Neurobehavioral
Assessments
Conducted
in
the
New
Zealand,
Faroe
Islands,
and
Seychelles
Islands
Studies
of
Methylmercury
Neurotoxicity
in
Children
David
C.
Bellinger
Professor
of
Neurology
Harvard
Medical
School
March
2005
Report
to
the
U.
S.
Environmental
Protection
Agency
Three
major
prospective
studies,
two
of
which
are
ongoing,
have
investigated
potential
neurotoxicity
of
low­
level,
chronic
methylmercury
(
MeHg)
exposure
from
seafood
consumption:
the
New
Zealand
study,
the
Seychelles
Child
Development
Study,
and
the
Faroe
Islands
study.
Among
the
epidemiological
studies
available,
these
are
the
ones
identified
by
the
National
Research
Council's
Committee
on
the
Toxicological
Effects
of
Methylmercury
as
the
most
rigorous
owing,
in
particular,
to
their
large
sample
sizes
and
their
prospective
designs,
including
contemporaneous
measurement
of
biomarkers
of
prenatal
MeHg
exposure.
All
three
studies
have
also
undergone
extensive
peer
review.
In
the
following
sections,
background
information
about
these
studies
is
provided,
including
the
methods
used
to
sample
the
target
population,
levels
of
mercury
exposure
in
the
study
samples,
and
the
neurobehavioral
tests
administered.

In
assembling
the
New
Zealand
sample,
Kjellstrom
et
al.
(
1989)
ascertained
the
fish
consumption
of
10,930
of
16,293
pregnant
women
in
the
study
area.
They
identified
935
women
who
reportedly
consumed
fish
at
least
3
times
per
week.
Hair
samples
were
obtained
from
these
women,
and
73
were
found
to
have
a
hair
mercury
(
Hg)
level
>=
6
µ
g/
g
(
or
ppm).
In
this
group,
the
mean
was
8.3
ppm,
with
a
range
of
6
to
86,
although
only
one
woman
had
a
level
greater
than
20
ppm.
Each
"
high"
Hg
woman
was
matched
to
3
controls,
based
on
hair
Hg,
reported
fish
consumption,
ethnic
group,
age,
smoking,
residence
time
in
New
Zealand,
and
child
sex.
One
control
was
a
woman
who
was
reported
eating
fish
3
or
more
times
per
week
but
whose
hair­
Hg
level
was
<
6
ppm,
thus
controlling
for
potential
confounding
of
hair­
Hg
level
and
characteristics
associated
with
frequent
fish
consumption.
Although
children
were
assessed
at
4
and
6
years
of
age,
only
the
data
collected
at
the
older
age
is
considered
in
our
analyses
because,
in
general,
the
reliability
and
validity
of
neurodevelopmental
testing
increases
with
child
age.
Table
1
lists
the
tests
administered
to
the
237
participating
children
at
the
6
year
evaluation,
and
indicates
the
general
functional
domain
each
is
considered
to
assess.

The
Faroe
Islands
investigators
assembled
a
birth
cohort
of
1,353
newborns
recruited
from
3
hospitals
over
a
21
month
period
in
1986­
1987.
In
1,022
women
(~
75%),
two
biomarkers
of
prenatal
MeHg
were
measured:
cord­
blood
Hg
and
maternal
hair
Hg
at
delivery.
The
mean
concentration
of
Hg
in
cord­
blood
for
917
children
assessed
at
age
7
was
22.6
ppb
(
inter­
quartile
range:
13.1
 
40.5
ppb,
full
range
0.9
 
351
ppb).
The
mean
concentration
of
Hg
in
maternal
hair
in
this
group
of
917
was
4.2
ppm
(
inter­
quartile
range:
2.6­
7.7
ppm,
full
range
0.2
 
39.1
ppm)
(
Budtz­
Jorgensen
et
al.
2004).
Neurodevelopmental
assessments
of
the
children
have
been
conducted
at
ages
7
and
14
years,
although
only
the
7­
year
results
have
been
published
(
Grandjean
et
al.,
1997).
Table
2
lists
each
test
administered
at
this
age
and
its
corresponding
general
functional
domain.

In
assembling
the
Seychelles
Child
Development
Study
sample,
investigators
obtained
hair
samples
from
779
pregnant
women
during
the
recruitment
period
(
representing
an
ascertainment
rate
of
approximately
50%).
After
application
of
exclusion
criteria,
the
study
sample
consisted
of
740
newborns.
Neurodevelopmental
assessments
were
conducted
when
the
children
were
6.5,
19,
29,
and
66
months,
and
at
9
years.
Only
data
from
the
9­
year
assessment
are
included
in
our
analyses
(
Myers
et
al.,
2003).
The
mean
maternal
hair
Hg
level
in
the
643
children
who
participated
in
this
assessment
was
6.9
ppm
(
SD
4.5)
(
Myers
et
al.,
2003).
Table
3
lists
the
tests
administered
at
this
age.
Several
of
the
tests
were
included
in
the
battery
in
order
to
increase
the
opportunities
for
direct
comparison
of
the
results
of
this
study
and
those
of
the
Faroe
Islands
study.
These
tests
are
identified
by
an
asterisk.

