SAP
Minutes
No.
2005­
04
A
Set
of
Scientific
Issues
Being
Considered
by
the
Environmental
Protection
Agency
Regarding:

PRELIMINARY
N­
METHYL
CARBAMATE
CUMULATIVE
RISK
ASSESSMENT
AUGUST
23
­
26,
2005
FIFRA
Scientific
Advisory
Panel
Meeting,
held
at
the
Holiday
Inn
­
Rosslyn
at
Key
Bridge,
Arlington,
Virginia
2
of
63
NOTICE
These
meeting
minutes
have
been
written
as
part
of
the
activities
of
the
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
(
FIFRA),
Scientific
Advisory
Panel
(
SAP).
The
meeting
minutes
represent
the
views
and
recommendations
of
the
FIFRA
SAP,
not
the
United
States
Environmental
Protection
Agency
(
Agency).
The
content
of
the
meeting
minutes
does
not
represent
information
approved
or
disseminated
by
the
Agency.
The
meeting
minutes
have
not
been
reviewed
for
approval
by
the
Agency
and,
hence,
the
contents
of
these
meeting
minutes
do
not
necessarily
represent
the
views
and
policies
of
the
Agency,
nor
of
other
agencies
in
the
Executive
Branch
of
the
Federal
government,
nor
does
mention
of
trade
names
or
commercial
products
constitute
a
recommendation
for
use.

The
FIFRA
SAP
is
a
Federal
advisory
committee
operating
in
accordance
with
the
Federal
Advisory
Committee
Act
and
established
under
the
provisions
of
FIFRA
as
amended
by
the
Food
Quality
Protection
Act
(
FQPA)
of
1996.
The
FIFRA
SAP
provides
advice,
information,
and
recommendations
to
the
Agency
Administrator
on
pesticides
and
pesticide­
related
issues
regarding
the
impact
of
regulatory
actions
on
health
and
the
environment.
The
Panel
serves
as
the
primary
scientific
peer
review
mechanism
of
the
EPA,
Office
of
Pesticide
Programs
(
OPP),
and
is
structured
to
provide
balanced
expert
assessment
of
pesticide
and
pesticide­
related
matters
facing
the
Agency.
Food
Quality
Protection
Act
Science
Review
Board
members
serve
the
FIFRA
SAP
on
an
ad
hoc
basis
to
assist
in
reviews
conducted
by
the
FIFRA
SAP.
Further
information
about
FIFRA
SAP
reports
and
activities
can
be
obtained
from
its
website
at
http://
www.
epa.
gov/
scipoly/
sap/
or
the
OPP
Docket
at
(
703)
305­
5805.
Interested
persons
are
invited
to
contact
Myrta
R.
Christian,
SAP
Designated
Federal
Official,
via
e­
mail
at
christian.
myrta@
epa.
gov.

In
preparing
the
meeting
minutes,
the
Panel
carefully
considered
all
information
provided
and
presented
by
the
Agency
presenters,
as
well
as
information
presented
by
public
commenters.
This
document
addresses
the
information
provided
and
presented
by
the
Agency
within
the
structure
of
the
charge.
3
of
63
TABLE
OF
CONTENTS
PARTICIPANTS...........................................................................................................
5
INTRODUCTION.........................................................................................................
7
PUBLIC
COMMENTERS............................................................................................
8
SUMMARY
OF
PANEL
DISCUSSION
AND
RECOMMENDATIONS...................
8
PANEL
DELIBERATIONS
AND
RESPONSE
TO
CHARGE.................................
11
REFERENCES................................................................................................................
.60
4
of
63
SAP
Minutes
No.
2005­
04
A
Set
of
Scientific
Issues
Being
Considered
by
the
Environmental
Protection
Agency
Regarding:

PRELIMINARY
N­
METHYL
CARBAMATE
CUMULATIVE
RISK
ASSESSMENT
AUGUST
23
­
26,
2005
FIFRA
Scientific
Advisory
Panel
Meeting,
held
at
the
Holiday
Inn
­
Rosslyn
at
Key
Bridge,
Arlington,
Virginia
Stephen
M.
Roberts,
Ph.
D.
Steven
G.
Heeringa,
Ph.
D.
FIFRA
SAP,
Session
Chair
FIFRA
SAP,
Session
Chair
FIFRA
Scientific
Advisory
Panel
FIFRA
Scientific
Advisory
Panel
Date:
October
13,
2005
Date:
October
13,
2005
Myrta
R.
Christian,
M.
S.
Designated
Federal
Official
FIFRA
Scientific
Advisory
Panel
Date:
October
13,
2005
5
of
63
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
Scientific
Advisory
Panel
Meeting
August
23
­
26,
2005
PRELIMINARY
N­
METHYL
CARBAMATE
CUMULATIVE
RISK
ASSESSMENT
PARTICIPANTS
FIFRA
SAP
Chairs
Stephen
M.
Roberts,
Ph.
D.,
Professor
&
Program
Director,
University
of
Florida,
Center
for
Environmental
&
Human
Toxicology,
Gainesville,
FL
Steven
G.
Heeringa,
Ph.
D.,
Research
Scientist
&
Director
for
Statistical
Design,
University
of
Michigan,
Institute
for
Social
Research,
Ann
Arbor,
MI
Designated
Federal
Official
Myrta
R.
Christian,
M.
S.,
FIFRA
Scientific
Advisory
Panel,
Office
of
Science
Coordination
and
Policy,
EPA
FIFRA
Scientific
Advisory
Panel
Members
Janice
E.
Chambers,
Ph.
D.,
D.
A.
B.
T.,
William
L.
Giles
Distinguished
Professor
&
Director,
Center
for
Environmental
Health
Sciences,
College
of
Veterinary
Medicine,
Mississippi
State
University,
Mississippi
State,
MS
H.
Christopher
Frey,
Ph.
D.,
Professor,
Civil
Engineering,
North
Carolina
State
University
Raleigh,
NC
Kenneth
M.
Portier,
Ph.
D.,
Associate
Professor,
Statistics,
Institute
of
Food
and
Agricultural
Sciences,
University
of
Florida,
Gainesville,
FL
FQPA
Science
Review
Board
Members
W.
Stephen
Brimijoin,
Ph.
D.,
Chair,
Department
of
Molecular
Pharmacology,
Mayo
Clinic
and
Medical
School,
Rochester,
MN
John
R.
Bucher,
Ph.
D.,
D.
A.
B.
T.,
Deputy
Director,
Environmental
Toxicology
Program,
National
Institute
of
Environmental
Health
Sciences,
National
Institutes
of
Health,
Research
Triangle
Park,
NC
6
of
63
Deborah
Cory­
Slechta,
Ph.
D.,
Director,
Environmental
and
Occupational
Health
Sciences
Institute,
Robert
Wood
Johnson
Medical
School,
University
of
Medicine
and
Dentistry
of
New
Jersey
and
Rutgers
State
University,
Piscataway,
NJ
Bernard
Engel,
Ph.
D.,
Professor,
Agricultural
&
Biological
Engineering,
Purdue
University,
West
Lafayette,
IN
Dale
Hattis,
Ph.
D.,
Research
Professor,
Center
for
Technology,
Environment
&
Development
(
CENTED),
George
Perkins
Marsh
Institute,
Clark
University,
Worcester,
MA
Chensheng
Lu,
Ph.
D.,
Assistant
Professor,
Department
of
Environmental
&
Occupational
Health,
Rollins
School
of
Public
Health,
Emory
University,
Atlanta,
GA
Peter
Macdonald,
D.
Phil.,
Professor
of
Mathematics
and
Statistics,
McMaster
University,
Hamilton,
Ontario,
Canada
Robert
W.
Malone,
Ph.
D.,
Agricultural
Engineer,
USDA­
ARS,
National
Soil
Tilth
Laboratory,
Ames,
IA
Nu­
May
Ruby
Reed,
Ph.
D.,
D.
A.
B.
T.,
Staff
Toxicologist,
Department
of
Pesticide
Regulation,
California
Environmental
Protection
Agency,
Sacramento,
CA
P.
Barry
Ryan,
Ph.
D.,
Professor,
Environmental
&
Occupational
Health,
Rollins
School
of
Public
Health,
Emory
University,
Atlanta,
GA
Tammo
S.
Steenhuis,
Ph.
D.,
Professor
of
Watershed
Management,
Cornell
University
Ithaca,
NY
7
of
63
INTRODUCTION
The
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
(
FIFRA),
Scientific
Advisory
Panel
(
SAP)
has
completed
its
review
of
the
preliminary
N­
methyl
carbamate
(
NMC)
cumulative
risk
assessment.
Advance
notice
of
the
meeting
was
published
in
the
Federal
Register
on
June
29,
2005.
The
review
was
conducted
in
an
open
Panel
meeting
held
in
Arlington,
Virginia,
on
August
23
­
26,
2005.
Dr.
Stephen
M.
Roberts
and
Dr.
Steven
G.
Heeringa
chaired
the
meeting.
Myrta
R.
Christian
served
as
the
Designated
Federal
Official.

