).
The bond between the alkyl radical and mercury can be broken by
hydroxy radicals (Suda & Hirayama, 1992) and other reactive oxygen
species (Suda & Takahashi, 1992), which attack ethylmercury more
readily than methylmercury.
In rats given a single dose of methylmercury, the 1:1 ratio of
organic: inorganic mercury was reached after 50 days in liver and
after 14 days in kidney, the main site of accumulation of inorganic
mercury. Inorganic mercury represented less than 4% of the mercury in
brain during the first 23 days after a single dose (Norseth &
Clarkson, 1970) and only 3.4% after prolonged daily treatment (Magos &
Butler, 1976).
Inorganic mercury in the brain is most likely formed by
decomposition in situ, as pretreatment with methylmercury did not
increase the concentration of mercury in the brains of rabbits treated
with inorganic mercury (Dock et al., 1994b). The slower clearance from
the brain may explain the 88% increase in the contribution of
inorganic mercury to total mercury in the brains of macaque monkeys
six months after a long exposure (Lind et al., 1988; Vahter et al.,
1994), and the concentration of inorganic mercury in the hypothalamus
was 6.6 times higher than that in monkeys exposed to inorganic mercury
for two to three months (Vahter et al., 1994).
The main site of decomposition of methylmercury is the intestinal
tract, where the portion secreted with bile or with cells shed from
the intestinal wall is decomposed and the remainder is reabsorbed
(Norseth & Clarkson, 1971). Most of the decomposition is carried out
by intestinal bacterial flora; disruption of this bacterial activity
by antibiotics prolonged the clearance half-time and decreased faecal
excretion in both rats and mice (Rowland et al., 1980, 1983). Both
demethylation and methylation occur. Caecal bacteria from rats
methylated 2.3% of inorganic mercury at a dose of < 0.1 µg/g and less
of doses > 0.1 µg/g (Rowland et al., 1977). Extrapolation of this
rate of synthesis to human populations with intakes of 4.6 µg of
inorganic mercury and 2.4 µg of organic mercury would add only 0.1 µg
to the daily intake of methylmercury.
The biliary excretion of methylmercury varies by species:
hamsters, rats, and mice excrete it readily, while guinea-pigs and
rabbits excrete 5 and 25 times less. The 'slow excretors' also
excreted significantly less reduced glutathione (GSH) than the 'fast
excretors' (Stein et al., 1988). Biliary excretion also depends on
age: in rats, the ability to excrete GSH and methylmercury develops
between 2 and 4 weeks of age (Ballatori & Clarkson, 1982).
2.1.5.2 Complexes with thiol radicals
Mercury compounds favour links with thiol ligands. Thiols of low
relative molecular mass offer transport, and those of high relative
molecular mass within cells offer anchorage for methylmercury,
targeting sulfydryl enzymes (Rothstein, 1970). The change from one
ligand to another ensures that diffusable thiol compounds can carry
methylmercury from the extracellular space to intracellular proteins
and vice versa. The high concentration of GSH within erythrocytes
suggests that exchange occurs between protein and GSH in these cells
(Rabenstein & Evans, 1978); however, other diffusable thiols, such as
cysteine, and unphysiological ones such as pencillamine,
N-acetylpenicillamine, and N-acetylcysteine, removed methylmercury
from erythrocytes when their concentration approached the
concentration of GSH (Kostyniak et al., 1975).
Complexing with L-cysteine or glutathione has particular
toxicological significance. In rats, injection of L-cysteine with
methylmercury changed the short-term distribution of the latter by
decreasing the plasma concentration and increasing the concentrations
in brain, liver, and kidney. D-Cysteine had similar effects but did
not increase the concentration of mercury in brain (Thomas & Smith,
1982). GSH had less effect on the brain uptake of methylmercury than
L-cysteine (Kerper et al., 1992). Increased uptake by brain was also
achieved by administration of a cysteine-supplemented diet (Farris et
al., 1977). The stimulatory effect of L-cysteine may be due to the
structural similarity between the methylmercury-cysteine complex and
L-methionine, which permits use of the same L-amino acid transport
system (Kerper et al., 1992). L-Cysteine can also increase the biliary
excretion of methylmercury, after a delay for conversion to GSH (Magos
et al., 1978).
GSH plays a key role in biliary secretion of methylmercury.
Consequently, compounds that deplete GSH decrease and compounds that
stimulate glutathione S-transferase increase the biliary excretion of
both GSH and methylmercury (Gregus & Varga, 1985). This effect is
modulated by the transport molecule ligandin (a glutathione
S-transferase): the biliary excretion of methylmercury was inhibited
by indocyanine green, a non-substrate ligand for ligandin (Magos et
al., 1979a). trans-Stilbene oxide, which induces glutathione
S-transferase activity, had no effect on the biliary excretion of GSH
or methylmercury (Gregus & Varga, 1985) but potentiated the effect of
GSH in rats (Magos et al., 1985b). The excreted methylmercury is
attached to GSH (Refsvik & Norseth, 1975) or to its metabolic product
(Ohsawa & Magos, 1974; Urano et al., 1988).
GSH also affects the renal handling of methylmercury. Depletion
of hepatic GSH by 1,2-dichlorobenzene resulted in a reduced blood GSH
concentration and a reduced renal content of methylmercury. Inhibition
of renal gamma-glutamyl-transpeptidase, a membrane enzyme that
catalyses the breakdown of GSH, decreased the renal deposition and
increased the urinary excretion of methylmercury. The sex difference
in the renal handling of methylmercury in some mouse strains may also
depend on the concentration and metabolism of renal GSH (Tanaka
et al., 1991)
Physiological agents that complex thiols do not affect urinary
excretion or whole-body clearance (Magos & Clarkson, 1976), while
unphysiological thiols of low relative molecular mass, such as
penicillamine, N-acetylpenicillamine, dimercaptosuccinic acid, and
dimercaptopropanesulfonate, accelerate clearance in both humans
(Clarkson et al., 1981) and experimental animals (Aaseth., 1976;
Magos, 1976; Planas-Bohne, 1981). Although clearance was accelerated
by dimercaptopropanol, this chelator consistently increased the
deposition of methylmercury in the brains of mice (Berlin et al.,
1965; Ogawa et al., 1976). Thiol-complexing agents mobilize
methylmercury more efficiently than inorganic mercury, as shown with
dimercaptosuccinic acid and penicillamine in mice (Friedheim & Corvi,
1975) and pencillamine in volunteers (Suzuki et al., 1976).
2.1.5.3 Interaction with selenium
Methylmercury has a greater affinity for selenium than for
sulfur, and the reaction of selenite with thiol groups such as GSH and
the sulfur radicals of proteins extends the possibility of
associations between selenium and methylmercury. The effects of
selenite cannot be extrapolated to the effects of selenium present in
food, however, because of differences in their decomposition to
selenide in food. Thus, selenite is twice as potent as
L-selenomethionine in forming volatile dimethylselenide (McConnel &
Roth, 1977) and four times as potent as selenium-enriched liver in
forming mercury-selenium complexes (Magos et al., 1983). The rate of
selenide formation from selenite is important in the formation of an
unstable adduct, bis(methylmercury)selenide, which carries
methylmercury through the blood-brain barrier (Magos et al., 1979b;
Naganuma & Imura, 1980). The availability of selenide for this
reaction is limited by methylation to dimethyl-selenide, which is
stimulated by methylmercury (Yonemoto et al., 1985).
Selenite can change the distribution of methylmercury in rats by
decreasing deposition in kidney and increasing deposition in brain,
independently of whether selenite is given 30 min before (Chen et al.,
1975), simultaneously with, or even days after methylmercury (Magos &
Webb, 1977). Both low (0.6 mg/L) and high (3 mg/L) concentrations of
selenium in drinking-water increased the uptake of mercury in the
brains of mice, the high dose more persistently than the low dose
(Wicklund Glynn & Lind, 1995). Within the brain, selenium did not
alter the subcellular distribution of methylmercury, while the mercury
compound shifted selenium from the cytosol to the mitochondrial
fraction (Prohaska & Ganther, 1977).
Selenite increased the fetal brain content of mercury without
increasing the fetal concentration, depending on the dose of
methylmercury given to mouse dams (Satoh & Suzuki, 1979). The
concentration of mercury in rat fetal brain was also unchanged by 1.3
mg/kg of selenite when the dams were given 1.6 or 4.8 mg/kg bw as
methylmercury on days 6-9 of gestation, although the fetal blood
concentration of mercury was increased by 1.4-and 2.8-fold,
respectively (Frederiksson et al., 1993). Seleno-L-methione had no
effect on the deposition of methylmercury in pups exposed prenatally,
but it increased the deposition of mercury in nearly all organs after
exposure during lactation (Nielsen & Andersen, 1992, 1995).
The role of selenite in the degradation of methylmercury is
amibiguous. Addition of selenite to liver homogenates for 90 min
increased the decomposition of phenylmercury but not of methylmercury
(Fang, 1974). Selenite in the presence of a reducing agent like GSH
(Iwata et al., 1982), cysteine (Baatrup et al., 1989), or hydroquinone
in vitro (Moller-Madsen & Danscher, 1991) induced some
decomposition, but the conditions were not physiological. Six months
after exposure of macaque monkeys to methylmercury for 12 months, the
concentration of inorganic mercury had declined, while the
concentration of selenium remained unchanged (Björkman et al., 1994),
which argues against a role for selenium in cleavage of methylmercury
in the brain. The final molar ratio of inorganic mercury to selenium
in these monkeys was 0.5. In contrast, persistent deposits with a 1:1
molar ratio were found in organs of rats treated with inorganic
mercury and hydrogen selenide (Groth et al., 1976), in the livers of
marine mammals (Koeman et al., 1973; Martoja & Viale, 1977), and in
the brains of retired mercury miners (Kosta et al., 1975).