These
three
populations
were
selected
for
study
in
large
part
because
fish
consumption
was
known
to
be
high.
Indeed,
the
levels
of
prenatal
exposure
to
mercury
reported
for
the
study
samples
recruited
were
higher
than
those
measured
in
the
general
U.
S.
population.
In
the
National
Health
and
Nutrition
Examination
Survey
(
NHANES)
for
1999­
2000,
the
mean
hair
Hg
level
in
women
16
to
49
years
of
age
was
0.47
ppm;
the
median
was
0.19
ppm;
the
75th
percentile
was
0.42
ppm,
the
90th
percentile
was
1.11
ppm;
and
the
95th
percentile
was
1.73
ppm
(
McDowell
et
al.,
2004).
The
geometric
mean
blood
Hg
level
(
NHANES,
1999­
2000)
was
1.02
ppb;
the
75th
percentile
was
2.07
ppb;
the
90th
percentile
was
4.84
ppb;
and
the
95th
percentile
was
7.13
ppb
(
Schober
et
al.,
2003).
This
blood
Hg
level
cannot
be
directly
compared
to
the
cord
blood
Hg
level
reported
in
the
Faroe
Islands
study,
however,
insofar
as
the
ratio
of
maternal
blood
Hg:
cord
blood
Hg
appears
not
to
1:
1,
as
was
formerly
assumed.
Stern
and
Smith
(
2003)
suggested
a
point
estimate
of
1:
1.7,
with
1:
3.5
as
the
upper
limit
of
the
95%
confidence
interval.
It
is
also
important
to
note
that
subgroups
of
the
U.
S.
population
that
are
high
fish
consumers
were
not
specifically
targeted
for
sampling
(
e.
g.,
subsistence
or
sport
fishermen).
Even
within
the
NHANES,
1999­
2000
sample,
however,
blood
Hg
levels
were
strongly
associated
with
level
of
reported
fish
consumption,
with
women
eating
more
than
9
fish
meals
per
month
having
blood
Hg
levels
that
were
generally
5
to
6
times
greater
than
the
levels
of
women
who
reported
not
consuming
any
fish
meals
(
Mahaffey
et
al.,
2004).

Why
did
the
test
batteries
used
in
the
three
studies
differ
so
much?
One
reason
appears
to
be
differences
in
the
criteria
the
three
sets
of
investigators
used
to
assemble
their
respective
test
batteries.
For
the
New
Zealand
study,
" 
tests
were
selected
in
accordance
with
the
recommendations
by
a
WHO
Expert
Group
on
indicators
of
lead
neurotoxicity
in
children
(
specifically
those
considered
"
essential
or
highly
recommended,"
supplemented
with
" 
specific
tests
that
are
generally
used
in
educational
psychology
in
New
Zealand"
(
Kjellstrom
et
al.,
1989,
p.
20).
For
the
Faroe
Islands
study,
"
The
selection
of
neuropsychological
tests 
was
based
on
consideration
of
the
literature
on
neuropathological
findings
in
Minamata
Disease
associated
with
fetal
and
early
childhood
exposure
to
methylmercury,
the
functional
deficits
seen
in
children
with
early
life
exposure
to
other
neurotoxicants
such
as
lead,
and
an
interest
in
sampling
specific
types
of
cognitive
processing
capacities
(
rather
than
obtaining
non­
specific
omnibus
measures
of
`
general
intelligence')"
(
White
et
al.,
1994).
For
the
Seychelles
Child
Development
Study,
"
A
review
of
the
literature
concerning
human
fetal
exposure
to
low
concentrations
of
methylmercury
revealed
that
adverse
effects
might
occur
in
one
or
more
of
the
following
eight
domains
of
developmental
functioning:
general
cognitive,
visual­
perceptual,
speech­
language,
visual
memory,
visual
attention,
neuromotorneurological
social­
emotional,
and
learning­
achievement.
Tests
were
then
selected
to
adequately
assess
each
of
these
domains"
(
Davidson
et
al.,
1995).
Given
these
statements,
it
is
not
surprising
that
the
assessments
applied
in
the
New
Zealand
study
were
primarily
so­
called
apical
or
omnibus
tests
(
e.
g.,
the
Wechsler
Intelligence
Scale
for
Children­
Revised)
that
integrate
performance
over
many
domains
or
that
test
highly
integrative
functions
such
as
academic
achievement
(
e.
g.,
reading,
mathematics).
The
approach
used
by
the
Seychelles
investigators
was
similar,
focusing
on
insuring
valid
assessment
of
broad
domains
of
function,
including
intelligence
and
academic
achievement.
The
approach
of
the
Faroe
Islands
investigators
was
fundamentally
different.
They
focused
on
assessment
of
specific
cognitive
functions
that
the
literature
on
brain­
behavior
relationships
in
MeHg
poisoning
cases
identified
as
the
areas
in
which
subtle
deficits
would
be
most
likely
to
occur
as
the
result
of
low­
level
MeHg
exposure.
Thus
these
investigators
were
interested
in
elucidating
the
neuropsychological
mechanisms
of
MeHg
neurotoxicity.