The
FIFRA
SAP
met
to
consider
and
review
the
preliminary
N­
methyl
carbamate
cumulative
risk
assessment.
The
Food
Quality
Protection
Act
of
1996
amended
both
FIFRA
and
FFDCA.
One
of
the
major
changes
is
the
requirement
that
EPA
considers
risk
posed
by
pesticides
acting
by
a
common
mechanism
of
toxicity.
For
such
groups
of
pesticides,
EPA's
Office
of
Pesticide
Programs
(
OPP)
has
treated
cumulative
risk,
under
FQPA,
as
the
risk
of
a
common
toxic
effect
associated
with
concurrent
exposure
by
all
relevant
pathways
and
routes.
The
N­
methyl
carbamate
pesticides
were
assigned
priority
for
tolerance
reassessment
early
during
the
process
of
FQPA
implementation.
OPP
established
the
N­
methyl
carbamate
pesticides
as
a
common
mechanism
group
in
February
2004
based
on
their
shared
ability
to
inhibit
AChE
by
carbamylation.
Those
pesticides
included
in
the
cumulative
risk
assessment
were
announced
in
the
February
FR
Notice.
OPP
has
proceeded
with
the
development
of
the
cumulative
risk
assessment
in
a
step
by
step
process
including
review
of
a
case
study
for
the
N­
methyl
carbamate
risk
assessment
in
February
2005
by
the
FIFRA
SAP.
Based
on
the
comments
from
the
SAP,
the
Agency
made
appropriate
revisions.
The
Agency
released
its
preliminary
cumulative
risk
assessment
for
the
N­
methyl
carbamate
pesticides
in
late
July
2005.
The
hazard
assessment
for
these
chemicals
involved
empirical
dose­
response
modeling
of
the
available
red
blood
cell
and
brain
cholinesterase
inhibition
and
recovery
data.
The
exposure
assessment
utilized
probabilistic
approaches
in
all
pathways
considered:
food,
drinking
water,
and
residential/
non­
occupational
for
various
population
subgroups
and
regions.
These
aspects
were
incorporated
into
a
preliminary
cumulative
risk
assessment
document,
which
the
SAP
reviewed
in
August
2005.

The
agenda
for
this
SAP
meeting
involved
an
introduction,
background,
and
detailed
presentations
of
the
several
issues
related
to
the
preliminary
N­
methyl
carbamate
cumulative
risk
assessment.
Issues
related
to
the
hazard
assessment
were
provided
by
Dr.
Anna
Lowit
(
Health
Effects
Division
[
HED],
Office
of
Pesticide
Programs
[
OPP],
EPA),
Dr.
R.
Woodrow
Setzer
(
Office
of
Research
and
Development
[
ORD],
National
Center
for
Computational
Toxicology,
EPA),
and
Dr.
Stephanie
Padilla
(
ORD,
National
Health
and
Environmental
Effects
Research
Laboratory,
EPA).
Issues
related
to
water
exposure
assessment
were
presented
by
Mr.
Nelson
Thurman
and
Dr.
Dirk
Young
(
Environmental
Fate
and
Effects
Division,
OPP,
EPA).
Dietary
Assessment
presentation
was
provided
by
Mr.
David
Hrdy
(
HED,
OPP,
EPA).
Residential
Assessment
issues
were
presented
by
Mr.
Jeff
Evans,
Dr.
Steve
Nako,
and
Mr.
Philip
Villanueva
(
HED,
OPP,
EPA).
Model
Results
8
of
63
Comparison
and
Cumulative
Risk
Assessment
presentations
were
provided
by
Mr.
Alan
Dixon
and
Mr.
David
Hrdy
(
HED,
OPP,
EPA)
,
respectively.
Finally,
the
Risk
Characterization
presentation
was
provided
by
Dr.
Anna
Lowit,
Mr.
Nelson
Thurman,
and
Mr.
David
Miller
(
OPP,
EPA)

Dr.
Clifford
Gabriel
(
Director,
Office
of
Science
Coordination
and
Policy,
EPA),
Mr.
Jim
Jones
(
Director,
Office
of
Pesticides
Programs,
EPA),
Dr.
Tina
Levine
(
Director,
Health
Effects
Division,
Office
of
Pesticide
Programs,
EPA)
and
Dr.
Steven
Bradbury
(
Director,
Environmental
Fate
and
Effects
Division,
Office
of
Pesticide
Programs)
offered
opening
remarks
at
the
meeting.

In
preparing
these
meeting
minutes,
the
Panel
carefully
considered
all
information
provided
and
presented
by
the
Agency
presenters,
as
well
as
information
presented
by
public
commenters.
This
document
addresses
the
information
provided
and
presented
at
the
meeting,
especially
the
response
to
the
Agency's
charge.

PUBLIC
COMMENTERS
Oral
statements
were
presented
as
follows:

On
behalf
of
Bayer
CropScience:
Iain
Kelly,
Ph.
D.,
Gary
Mihlan,
Ph.
D.,
and
Abraham
Tobia,
Ph.
D.

On
behalf
of
Carbamate
Working
Group:
Harvey
Clewell,
Director,
Center
for
Human
Health
Assessment,
CIIT
Centers
for
Health
Research
Jane
D.
McCarty,
DABT,
Chair,
Toxicology
Sub­
team
of
CWG,
Technical
Leader
for
Toxicology,
FMC
Corporation
On
behalf
of
DuPont
Crop
Protection:
Ralph
L.
Warren,
Ph.
D.

On
behalf
of
the
American
Bird
Conservancy:
Michael
Fry,
Ph.
D.

SUMMARY
OF
PANEL
DISCUSSION
AND
RECOMMENDATIONS
The
Panel
addressed
a
total
of
eleven
questions
regarding
the
preliminary
cumulative
risk
assessment
for
the
N­
methyl
carbamates.
In
its
responses,
the
Panel
repeatedly
commended
the
Agency
for
its
progress
in
developing
a
cumulative
risk
assessment
based
on
good
empirical
data,
sound
technology
and
proper
statistical
methodology.
Overall
there
was
strong
support
for
the
Agency's
approach
and
for
the
9
of
63
document
that
was
under
review.
Although
there
were
many
suggestions
for
improving
the
clarity
and
transparency
of
the
document,
the
Panel
raised
no
issues
that
represent
major
stumbling
blocks
in
the
path
towards
a
final
risk
assessment.
Nonetheless,
a
number
of
points
were
felt
to
deserve
serious
consideration
or
reconsideration.
These
points
are
elaborated
in
later
sections
of
this
report
but
the
most
significant
of
them
are
highlighted
here
under
the
headings:
Hazard,
Water,
Food,
Residential,
and
Integration.

HAZARD
EPA's
choice
of
oxamyl
as
the
index
chemical
provoked
extensive
discussion.
The
Panel
recognized
that
the
diversity
of
chemical
structure,
toxicity,
and
metabolic
half­
lives
among
the
N­
methyl
carbamates
makes
it
nearly
impossible
to
identify
one
single
compound
that
might
be
considered
truly
representative
of
the
group.
Many
of
the
Panel
members
agreed
with
the
selection
of
oxamyl
as
the
index
chemical.
However,
several
reasons
for
choosing
carbaryl
over
oxamyl
were
articulated.
These
included
carbaryl's
wider
use,
including
residential
applications,
the
availability
of
metabolic
studies,
and
the
progress
toward
a
PBPK
model
for
this
compound
that
will
inform
future
risk
assessments.
Because
the
Panel
did
not
reach
a
true
consensus
on
the
issue,
the
EPA
is
advised
to
consider
carefully
the
discussion
summarized
later
in
this
report,
and
thoroughly
reconsider
its
choice
of
index
chemical.

WATER
The
Panel
was
in
favor
of
the
Agency's
plan
to
account
for
variable
rates
of
pesticide
degradation
at
different
soil
depths
and
with
setback
distances
from
field
to
well.
As
for
comparisons
of
the
three
models
under
consideration
as
predictors
of
pesticide
residues
in
drinking
water,
the
Panel
favored
evaluating
models
in
light
of
the
mass
balances
of
water
and
pesticide
and
to
consider
the
hydrology
carefully,
especially
the
effect
of
wells
used
for
irrigation.
The
Panel
also
urged
the
Agency
to
consider
scenarios
in
which
a
pesticide
application
is
quickly
followed
by
rainfall
that
leads
to
preferential
flow
to
shallow
groundwater.
Suggestions
were
offered
on
several
other
points,
including
the
potential
use
of
informed
individuals
to
identify
small
local
areas
of
high
use
and
great
leaching
potential
that
might
lead
to
elevated
risks
of
exposure.