2.2 Toxicological studies
2.2.1 Acute toxicity
Although methylmercury is almost completely absorbed from the
gastrointestinal tract, the enteric and parenteral LD50 values differ
widely. Thus, the LD50 values 24 h after intraperitoneal injection
were 9.5 mg/kg bw in rats, 20 mg/kg bw in hamsters, and > 14 mg/kg bw
in squirrel monkeys; after 30 days, the values were 8.1, 12, and
3.8-5.1 mg/kg bw, respectively. In these three species, most of the
deaths occurred within the first 24 h (Hoskins & Hupp, 1978). The
LD50 values after oral administration were 25 mg/kg bw in old rats
(450 g bw) and 40 mg/kg bw in young rats (200 g bw) (Lin et al.,
1975).
The difference in the LD50 values after administration by these
two routes may reflect the corrosive effect of methylmercury at the
site of contact. The risk of corrosion depends on the route of
administration, and it decreases in the order intraperitoneal =
subcutaneous > intubation > dose given in a small volume of water or
juice > supplemented food. The importance of corrosion is supported
by the dramatic reaction seen in rats 15 min after intraperitoneal
injection of doses > 3 mg/kg bw: the animals became lethargic, with
drooping heads and dulled eyes; some died after developing dyspnoea,
spasticity, and loss of the ability to walk. Animals given < 3 mg/kg
bw became drowsy but survived (Hoskins & Hupp, 1978). The rapidity of
this reaction suggests a corrosive effect and shock. The corrosive
effect is also reflected by the ulcerative oesophagitis seen in cats
dosed orally with methylmercury at 1.3 mg/kg bw per day (Davies &
Nielsen, 1977) and by the necrosis, oedema, and congestion of the
stomach in dogs dosed with 0.43 or 0.64 mg/kg bw per day (Davies et
al., 1977) and in pigs given 0.64 or 0.86 mg/kg bw per day (Davies et
al., 1976). All the affected animals also had renal and hepatic
damage.
These local effects should be taken into consideration in
interpreting more subtle outcomes. In rats dosed orally for two days
with 12 mg/kg bw per day as mercury or 8 mg/kg bw per day as
methylmercury chloride, decreased wakefulness and increased slow-wave
sleep peaked three days after the second dose, while the brain mercury
concentration peaked after nine days (Arito & Takahashi, 1991). A
confounding role of gastrointestinal inflammation could not be ruled
out. A reduced ability of mice to stand on their hind legs and to
move, seen 1 h but not 72 h after an intraperitoneal injection of 10
mg/kg bw as methylmercury chloride (Salvaterra et al., 1973),
indicates peritoneal irritation and possibly peritonitis rather than
systemic toxicity.
2.2.2 Renal and hepatic toxicity
Mice
Experiments in mice suggest that males are more sensitive than
females. In long-term experiments, the incidence of chronic
nephropathy was higher in male than in female mice given diets
containing 10 mg/kg (equivalent to 1.5 mg/kg bw per day), and only
male mice had nephropathy when given 2 mg/kg in the diet (Hirano et
al., 1986; Mitsumori et al., 1990).
Rats
Clinical signs of neurotoxicity induced by methylmercury are
always accompanied by renal damage. In female rats dosed orally five
times a week with 0.84 mg/kg bw as methylmercury dicyandiamide for
8-12 weeks, the renal cortex was extensively damaged, with desquamated
cells in the tubules and inflammatory reactions and fibrosis in the
surrounding area. At the end of treatment, mild ataxia was observed in
some but not all animals (Magos & Butler, 1972). Male rats receiving
0.8 mg/kg bw per day as methylmercury chloride developed severe
diarrhoea and loss of appetite after two to four days, and necropsy
after 10 days of treatment showed ultrastructural changes in the pars
recta of the proximal tubules (Ware et al., 1975). Differences in the
time of onset of renal damage rather than in its severity were seen in
male rats given 8.5 or 1.7 mg/kg bw per day as methylmercury
hydroxide: renal lesions, mainly in the proximal tubules, were seen
one day after the last high dose and six days after the low dose
(Klein et al., 1973).
In comparisons of renal toxicity, male rats were usually more
sensitive than females, as indicated by increased serum creatinine
concentration and decreased bromosulphthalein excretion after single
oral doses of methylmercury chloride ranging from 4 to 40 mg/kg bw
(Yasutake et al., 1991); proteinurea in rats fed diets containing
methylmercury chloride at 0.5, 2.5, or 25 mg/kg for 12 weeks
(Verschuuren et al., 1976a); and deaths and renal lesions in rats
given 0.05 or 0.25 mg/kg bw per day as methylmercury chloride in food
(Munro et al., 1980). No difference in renal morphology was found
between the sexes after exposure to 2.5 mg/kg of diet for two years
(Verschuuren et al., 1976b), although the results suggested that
females were more sensitive than males to diets containing 2 mg/kg of
methylmercury chloride (equivalent to 0.2 mg/kg bw per day) for 84 or
142 days (Fowler, 1972).
The effect of methylmercury on the liver can be rapid and
lasting. Ultrastuctural changes were detected in the liver 1 h after a
single subcutaneous dose of 8 mg/kg bw as methylmercury chloride to
male rats, which developed into cytoplasmic degeneration during the
first day (Desnoyers & Chang, 1975). Similar changes were seen in the
livers of cats fed tuna fish containing 0.3-0.5 mg/kg of mercury daily
for 7-11 months (Chang & Yamaguchi, 1974).
2.2.3 Anorexia
A frequent response to methylmercury in experimental animals is
anorexia resulting in decreased weight gain or even loss of weight.
Anorexia precedes the neurological signs of methylmercury poisoning in
rats (Hunter et al., 1940; Herman et al., 1973), rabbits (Jacobs et
al., 1977), guinea-pigs (Falk et al., 1974), and mice (McDonald &
Harbison, 1977). In cats (Davies & Nielsen, 1977) and non-human
primates (Shaw et al., 1975; Evans et al., 1977), anorexia occurred
only after the onset of disorders of coordination and vision.
2.2.4 Neurotoxicity
2.2.4.1 Small rodents
The species studied most extensively for neurotoxicity is the
rat, and few experiments have been conducted on other species.
Guinea-pigs were used to study the effect of methymercury on the
cochlea. Five weekly doses of 1.7 mg/kg bw as methylmercury hydroxide
for two to six weeks decreased body weight and locomotor activity and
damaged the outer rows of hair cells of the cochlea at two-and-a-half
turns from the cochlear base. The auditory tract was not damaged (Falk
et al., 1974). In a follow-up study, the damage was localized to the
sensory nerve end (Konishi & Hamrick, 1979).
The role of the granular layer of the cerebellum and the
posterior root fibres as a target of methylmercury was identified in
rats 60 years ago. The report of this study also described the
clinical course of severe poisoning as weight loss, ataxia, paralysis,
and death (Hunter et al., 1940). Axoplasmic and myelin degeneration of
posterior root fibres was produced by daily doses of 0.8 mg/kg bw as
methylmercury chloride (Chang & Hartmann, 1972), while the ventricular
root fibres and the dorsal root nerves remained intact after
administration of 1.6 mg/kg bw per day (Yup & Chang, 1981). The
vulnerability of dorsal root ganglia was explained by the extent and
duration of inhibition of amino acid incorporation (Cavanagh & Chen,
1971).
Female rats dosed orally five times a week with 0.84, 1.68, or
3.32 mg/kg bw as methylmercury dicyandiamide showed signs of ataxia
after 39, 25, and 10-12 doses, respectively. Owing to progressive
weakness, animals at the intermediate dose had to be sacrificed after
34 doses and those at the highest dose after 16 doses, when their
brain mercury concentrations were 10 and 16 µg/g, respectively.
Histological examination of the brain showed necrosis in the granular
layer of the cerebellum and, at the two higher doses, oedema in the
white matter; at the low dose, only mild damage involving few cells
was seen even after 59 doses, when the brain mercury concentration was
about 5 µg/g (Magos & Butler, 1972). The lack of damage at this dose
was corroborated by the lack of gross or histological effects in male
rats dosed with 0.8 mg/kg bw per day as methylmercury chloride for 28
weeks. A dose of 4 mg/kg bw per day caused weight loss, early toxic
signs, and decreased protein synthesis in the granule neurones. These
pathological changes were precipitated when the brain mercury
concentration was 3-6 µg/g. At higher concentrations, Purkinje cells
were also affected (Syversen, 1982).
After subcutaneous administration of 2 mg/kg bw as methylmercury
hydroxide five times a week to male rats, the first signs of
coordination disorder were detected after the fifteenth dose. The
earliest morphological change in the central nervous system was
shrinkage of cells and nuclear disintegration in the cerebellar
internal granular layer, the vermis, and the depth of the folia of the
hemispheres. These early changes progressed to more widespread
necrosis involving the putamen, corpus striatum, and the visual cortex
in the occipital lobe. The most severely affected parts of the
peripheral nervous system were the dorsal roots and the sciatic and
sural nerves (Herman et al., 1973). The concentration of mercury on
the fifteenth day was about 30% lower in the cerebral cortex than in
the cerebellum or spinal ganglia, but by day 29 the concentrations in
each of these tissues was nearly 14 µg/g (Somjen et al., 1973b).
Damage to the nervous system resulted in the formation of
autoantibodies that reacted with neurotypic and gliotypic proteins,
such as glial fibrillary acid protein, after seven days of exposure to
13 or 26 mg/kg of mercury as methylmercury chloride in the diet
(El-Fawal et al., 1996).