Despite
the
differences
in
assessment
philosophy
that
motivated
the
test
choices
of
the
investigators
who
conducted
the
three
studies,
some
overlap
does
exist
in
the
specific
instruments
used.
The
New
Zealand
and
Seychelles
Islands
studies
both
administered
the
Wechsler
Intelligence
Scale
for
Children
(
WISC),
an
IQ
test,
although
the
older
WISC­
R
was
used
in
New
Zealand
and
the
WISC­
III
in
the
Seychelles.
In
the
Faroe
Islands
study,
three
of
the
13
subtests
of
the
WISC­
R
were
administered,
providing
the
only
overlap
between
the
tests
administered
in
this
study
and
in
the
New
Zealand
study.
As
noted,
the
battery
of
tests
administered
to
the
Seychellois
children
at
the
9­
year
assessment
was
enriched
with
tests
that
were
administered
to
the
Faroese
children
at
7­
years
in
order
to
enhance
study
comparability.
Thus,
the
overlap
is
greater
for
these
two
studies
than
for
other
pairs.
Nevertheless,
for
2
of
the
4
tests
administered
in
both
studies
(
California
Verbal
Learning
Test­
Children,
Boston
Naming
Test,
finger
tapping,
continuous
performance
test),
comparability
is
less
than
it
would
appear.
Different
procedures
were
used
in
the
two
studies
to
assess
finger
tapping.
Similarly,
the
Continuous
Performance
Tests
(
CPT)
used
differed
considerably.
Each
test
assesses
the
broad
domain
of
attention,
but
the
one
used
in
the
Faroe
Islands
assesses
primarily
vigilance
(
response
to
a
target
but
not
to
non­
targets),
while
the
one
used
in
the
Seychelles
Islands
assesses
primarily
response
inhibition
(
response
to
non­
targets
but
not
to
targets).
These
neuropsychological
processes
might
be
differentially
vulnerable
to
MeHg.
Moreover,
likely
differences
in
the
psychometric
characteristics
of
the
two
CPT
tasks
used
would
also
reduce
the
validity
of
comparisons
of
findings
in
the
two
studies.
The
only
test
for
which
scores
are
available
in
all
three
studies
is
the
Wechsler
Intelligence
Scale
for
Children,
although
a
different
version
of
this
test
(
was
used
in
the
Seychelles
Islands
(
WISC­
III)
than
in
New
Zealand
or
the
Faroe
Islands
(
both
WISC­
R).
As
part
of
the
standardization
of
the
WISC­
III,
however,
both
versions
were
administered
to
approximately
200
children.
The
correlation
between
the
Full­
Scale
IQ
scores
on
the
two
was
0.89,
although
children's
scores
averaged
5
points
higher
on
the
WISC­
R.
Such
an
upward
drift
over
time
in
the
mean
score
on
a
test
occurs
with
sufficient
frequency
to
have
been
given
a
name,
the
"
Flynn
effect"
(
Flynn,
1987).
The
WISC­
R
and
WISC­
III
appear
to
measure
the
same
constructs
and
generate
scores
with
similar
dispersion,
providing
little
reason
to
believe
that
scores
on
one
version
of
the
WISC
will
be
more
sensitive
to
MeHg
than
are
scores
on
the
other
version,
or
that
features
of
the
dose­
effect
relationship
will
differ.
In
the
Faroe
Islands
study,
only
three
of
the
13
subtests
of
the
WISC
were
administered
due
to
the
investigators'
conclusion
that
a
MeHg­
associated
deficit
in
an
apical
score
such
as
Full­
Scale
IQ
provides
relatively
little
insight
into
the
specific
nature
of
MeHg's
neuropsychological
effects
on
children.
As
described
in
a
later
section,
it
is
possible,
however,
to
estimate
a
Full­
Scale
IQ
score
from
scores
on
the
subtests
that
were
administered
in
the
Faroe
Islands
study.