FOOD
Panel
recommendations
included
the
following.
The
Agency
is
urged
to
make
a
more
detailed
analysis
of
food
exposure
and
identify
specific
food­
pesticide
combinations
that
are
major
contributors
to
risk.
A
sensitivity
analysis
should
be
conducted
to
determine
how
the
choice
of
assumed
values
for
"
non­
detects"
affects
the
estimated
exposure.
The
Agency
should
evaluate
the
Carbamate
Market
Basket
Residue
Monitoring
Study
and
its
implications
for
cumulative
risk
assessment
(
particularly
with
respect
to
single
item
vs.
composite
samples).
The
Agency
also
should
investigate
the
effect
of
seasonal
residues
and
consumption
patterns
on
the
cumulative
assessment,
especially
with
10
of
63
regard
to
individuals
whose
diet
is
heavily
weighted
towards
certain
food
sources.

RESIDENTIAL
Several
comments
were
offered
on
limitations
in
the
REJV
database,
on
ways
to
supplement
or
improve
this
database,
and
on
ways
to
use
a
larger
fraction
of
the
data
in
it.
Panel
statisticians
were
united
in
opposing
the
creation
and
use
of
uniform
distributions
and
to
data
truncation
except
in
extraordinary
cases,
as
when
physical
factors
constrain
the
range
of
possible
values.
Another
issue
that
received
much
discussion
was
the
uncertainty
associated
with
assessment
of
exposure
from
hand­
to­
mouth
behavior
in
small
children.
The
Panel
agreed
that
the
macro­
activity
approach
in
the
current
document
will
overestimate
exposure,
but
there
was
no
consensus
on
the
proper
solution
to
this
problem.
At
least
one
Panel
member
argued
for
deleting
this
component
from
the
assessment.
Reasons
given
for
such
a
step
were
that
overall
exposure
from
food
and
water
is
more
important,
that
the
dermal
route
already
accounts
for
residential
exposure
in
part,
and
that
the
hand­
to­
mouth
data
are
more
likely
to
propagate
uncertainty
than
reduce
it.
This
argument
should
not
be
dismissed
"
out
of
hand".
In
any
case
the
Agency
can
consider
the
Panel's
constructive
suggestions
for
mitigating
the
problems
involved
in
properly
assessing
children's
exposure
risks
from
mouthing
behavior
(
see
detailed
response
to
R2).

INTEGRATION
With
regard
to
this
topic
the
Panel
again
gave
the
Agency
high
marks
for
the
great
improvement
in
its
latest
cumulative
risk
assessment
document.
Two
integrative
aspects
inspired
extensive
thought
and
discussion
at
the
present
meeting.
The
first
of
these
was
the
question
as
to
whether
BMD10
values
should
be
based
exclusively
on
estimated
peak
levels
of
AChE
inhibition.
The
Panel
considered
an
alternative
strategy
to
accommodate
the
possibility
that
duration
of
action
also
is
important,
particularly
in
regard
to
developmental
toxicity.
An
approach
recommended
for
further
consideration
would
multiply
the
standard
relative
potency
factor
(
RPF)
for
each
carbamate
by
the
associated
half­
time
for
recovery
of
brain
AChE
with
that
agent.
The
effect
would
be
to
increase
the
RPF
values
for
compounds
with
relatively
slow
metabolic
clearance.

While
considering
the
recovery
half­
life
for
inhibition
of
brain
AChE
in
rodents,
the
Panel
also
recognized
a
problem
that
may
arise
in
extrapolating
to
humans.
When
standard
inter­
species
scaling
factors
are
applied
to
some
compounds,
the
resulting
halflives
may
violate
one
of
the
basic
assumptions
of
the
cumulative
risk
assessment
now
envisaged
by
the
Agency.
That
is
to
say,
cholinesterase
inhibition
by
the
N­
methyl
carbamates
in
humans
may
reverse
slowly
enough
to
cumulate
from
one
day
to
the
next.

Inhibition
half­
lives
also
were
considered
in
a
final
Panel
discussion
focusing
on
the
timing
of
water
consumption
events.
The
present
practice
of
lumping
such
events
into
a
single
occasion
appears
to
be
conservative.
When
water
consumption
is
distributed
across
the
day,
and
half­
lives
for
reversal
of
inhibition
are
taken
into
account,
peak
inhibition
of
11
of
63
brain
AChE
is
predicted
to
be
substantially
smaller
than
currently
estimated.
Since
exposure
to
water­
borne
pesticide
is
the
major
contributor
to
risk
in
certain
geographical
settings,
the
regulatory
impact
of
a
decision
to
distribute
or
lump
consumption
could
be
significant.

PANEL
DELIBERATIONS
AND
RESPONSE
TO
CHARGE
The
specific
issues
addressed
by
the
Panel
are
keyed
to
the
Agency's
background
documents,
references,
and
the
Agency's
charge
questions.

Questions
HAZARD
EPA's
hazard
and
dose­
response
chapter
(
I.
B)
and
associated
appendices
(
II.
B.
1­
6)
of
the
Preliminary
Cumulative
Risk
Assessment
describe
the
application
of
the
Relative
Potency
Factor
(
RPF)
method
to
the
N­
methyl
carbamate
pesticides.
These
documents
a)
outline
the
steps
in
developing
the
dose­
response
relationships
for
each
pesticide
and
its
capacity
to
inhibit
AChE
in
rats;
b)
describe
the
data
used
in
the
assessment;
c)
summarize
the
empirical
dose­
response
modeling
which
provides
the
basis
for
the
relative
potency
factors
(
RPFs),
points
of
departure
(
PoDs),
and
estimates
of
AChE
inhibition
half
life;
and
d)
provide
the
rationale
for
selecting
oxamyl
as
the
index
chemical.

HAZARD
QUESTION
#
1
Empirical
Dose­
Response
and
Time
Course
Modeling
At
the
February,
2005
meeting
of
the
FIFRA
SAP,
EPA
proposed
an
empirical
model
for
use
in
the
cumulative
risk
assessment
of
the
N­
methyl
carbamates.
This
model
contains
a
dose­
response
and
a
time
to
recovery
component.
Based
on
the
comments
from
the
Panel
and
following
experience
with
its
application
EPA
made
some
modifications
to
this
proposed
model.
EPA
has
applied
this
revised
empirical
model
to
the
available
RBC
and
brain
cholinesterase
data
for
the
N­
methyl
carbamates.
BMD
and
BMDL
estimates
provided
in
the
preliminary
assessment
were
derived
from
cholinesterase
data
from
multiple
studies
and
in
some
cases,
using
different
cholinesterase
measurement
techniques.

H1a.
Please
comment
on
the
mathematical/
statistical
approach
to
modeling
cholinesterase
data
used
to
estimate
benchmark
dose
values
and
time
to
halflife
recovery
in
the
preliminary
cumulative
risk
assessment.
Please
address
biological
and
mathematical/
statistical
considerations
in
your
response.

Response
12
of
63
The
Panel
was
in
consensus
that
EPA
has
used
best
available
statistical
methodology
to
fit
the
proposed
empirical
model
for
cholinesterase
concentration
and
time
course
data.
It
was
also
agreed
that
the
presented
analysis
demonstrated
implementation
of
the
comments
suggested
in
the
February
2005
SAP.

There
was
significant
discussion
about
whether
the
model­
estimated
BDM10
(
or
BMDL10)
values
are
the
appropriate
values
to
be
used
in
the
determination
of
Relative
Potency
Factors
(
RPFs).
A
proposal
that
the
RPFs
be
modified
to
take
into
account
the
apparently
substantial
differences
in
estimated
half­
lives
for
reversal
of
brain
AChE
inhibition
among
the
carbamates
was
discussed
several
times
in
the
Panel's
three­
day
deliberations.
At
this
initial
stage
of
discussion
Panel
members
generally
agreed
that
the
peak
inhibition
level
is
likely
to
be
the
most
relevant
measure
of
internal
dose
for
producing
gross
acute
toxic
effects.
It
was
understood
that
the
current
model
aggregates
all
of
the
exposures
projected
for
an
individual
in
the
course
of
a
day
by
summing
the
RPF
converted
concentrations.
There
was
concern
that
aggregating
chemicals
with
very
different
recovery
half­
lives
may
not
properly
capture
true
exposure.
Final
Panel
recommendations
on
this
topic
are
included
in
the
response
to
question
I1.

A
number
of
other
observations
relating
to
the
mathematical/
statistical
aspects
of
the
exposure
model
were
discussed
and
are
summarized
below.