The dose-effect relationship for neurotoxicity differs in male
and female rats, as suggested by the response to four or five doses of
8 mg/kg bw as methylmercury chloride given by intubation. The flailing
reflex (rotation of the lower body when the animal is held loosely
under the forelimbs), hind-leg crossing, and damage in the granular
layer indicated that female rats were more affected than males. The
difference may be due to the 24-40% higher concentration of mercury in
the brains of female than of male rats (Magos et al., 1981). A
difference was seen also with age: in young rats, an intraperitoneal
injection of 8 mg/kg bw which resulted in approximately 2 µg/g of
brain, the morphological changes in the cerebellar hemisphere were
subtle, including nuclear swelling and increased heterochromatin, were
reversible, and were not accompanied by the formation of dark cells in
the granular layer. In older rats, 'dark cells' were a conspicious
feature (Syversen et al., 1981)
The relative merit of electrophysiology in the grading of
methylmercury poisoning was the subject of several investigations
(Somjen et al., 1973a; Fehling et al., 1975; Miyama et al., 1983). The
results were not consistent, and its advantage over simple clinical
observations remains questionable.
In male mice, dietary intake of methylmercury caused no loss of
body weight when the concentration was 10 mg/kg of diet as
methylmercury chloride, but 32 mg/kg of diet caused loss after 30 days
and 100 mg/kg of diet resulted nearly immediately in weight loss. Mice
at the intermediate dose showed signs of coordination disorders which
coincided with the onset of weight loss. A slight effect on head
positioning was observed earlier. No neurological signs were seen in
mice at the lowest dose (Suzuki & Miyama, 1971). In wild mice exposed
to methyl-mercury in their natural diet, the concentration of mercury
in hair correlated with deviant behaviour and decreased ability to
swim (Burton et al., 1977).
2.2.4.2 Non-human primates
In the first experiment performed on the toxicity of
methylmercury, one female macaque monkey was exposed by inhalation to
a concentration in air that was so high that it caused respiratory
irritation. The clinical events were ataxia, prostration, inability to
eat, and death after 21 days. The main damage to the nervous system
was in the sciatic nerve, posterior root ganglia, brain stem, and
cerebrum, where the frontal and occipital cortices were equally
affected while the cerebellar cortex was spared (Hunter et al., 1940).
Rhesus macaque and mulatta monkeys were given methylmercury in
pellets at doses of 0.01, 0.03, 0.1, or 0.3 mg/kg bw per day for 52
months or until signs of severe poisoning appeared. Exposure to the
lowest dose had no effect on body weight, and 0.03 mg/kg bw per day
depressed body-weight gain only marginally; however, daily doses of
0.1 or 0.3 mg/kg bw per day caused severe poisoning, with ataxia,
visual disturbances, blindness, tremor, spasms, paralysis, and death
or moribund condition at six and three months, respectively. At the
two lower doses, the average monthly concentrations of mercury in
blood between 12 and 52 months were 0.4 and 1.1 µg/ml. The mean
concentrations of mercury in the occipital lobe after 52 months were
0.7 µg/g and 2.3 µg/g at these doses; the cerebellum contained 10-26%
less than the occipital lobe. Histological examination of the brain
showed no abnormalities. In monkeys at the two highest doses, the
concentrations in the occipital lobe were 13 and 24 µg/g; the most
severe lesions were found in the occipital cortex and, within it, in
the primary visual cortex. No changes were seen in the cerebellar
cortex (Ikeda et al., 1973; Kawasaki et al., 1986). In this study, the
threshold dose of methylmercury that induced toxic effects after
long-term exposure was 0.03-0.1 mg/kg bw per day, resulting in
concentrations of mercury of 2-13 µg/g in brain and 0.7-21 µg/ml in
blood. The NOEL was 0.01 mg/kg bw per day.
In corroboration of this finding, daily doses of 0.068-0.085
mg/kg bw as methylmercury hydroxide given orally through a hypodermic
syringe for 68 weeks caused clinical signs or cortical lesions in only
3 of 15 macaque monkeys; the blood concentration of mercury was about
2 µg/ml. A daily dose of 1.1 mg/kg bw, resulting in a maximum blood
concentration of about 3 µg/g, led to death or a moribund condition.
The vision of these three monkeys was also affected (Finocchio et al.,
1980). Another study with macaque monkeys showed loss of peripheral
vision at blood concentrations > 2.2 µg/ml (Luschei et al., 1977).
These concentrations of mercury in blood correspond to 6-10 µg/g in
brain.
In squirrel monkeys, the visual cortex was invariably damaged,
and extension to adjacent areas increased with duration of exposure
and increasing brain mercury concentration. In one monkey who was
blinded, the maximum concentration in the occipital lobe 40 days after
termination of exposure was 6.6 µg/g (Berlin et al., 1975b), which,
assuming a 47-day half-time in brain, corresponds to 12 µg/g at the
end of treatment.
In macaque monkeys exposed for a long time but with no clinical
signs of poisoning, sensitivity to visual stimuli of low luminiscene
was reduced when the concentration of mercury was 2.6 µg/ml in blood
and 8.7 µg/g in the primary visual cortex. The concentrations were
somewhat higher in the calcarine side of the primary visual cortex and
in the lateral geniculate of the optic thalamus, where the optic
neurons from the retina are projected onto the primary visual cortex,
than in the lateral side of the occipital cortex. The borderline
between effect and no effect was not sharp: unaffected monkeys had
slightly higher concentrations of mercury in blood and in the
occipital cortex than monkeys that were moderately affected, although
they also were exposed for a shorter time. Constricted vision field,
somaesthetic impairment, and ataxia usually occurred together (Evans
et al., 1977),
The lesions seen in the brains of rhesus monkeys given
methylmercury hydroxide in fruit juice differed when it was given for
6-17 days or for 6.3-12 months. Two of four monkeys exposed acutely
and two of six exposed chronically had no clinical or histological
signs of poisoning. The maximum concentrations of mercury in blood
were < 1.1 µg/ml in unaffected monkeys and > 2.0 µg/ml in affected
monkeys. The two acutely poisoned monkeys had a blood mercury
concentration of 11 µg/ml. After acute poisoning, the most evident
histological lesions were seen in the lateral geniculate nucleus and
in large neurons in several areas. The cerebral and cerebellar
cortices, including the calcarine and insular cortices, were not
involved. Animals exposed chronically had damage to the cerebral
cortex which was maximal around the calcarine and lateral cerebral
sulci (Shaw et al., 1975).
The entry of methylmercury into brain, even at concentrations
below those that cause damage, precipitates an increase in the
reactive glial population. A similar reaction was produced after
infusion of inorganic mercury (Charleston et al., 1994).
2.2.4.3 Domestic animals
Few experiments have been reported on the toxicity of
methylmercury in domestic animals, and even fewer that would allow an
approximation of the threshold toxic dose. The oral dose that had no
effect was 0.1-0.2 mg/kg bw per day for calves exposed for 91 days
(Herigstad et al., 1972), 0.19-0.35 mg/kg bw per day for pigs exposed
for 60 days (Triphonas & Nielsen, 1973), 0.06-0.12 mg/kg per day for
dogs exposed for 55 days (Davies et al., 1977), and < 0.25 mg/kg bw
per day for cats exposed for 90 days, whether present naturally in
fish or added in pure form to the diet (Charbonneau et al., 1974). In
the central nervous system, the damage was more extensive in the
cerebellar granular layer in calves and cats and in the cerebral
cortex in pigs and dogs.
In rabbits given one to four doses of mercury at 5.8 mg/kg bw as
methyl-mercury acetate, the most sensitive areas of the nervous system
were the dorsal root and trigeminal ganglia, which showed degeneration
after two doses. In the cerebral and cerebellar cortices, damage was
seen two days after four daily doses. The more severely affected areas
were the II, III, and IV layers of the cerebral cortex and the
molecular and granular layers of the cerebellar cortex, where mainly
the cells of the small neurones, including granule and basket cells,
were damaged and the Purkinje cells spared (Jacobs et al., 1977), as
in a human case of methylmercury poisoning (Hunter & Russel, 1954).
2.2.5 Reproductive and developmental toxicity (other than
neurotoxicity)
Mice
Treatment of male mice on seven consecutive days with
methylmercury chloride at doses of 1, 2.5, or 5 mg/kg bw per day
before mating with virgin females had no effect on fertility or
postimplantation losses but marginally reduced the number of viable
embryos (Khera, 1973). Intraperitoneal injection of male mice with 8.5
mg/kg bw per day as methylmercury hydroxide and serial matings with
young virgin females increased the number of dead implants during the
first 7.4 days in one strain but not in another. The same treatment of
females of the unresponsive strain slightly reduced the total number
of live implants (Suter, 1975).
Exposure to methylmercury chloride prolonged the length of the
menstrual cycle by 11% in mice fed 3.2 mg/kg of diet and by 27% at 6.4
mg/kg of diet. Exposure from 30 days before mating to day 18 of
gestation decreased maternal weight gain at the high dose. The loss
due to resorptions and deaths increased from 7.1% in the control group
to 12% in mice at 3.2 mg/kg of diet and to 44% at 6.4 mg/kg of diet.
The weight of fetuses on day 18 of gestation was also lowered. Both
doses increased the frequency of malformations to 17% of fetuses at
the low dose and 56% at the high dose (Nobunaga et al., 1979). A
follow-up experiment confirmed that exposure to 3.2 mg/kg of mercury
in the diet can cause postimplantation loss in some pregnant mice
(Satoh & Suzuki, 1983). If their food consumption is assumed to be 150
g/kg bw per day, the daily doses of methylmercury were 0.48 and 0.96
mg/kg bw. When selenite was added to the drinking-water in these two
studies, no effect was seen on postimplantation loss but the number of
malformations was increased, at least at the high dose.
Treatment of mouse dams on days 6-17 of gestation with 5 mg/kg bw
per day as methylmercury chloride by intubation reduced the number of
live pups, and pups born live died within two days. The number of live
pups and survival were not affected by 1 mg/kg bw per day, but there
was transitory inhibition of cerebellar cellular migration from the
external granular layer (Khera & Tabacova, 1973). When methylmercury
chloride was given orally on days 6-13 of gestation, the lowest dose
of 2 mg/kg bw per day caused only a few malformations, 4 mg/kg bw per
day decreased fetal weights and caused a large increase in the
frequency of malformations, and 4.8 mg/kg bw per day also increased
postimplantation loss (Fuyuta et al., 1978).