In
some
respects,
it
is
fortunate
that
Full­
Scale
IQ
is
the
single
score
that
can
be
compared
across
the
three
studies
and
used
as
the
basis
for
estimating
the
benefits
of
Hg
reductions.
It
is
a
composite
index
that
averages
a
child's
performance
across
many
functional
domains,
providing
a
good
overall
picture
of
cognitive
health.
Despite
some
debate
over
the
usefulness
and
meaning
of
the
concept
of
IQ,
an
extensive
body
of
data
documents
the
predictive
validity
of
Full­
Scale
IQ,
as
measured
at
school­
age,
and
late
outcomes
such
as
academic
and
occupational
success
(
Neisser
et
al.,
1996).
Finally,
whereas
the
economic
import
of
changes
in
children's
scores
on
most
neuropsychological
tests
are
unknown,
methods
are
readily
available
for
valuing
shifts
in
IQ
and
thus
conducting
a
benefits
analysis
of
interventions
that
shift
the
IQ
distribution
in
a
population.

It
is
important
to
acknowledge,
however,
that
Full­
Scale
IQ
might
not
be
the
cognitive
endpoint
that
is
most
sensitive
to
prenatal
MeHg
exposure.
If
it
is
not,
a
benefits
analysis
resting
solely
on
IQ
would
underestimate
the
beneficial
changes
in
cognition
that
would
be
expected
to
result
from
reductions
in
MeHg
exposure.
The
three
studies
do,
in
fact,
provide
evidence
that
this
might
be
the
case.
Significant
inverse
associations
were
found,
in
both
the
New
Zealand
and
Faroe
Islands
studies,
between
prenatal
mercury
levels
and
cognitive
endpoints
other
than
IQ.
If,
as
the
Faroe
Islands
investigators
surmise,
the
effects
of
MeHg
are
focal,
affecting
only
specific
cognitive
functions,
using
Full­
Scale
IQ
as
the
primary
endpoint
in
a
benefits
analysis
might
underestimate
the
costs.
This
is
because
performances
on
many
diverse
functions
are
aggregated
in
computing
a
Full­
Scale
IQ
score.
If
MeHg
impairs
performance
on
only
certain
of
the
functions,
the
magnitude
of
the
change
in
Full­
Scale
IQ
score
per
unit
change
in
the
MeHg
biomarker
would
be
misleading
as
an
index
of
the
effects
of
MeHg,
overestimating
the
impact
on
the
functions
that
are
insensitive
to
MeHg
and
underestimating
the
impact
on
the
functions
that
are
sensitive
to
MeHg.
In
someone
with
deficits
that
are
severe
but
focal,
therefore,
Full­
Scale
IQ
would
be
misleading.
Moreover,
it
is
well­
known
that
cognitive
well­
being
might
be
severely
compromised
in
an
individual
whose
Full­
Scale
IQ
is
average
or
even
well­
above
average.
The
criterion
most
frequently
used
to
identify
children
with
learning
disabilities
for
the
purposes
of
assignment
to
special
education
services
is
a
discrepancy
between
IQ
and
achievement.
Specifically,
the
child's
achievement
in
reading,
math,
or
other
academic
area
is
significantly
lower
than
what
would
be
expected,
given
his
or
her
Full­
Scale
IQ.

Even
if
it
were
true
that
IQ
is
the
most
sensitive
cognitive
endpoint
with
respect
to
prenatal
MeHg
exposure,
a
calculation
of
the
benefits
of
a
reduction
in
mercury
exposure
that
is
based
solely
on
IQ
changes
will
almost
certainly
underestimate
those
benefits.
As
noted,
it
is
possible
for
an
individual
to
have
significant
cognitive
impairments
that
are
not
reflected
in
the
Full­
Scale
IQ
score.
For
instance,
two
of
the
most
sensitive
endpoints
in
the
Faroe
Islands
study
were
the
Boston
Naming
Test
(
BNT),
which
assesses
word
retrieval,
and
the
California
Verbal
Learning
Test­
Children
(
CVLT­
C),
which
assesses
the
acquisition
and
retention
of
information
presented
verbally.
Depending
on
the
severity
of
the
deficits,
a
child
who
has
deficits
in
either
of
these
skills
could
be
at
a
considerable
disadvantage
in
the
classroom
setting
and
at
substantial
educational
risk.
Neither
of
these
abilities
is
directly
assessed
by
the
WISC­
R
or
WISC­
III,
however,
and
so
do
not
explicitly
contribute
to
a
child's
IQ
score.
According
to
the
manual
for
the
CVLT­
C,
the
correlation
between
a
child's
scores
on
the
WISC­
R
subtest
Vocabulary
and
on
"
List
A,
Trials
1­
5,"
the
primary
CVLT­
C
index
of
immediate
free
recall,
was
0.33
for
5­
8
year
olds.
The
variance
shared
by
these
two
tests,
therefore,
is
only
about
11%.
This
dissociation
led
the
test
developers
to
conclude
that,
" 
the
CVLT­
C
evaluates,
in
large
part,
a
domain
of
cognition
different
from
verbal
I.
Q."
(
Delis
et
al.,
1994,
p.
91).
Unfortunately,
the
unavailability
of
the
kind
of
detailed
information
that
is
available
on
the
costs
associated
with
IQ
deficits
would
make
it
difficult
to
base
a
benefits
calculation
on
tests
such
as
the
BNT
and
CVLT­
C,
or
even
to
include
them
along
with
IQ
in
such
a
calculation.
The
unavoidable
implication,
therefore,
is
that
a
calculation
that
does
not
include
them
will
underestimate
the
benefits
of
MeHg
reductions.
As
noted,
the
definition
of
a
"
learning
disability"
usually
used
to
identify
children
in
need
of
remedial
services
is
academic
achievement
that
is
lower
than
that
which
would
be
expected
on
the
basis
of
the
child's
IQ.
If
the
costs
of
an
exposure
were
calculated
solely
on
the
basis
of
reduced
IQ,
an
exposure­
related
increase
in
the
prevalence
of
children
with
a
learning
disability
would
not
add
anything
to
the
total
costs.
Nevertheless,
in
the
absence
of
the
data
needed
to
calculate
the
benefits
of
increases
in
scores
on
tests
such
as
the
CVLT­
C
or
BNT,
estimating
the
benefits
solely
on
the
basis
of
IQ
remains
the
best
option.