 
There
is
uncertainty
associated
with
the
model
building
process
and
decisions
made
regarding
model
parameterization.
The
model
used
is
non­
linear
and
the
available
data
are
often
inadequate
to
estimate
all
model
parameters
or
the
available
data
do
not
support
the
more
complex
model
forms.
In
a
number
of
situations
the
model
is
simplified
by
either
simplifying
the
parameterization
or
by
specifying
some
parameters
as
known
constants.
While
these
changes
are
documented,
the
impacts
of
these
changes
on
the
final
estimate
of
the
BMD10
or
BMDL10
are
not
discussed.
It
was
recognized
that
this
may
be
part
of
the
sensitivity
analysis
performed
on
the
final
estimating
models.
 
The
estimating
model
incorporates
random
effect
terms
to
account
for
study­
tostudy
and
animal­
to­
animal
differences
in
background
cholinesterase
levels.
Histograms
of
the
estimated
random
effects
(
the
Best
Linear
Unbiased
Predictors)
would
help
demonstrate
the
appropriateness
of
distributional
assumptions
for
the
random
effects
as
well
as
help
identify
influential
("
outlier")
individuals
or
studies.
 
The
initial
fit
of
model
parameters
was
via
a
graphical
technique
followed
by
a
few
maximum
likelihood
iterations.
Given
the
expected
correlation
among
model
parameters,
minor
changes
in
initial
estimates
can
result
in
significantly
different
final
parameter
estimates.
Assurances
need
to
be
provided
that
the
estimates
presented
are
the
maximum
likelihood
values
or
are
very
close
to
them.
 
At
least
one
Panel
member
suggested
that
in
addition
to
helping
implement
the
constraints
on
model
parameters,
the
transformation
of
the
parameters
to
log
scale
13
of
63
is
likely
to
reduce
the
correlation
among
parameters
and
make
for
a
model
that
is
easier
to
fit.
 
EPA
chose
BMD10,
which
is
the
central
tendency
estimate
of
the
benchmark
dose
at
which
there
is
a
10
percent
response
level,
as
a
reference
point
for
developing
Relative
Potency
Factors
(
RPFs).
The
lower
limit
is
not
clearly
defined
but
is
presumably
the
2.5th
percentile
of
the
confidence
interval.
This
lower
limit
is
used
as
the
point
of
departure
(
PoD)
for
extrapolating
risk.
Most
studies
are
reported
as
able
to
detect
a
benchmark
response
(
BMR)
of
10%
that
is
significantly
different
from
zero.
There
was
a
question
about
what
other
measures
of
effect
can
be
examined
by
the
model
(
e.
g.,
a
BMD20
or
peak
concentration)
and
how
such
a
choice
would
change
the
RPFs.
 
The
measurements
of
cholinesterase
inhibition
may
tend
to
underestimate
the
actual
inhibition,
because
recovery
can
occur
during
the
dilution
and
prolonged
incubation
of
various
methods.
Reducing
dilution,
shortening
the
incubation
time,
and
lowering
the
assay
temperature
are
known
to
limit
the
decarbamylation
of
inhibited
enzyme.
These
modifications
are
not
yet
universal,
although
EPA
scientists
have
published
them
in
the
open
literature
(
Nostrandt
et
al.,
1993).
At
the
previous
meeting,
EPA
reported
that
their
inhibition
measurements
with
the
modified
Ellman
method
and
with
the
radiometric
method
generally
agreed
with
registrant
data
on
the
N­
methyl
carbamates.
The
Panel
encourages
EPA
to
publish
these
results
to
reduce
the
remaining
uncertainty
on
this
issue.
 
The
Panel
discussed
whether
the
available
dose­
response
data
actually
capture
the
peak
cholinesterase
inhibition.
While
researchers
target
the
peak,
it
is
typically
not
known
if
the
first
measurement
is
taken
at
the
peak
or
shortly
after.
The
question
is
how
much
error
is
introduced
into
estimates
of
BMD10
and
recovery
rates
when
a
dose­
response
model
is
fitted
to
data
obtained
after
the
moment
of
peak
inhibition.
 
The
main
focus
of
the
modeling
was
on
the
oral
exposure
route.
To
the
extent
that
data
are
lacking
for
the
other
routes
of
exposure,
one
Panel
member
wondered
how
one
might
characterize
at
least
semi­
quantitatively
the
dose­
response
and
time­
response
relationships.
Another
Panel
member
suggested
that
some
sort
of
Bayesian
analysis,
using
prior
distributions
developed
via
information
gleaned
from
experts,
might
allow
extension
of
results
to
other
groups
such
as
the
very
young
or
the
elderly
for
which
appropriate
data
are
not
available.
This
issue
is
partly
revisited
in
the
Panel's
response
to
question
I1.

H1b.
Please
comment
on
the
adequacy,
clarity,
and
transparency
of
the
documentation
provided
for
the
empirical
dose­
response
and
time
course
modeling.

Response
In
general
the
Panel
found
the
documentation
of
the
model
a
little
too
terse
and
overly
dependent
on
a
reader's
previous
experience
with
the
model
and
associated
topics.
14
of
63
The
recommendations
listed
below
summarize
the
Panel
discussion
and
should
help
clarify
model
components
and
improve
transparency.

 
The
report
only
presents
the
model
without
justifying
or
supporting
its
choice.
 
Key
assumptions
of
the
model
should
be
listed
along
with
arguments
that
support
them.
In
particular,
error
distribution
assumptions
should
be
documented.
 
Dr.
Setzer's
slide
presentation
was
better
organized
and
more
useful
in
many
ways
than
the
written
report.
The
written
presentation
of
the
full
model
would
benefit
from
a
reversed
order,
beginning
with
an
overview
of
the
full
inhibition
model
(
Eq.
5)
and
then
proceeding
to
particular
components
in
more
detail.
This
was
the
order
of
the
oral
presentation,
which
the
Panel
found
easier
to
grasp.
 
More
graphics
would
help
to
demonstrate
the
forms
possible
for
the
model
components.
For
example,
to
illustrate
the
non­
linearity
of
the
dose­
response
model
for
inhibition
as
a
function
of
 ,
the
Panel
recommends
Figure
1
(
see
below),
similar
to
one
of
the
graphs
shown
by
EPA.
 
Inhibition
as
a
function
of
time
as
given
by
Equation
2
could
be
usefully
illustrated
as
in
Figure
2.
This
figure
also
demonstrates
that
inhibition
reaches
a
peak
at
time
T*,
representing
a
competition
between
an
exponential
increase
in
inhibition
after
dosing
and
an
exponential
decrease
in
inhibition
associated
with
recovery.
 
A
detailed
explanation
and
justification
for
the
full
inhibition
model
should
be
included
in
the
report.
 
The
simplified
model
(
Eq.
4)
could
be
illustrated
with
the
graph
given
in
Figure
3.
Use
of
the
(
t­
 )
term
presumes
that
a
unit
response
occurs
at
time
t
=
 .
This
model
seems
to
presume
that
cholinesterase
inhibition
occurs
instantaneously,
followed
by
rapid
recovery.
 
In
comparing
the
two
models
it
can
be
shown
that
if
TA
(
half­
life
for
the
process
of
inhibition)
is
very
short
compared
to
TR
(
half
life
for
the
process
of
recovery),
the
full
model
and
the
simplified
model
give
essentially
equivalent
results.
This
can
be
illustrated
with
Figure
4.
 
The
document
needs
to
indicate
clearly
which
parameters
have
been
estimated
and
which
have
been
held
constant.
It
should
also
show
whether
or
not
the
values
applied
for
constants
in
cases
where
a
parameter
cannot
be
estimated
are
at
least
somewhat
similar
to
the
values
obtained
when
the
parameter
can
be
estimated.
 
Documentation
of
R
scripts
is
in
the
help
files
but
absent
from
the
scripts
themselves.
One
Panel
member
suggested
that
this
documentation
should
also
be
included
as
comments
in
the
scripts
themselves.
 
It
is
suggested
that,
when
examples
are
given
in
the
document,
the
background
and
justification
for
the
example
should
also
be
provided.
 
Although
the
Panel
recognizes
that,
empirically,
there
may
be
a
level
below
which
cholinesterase
activity
cannot
drop,
regardless
of
pesticide
dose,
some
comment
is
needed
to
justify
incorporating
this
feature
into
the
model
(
p
33).
15
of
63
 
The
second
bullet
on
page
33
could
be
written
in
a
less
confusing
manner
to
explain
how
the
model
can
account
for
the
possibility
that
data
were
not
collected
at
the
time
of
peak
inhibition.
 