Postimplantation loss was not observed when dams were given 3, 5,
or 7 mg/kg bw per day as methylmercury chloride subcutaneously on days
13-15 of gestation, but the postnatal survival rates were 30%, 22%,
and 0, respectively (Nishikido et al., 1987). The outcome was similar
when mouse dams were given 16 mg/kg bw per day as methylmercury
chloride orally on one of days 13-17 of gestation. Postimplantation
loss was slight or nil, but only 11% of the liveborn pups survived for
eight weeks, apparently as a consequence of their inability to suck.
The spontaneous locomotor activity of live pups was depressed between
3 and 8 weeks, they had defects in righting and tail position, and, at
the end of 8 weeks, they had smaller brains than controls (Inouye et
al., 1985). Starvation, undernourishment, and the consequent general
weakness were probable contributory factors.
Rats
Long-term intake of 2.5 mg/kg of diet as methylmercury chloride
increased testicular but not ovarian weights in rats (Verschuuren et
al., 1976c). Exposure of males and females had no effect on fertility,
but the viability of their offspring was impaired (Verschuuren et al.,
1976b). If their daily food consumption is assumed to be 100 g/kg bw,
the daily dose was 0.25 mg/kg bw. Exposure of female rats to
methylmercury chloride at 8 mg/kg of diet from weaning until delivery
did not affect litter size, the frequency of stillbirths, birth
weight, survival, or weight gain up to weaning (Nixon, 1977).
The mating success of male rats declined by seven days after oral
treatment with methylmercury chloride at 2.5 or 5 mg/kg bw per day but
not at 1 mg/kg bw per day. The number of viable embryos per litter
decreased transiently when 2.5 or 5 mg/kg bw per day was given for
seven days, 1 mg/kg bw per day for 35 days, or 0.5 mg/kg bw per day
for 90 days (Khera, 1973).
Oral treatment of female rats with 0.25 mg/kg bw per day as
methylmercury chloride from weaning had no apparent adverse effect on
fetuses, and the only abnormality seen postnatally was eyelid lesions
associated with hardening of the lachrymal glands. A dose of 0.05
mg/kg bw per day had no effect (Khera & Tabacova, 1973).
Higher doses were usually used when administration was restricted
to the gestation period. In rats given 2, 4 or 6 mg/kg bw per day as
methylmercury chloride orally on days 7-14 or 18-20 of gestation,
resorptions, deaths, and malformations were observed at 6 mg/kg bw per
day. Malformations consisting mostly of cleft palate and vertebral
defects were seen in the offspring of dams at the two higher doses
(Fuyuta et al., 1978). When a single oral dose of 8, 16, or 24 mg/kg
bw as methylmercury chloride was given orally in saline on day 7 of
gestation, maternal body weight declined at all doses. The decrease in
the number of live fetuses on day 20 of gestation was 60% at 8 mg/kg
bw and > 90% at 16 mg/kg bw. A dose-dependent decrease in
ossification centres was seen. The concentrations of mercury in
maternal brain were 2.6, 9, and 21 µg/g, and those in fetal brain were
3.5, 11, and 15 µg/g (Lee & Han, 1995).
Hamsters
In hamsters, a single subcutaneous dose of 6.4 mg/kg bw as
methylmercury chloride on day 3, 5, or 9 of gestation caused some
maternal deaths, a high incidence of resorptions, decreased fetal
weights, and moderate to severe malformations, consisting mainly of
clubfoot and hydrocephalus. A dose of 1.6 mg/kg bw had no visible
effect on dams or offspring, but when given on days 1-14 of gestation
it increased the numbers of maternal deaths, resorptions, and
malformations although it did not decrease fetal weights (Harris et
al., 1972).
Non-human primates
In macaque monkeys, a daily oral dose of 50 or 70 µg/kg bw as
methylmercury hydroxide in fruit juice for 20 weeks decreased sperm
motility and increased the frequency of abnormal sperm tail forms,
with no significant change in serum testosterone concentration or
testicular morphology (Mohamed et al., 1987). In females given doses
including 90 µg/kg bw per day, exposure did not affect the menstrual
cycle, conception rate, or size of offspring at birth, but a maternal
blood concentration > 1.5 µg/ml decreased the number of viable
deliveries (Burbacher et al., 1988), and a concentration > 2 µg/ml
was toxic to the dams (Burbacher et al., 1984, 1988).
2.2.5 Developmental neurotoxicity
2.2.5.1 Exposure in utero
Mice
Mouse dams were given methylmercury hydroxide subcutaneously as a
single dose of 5.1, 6.8, or 10 mg/kg bw on day 10 of gestation. As the
rate of mortality of the neonates of dams treated with 10 mg/kg was
high, an additional group of dams were given 3.4 mg/kg bw on days
10-12. Central latency in the open-field behaviour test was increased
in the pups of dams given 10 mg/kg bw as a single or three divided
doses. Locomotion was decreased by exposure to 6.8 and 10 mg/kg bw at
postnatal day 24 but not at day 44 (Su & Okita, 1976). The righting
reflex and walking ability lagged behind those of controls
consistently in the pups of dams given 6.8 mg/kg bw subcutaneously on
day 9, but the difference was not significant (Satoh et al., 1985b).
Rats
Methylmercury given at a dose of 3.2 mg/kg bw to pregnant rats on
day 8 of gestation caused no significant change in the appearance of
pups, but samples of calcarine and cerebellar cortices, especially in
the granule cells of the cerebellum, showed focal weakening of the
nuclear envelope, myelin figure formation, focal cytoplasmic
degradation, and phagocytosis of cellular debris by macrophages (Chang
et al., 1977). In pups of hamster dams given methylmercury chloride
either as a single dose of 8 mg/kg bw on day 10 of gestation or 1.6
mg/kg bw daily on days 10-15, the early cerebellar changes were seen
in the external granular layer and in more extensively differentiated
neural elements in the molecular and internal granular layer (Reuhl et
al., 1981a). The sequelae of the early injuries, such as astrogliosis,
may have had clinical or physiological significance when the pups
reached the age of 275-300 days (Reuhl et al., 1981b).
When methylmercury chloride was given by gavage to rat dams at
doses of 0.02, 0.04, 0.4, or 4 mg/kg bw on days 6-9 of gestation, the
highest dose impaired swimming behaviour at 4-35 days of age, and the
doses of 0.04 and 4 mg/kg bw increased passiveness and decreased
habituation to an auditory startle 60-210 days postnatally. The
histological changes seen at the highest dose were mainly in the
dendritic spines of the pyramidal neurones (Stoltenberg-Didinger &
Markwort, 1990). Methylmercury chloride given in apple juice on the
same days of gestation at a dose of 1.6 mg/kg bw per day caused no
change in the clinical markers of adverse effects up to weaning. No
deficits in behavioural function, such as spatial learning in a
circular bath and maze learning for food reward, were seen at four to
five months of age (Frederiksson et al., 1996).
The effect of methylmercury on locomotion is ambiguous. When
given orally to rat dams on day 8 or 15 of gestation at a dose of 4 or
6.4 mg/kg bw, no consistent changes in spontaneous locomotor activity
were seen 4, 8, or 15 days postnatally. Activity was increased on
postnatal day 4 when the low dose was given on day 8 of gestation, on
postnatal day 8 when the low dose was given on day 15 or the high dose
on day 8 of gestation, and on postnatal day 15 when the low dose was
given on day 15 of gestation (Eccles & Annau, 1982a). The higher dose
given on day 15 did not affect locomotor activity on postnatal day 14,
21, or 60 (Cagiano et al., 1990), but 8 mg/kg bw given on day 4 of
gestation depressed locomotor activity at 110-140 days of age.
Avoidance learning was also depressed (Schalock et al., 1980).
The effect of 4 or 6.4 mg/kg bw as methylmercury chloride given
to rats on day 8 or 15 of gestation on two-way avoidance was tested at
nine weeks of age. Exposure increased the number of trials required to
reach the preset criterion; however, owing to variations within
groups, only the higher dose given on day 8 induced a significant
difference in reacquisition and both doses given on day 15 for
acquisition (Eccles & Annau, 1982b). The higher dose given on day 15
decreased the number of cortical muscarinic receptors by 53% in
14-day-old pups and by 21% in 21-day-old pups. Recovery from this
defect was complete at the age of 60 days, but the results of a
passive avoidance test even a few days before that age indicated
learning and memory deficits (Zanoli et al., 1994).
Non-human primates
Female macaque monkeys were given methyl-mercury in apple juice
at 0.04 or 0.06 mg/kg bw per day for 198-747 days before mating.
Infants were separated from their mothers at birth and were tested 210
and 220 days after conception (50-60 days after birth). The test
indicated randomness in visual attention to novel stimuli. The mean
maternal blood concentrations at birth were 0.84 and 1.04 µg/ml, and
the blood concentrations of the offspring were 0.88 and 1.7 µg/ml
(Gunderson et al., 1988). The offspring of squirrel monkeys exposed to
methylmercury during the second half or the last third of pregnancy,
with maternal blood concentrations of 0.7-0.9 mg/ml, showed reduced
sensitivity to changes in the source of reinforcement, indicating
learning impairment at five to six years of age (Newland et al.,
1994).
2.2.5.2 Exposure in utero and postnatally
Mice
Mouse dams were exposed to 3.2 mg/L as methylmercury in
drinking-water from mating to parturition, and their pups were further
exposed up to postnatal day 30 during lactation. The litter weights of
exposed and control pups were similar. Males but not females showed
some decrease in the width of the external granular layer in a region
of the inferior lobe of the cerebellum on postnatal day 7, but not
later. The density of migrating cells in the molecular layer was also
decreased (Markowski et al., 1998).