The
Faroe
Islands
investigators
identified
a
concern
about
a
portion
of
the
data
they
collected
on
the
WISC­
R
subtest,
Similarities.
After
data
collection
had
begun,
a
decision
was
made
that,
for
logistical
reasons,
the
technician
would
administer
this
subtest
rather
than
the
psychologist.
Whereas
the
association
between
cord­
blood
Hg
and
Similarities
score
was
not
significant
in
data
for
the
cohort
as
a
whole,
it
was
significant
when
the
analysis
was
limited
to
the
data
collected
by
the
psychologist.
This
suggests
that
in
the
cohort
as
a
whole
the
Similarities
data
are
"
noisy,"
and
that
this
noise
would
carry
over
into
the
estimates
of
the
children's
Full­
Scale
IQ.
This
would,
possibly,
result
in
an
attenuation
of
the
association
between
prenatal
Hg
and
estimated
Full­
Scale
IQ.
One
option
for
addressing
this
would
be
to
estimate
the
Full­
Scale
IQ
scores
of
the
Faroese
children
on
the
basis
of
the
Block
Design
and
Digit
Span
scores
alone.
This
ad
hoc
solution
would
be
unsatisfactory
insofar
as
it
was
developed
only
after
the
results
of
data
analyses
were
known.
Another
solution
would
be
to
include
in
the
analyses
only
those
children
to
whom
the
psychologist
administered
the
Similarities
subtest.
This
would
be
unsatisfactory
because
it
would
result
in
a
greatly
diminished
sample
size.

Because
only
3
WISC­
R
subtests
were
administered
in
the
Faroe
Islands
study
(
Similarities,
Block
Design,
and
Digit
Span),
it
is
necessary
to
consider
whether
a
valid
Full­
Scale
IQ
can
be
estimated
from
them
and
compared
to
the
IQ
scores
measured
in
the
New
Zealand
and
Seychelles
Islands
studies.
The
WISC­
R
includes
10
core
subtests
and
3
supplementary
subtests.
Similarities
and
Block
Design
are
core
subtests,
and
Digit
Span
a
supplementary
subtest.
The
WISC­
R
was
standardized
on
a
nationallyrepresentative
sample
of
U.
S.
children
ages
6
to
16
years.
Based
on
subtest
scores,
Sattler
(
1988)
identified
the
pair,
triad,
quartet,
etc.
of
subtests
that
provides
the
most
valid
estimate
of
Full­
Scale
IQ.
Only
the
10
core
subtests
were
considered
in
this
exercise.
Of
the
45
possible
combinations
of
2
core
subtests
(
i.
e.,
10
subtests
taken
2
at
a
time),
the
combination
of
Similarities
and
Block
Design,
the
two
core
subtests
administered
in
the
Faroe
Islands
study,
ranked
3rd
in
the
magnitude
of
the
validity
coefficient
(
0.885).
The
top­
ranked
combination
was
Vocabulary
and
Block
Design
(
0.906).
The
combination
ranked
2nd
was
Information
and
Block
Design
(
0.888).
It
is
reasonable
to
expect
that
taking
into
account
Digit
Span
scores,
the
supplementary
subtest
administered,
will
increase
the
validity
coefficient.
The
results
of
this
exercise
engenders
confidence
that
combining
the
scores
of
the
Faroese
children
on
Similarities,
Block
Design
and
Digit
Span
will
provide
valid
estimates
of
their
Full­
Scale
IQ
scores
that
can
be
compared
to
the
Full­
Scale
IQ
scores
for
the
New
Zealand
and
Seychellois
children.