Clarification
should
be
given
as
to
whether
the
distributions
described
on
pages
36
and
later
are
intended
to
represent
uncertainty
or
inter­
individual
variability.
For
example,
when
there
is
more
than
one
study,
for
each
ID
there
is
assumed
to
be
a
specific,
DR
(
or
ln(
DR)),
and
that
these
values
are
assumed
to
follow
a
normal
distribution.
The
text
indicates
that
this
distribution
may
be
different
for
the
different
sexes.
Is
this
true
and
what
are
the
implications
for
the
model?
 
The
modeling
steps
are
summarized
on
pages
37­
38,
but
quantitative
examples
of
how
these
steps
are
implemented
would
really
help
the
reader.
 
More
discussion
is
needed
on
the
potential
dependence
of
recovery
half­
life
on
dose.
The
implications
of
this
dependence
for
the
final
use
of
the
model
should
be
discussed.
 
During
the
public
comment
period,
a
representative
of
the
American
Bird
Conservancy
raised
the
concern
that
half­
lives
associated
with
dermal
exposure
could
be
much
longer
than
those
for
the
oral
route
(
Table
I.
B.
6).
EPA
should
consider
this
comment
and
possibly
revise
the
dermal
exposure
assessment
appropriately.
 
There
is
some
confusion
as
to
when
female
data
are
used
in
the
model
and
when
not.
Specifically,
what
is
done
when
there
is
a
significant
gender­
specific
sensitivity,
and
how
does
this
decision
affect
the
overall
model
uncertainty?
 
An
argument
is
made
that
BMD10
values
for
brain
AChE
are
more
suitable
health
endpoints
than
RBC­
based
BMD10'
s,
(
Figure
I.
B.
3
 
the
label
for
this
figure
should
identify
the
oral
route
of
exposure).
For
many
of
the
lower
potency
chemicals,
including
thiodicarb
and
carbaryl,
it
does
appear
that
brain
AChE
data
provide
endpoints
as
conservative
or
more
conservative
than
RBC
data.
For
the
potent
agents
aldicarb
and
carbofuran,
however,
the
BMD10
based
on
RBC
enzyme
is
lower
than
one
based
on
brain
AChE.
If
that
is
truly
the
case,
what
does
it
say
about
the
overall
conservativeness
of
the
process?
 
In
presenting
a
series
of
equations
it
would
be
helpful
to
number
them
all,
to
define
the
x
and
y
side
of
each
one,
to
specify
units,
to
include
a
nomenclature
box,
and
to
minimize
confusion
by
using
brackets.
 
When
considering
alternatives,
it
would
be
useful
to
specify
the
criteria
for
a
good,
adequate,
or
acceptable
model,
specify
what
is
expected
from
each
model,
and
specify
how
the
performance
of
each
model
will
be
assessed.
 
The
main
document
should
include
better
references
to
and
descriptions
of
the
material
in
the
appendices.
 
A
flow
diagram
is
needed
to
show
how
the
final
models
are
derived.
An
example
could
be
given,
followed
by
a
reference
to
the
appropriate
appendix
(
II.
B.
2).
There
are
many
files
in
this
appendix
but
they
appear
to
follow
a
common
methodology.
Some
explanation
of
the
purpose
of
each
step
in
the
analysis
would
16
of
63
help.
Also,
there
should
be
a
summary
of
the
results
for
each
of
these
cases,
so
that
a
reader
can
reproduce
the
final
dose­
response
and
time
response
equations.
 
Graphical
or
other
appropriate
quantitative
information
is
needed
to
support
the
assertion
that
the
"
nonstandard"
modified
Ellman
method
is
reliable
(
p.
30).
 
On
pages
32­
33
several
key
assumptions
for
the
empirical
dose­
response
and
time
course
models
are
listed.
This
is
very
helpful.
However,
to
be
fully
convincing,
empirical
observations
that
support
these
key
assumptions
should
be
provided
to
show
that
they
were
derived
based
on
interpretation
of
data.
 
The
documentation
of
the
dose­
response
model
for
inhibition
could
be
more
clearly
explained,
as
for
example,
by
defining
the
response
g(
d).
 
Units
should
be
given
for
all
variables.
 
Equation
1
lacks
clarity.
For
example,
is
the
log
in
the
exponent
a
log
10?
Does
this
log
take
as
its
argument
both
terms,
including
the
one
raised
to
the
 
power?
Probably
not.
This
could
be
made
clearer
with
an
appropriate
equation
editor.
 
It
also
would
help
to
have
an
explanation
up
front
as
to
what
general
modeling
approaches
were
used
with
respect
to
pesticide
and
exposure
route.
For
example,
after
slogging
through
Equations
1­
5,
the
reader
is
told
that
other
modeling
approaches
were
used.
Each
and
every
model
actually
used
should
be
documented,
not
just
alluded
to.
A
sensitivity
analysis
of
each
generic
type
of
model
should
be
provided
to
illustrate
(
and
explain/
justify)
the
key
behaviors
of
these
models.
 
The
final
parameterized
version
of
the
model
should
be
given
explicitly,
e.
g.,
on
page
36.
 
Discussion
of
the
biological
aspects
of
inhibition
and
recovery
would
be
very
helpful
 
to
establish,
perhaps,
the
biological
plausibility
of
the
models
used.
 
Clear
discussion
of
data
quality
objectives
would
help
 
for
example,
what
constitutes
a
"
good
model"
in
terms
of
precision
or
accuracy
relative
to
available
data,
and
are
the
models
intended
to
be
somewhat
qualitative
in
simply
describing
trends
rather
than
providing
exact
estimates?
 
A
footnote
on
table
I.
B.
6
would
help
the
reader
understand
what
the
values
in
parentheses
are
for
some
of
the
chemicals.
 
The
point
of
Figure
I.
B.
3
would
be
better
illustrated
if
RBC
and
brain
cholinesterase
BMD10'
s
were
shown
separately
for
each
chemical,
side
by
side,
rather
than
in
the
overlapping
graph
presented.
 
A
user's
guide
to
DRutils
should
be
provided.
 
The
issue
of
what
happens
to
model
estimates
when
data
are
collected
after
the
peak
should
be
addressed
explicitly.
It
is
worth
trying
to
detect
differences
in
slopes
calculated
with
and
without
the
first
two
points.
One
must
determine
the
sensitivity
of
bias
in
this
estimation
and
also
deal
with
the
issue
that
the
third
data
point
may
not
be
well
estimated.
 
There
should
be
a
discussion
of
how
often
the
gamma
parameter
(
 )
does
differ
statistically
from
1.
17
of
63
 
It
should
be
pointed
out
that
the
present
model
is
rooted
in
the
mechanistic
construction
of
previous
models
for
the
OP
anticholinesterases.

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
10
20
30
40
50
Dose
(
mg/
kg)
Max.
Fraction
Inhibition
g
=
1,
P
=
0.1
g
=
1,
P
=
0.4
g
=
2,
P
=
0.1
g
=
2,
P
=
0.4
Figure
1
Sensitivity
of
the
dose
response
model
equation
to
different
levels
of
 
for
DR=
3
and
R=
0.1.
18
of
63
0
0.2
0.4
0.6
0.8
1
1.2
0
0.5
1
1.5
2
2.5
time,
t
[
units?]
ChE
Inhibition
TA
=
0.1
TA
=
0.2
TA
=
0.5
TA
=
0.9
Co
=
f(
TA,
TR),
TR
=
1
Figure
2
Impact
of
different
Ta
values
on
the
time
pattern
of
response.

0
0.2
0.4
0.6
0.8
1
1.2
0
0.5
1
1.5
2
2.5
time,
t
[
units?]
ChE
Inhibition
TR
=
0.1
TR
=
0.2
TR
=
0.5
TR
=
1.0
First
ChE
Measurement
Assumed
at
t
=
0.5
Figure
3
Impact
of
different
Tr
values
on
the
time
pattern
of
response
using
the
simplified
recovery
model.
19
of
63
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0
0.5
1
1.5
2
2.5
time,
t
[
units?]
ChE
Inhibition
Model
1:
TA
=
0.1,
TR
=
1
Model
2:
TR
=
1.0
Figure
4
Comparison
of
the
full
model
and
simplified
model
for
recovery
over
time.

HAZARD
QUESTION
#
2
Selection
of
the
Index
Chemical
EPA's
cumulative
risk
assessment
guidance
indicates
that
the
index
chemical
should
be
selected
based
on
the
availability
of
a
high
quality
toxicity
database
for
the
common
mechanism
endpoint.
The
selection
of
the
index
chemical
is
an
important
step
in
the
cumulative
risk
assessment;
the
BMD
for
oxamyl
was
used
to
calculate
RPFs
and
the
BMDL
for
oxamyl
was
used
as
the
PoD
for
extrapolating
cumulative
risk.

H2.
Please
comment
on
the
rationale
provided
for
the
selection
of
the
index
chemical.
Should
any
additional
factors
be
included
in
the
rationale
for
the
selection
of
oxamyl
as
the
index
chemical?