Rats
The concentration of mercury was 10 µg/ml in blood and 1.4 µg/g
in the brains of rat pups exposed throughout gestation and lactation
via their dams and directly to the same concentration of mercury as
methyl-mercury chloride at 3.9 mg/kg of diet. No adverse effect was
seen clinically or histologically in the brain, even on morphological
maturation of neurons and astrocytes. The only deviation from control
values was an increase in noradrenaline activity in the cerebellum
(Lindström et al., 1991).
In a crossfostering study, rat dams received an approximate daily
dose of 2.5 mg/kg bw of methylmercury chloride in drinking-water, and
an additional group of pups was exposed only through drinking-water on
postnatal days 21-30. Only pups exposed prenatally or directly after
weaning had deficits in learning ability at day 30, and the effect
lasted at least until postnatal day 50 (Zenick, 1974). At day 30, the
action potential in the visual cortex was decreased in each group
(Zenick, 1976). Offspring of rat dams exposed to 1.2 or 4 mg/L as
methylmercury in drinking-water from two weeks before mating to
weaning showed a dose-dependent deficit in correct response when the
feedback was tactile kinetic perception. The methylmercury intake of
the dams was 0.12 or 0.25-51 mg/kg bw per day before and during
gestation and 0.22-0.38 or 0.53-0.95 mg/kg bw per day during lactation
(Elsner, 1991).
Non-human primates
Macaque monkey dams were given methylmercury chloride orally
three times a week at doses of 10, 25, or 50 µg/kg bw and were mated
when their blood concentration of mercury approached 90% of the
steady-state level. Treatment was continued during gestation. Infants
were separated from their mothers immediately after birth and given
the same dose of methylmercury until they were four years of age. The
blood mercury concentration of the infants at birth was 0.45-2.7
µg/ml, and their steady-state concentration reached 0.22-0.78 µg/ml.
Two of the five animals were severely intoxicated and could not be
tested for spatial or temporal visual function. Some inconsistent
defects in spatial vision at high and low luminescence and in
temporal vision at high luminiscence were seen. Temporal vision at low
luminescence was better in exposed than in control monkeys (Rice &
Gilbert, 1990). The auditory response was tested in the same monkeys
when they were 11 and 19 years of age, 7-15 years after the end of
exposure. The deterioration in hearing between 11 and 19 years was
more pronounced in exposed than in control monkeys. In 19-year-old
monkeys, the thresholds for all frequencies were higher in exposed
than in control monkeys (Rice, 1998).
2.2.5.3 Exposure after parturition
Mice
A single oral dose of 8 mg/kg bw as methylmercury chloride to
two-day-old mice resulted in a brain mercury concentration of 2.7
µg/g, reductions in the number of cells and the percentage of late
mitotic figures, and an increase in cells with reduced nuclear
diameter (Sager et al., 1982). A smaller reduction in cell numbers in
the granular layer of the cerebellum was seen after 4 mg/kg bw, but
the number and proportion of late mitotic figures (anaphase) remained
significantly lower than in controls at day 19 post partum (Sager et
al., 1982).
Rats
Two-day-old rats were given methylmercury chloride at a dose of 5
mg/kg bw per day subcutaneously until loss of body weight and signs of
neurological impairment, including impaired vision (weak visual
placing response), became evident. At this point, the animals were
necropsied, and their visual cortices were studied. Degenerating
neurons were concentrated mainly in layer IV and were scattered in
layers II, III, V, and VI (O'Kusky, 1985). The same treatment led to
the spastic dyskinetic syndrome and to decreases in the specific
activity of glutamic acid decarboxylase (which synthesizes
gamma-aminobutyric acid) in the occipital cortex (O'Kusky et al.,
1988a) and in the tissue concentrations of serotonin and dopamine and
their metabolites (O'Kusky et al., 1988b). As the dose used in these
studies was high, pain and local irritation could have been
confounding factors at least on the concentrations of catecholamines
and serotonin In addition, renal function was certainly affected, as
even 0.85 mg/kg bw given as methylmercury hydroxide subcutane-ously
daily from the day of birth impaired the kidneys (Slotkin et al.,
1985).
Non-human primates
Four infant monkeys were given methylmercury chloride at 0.5
mg/kg bw per day orally for 28-29 days from day 1 after birth. They
progressively became ataxic, blind, and comatose and were necropsied
at 35-43 days. Histopathological changes were marked in the cerebrum
and less severe in the cerebellum, where the Purkinje and granular
cells appeared normal (Willes et al., 1978). In an experiment in which
much lower doses were used, the blood mercury concentration of five
newborn macaque monkeys given 0.05 mg/kg bw per day orally in fruit
juice peaked at 1.2-1.4 µg/ml and declined after weaning to 0.6-0.9
µg/ml. At three to four years of age, they had impaired spatial vision
at both high and low luminescence (Rice & Gilbert, 1982). Continuation
of exposure to the same daily doses until seven years of age resulted
in an increased auditory threshold in four of five monkeys, at high
frequency (25 000 Hz) in one monkey and at middle frequencies in three
monkeys (Rice & Gilbert, 1992). When these monkeys reached the age of
13 years, some appeared to be clumsy and hesitant in large exercise
cages. Exposed monkeys required more time to collect 10 objects from a
recessed square and had a higher vibration threshold, but their motor
response time was normal (Rice, 1996).
2.2.7 Carcinogenicity
In mice exposed for two years to methylmercury at 0.4, 2, or 10
mg/kg of diet, renal tumours developed only in males at 10 mg/kg of
diet. The incidence of renal adenoma was 8% and that of carcinoma was
22% in B6C3F1 mice, the corresponding figures being 5% and 17% in ICR
mice. Nearly all male and female B6C3F1 mice at 10 mg/kg of diet had
chronic nephropathy, while only 78% male and 43% female ICR mice had
this pathological change (Mitsumori et al., 1990). The chronic insult
to the kidney may have contributed to the induction of cancer. In
groups of male and female rats fed diets containing methylmercury at
0.1, 0.5, or 2.5 mg/kg of diet for two years, the incidence of
pathological lesions and tumours was not increased (Verschuuren et
al., 1975c).
2.2.8 Immunomodulation
Exposure of mice to methylmercury in the diet at 3.2 mg/kg did
not affect body weight or kidney, liver, or spleen weight, but the
weight of the thymus and the number of thymocytes decreased by 22 and
50%, respectively. The lymphoproliferative response to T-and B-cell
mitogens was increased in both thymus and spleen, and natural killer
cell activity was decreased by 44% in spleen and by 75% in blood
(Ilbäck, 1991). In a cross-fostering study, when rat dams were exposed
to mercury as methylmercury at 3.2 mg/kg of diet 11 weeks before
mating, no consistent alterations were seen in the body weight,
lymphoid organ weights, or cell number and lymphoproliferative
response to Bcell mitogens of pups at the age of 15 days. The response
of thymocytes to T-cell mitogens increased by 30-48% and that of
splenocytes decreased by 12% only in pups exposed throughout
lactation; natural killer cell activity decreased by 42% after both
doses (Ilbäck et al., 1991). Mouse dams and pups exposed to
methylmercury at a concentration of 0.5 or 5 mg/kg of diet from 10
weeks before mating until weaning showed inconsistent effects on the
immune system (Thuvander et al., 1996).
2.2.9 Extrapolation between species
The doses used in experiments in rats and mice frequently
exceeded 1 mg/kg bw per day, which is within the range of the intake
of the Iraqi patients with severe methylmercury poisoning (Bakir et
al., 1973) and 140-330 times higher than the daily intake that would
cause paraesthesia in 5% of a population (WHO, 1976). Thus,
extrapolation of doses on the basis of body weight clearly results in
nonsense values. Table 3 shows that the clearance half-times of
methylmercury increase with mass, in the order mouse < rat < macaque
monkey < human, and that the concentrations in the whole body after a
unit dose follow the same order. The data in the Table are based on
the assumption that clearance from blood approximates clearance from
the whole body, and the clearance half-times are given as the daily
clearance in percent of the whole-body burden. The units for dose and
concentration are identical. As the half-time increases with body
mass, the concentration in the body after identical doses of
methylmercury also increases with body mass. Thus, allometric
extrapolation based on surface area can be used. The method of
correction, for example from the dose for mouse to the human dose, is:
human dose = mouse dose (0.3:70)0.3
With this allometric extrapolation, the dose in mice multiplied by
0.098 gives the equivalent human dose, while the multiplication factor
for the dose in rats is 0.20 and that for macaques is 0.42. The effect
of this extrapolation on steady-state concentrations is shown in Table
4.
2.3 Observations in humans
Investigations of the possible neurodevelopmental effects of
prenatal exposure to methylmercury followed a sequence similar to
those of other neurotoxic exposures: case series of children who
manifested clinical signs of poisoning and then prospective cohort
studies of asymptomatic children considered to have 'low' exposure or
at least exposure lower than that at which clinical signs and symptoms
appear. The populations chosen were mostly those known to consume
large amounts of fish, which contain variable amounts of methylmercury
(see section 3). The goal of the latter studies is largely to
determine whether a dose-response relationship can be identified for
adverse neurodevelopmental effects associated with exposure to
methylmercury, in order to assess the significance to public health of
exposure in various populations.