The
magnitude
of
the
changes
in
neurobehavioral
test
scores
associated
with
increases
in
neurotoxicant
biomarker
levels
tend
to
be
small.
In
the
case
of
lead
and
children's
IQ
scores,
for
instance,
most
estimates
cluster
around
the
figure
of
a
2
to
3
point
IQ
deficit
for
each
increase
of
10
µ
g/
dL
(
IPCS,
1995).
Whether
such
a
change
has
any
practical
import
for
a
child
or
for
the
population
has
been
a
topic
of
spirited
debate.
Some
have
argued
that
it
is
not,
that
it
is
"
within
the
noise"
because
it
is
smaller
than
the
typical
standard
error
of
measurement
(
SEM)
for
Full­
Scale
IQ
(
e.
g.,
the
average
SEM
for
the
WISC­
III
Full­
Scale
IQ
is
3.2).
Others
have
argued
that
this
statistic
is
irrelevant
with
respect
to
the
issue
of
shifts
in
the
mean
score
for
a
population
(
Bellinger,
2004).
In
many
respects,
the
issues
surrounding
the
public
health
implications
of
neurotoxicantassociated
shifts
in
a
test
score
are
identical
to
those
germane
to
other
topics
in
environmental,
nutritional,
and
chronic
disease
epidemiology.
Rose
and
Day
(
1990),
for
instance,
showed
that
the
correlation
is
very
high
(
approximately
0.9)
between
the
mean
values
of
a
health
indicator
within
a
population
(
e.
g.,
systolic
blood
pressure,
body
mass
index)
and
the
percentage
of
members
in
the
population
with
values
on
the
indicator
that
meet
diagnostic
criteria
for
the
corresponding
disease
(
e.
g.,
hypertension,
obesity).
Moreover,
the
slope
of
the
relationship
between
mean
value
and
percent
meeting
disease
criteria
is
quite
steep,
suggesting
that
a
small
difference
between
two
populations
in
mean
value
can
reflect
substantial
differences
between
populations
in
the
prevalence
of
cases
of
disease.
For
instance,
based
on
the
observed
relationship,
Rose
estimated
that
a
reduction
of
1
mmHg
in
mean
systolic
blood
pressure
in
a
population
would
be
associated
with
a
reduction
of
about
1%
in
the
prevalence
of
hypertension.
If
the
baseline
prevalence
of
hypertension
in
the
population
had
been
15%,
a
reduction
in
the
mean
systolic
blood
pressure
of
5
mmHg,
which
could
well
be
within
the
SEM
for
an
individual,
would
result
in
a
new
population
prevalence
of
10%,
a
reduction
of
33%.
On
the
basis
of
such
reasoning,
Rose
(
1981)
argued
that
equal
benefits
would
be
expected
to
result
from
a
"
high
risk"
strategy
involving
the
administration
of
anti­
hypertensive
drugs
to
everyone
with
a
diastolic
blood
pressure
>
105
mmHg
and
a
"
population"
strategy
involving
an
intervention
that
produced
a
downward
shift
of
2­
3
mmHg
in
the
mean
blood
pressure
in
the
population.
This
is
illustration
of
what
he
termed,
"
the
prevention
paradox,"
namely
that
"
A
preventive
measure
which
brings
much
benefit
in
the
population
offers
little
to
each
participating
individual"
(
Rose,
1985).
Although
Rose's
developed
his
principles
on
the
basis
of
cross­
sectional
data
comparing
different
population,
recent
prospective
studies,
including
a
randomized
trial,
have
confirmed,
within
a
given
population,
that
small
shifts
in
the
central
tendency
of
the
distribution
of
a
health
indicator
is
associated
with
substantial
changes
in
the
tails
of
the
distribution,
specifically
in
terms
of
the
proportion
the
population
with
values
of
the
indicator
in
the
range
within
which
individuals
become
"
cases"
of
disease.

It
is
also
important
to
recognize
that
a
shift
in
the
mean
value
of
a
response
parameter
in
a
population
does
not
mean
that
the
response
parameter
of
each
individual
within
the
population
shifted
by
the
same
amount.
The
mean
shift
is
a
point
estimate,
the
weighted
average
of
the
shifts
among
individuals
who
vary
in
their
sensitivity
to
MeHg.
The
shift
might
be
considerably
greater
than
the
point
estimate
among
individuals
who
are
particularly
sensitive
to
MeHg
and
considerably
less
among
individuals
who
are
not.

With
respect
to
neurotoxicant
exposures,
this
phenomenon
has
been
explored
most
with
regard
to
lead
exposure
in
children.
Needleman
et
al.
(
1982)
showed
that
whereas
the
median
verbal
IQ
scores
of
"
high"
lead
and
"
low"
lead
children
differed
only
modestly,
the
frequency
of
low
scores
(
i.
e.,
<
80)
in
the
high
group
was
double
that
in
the
low
lead
group.