Response
The
preliminary
N­
methyl
carbamate
cumulative
risk
assessment
designates
oxamyl
as
the
index
chemical
for
calculating
Relative
Potency
Factors
(
RPFs)
for
the
other
nine
pesticides
and
for
pesticide
mixtures
using
an
additive
approach.
In
principle
any
of
the
ten
pesticides
could
be
selected
as
the
index
chemical,
but
the
EPA
proposes
that
the
optimal
choice
is
a
pesticide
with
robust
experimental
data
on
an
endpoint
of
interest
(
in
this
case
brain
AChE
activity)
from
studies
with
all
relevant
routes
of
exposure.
20
of
63
The
Panel
recognized
the
strengths
of
the
oxamyl
database
and
generally
agreed
with
the
rationale
for
its
selection
as
the
index
chemical.
Oxamyl
is
one
of
only
two
chemicals
for
which
the
database
includes
studies
with
all
three
major
routes
of
exposure:
oral,
dermal,
and
inhalational.
Nonetheless,
several
Panel
members
argued
that
carbaryl
should
be
considered
for
reasons
addressed
below.

The
oral
exposure
database
for
oxamyl
comprises
three
acute
rat
dosing
registration
studies
and
the
EPA
NHEERL
rat
dose­
response
and
time
course
studies.
The
doses
used
covered
a
wide
range,
and
whole
or
half
brain
ChE
data
were
available
from
all
four
studies.
Recovery
data
also
are
available
although
it
is
not
stated
how
many
studies
included
this
endpoint.
Calculated
BMD10
values
for
brain
ChE
activity
differed
statistically
between
the
sexes,
but
the
differences
were
not
biologically
important,
and
the
95%
confidence
intervals
were
small.
The
relative
potency
of
oral
oxamyl
for
blood
and
brain
ChE
inhibition
is
fairly
high
among
the
10
pesticides
(
Fig.
1.
B.
3)
a
possible
negative
for
its
selection
if
this
introduces
a
systematic
bias
when
used
to
estimate
RPFs
for
pesticides
with
markedly
different
pharmacokinetics.
Adding
to
this
concern
was
that
the
acute
oral
dose
response
curve
for
brain
ChE
inhibition
for
oxamyl
did
not
fully
parallel
those
of
the
other
pesticides
in
the
NHERRL
data.
With
regard
to
the
chemistry
of
the
pesticides,
six
of
the
compounds
have
an
aromatic
or
heterocyclic
ring
whereas
four
do
not,
oxamyl
among
them.
Thus,
although
hardly
an
outlier,
oxamyl
differs
chemically
from
60%
of
the
compounds
in
the
common
mechanism
group.
In
contrast,
and
a
potential
plus,
the
water
solubility
of
oxamyl
is
intermediate
for
the
set
of
common
mechanism
compounds.
Oxamyl
uses
are
limited
to
agricultural
applications,
and
human
exposure
is
projected
through
food
or
drinking
water.
Oxamyl
has
no
residential
use.

For
carbaryl,
five
oral
administration
studies
are
included
in
the
database.
Recovery
half­
life
estimates
for
brain
ChE
after
oral
dosing
with
carbaryl
were
somewhat
slower
than
for
oxamyl,
and
were
shown
to
increase
with
dose.
The
availability
of
information
about
the
dose
dependency
of
recovery
with
carbaryl
contrasts
favorably
with
the
single
dose
recovery
data
for
oxamyl.
However
carbaryl's
relatively
low
toxicity
was
considered
a
drawback
for
an
index
chemical
in
a
group
of
pesticides
that
are
generally
one
or
two
orders
of
magnitude
more
toxic.
The
low
toxicity
suggests
that
detoxication
and
clearance
factors
are
more
important
for
carbaryl
than
for
the
other
pesticides.
Therefore
using
carbaryl
as
the
index
chemical
might
distort
the
assessment
of
the
more
toxic
compounds
whose
detoxication
and
clearance
are
likely
less
important.

Two
dermal
studies
were
available
for
oxamyl,
both
in
the
rabbit.
The
BMD10
values
were,
as
expected,
much
higher
than
those
from
the
oral
studies,
probably
owing
to
kinetic
and/
or
species
differences.
Unfortunately
the
studies
on
rabbits
used
only
dermal
exposure,
so
that
we
cannot
assess
species
differences
in
response
by
comparable
routes,
or
compare
dermal
exposure
with
oral
exposure
in
a
single
species.
This
deficiency
weakens
the
case
for
oxamyl
as
the
index
chemical.
For
carbaryl,
on
the
other
hand,
the
one
dermal
dosing
study
was
performed
in
the
rat,
allowing
direct
comparison
with
oral
gavage
studies
in
that
animal.
The
document
states
that
the
resulting
data
were
sufficient
21
of
63
to
calculate
BMD10
values
for
RBC
and
brain
ChE
inhibition.

One
study
of
acute
inhalation
exposure
is
available
for
oxamyl
and
propoxur
(
appendix
II.
B.
2).
None
exists
for
carbaryl,
which
is
unfortunate
since
this
agent
has
residential
uses
that
may
generate
exposures
by
that
route.
A
PBPK
model
under
development
might
eventually
help
to
estimate
BMD10
for
inhalational
exposure
to
carbaryl.
Meanwhile,
only
oxamyl
and
propoxur
have
been
studied
by
all
routes
and
the
latter
suffers
from
an
unusually
large
confidence
interval
in
the
BMD10
data
for
brain
ChE
in
acute
oral
studies.

In
light
of
all
available
data,
the
Panel
generally
agreed
it
is
appropriate
to
use
oxamyl
as
the
index
chemical.
If
the
Agency
goes
forward
with
this
choice,
however,
there
are
a
number
of
underlying
assumptions
that
should
be
pointed
out,
succinctly
and
explicitly,
as
an
aid
to
understanding.

All
ten
pesticides
in
this
class
act
on
ChE
by
carbamylating
its
active
site
serine
and
modifying
the
enzyme
in
exactly
the
same
way.
Therefore,
the
intrinsic
recovery
rate
should
be
identical
for
each
agent,
quite
independent
of
the
leaving
group.
Differences
among
recovery
rates
experimentally
observed
in
vivo
must
reflect
differences
in
the
persistence
of
residual
unreacted
pesticide,
available
to
inhibit
the
enzyme.
To
approach
this
issue
mathematically
requires
knowing
the
time
course
of
the
pesticide
levels.
Some
of
the
necessary
information
may
be
available
in
the
absorption,
distribution,
metabolism,
and
excretion
(
ADME)
data
developed
for
these
agents.
Alternatively,
rough
estimates
of
the
residual
active
pesticide
may
be
obtained
from
the
degree
to
which
enzyme
recovery
is
delayed
when
compared
to
the
recovery
from
oxamyl,
the
agent
with
the
shortest
recovery
half­
life
in
vivo.
It
should
be
remembered
that
the
recovery
rates
reflect
an
integration
of
the
parent
pesticide
elimination
kinetics
and
the
constant
rate
for
reversal
of
enzyme
carbamylation.
This
is
why
"
recovery
half­
life"
has
a
complex
meaning.
None
of
these
considerations,
however,
suggest
that
the
recovery
information
has
been
incorrectly
incorporated
into
the
current
assessment.

The
calculated
BMDs
for
brain
ChE
inhibition
are
based
on
applied
doses.
While
this
approach
is
fine
for
the
purposes
of
this
risk
assessment,
it
may
again
assist
in
understanding
if
the
BMDs
derived
by
the
three
routes
of
administration
are
put
into
perspective.
The
BMDs
for
oxamyl
are
essentially
the
same
for
the
oral
and
inhalation
routes
of
exposure,
but
are
much
higher
for
the
dermal
route.
This
is
what
would
be
expected
if
acute
oral
exposure
and
inhalation
exposures
both
resulted
in
rapid
uptake
and
distribution
to
the
brain.
The
much
higher
BMD
for
dermal
studies
probably
reflects
the
slower
kinetics
of
absorption,
although
the
species
problem
cited
above
makes
it
impossible
to
say
this
with
certainty
based
on
these
data.
These
arguments
clarify
why
it
is
best
to
have
information
from
all
three
routes
of
exposure
when
selecting
an
index
chemical,
rather
than
relying
on
a
possibly
superior
dataset
in
studies
of
oral
exposure
alone,
as
with
carbaryl.
These
considerations
also
support
the
use
of
the
data
from
oral
studies
for
pesticides
for
which
dermal
or
inhalation
data
are
absent,
as
an
adequately
protective,
or
22
of
63
conservative
approach
to
the
cumulative
risk
assessment.
In
this
regard,
it
is
better
to
use
the
oral
RPFs,
rather
than
the
BMD10
(
as
done
for
Table
1.
B.
8)
to
derive
RPFs
for
other
routes
of
exposure
on
which
we
lack
data.