Table 3. Clearance half-times and whole-body concentrations of methylmercury
Species Body Half-time Clearance Ratio of concentration:daily dosea
weight (days) (% body
(kg) burden) 10 days after After 10 At steady
single dose doses state
(long-term
exposure)
Human 70 52 0.014 0.87 9.4 75
Macaque 4 25 0.028 0.76 8.7 36
Rat 0.35 12 0.058 0.56 7.6 17
Mouse 0.03 7 0.099 0.37 6.3 10
a Same units as daily dose
2.3.1 Case series
The mass poisoning of persons living near Minamata Bay in Japan
in the 1950s first raised awareness of the severe neurological
sequelae associated with methylmercury poisoning, particularly when
experienced prenatally. The primary route of exposure in this episode
was the consumption of fish contaminated by methylmercury, which
bioaccumulated up the aquatic food chain. According to Harada (1995),
all children identified as suffering from the most severe form of
congenital Minamata disease showed mental retardation, primitive
reflexes, cerebellar ataxia, disturbances in physical growth,
dysarthria, and limb deformities, and most showed hyperkinesis (95%),
hypersalivation (95%), seizures (82%), strabismus (77%), and pyramidal
signs (75%). The incidence of cerebral palsy among children with the
disease was also increased, involving 9% of 188 births in three
villages. Some of the signs and symptoms, such as paroxysmal events,
hypersalivation, primitive reflexes, and ataxia, abated somewhat in
subsequent years, although others such as reduced intelligence and
dysarthria did not. Most patients with the severe form of the disease
were unable to function successfully in society. The mothers of many
affected children experienced only transient paresthesia, indicating
that fetal vulnerability exceeds that of mature individuals. Although
measurements of the body burden of mercury were not available until
several years after the episode, analyses of the mercury
concentrations in archived umbilical cord tissue from patients with
congenital Minamata disease suggest that the mean concentration in
maternal hair may have been approximately 41 µg/g (25-75th percentile:
20-59) (Akagi et al., 1998). The uncertainty associated with this
estimate is likely to be substantial, however, as case ascertainment
was undoubtedly incomplete, particularly among individuals who
suffered milder forms of the disease. For example, even if cases of
known disease are excluded, the prevalence of mental retardation among
children born between 1955 and 1958 in the contaminated area was 29%,
which is far higher than would have been expected and suggests that
congenital Minamata disease was not diagnosed in many children with
less severe forms. Thus, these data cannot provide precise estimates
of the minimum concentration of methylmercury required to produce this
disease.
A second episode of mass methylmercury poisoning occurred in Iraq
in the early 1970s, when seed grain treated with a fungicide
containing this compound was ground into flour and consumed and
resulted in 600 deaths and 6000 cases of methylmercury poisoning.
Thus, the exposure was probably more acute and involved higher doses
than those experienced by the persons living around Minamata Bay. The
results of early studies of the most severely affected children who
were exposed during fetal development were concordant with those in
Minamata: the children manifested severe sensory impairment
(blindness, deafness), general paralysis, hyperactive reflexes,
cerebral palsy, and impaired mental development (Amin-Zaki et al.,
1974). Several follow-up studies of the exposed population were
conducted. Marsh et al. (1987) identified 81 children who were in
utero at the time of the episode and collected information on their
neurodevelopmental outcomes from two sources: neurological examination
of each child and an interview with the mother about the age at which
the child achieved standard developmental milestones such as walking
and talking. The maximum concentrations of mercury in maternal hair
during the pregnancy, which were used as the index of fetal exposure,
ranged from 1 to 674 µg/g. Developmental retardation was defined as a
child's failure to walk a few steps unaided by 18 months of age or to
say two or three meaningful words by 24 months of age. A point system
was devised for the neurological examination, a score > 3 indicating
a definite abnormality. The prevalences of these indicators of poor
outcome were related to the concentrations of mercury in maternal
hair. The most frequent neurological findings were increased limb tone
and deep-tendon reflexes with persistent extensor plantar responses;
ataxia, hypotonia, and athetoid movements were also reported. Boys
appeared to be more severely affected than girls. Seven of the 28
children with the highest exposure and none of the 53 children with
lower exposure had had seizures.
Additional analyses of this data set were performed to identify
more precisely the shape of the dose-response relationship and, in
particular, the threshold for adverse neurodevelopmental effects, if
indeed one exists. Cox et al. (1989) obtained more accurate estimates
of peak exposure during pregnancy by applying an X-ray fluorescent
method to single strands of maternal hair. Using logit, hockey-stick,
and non-parametric kernel smoothing methods, they estimated a
population threshold of around 10 µg/g for the outcomes investigated.
The uncertainty associated with this estimate is heavily dependent,
however, on the estimated background prevalence of the poor outcomes.
For example, the upper bound of the 95% confidence interval for motor
retardation increases from 14 to 190 µg/g if the estimate of
background prevalence is changed from 0 to 4%. For neurological
abnormality, the upper bound of the 95% confidence interval for the
threshold estimate was 287 µg/g when a 9% background prevalence was
assumede. In later re-analyses of these data, Crump et al. (1995) and
Cox et al. (1995) demonstrated that the estimate of threshold depends
on the model used and is sensitive to the definition of abnormality.
In the case of delayed walking, the estimate was influenced by the
only four cases of delayed walking among the children of women whose
hair concentration of mercury was < 150 µg/g. The statistical
variability of the estimates of threshold appears likely to be
considerably greater than that of Cox et al. (1989). Crump et al.
(1995) concluded that the data from the Iraqi episode do not provide
convincing evidence of any adverse neurodevelopmental effect of
methylmercury at concentrations in maternal hair < 80 µg/g.
In evaluating these data, it is important to note that the
interviews were conducted when the mean age of the children was 30
months, but some of the children must have been considerably older at
this time, as the age at which children in the sample were reported to
have walked or talked was as much as 72 months. In addition, the birth
dates were generally not accorded significance, and maternal
recollection of the ages at which their children achieved milestones
were based on external events such as religious holidays. The extent
of the imprecision of these data is suggested by the strong digit
Table 4. Steady-state concentrations of methylmercury in humans after allometric
extrapolation of unit doses from three experimental species and comparison of
human and animal steady-state concentration ratios
Animal Concentration Human concentration Ratio at steady state
after equal doses (based on mass)
(based on surface)
Macaque 28 0.09 2.1
Rat 12 0.9 4.4
Mouse 6 1.0 7.5
Steady-state concentrations in animals are shown in the last column of Table 3.
preferences in the mothers' responses. For instance, an even number of
months was given for the estimated age at walking for 70 of the 78
children and for the estimated age at talking for 70 of 73 children;
75% of the estimates were multiples of six months. Finally, the extent
of selection bias in this cohort cannot be characterized because the
size of the base population from which it was drawn and the referral
mechanism that brought mothers and children to medical attention are
both unknown. For instance, women who knew that they had consumed
large amounts of contaminated grain and were concerned about their
children's welfare may have come forward, while women who consumed
equally large amounts of contaminated grain but whose children were
developing well may not. This issue is critical, because calculation
of a threshold requires a denominator (the size of the exposed
population) and the background prevalence of the adverse outcome in
order to estimate the 'added risk' associated with the exposure of
interest. In this regard, the background prevalence of developmental
abnormality appears to have been extremely high among the Iraqi
children who participated in the follow-up studies. The prevalence of
delayed walking among children whose mothers had concentrations of
mercury in hair < 10 µg/g, who can be viewed essentially as a control
group for estimating background prevalence, was 36% (11/31). In
contrast, in the population of children in the United States on whom
the Bayley scales of infant development were standardized (Bayley,
1969), the prevalence of delayed walking by this criterion was
approximately 5%. Similarly, the prevalence of delayed talking among
the Iraqi children was 22% (6/27), whereas 95% of 24-month-old
children in the standardization sample of the MacArthur communicative
development inventory were saying 50 words or more (Fenson et al.,
1993).
2.3.2 Childhood development
2.3.2.1 Neurological status
McKeown-Eyssen et al. (1983) studied 234 Cree children aged 12-30
months living in four communities in northern Quebec, Canada, whose
prenatal exposure to methylmercury was estimated on the basis of
maternal hair samples. Hair samples were collected from 28% of the
mothers during pregnancy, but the prenatal exposure of the rest of the
cohort was estimated from hair segments assumed to date from the
period of the pregnancy. The exposure index was the maximum
concentration of mercury in the segment of hair corresponding most
closely to the period from one month before conception to one month
after delivery. The mean concentration of mercury in maternal hair was
approximately 6 µg/g; it exceeded 20 µg/g in 6% of samples. One of
four paediatric neurologists who were unaware of the child's status of
exposure measured height, weight, and head circumference, identified
dysmorphology, and conducted a neurological examination, assessing
coordination, cranial nerves, and muscle tone and reflexes. The
neurologist then made a summary clinical judgement about the presence
of a neurological abnormality. None of the children was judged to have
an abnormal physical finding, but 3.5% of the boys ( n = 4) and 4.1%
of the girls ( n = 5) were considered to have a neurological
abnormality. The most frequent abnormality involved tendon reflexes,
which was seen in 11% of boys ( n = 13) and 12% of girls ( n = 14).
The only neurological finding that was significantly associated with
prenatal exposure to methylmercury, either before or after adjustment
for confounding, was abnormal muscle tone ( n = 2; increased tone in
legs only) or reflexes ( n = 13; five with isolated decreased
reflexes, six with generalized decreases, and two with generalized
increases) in boys ( p = 0.05). The risk for abnormal tone or
reflexes increased seven times with each 10-µg/g increase in prenatal
exposure to methylmercury (95% confidence interval, 1-51). After log
transformation of prenatal exposure, however, the p value for this
association increased to 0.14. When exposure was categorized, the
prevalence of tone or reflex abnormality did not increase in a clear
dose-related manner across the categories. In girls, the only
association identified was an unexpected inverse relationship between
prenatal exposure to methylmercury and incoordination, with a 60%
decrease in the probability of incoordination for each 10-µg/g
increment (odds ratio, 0.3; 95% confidence interval, 0.1-0.9;
p = 0.02). The authors noted several caveats with regard to the one
significant adverse association identified: the abnormalities of
muscle tone and reflexes in boys were isolated, mild, and of doubtful
clinical importance; the finding is not consistent with previous
results which suggest that increased exposure is expressed as severe
generalized neurological disease, including increases in tone and
reflexes; there was no coherent dose-response relationship; and there
was no consistency between the sexes. The finding may reflect chance,
lack of normality in the distribution of the exposure index, or
residual confounding.