NOTE:
This
report
has
been
peer­
reviewed
in
accordance
with
EPA
guidelines.
Peer
review
comments
and
responses
are
presented
elsewhere
(
U.
S.
EPA
2005).
References
Bellinger
DC.
(
2004).
What
is
an
adverse
effect?
A
possible
resolution
of
clinical
and
epidemiological
perspectives
on
neurobehavioral
toxicity.
Environmental
Research,
95,
394­
405.

Budtz­
Jorgensen
E,
Grandjean
P,
Jorgensen
PJ,
Weihe
P,
Keiding
N.
(
2004).
Association
between
mercury
concentrations
in
blood
and
hair
in
methylmercury­
exposed
subjects
at
different
ages.
Environmental
Research;
95(
3):
385­
93.

Davidson
PW,
Myers
GJ,
Cox
C,
et
al.
(
1995).
Neurodevelopmental
test
selection,
administration,
and
performance
in
the
main
Seychelles
Child
Development
Study.
Neurotoxicology
16:
665­
676.

Davidson
PW,
Myers
GJ,
Cox
C,
Axtell
C,
Shamlaye
C,
Sloane­
Reeves
J,
Cernichiari
E,
Needham
L,
Choi
A,
Wang
Y,
Berlin
M,
Clarkson
TW
(
1998).
Effects
of
prenatal
and
postnatal
methylmercury
exposure
from
fish
consumption
on
neurodevelopment:
outcomes
at
66
months
of
age
in
the
Seychelles
Child
Development
Study.
Journal
of
the
American
Medical
Association
Aug
26;
280(
8):
701­
7.

Delis
DC,
Kramer
JH,
Kaplan
E,
Ober
BA.
CVLT­
C.
California
Verbal
Learning
Test.
Children's
version.
Manual.
San
Antonio:
The
Psychological
Corporation,
1994.

Flynn
JR.
(
1987).
Massive
IQ
gains
in
14
nations:
What
IQ
tests
really
measure.
Psychological
Bulletin
101:
171­
191.

Grandjean
P,
Weihe
P,
White
RF,
Debes
F.,
Araki
S.,
Yokoyama
K.,
et
al.
(
1997).
Cognitive
deficit
in
7­
year­
old
children
with
prenatal
exposure
to
methylmercury.
Neurotoxicology
and
Teratology,
19,
417­
428.

International
Programme
on
Chemical
Safety.
(
1995).
Environmental
Health
Criteria
165:
Inorganic
Lead.
Geneva:
World
Health
Organization.

Kjellstrom
T,
Kennedy
P,
Wallis
S,
Stewart
A,
Friberg
L,
Lind
B,
et
al.
(
1989).
Physical
and
mental
development
of
children
with
prenatal
exposure
to
mercury
from
fish.
National
Swedish
Environmental
Protection
Board
Report
No.
3642.

Mahaffey
KR,
Clickner
RP,
Bodurow
CC.
(
2004).
Blood
organic
mercury
and
dietary
mercury
intake:
National
Health
and
nutrition
examination
survey,
1999
and
2000.
Environmental
Health
Perspectives,
112(
5),
562­
570.
McDowell
MA,
Dillon
CF,
Osterloh
J,
Bolger
PM,
Pellizzari
E,
Fernando
R,
Montes
de
Oca
R,
Schober
SE,
Sinks
T,
Jones
RL,
Mahaffey
KR.
(
2004).
Hair
mercury
levels
in
U.
S.
children
and
women
of
childbearing
age:
reference
range
data
from
NHANES
1999­
2000.
Environmental
Health
Perspectives
Aug;
112(
11):
1165­
71.

Myers
GJ,
Davidson
PW,
Cox
C,
Shamlaye
CF,
Palumbo
D,
Cernichiari
E,
et
al.
(
2003).
Prenatal
methylmercury
exposure
from
ocean
fish
consumption
in
the
Seychelles
child.
Lancet,
361,
1686­
1692.

Needleman
HL,
Leviton
A,
and
Bellinger
D.
(
1982).
Lead­
associated
intellectual
deficits.
New
England
Journal
of
Medicine
306,
367.

Neisser
U,
Boodoo
G,
Bouchard
TJ,
et
al.
(
1996).
Intelligence:
Knowns
and
unknowns.
American
Psychologist
51:
77­
101.

Rose
G.
(
1985).
Sick
individuals
and
sick
populations.
International
Journal
of
Epidemiology
14,
32­
38.

Rose
G.
(
1981).
Strategy
of
prevention:
Lessons
from
cardiovascular
disease.
British
Medical
Journal
282,
1847­
1851.

Rose
G,
and
Day
S.
(
1990).
The
population
mean
predicts
the
number
of
deviant
individuals.
British
Medical
Journal
301,
1031­
1034.

Sattler
JM.
(
1998).
Assessment
of
Children,
3rd
Edition.
San
Diego:
Jerome
M.
Sattler
Publisher.