Other
arguments
were
put
forth
and
supported
by
some
of
the
Panel,
in
favor
of
selecting
carbaryl
as
the
index
chemical.
These
points
follow.

The
cumulative
risk
assessment
guidance
document
(
USEPA,
2002a)
states
that
the
criteria
for
an
index
chemical
should
be
high
quality
dose­
response
data
on
the
common
mechanism
endpoint,
preferably
with
each
exposure
route,
and
a
toxicology
that
resembles
other
agents
in
the
common
mechanism
group.
Even
so,
the
selection
of
the
index
chemical
should
consider
real­
world
uses.
It
is
understood
that
use
of
an
index
chemical
with
imperfect
toxicology
data
may
introduce
error
and
uncertainty
into
the
estimation
of
cumulative
risk.
However,
the
PoD
for
the
index
chemical
will
be
used
to
extrapolate
risk
to
exposure
levels
anticipated
in
the
human
population.
Therefore
the
selection
guidelines
for
the
index
chemical
should
take
into
account
both
the
potency
of
the
agent
and
the
range
of
uses
for
which
it
is
registered.

Oxamyl
has
very
limited
use
in
agriculture,
and
has
no
residential
applications.
Regardless
of
their
quality,
the
toxicological
data
from
dermal
and
inhalational
studies
of
oxamyl,
and
to
a
lesser
extent
the
oral
studies,
indicate
that
this
compound
contributes
little
to
the
cumulative
carbamate
risk
estimates
in
the
human
population.
Almost
all
the
NMC
cumulative
risk
would
result
from
exposures
to
pesticides
other
than
oxamyl,
over
dose
ranges
such
that
the
responses,
either
brain
or
RBC
ChE
effects,
would
need
to
be
extrapolated
from
the
oxamyl
dose­
response
model.
Thus,
greater
error
and
uncertainty
may
actually
be
introduced
into
the
cumulative
risk
estimates
if
oxamyl
is
chosen
as
the
index
chemical.

The
wide
use
of
carbaryl,
compared
to
other
NMC
members,
both
in
agricultural
and
residential
environments
should
make
carbaryl
a
favorable
candidate
for
the
index
chemical.
Because
of
its
wider
use,
selecting
carbaryl
as
the
index
chemical
would
require
less
data
conversion
in
the
cumulative
carbamate
exposure
assessment,
which
should
minimize
the
inherent
error
and
uncertainty
associated
with
the
cumulative
risk
assessment
involving
relative
potency
factors.
In
addition,
the
development
of
a
PBPK/
PBPD
model
for
carbaryl
has
been
encouraged
by
the
Agency
and
is
currently
being
undertaken.
It
was
intuitively
apparent
to
some
Panel
members
that
the
selection
of
the
index
chemical
for
the
derivation
of
relative
potency
factors
should
parallel
the
effort
of
PBPK/
PBPD
model
development.

HAZARD
QUESTION
#
3
Selection
of
Brain
ChE
data
for
developing
RPFs
and
PoDs
EPA
has
used
data
for
brain
ChE
as
the
basis
for
the
RPFs
and
PoDs.
The
23
of
63
rationale
for
this
selection
was
provided
in
I.
B.

H3.
Please
comment
on
the
rationale
provided
for
the
selection
of
the
brain
ChE
as
the
basis
for
RPFs
and
PoDs
in
the
preliminary
cumulative
risk
assessment.
Should
any
additional
factors
be
considered?

Response
The
Panel
found
a
compelling
case
for
using
brain
cholinesterase
as
the
endpoint
in
determining
relative
potency
factors.
Brain
AChE
is
abundant
and
critical
to
normal
physiologic
function.
Brain
tissue
is
readily
removed
and
homogenized
in
a
reproducible
manner.
Lastly,
inhibition
of
this
enzyme
is
not
simply
an
index
of
exposure
but
is
an
integral
portion
of
the
common
mechanism
of
toxicity
for
N­
methyl
carbamates.

The
alternatives,
to
mention
them
briefly,
would
be
to
use
1)
AChE
inhibition
in
blood
(
i.
e.,
RBCs);
2)
AChE
inhibition
in
a
peripheral
tissue
such
as
muscle,
nerve,
gut,
or
heart;
or
3)
clinical
signs
and
behavioral
disturbances.
Each
of
these
alternatives
has
serious
drawbacks.
RBC
AChE
is
difficult
to
assay
and
has
no
known
physiologic
function.
Inhibition
of
this
enzyme
is
therefore
at
best
a
surrogate
for
the
actual
mechanism
of
toxicity.
Likewise,
although
AChE
inhibition
in
peripheral
tissues
might
ultimately
provide
a
sensitive
and
direct
index
of
toxicity,
there
are
no
extensive
data
to
support
this
concept
as
yet,
and
the
accurate
dissection
and
assay
of
such
tissues
requires
care
and
skill.
Finally,
when
it
comes
to
clinical
and
behavioral
observations,
although
these
measures
are
relevant
and
qualitatively
informative,
as
endpoints
they
are
more
subjective
and,
generally
speaking,
less
sensitive
than
biochemical
determinations.

Further
support
for
focusing
on
brain
AChE
comes
from
NHEERL
data
summarized
in
the
current
document.
These
data
showed
that
BMD10
values
for
AChE
inhibition
in
the
brains
of
carbamate­
dosed
rats
were,
on
average,
as
low
as
those
derived
by
measuring
the
enzyme
in
RBCs.
Furthermore,
as
an
Agency
expert
explained
at
the
meeting,
confidence
limits
on
a
brain
BMD10
are
as
tight
as
those
on
red
cells,
despite
the
advantage
of
repeated
measures
analysis
in
the
latter
case.
This
statistical
advantage
is
probably
offset
by
the
low
activity
of
AChE
in
rat
red
cells
and
the
problems
of
accurate
quantitation
in
the
presence
of
hemoglobin.
In
any
case,
methods
now
under
development
may
further
increase
the
precision
of
brain
AChE
determinations
with
rapidly
reactivating
inhibitors
and
increase
the
utility
of
this
metric.

One
Panel
member
suggested
establishing
BMD10
values
for
inhibition
of
AChE
in
multiple
sub­
regions
of
brain,
rather
than
in
whole
brain
or
brain
hemispheres.
Such
information
could
lead
to
RPFs
and
PoDs
based
on
the
most
sensitive
target
area.
If
brain
regions
are
eventually
found
to
differ
sharply
in
vulnerability
to
pesticides,
a
regional
analysis
will
become
essential.
At
present,
however,
there
are
good
reasons
for
going
forward
with
whole
brain
AChE.
First
there
is
little
evidence
to
suggest
large
regional
variations
of
AChE
inhibition
in
brain
after
systemic
exposure
to
a
carbamate
24
of
63
anticholinesterase.
In
one
of
the
few
papers
to
address
the
issue
(
Hammond
et
al.,
1996)
brains
were
micro­
dissected
1.5
hr
after
rats
were
gavaged
with
carbaryl
(
50
mg/
kg).
The
results
showed
a
small
variation
from
the
most
sensitive
regions
(
cortex
and
striatum
at
55%
inhibition)
to
the
least
sensitive
regions
(
inferior
colliculus
and
hippocampus
at
about
40%
inhibition).
It
is
worth
noting
that
the
more
sensitive
areas
together
account
for
a
large
fraction
of
the
total
brain
AChE
(
the
total
mass
of
cortex
is
large,
and
striatum
contains
disproportionately
high
concentrations
of
the
enzyme).
Thus,
AChE
inhibition
in
whole
brain
will
not
be
drastically
lower
than
in
the
"
sentinel
regions".
A
second
reason
for
accepting
measurements
in
whole
brain
samples
is
that,
even
in
skilled
hands
and
within
a
single
laboratory,
regional
dissection
can
lead
to
data
with
much
greater
variability
than
data
from
whole
brains.
Variability
across
multiple
performance
sites
is
likely
to
be
unacceptably
high,
even
if
EPA's
time
frame
allowed
for
a
data
call­
in.
In
summary
these
considerations
speak
decisively
in
favor
of
the
Agency's
plan
to
use
existing
data
on
whole
brain
AChE
as
its
basic
metric.
The
only
qualification
to
be
added
is
an
assumption
that
the
database
is
confined
to
studies
for
which
the
Agency
has
documented
Standard
Operating
Procedures
(
SOPs)
that
demonstrate
care
to
minimize
distortions
caused
by
rapid
reactivation
of
carbamylated
enzyme.