A study was conducted in Mancora, a fishing community on the
northern coast of Peru, in which hair samples and clinical data were
obtained for 131 infant-mother pairs. The mean concentration of
mercury in maternal hair was 7 µg/g (range, 0.9-28), with an average
peak of 8.3 µg/g. The small difference between the mean and the peak
was probably due to the stability of the fish consumption of the
mothers. The major outcomes measured were anthropmorphic end-points
(birth weight, head circumference, and height), maternal reports of
infant development (age at which the infant sat, stood, walked, and
talked), and neurological status (Marsh et al., 1995). The specific
elements of the neurological assessment conducted and the age at which
the infants were examined were not described. Tone was decreased in
two children, limb weakness was seen in one child, reflexes were
decreased in one child and increased in four, and an abnormal Babinski
reflex was seen in one child; increased tone, primitive reflexes, and
ataxia were not observed. None of the signs was significantly
associated with either the mean or the peak concentration of mercury
in maternal hair.
In the pilot phase of a cross-sectional cohort study of child
development in the Seychelles, 789 infants aged 5-109 weeks were
evaluated by one paediatric neurologist who was unaware of their
exposure status (Myers et al., 1995a). The mean concentration of
mercury in maternal hair was 6.1 µg/g (range, 0.6-36 µg/g). The
features assessed included mental status, attention, social
interactions, vocalization, behaviour, coordination, posture and
movements, cranial nerves II-XII, muscle strength and tone, primitive
and deep-tendon reflexes, plantar responses, and age-appropriate
abilities such as rolling, sitting, pulling to stand, walking, and
running. The statistical analyses focused on three end-points,
selected because of their apparent sensitivity to prenatal exposure to
methylmercury in the studies in Iraq and the Cree population: overall
neurological status, increased muscle tone, and deep-tendon reflexes
in the extremities. The result of the overall examination was
considered to be 'abnormal' if any findings judged to be pathological
were present, including abnormalities of cranial nerves (pupils,
extraocular muscles, facial or tongue movement, swallowing, or
hearing), increase or decrease in muscle tone or deep-tendon reflexes,
incoordination, involuntary movements, or poorly developed speech or
functional abilities. Findings that were considered to be neither
normal nor pathological were categorized as 'questionable'. Because
the frequency of abnormal findings was low (2.8%), the questionable
(11%) and abnormal categories were combined. No association was found
between the concentration of mercury in maternal hair and questionable
or abnormal results, the frequency ranging from 16% for hair
concentrations of 0-3 µg/g to 12% for hair concentrations > 12 µg/g.
The frequency of abnormalities of limb tone or deep-tendon reflexes
was about 8% and did not vary in a dose-dependent manner with the
concentration of mercury in maternal hair.
In the main study, which involved a cohort of 735 children, one
paediatric neurologist who was unaware of the exposure status of the
children conducted essentially the same neurological examination that
had been used in the pilot study but when the participants were 6.5
months old (Myers et al., 1995b). The results of the overall
examination were considered to be abnormal or questionable if changes
in muscle tone, deep-tendon reflexes, or other neurological features
were pathological or the examiner considered that a child's functional
abilities were not appropriate for his or her age. Abnormal or
questionable neurological scores were found for 3.4% ( n = 25) of the
children, a frequency too low to permit statistical analysis. For both
limb tone and deep-tendon reflexes, the frequency of abnormalities was
2%; questionable limb tone was found in approximately 20% of the
children and questionable deep-tendon reflexes in approximately 15%.
For neither limb tone nor deep-tendon reflexes was the frequency of
abnormal or questionable findings significantly associated with the
concentration of mercury in maternal hair.
In a study carried out in the Faroe Islands (Denmark), a
functional neurological examination was administered to children at
the age of 7 years as part of a general physical examination. The
examination focused in particular on motor coordination and
perceptual-motor performance (Dahl et al., 1996). The tests for
coordination included rapid pronation or supination, reciprocal
coordination (alternately closing and opening the fists), and finger
opposition (touching the pulpa of the thumb with the pulpa of the
other fingers of the same hand). The perceptual-motor tests included
catching a 15-cm ball thrown from a distance of 4 m, finger agnosia,
and double finger agnosia. The results were scored as automatic or as
questionable or poor. Exposure to mercury was evaluated on the basis
of the concentration in maternal hair at delivery, in umbilical cord
blood, and in children's hair obtained at about 12 months of age
Exposure to mercury was not significantly associated with the number
of tests on which a child's performance was considered to be
'automatic', as < 60% of the children achieved such a score on the
tests for reciprocal motor coordination, simultaneous finger movement,
and finger opposition; however, children with questionable or poor
performance for finger opposition had had significantly higher mean
exposure to mercury than children with automatic performance (24
versus 22 µg/L; p = 0.04) (Grandjean et al., 1997).
2.3.2.2 Developmental milestones
The association between the achievement of developmental
milestones and prenatal exposure to methylmercury was evaluated in the
main cohort of the study in the Seychelles (Myers et al., 1997; Axtell
et al., 1998). Information on the ages at which a child was able to
walk without support and to say words other than 'mama' or 'dada' was
elicited by means of an interview with the person with whom the child
spent five or more nights per week (the 'caregiver'), conducted at an
evaluation at 19 months. Such information was available for 738 of the
779 children. The statistical approaches explored included standard
multiple regression in which age at achievement of a milestone was
log-transformed, hockey-stick models to estimate the threshold
concentration of mercury in maternal hair associated with delay in
achieving a milestone, and logistic regression analyses of 'delayed
walking', a binary variable in which an abnormal response was defined
as > 14 months. Prenatal exposure to methylmercury was estimated
from the total mercury in the single longest segment of hair dating
from the index pregnancy; the mean concentration was 5.8 µg/g (range,
0.5-27), and 22% of the children had been exposed to > 10 µg/g. The
mean age at which a child was considered to talk was not significantly
associated with the concentration of mercury in maternal hair in any
of the models tested. In regression analyses stratified by sex, a
positive association was found between age at walking and exposure to
mercury for boys ( p = 0.043) but not for girls. The interaction term
'mercury × sex' in analyses of the complete cohort was not
statistically significant. The magnitude of the delay in the age at
which boys walked--a 10-µg/gincrease in the concentration of mercury
in maternal hair associated with an approximately two-week delay in
walking--was viewed by the authors as clinically insignificant, and
the association was not significant when four statistical outliers
were excluded from the analysis. Hockey-stick models provided no
evidence of a threshold, as the fitted curves were essentially flat. A
child's risk for 'delayed walking' was not associated with the
concentration of mercury in maternal hair.
In a re-analysis of these data, Axtell et al. (1998) used
semiparametric generalized additive models with smoothing techniques
to identify lack of linearity. These models are less restrictive than
those used by Myers et al., which require strong assumptions about the
true functional form of a relationship. The major finding of Axtell et
al. was that the association between age at walking and the
concentration of mercury in maternal hair in boys was not linear,
walking being achieved at a later age as exposure increased from 0 to
7 µg/g but at a slightly earlier age as the concentration increased
beyond 7 µg/g. The size of the effect associated with the increase
from 0 to 7 µg/g was small, corresponding to a delay of less than one
day in the achievement of walking. Because no clear dose-response
relationship was seen at concen-trations > 7 µg/g, the authors
considered that the association found at lower concentrations did not
reflect a causal effect of mercury on the age at walking.
Data on developmental milestones were also collected in the study
in Peru (Marsh et al., 1995). The ages of the children at the time the
mothers were questioned about these events was not stated, although
the study was conducted prospectively and data were apparently
collected throughout the women's visits to postnatal clinics.
Regression analyses, including analyses stratified by sex, showed no
significant association between the concentration of mercury in
maternal hair and the ages at which the children sat, stood, walked,
or talked. The rates of developmental retardation were substantial,
especially for speech (13/131), although the criteria used to define
this outcome were not stated. The children's birthweight, height, and
head circumference were also unrelated to the concentration of mercury
in maternal hair.
The ages at which children achieved motor milestones were
investigated in a birth cohort of 1022 children born during a 21-month
period in 1986-87 in the Faroe Islands (Grandjean et al., 1995a). The
data were obtained from interviews with the mothers and from the
observations of district health nurses who had visited the homes of
the children on several occasions during their first year of life.
Complete data were available for 583 children (57% of the cohort).
Three motor milestones commonly achieved between 5 and 12 months of
age were selected for analysis: sitting without support, crawling, and
standing with support. The age at achievement of the three milestones
was not significantly associated with the concentration of mercury in
cord blood or maternal hair, but a significant negative association
was found between the age at achievement of all three milestones and
the concentration of mercury in children's hair at 12 months. The
authors concluded that this association reflected residual confounding
by duration of breast-feeding, since nursing was associated with both
higher hair mercury concentrations in children at 12 months of age and
more rapid achievement of milestones. This finding suggests that the
beneficial effects of nursing on early motor development are
sufficient to compensate for any slight adverse impact that prenatal
exposure to low concentrations of methylmercury may have on these
end-points.
2.3.2.3 Early development
In the study of Cree people reported by McKeown-Eyssen et al.
(1983), the Denver developmental screening test (Frankenburg et al.,
1981)was administered to all children aged 12-30 months. The scores
were reported as the percentage of items passed on each subscale
(gross motor, fine motor, language, personal and social) and on the
entire test. Although quantitative estimates of the associations
between test scores and the concentration of mercury in maternal hair
(mean, 6 µg/g; 6% > 20 µg/g) were not provided, the authors reported
that they found no significant association compatible with an adverse
effect of methylmercury, before or after adjustment for confounding
variables.