Schober
SE,
Sinks
TH,
Jones
RL,
Bolger
PM,
McDowell
M,
Osterloh
J,
et
al.
(
2003).
Blood
mercury
levels
in
US
children
and
women
of
childbearing
age,
1999­
2000.
Journal
of
the
American
Medical
Association,
289(
13),
1667­
1674.

Stern
AH,
Smith
AE.
(
2003).
An
assessment
of
the
cord
blood:
maternal
blood
methylmercury
ratio:
Implications
for
risk
assessment.
Environmental
Health
Perspectives
111:
1465­
1470.

U.
S.
EPA
(
2005).
Documentation
of
Peer
Review
for
EPA's
Estimation
of
an
Integrated
Dose­
Response
Function
for
Intelligence
Quotient
(
IQ)
and
Prenatal
Exposure
to
Mercury.
National
Center
for
Environmental
Economics,
U.
S.
Environmental
Protection
Agency.

Wechsler
D.
(
1991).
WISC­
III
Manual.
San
Antonio:
The
Psychological
Corporation.

White
RF,
Debes
F,
Dahl
R,
Grandjean
P
(
1994).
Developmental
and
field
testing
of
a
neuropsychological
test
battery
to
assess
the
effects
of
methylmercury
in
the
Faroe
Islands.
In:
Proceedings
of
the
International
Symposium
on
Assessment
of
Environmental
Pollution
and
Health
Effects
from
Methylmercury
(
pp.
127­
140).
University
of
Kumamoto,
Kumamoto,
Japan.
Table
1
Neurobehavioral
Tests
Administered
at
the
6­
Year
Evaluations
in
the
New
Zealand
Study
Test
Primary
Domain
Assessed
Wechsler
Intelligence
Scale
for
Children­
Revised
General
intelligence
McCarthy
Scales
of
Children's
Abilities
General
development
Test
of
Language
Development
General
verbal
skills
Peabody
Picture
Vocabulary
Test
Receptive
language
Clay
Reading
Diagnostic
Survey
Reading
Burt
Word
Recognition
Test
Single
word
reading
Key
Math
Diagnostic
Arithmetic
Test
General
math
skills
Everts
Behavioural
Rating
Scale
Behavior
disorders
Table
2
Neurobehavioral
Tests
Administered
at
the
7­
Year
Evaluations
in
the
Faroe
Islands
Study
Test
Primary
Domain
Assessed
Wechsler
Intelligence
Scale
for
Children­
Revised
(
selected
subtests)
Digit
span
Similarities
Block
Design
Short­
term
memory
Abstract
verbal
reasoning
Constructional
praxis
Bender­
Gestalt
Test
Visual­
motor
integration
California
Verbal
Learning
Test­
Children
Verbal
learning
and
memory
Boston
Naming
Test
Confrontational
naming
Tactual
Performance
Test
Nonverbal
memory
Neurobehavioral
Evaluation
System
(
NES)
(
selected
tests)
Finger
tapping
Hand­
eye
coordination
Continuous
performance
test
Motor
speed
Hand­
eye
coordination
Vigilance
Profile
of
Mood
States
Mood
Child
Behavior
Checklist
(
selected
items)
Behavior
disorders
Table
3
Neurobehavioral
Tests
Administered
at
the
9­
Year
Evaluations
in
the
Seychelles
Islands
Study
Test
Primary
Domain
Assessed
Wechsler
Intelligence
Scale
for
Children­
Third
Edition
General
intelligence
California
Verbal
Learning
Test­
Children*
Verbal
learning
and
memory
Boston
Naming
Test*
Confrontational
naming
Finger
tapping*
a
Motor
speed
Continuous
performance
test*
a
Vigilance
Developmental
Test
of
Visual­
Motor
Integration
Visual­
motor
integration
Bruininks­
Oseretsky
Test
of
Motor
Proficiency
(
selected
subtests)
Gross
and
fine
motor
skills
Grooved
Pegboard
Manual
dexterity
Trail­
Making
Test
Visual
tracking
and
executive
function
Woodcock­
Johnson
Tests
of
Achievement
(
selected
subtests)
Letter­
Word
Identification
Applied
Math
Single
word
reading
Quantitative
problem­
solving
Wide
Range
Assessment
of
Memory
and
Learning
Design
Memory
subtest
Visual
memory
Haptic
Discrimination
Test
Cross­
modal
integration
Child
Behavior
Checklist
Behavioral
disorders
Connors'
Hyperactivity
Index
ADHD
screener
 
tests
presumably
added
to
the
battery
in
order
to
increase
the
comparability
of
the
data
collected
in
the
Faroe
Islands
study
at
age
7
years
and
in
the
Seychelles
Islands
study
at
age
9
years
a.
Although
the
batteries
used
in
both
the
Faroe
Islands
and
Seychelles
Islands
studies
included
a
finger
tapping
task
and
a
continuous
performance
test,
the
specific
procedures
used
in
the
two
studies
were
not
the
same,
reducing
the
comparability
of
the
scores
obtained.