Kjellström et al. (1986, 1989) studied a cohort of children in
New Zealand whose prenatal exposure to methylmercury was estimated on
the basis of maternal hair samples and dietary questionnaires
collected during the pregnancy. Although nearly 11 000 women
participated in the initial phase during which information on exposure
was obtained, Kjellström et al. focused on 935 women who had reported
eating fish more than three times per week during the pregnancy. The
74 children of 73 women whose concentration of mercury in hair was
> 6 µg/g were considered to have had heavy exposure to mercury. Three
controls were matched to each of these children on the basis of ethnic
group, sex, maternal age, maternal smoking, area of maternal
residence, and the duration of maternal residence in New Zealand. The
concentration of mercury in maternal hair was 3-6 µg/g for one of each
of the controls, whose mother ate fish more than three times per week,
and 0-3 µg/g for the other two, whose mothers had a lower consumption
of fish. There were 57 fully matched sets of four children and four
incomplete sets, for a cohort of 237 children. Evaluations conducted
when the children were four years of age indicated that about 50% of
children with heavy exposure to mercury and 17% of the children in the
control group had an abnormal or questionable result on the Denver
developmental screening test.
In the pilot phase of the Seychelles study, a revised version of
the Denver test was administered to 789 children aged 1-25 months by
one examiner who was unaware of their exposure status (Myers et al.,
1995a). No association was found between the concentration of mercury
in maternal hair during pregnancy (mean, 6.6 µg/g) and the results on
the test when normal and questionable results were combined in the
conventional manner, although the prevalence of abnormal findings was
so low (three children, < 1%) that statistical analysis was not
meaningful. When abnormal and questionable (n = 65; 8%) results were
grouped, as was done in the study in New Zealand (Kjellström et al.,
1986), a higher concentration of mercury in maternal hair was
significantly (p = 0.04; one-tailed test) associated with a poor
outcome. This result was largely attributable to the higher frequency
of abnormal or questionable results (13%) among children with the
heaviest exposure to mercury (> 12 µg/g), in contrast to the
frequency of approximately 7% among children in each of the other four
groups (0-3, 3-6, 6-9, and 9-12 µg/g).
In the main study in the Seychelles, the revised Denver
developmental screening test was administered to a cohort of 740
children (mean concentration of mercury in maternal hair during
pregnancy, 5.9 µg/g; interquartile range, 6 µg/g) aged 6.5 months by
one examiner who was unaware of their exposure status (Myers et al.,
1995b). The frequency of results considered to be abnormal (three
children; 0.4%) or questionable was very low (11; 1.5%), precluding
meaningful statistical analysis. The Fagan test of infant intelligence
(Fagan, 1987), an assessment of visual recognition memory or novelty
preference, was also administered to 723 of the children at the same
age. The mean percent novelty preference in the entire cohort was 60%,
which is similar to that observed in other cohorts, and varied by <
1% across categories of concentration of mercury in maternal hair. The
index of perfor-mance on visual attention (the time required to reach
visual fixation criterion in familiarization trials) was also
unrelated to the concentration of mercury.
The Bayley scales of infant development were administered to
children in this cohort at the ages of 19 months ( n =738) and 29
months ( n =736) by examiners who were unaware of their exposure
status (Davidson et al., 1995a). Six items of the infant behaviour
record, a rating scale, were also completed by the examiner for
children aged 29 months, to assess activity, attention span,
responsiveness to the examiner, response to the caregiver,
cooperation, and general emotional tone. The Bayley scales yield two
primary scores: the mental development index and the psychomotor
development index. At both ages, the scores for mental development
were similar to the expected mean for children in the United States,
but the children's psychomotor development index scores were markedly
higher: at 19 months, approximately 200 of the children achieved the
highest possible score. Accordingly, the psychomotor development
scores at both ages were expressed as a binary variable, dividing the
distribution at the median score. The mental development scores were
not significantly associated with the concentration of mercury in
maternal hair during pregnancy. Similar results were obtained in a
secondary analysis that included only children with the lowest (< 3
µg/g) or highest (> 12 µg/g) concentration of mercury in maternal
hair. The scores at 19 months for the items on the mental development
scale designed to assess perceptual skills, dichotomized because of
the skewing, were not associated with exposure to mercury. The scores
for this index at 29 months could not be evaluated because of 'a
pronounced celling effect'. A psychomotor development index score
below the median was not significantly associated with the
concentration of mercury in maternal hair in the full logistic
regression model, but was associated with this exposure index
( p = 0.05) in a reduced model in which adjustment was made for a
smaller number of covariates selected a priori. A secondary analysis
of the psychomotor development index scores of children with the
lowest and highest exposure to mercury was not conducted because
statistical significance was not achieved in the full logistic
regression model. In analyses of the six items for infant behaviour,
the concentration of mercury in maternal hair was significantly
associated only with the examiner's ratings of children's activity
during the test session and only in boys. The score decreased by one
point (on a nine-point scale) for each 10 µg/g.
In general, the use of screening tests such as the Denver test in
studies of neurobehavioural toxicology is not recommended because of
their insensitivity to variations within the range of normal
performance (Dietrich & Bellinger, 1994). More detailed instruments,
such as the Bayley scales of infant development, have proven to be
sensitive to prenatal exposure to a variety of neurotoxicants,
including lead (Bellinger et al., 1987; Dietrich et al., 1987;
Wasserman et al., 1992) and polychlorinated biphenyls (PCBs; Rogan &
Gladen, 1991; Koopman-Esseboom et al., 1996). The only study of
exposure to mercury in which the Bayley scales were administered was
the study in the Seychelles, which found no significant association
between children's scores and prenatal exposure. It is notable that
the psychomotor development index scores were so high in this cohort
that the distribution had be split at the median and analysed as a
categorical variable. The median value was not provided by Davidson et
al. (1995a) but was derived from a figure in the paper of Davidson et
al. (1995b); it appeared to be approximately 130, or two standard
deviations above the expected population mean. The mental development
index scores were close to the expected population mean. It is
questionable, however, whether statistical analysis of scores > 130
versus < 130 allows assessment of a potential adverse effect of
prenatal exposure to mercury on early motor development.
2.3.2.4 Development later in childhood
In the study in New Zealand, the 237 children in the 57 fully
matched groups participated in a follow-up evaluation of
neurodevelopmental status at 6 years of age (Kjellström et al., 1989).
The mean concentration of mercury in maternal hair in the group with
heavy exposure was was 8.3 µg/g (range, 6-86 µg/g; all but 16, 6-10
µg/g). Extensive information was collected on possible confounding
factors such as social class, medical history, and nutrition. A
battery of 26 psychological and scholastic tests was administered to
assess general intelligence, language development, fine and gross
motor coordination, academic attainment, and social adjustment.
Multiple regression analyses were conducted of five primary
end-points: language development and spoken language; the revised
Wechsler intelligence scale for children for both performance and
fill-scale intelligence quotient (IQ) (Wechsler, 1974); the McCarthy
scales of children's perceptual performance abilities; and the
McCarthy scale of motor ability (McCarthy, 1972). In addition, robust
regression methods were applied, involving the assignment of weights
to observations, depending on their position within the distribution.
In these analyses, the concentration of mercury in maternal hair was
associated with poorer scores (p values ranging from 0.0034 to 0.074)
for full-scale IQ, language development, visual and spatial skills,
and gross motor skills. The findings in the unweighted regression
analyses were similar in direction although generally less
significant. The poorer mean scores of the children with heavier
exposure to mercury were largely attributable to those whose mothers
had concentrations in their hair < 10 µg/g, for whom the mean average
concentration during pregnancy was 13-15 µg/g and the mean peak
monthly concentration was about 25 µg/g. The concentration of mercury
in maternal hair accounted for relatively little variance in the
outcome measures and generally less than covariates such as social
class and ethnic group.
In additional analyses of this data set, Crump et al. (1998)
found that when the concentration of mercury in maternal hair was
expressed as a continuous rather than as a binary variable, none of
the 26 scores was associated with exposure to mercury at p < 0.10.
The results were heavily influenced, however, by the results of a
child whose mother had a concentration of mercury of 86 µg/g--more
than four times the next highest level--despite the fact that the
child's scores were not outliers by the usual criteria. When the data
for this child were excluded, the scores on six end-points (Clay
reading concepts or reading letters [Clay, 1979], McCarthy general
cognitive index, McCarthy perceptual performance index, grammar
completion, and grammar understanding) were inversely associated with
the concentration of mercury in maternal hair at the 10% level.
Several features of the study in New Zealand are noteworthy,
including the efforts made to collect data on potential confounding
variables and the broad battery of standardized outcome measures,
administered by trained examiners. In contrast to the acute, heavy
exposure of the Iraqi population, the cohort in New Zealand received
chronic, low exposure, which was probably fairly constant over time,
reflecting well-established food consumption patterns. In addition,
the concentration of mercury in maternal hair was measured
prospectively.
As part of the pilot phase of the study in the Seychelles,
children who reached the age of 66 months underwent developmental
assessment (Myers et al., 1995c). Of the 247 eligible children, 217
(88%) were given a battery of tests consisting of the McCarthy scales
of children's abilities, a preschool language scale, and two subtests
of the Woodcock-Johnson tests of achievement (Woodcock & Johnson,
1989): letter and word identification and applied problems. The median
concentration of mercury in maternal hair in 73 children whose mothers
had had a concentration of mercury in their hair > 9 µg/gor < 4
µg/g was 7.1 µg/g(range, 1-36). The rate of missing values was
substantial for some end-points: e.g. 34% for the general cognitive
index of the McCarthy scales. Increased concentrations of mercury in
maternal hair were associated with significantly lower general
cognitive scores ( p = 0.024), the scores declining approximately
five points between the lowest (< 3 µg/g) and the highest (> 12
µg/g) categories of exposure. A similar association was found on the
perceptual performance subscale of the McCarthy scales ( p = 0.013).
The children's scores on the auditory comprehension scale of the
preschool language test were also inversely associated with the
concentrations of mercury in maternal hair ( p = 0.0019), the scores
declining approximately 2.5 points across the range of exposure to
mercury. In additional analyses, exclusion of several outliers or
influential data points reduced the estimates of the effect of mercury
substantially, sometimes to nonsignificance. It is important to note
that in the pilot phase of the study information was not collected on
several key variables that frequently confound the association between
exposure to neurotoxicants and child development, specifically
socioeconomic status, maternal intelligen