Turning
to
issues
of
communication
and
presentation,
the
Panel
finds
a
need
for
more
detail
in
the
Preliminary
Cumulative
Risk
Assessment
and
also
a
need
for
additional
references
to
support
the
Agency's
rationale
and
conclusions.
For
example,
part
of
the
rationale
given
on
page
27
for
choosing
brain
AChE
inhibition
as
the
basis
for
the
point
of
departure
is
that
this
metric
is
as
sensitive
as
behavioral
measures
of
toxicity,
or
more
sensitive.
That
statement
is
justified
with
reference
to
internal
EPA
studies
summarized
in
Appendix
II.
B.
5.
These
studies
concluded
that
traditional
clinical
measures
of
cholinergic
signs
(
salivation,
lacrimation,
urination,
and
defecation,
abbreviated
as
SLUD)
are
less
sensitive
than
effects
on
motor
activity,
and
that
these
in
turn
are
less
sensitive
than
inhibition
of
blood
and
brain
ChE.
The
underlying
data
directly
support
the
concept
that
adverse
effects
of
ChE
inhibition
in
the
peripheral
and
central
nervous
systems
are
adequately
prevented
by
preventing
effects
on
brain
ChE.
Appendix
II.
B.
5,
however,
reports
only
locomotor
activity
and
toxic
signs
for
each
of
7
carbamates.
It
states
that
correlations
between
behavioral
outcomes
and
changes
in
cholinesterase
were
analyzed
but
never
shows
them.
Explicit
presentation
of
this
information
would
strengthen
the
Agency's
case
for
its
choice
of
metric.
The
case
would
be
further
improved
by
fuller
reference
to
the
EPA
data
in
Appendix
II.
B.
5
in
which
the
levels
and
variance
of
RBC
and
brain
AChE
inhibition
are
directly
compared.

Elsewhere
the
document
needs
better
referencing.
For
example,
page
27
provides
a
general
discussion
of
endpoints
in
toxicology
studies
with
N­
methyl
carbamates.
Part
of
this
discussion
simply
states
without
reference
that
behavioral
measures
often
lack
standardization,
are
variable,
and
less
sensitive
to
disturbance
by
carbamates
than
are
measures
of
peripheral
or
central
AChE
activity.
Also
largely
unreferenced
is
the
discussion
of
the
difficulties
associated
with
measuring
ChE
inhibition
in
the
peripheral
nervous
system,
and
the
problems
with
assays
of
whole
blood
that
do
not
distinguish
25
of
63
activity
from
AChE
in
red
blood
cells
and
butyrylcholinesterase
in
plasma.

WATER
WATER
QUESTION
#
1
Revised
Conceptual
Model
for
Ground
Water
Based
on
recommendations
of
the
February
2005
SAP,
OPP
revised
its
ground
water
modeling
approach
to
estimate
pesticide
concentrations
in
the
upper
meter
of
a
fixed
saturated
zone
(
ground
water)
that
starts
at
3.5
m
below
the
surface.
The
Agency
has
included
two
additional
adjustments
to
the
original
conceptual
model
since
the
earlier
SAP.
The
models
consider
variable
degradation
rates
with
depth
and
account
for
setback
distances
between
the
well
and
the
application
area
by
using
lateral
velocity
to
estimate
the
additional
travel
time
for
a
pesticide
to
reach
the
well.

W1.
Please
comment
on
the
Agency's
revisions
to
the
ground
water
modeling
approach
to
account
for
variable
degradation
rates
with
depth
and
varying
setback
distances
between
the
well
and
treated
fields.

Response
EPA
has
made
great
progress
in
modeling
pesticide
movement
in
the
vadose
and
ground
water
zones.
The
revised
ground
water
modeling
approach
provides
a
more
realistic
representation
of
conditions
than
was
used
for
the
February
2005
SAP.
The
revised
modeling
is
still
adequately
conservative.
Although
the
model
designers
did
not
correct
all
errors
found
in
the
computer
code,
remaining
problems
are
small
and
overall
results
are
impressive.

For
the
simulated
profiles
(
high
conductivity
soil
with
no
ponding
of
water
on
a
restrictive
layer)
a
degradation
rate
established
for
aerobic
metabolism
in
lab
studies
is
reasonable
in
the
top
25cm
of
soil.
For
the
zone
from
25
cm
to
1
m
the
Panel
also
agrees
that,
unless
there
are
data
to
the
contrary,
the
best
assumption
is
a
linear
decrease
from
the
aerobic
metabolism
rate
to
the
abiotic
degradation
rate.
On
this
point,
however,
the
Agency
should
check
the
report
of
Ou
et
al.
(
1988),
who
measured
degradation
rates
with
depth
in
Florida.

The
concentration
profile
predicted
by
the
model
is
quite
"
peaky".
Data
from
Long
Island
(
Steenhuis
et
al.,
1987)
indicated
quite
a
bit
of
spatial
variation
in
the
movement
of
the
chemical
with
slow
and
fast
paths,
giving
a
less
peaky
profile.
This
means
that
the
average
of
the
predictions
is
likely
more
realistic
than
the
temporal
concentration
profile.
Smoother
profiles
can
also
be
simulated
by
running
the
model
several
times
with
different
fluxes,
each
passing
through
a
portion
of
the
soil
profile.
26
of
63
The
ground
water
routine
is
refreshingly
simple
and
represents
processes
well.
However,
EPA
is
advised
to
consider
the
following
points.

A
value
of
0.15
m/
day
seems
to
be
at
the
high
end
of
groundwater
velocities
in
"
real"
aquifers
for
water
flowing
with
a
natural
gradient,
but
it
probably
underestimates
the
velocity
in
aquifers
in
valley
bottoms
with
rivers.
The
assumption
is
sound
that
the
presence
of
domestic
wells
does
not
affect
travel
time
of
pesticides.
However,
wells
used
for
irrigation
might
alter
flow
patterns
in
surrounding
ground
water
to
a
larger
degree,
and
the
induced
velocity
fields
could
be
many
times
greater
than
the
natural
gradient­
induced
flows.
This
might
explain
the
high
pesticide
levels
in
the
Florida
data
set,
even
in
wells
with
a
substantial
offset.
Since
the
MOE
for
central
Florida
sites
is
strongly
influenced
by
the
carbamate
concentrations
in
drinking
water,
it
is
important
to
decide
how
to
handle
the
effect
of
irrigation
wells
on
groundwater
concentration.

Although
the
proposal
to
average
pesticide
concentration
over
a
1­
meter
depth
interval
is
technically
correct,
a
plan
to
screen
for
pesticide
just
below
the
surface
of
the
ground
water
is
physically
unrealistic
for
wells
with
an
offset.
Consider
a
residential
drinking
well
that
is
located
300
ft
from
the
edge
of
a
field,
with
water
flowing
under
a
natural
gradient
of
0.15
m/
day.
At
this
rate,
water
will
take
600
days
to
travel
from
field
to
well.
Along
the
way,
clean
water
recharges
and
will
push
the
pesticide
deeper.
Assuming
a
recharge
rate
of
36
cm/
year,
a
porosity
of
0.4,
and
no
diffusion,
chemicals
eluted
from
the
field
will
be
2.4
m
below
the
top
of
the
aquifer
by
the
time
they
reach
the
well.
Therefore
pesticide
levels
should
be
screened
at
least
2.4
m
below
the
ground
water.
Because
the
current
model
assumes
that
the
wells
are
in
fact
screened
deeper
than1
m,
only
the
wording
in
the
document
needs
to
be
changed.
A
problem
that
is
based
more
in
reality,
on
the
other
hand,
is
that
some
diffusion,
dispersion,
mixing,
and
dilution
with
recharge
will
occur
throughout
the
setback
distance.
To
deal
with
this
problem
EPA
may
want
to
use
a
simple
model
(
e.
g.,
KYSPILL
­
developed
by
Sergio
Serrano,
sserrano@
temple.
edu)
to
estimate
changes
in
pesticide
burdens
along
the
path
from
source
to
well.
This
will
assure
that
the
equations
proposed
are
reasonable.
In
addition
the
model
can
be
used
to
assess
the
effects
of
pumping­
induced
flows
by
irrigation
wells.

In
summary,
the
Panel
supports
the
approach
to
identifying
the
spatially
variable
nature
of
pesticides
reaching
groundwater.
OPP
should
continue
to
pursue
this
approach.

Other
minor
comments.

It
was
assumed
that
Aldicarb
degrades
by
a
first
order
process.
In
reality
Aldicarb
degrades
to
some
byproducts
that
are
just
as
toxic.
Therefore
the
latter
process
should
be
taken
into
account
and
can
be
simulated
by
using
the
first
order
degradation
rates
of
the
byproducts.
Liu
et
al
(
2003)
have
shown
that
Aldicarb
degradation
is
much
faster
than
the
degradation
of
its
byproducts.
Therefore
an
assumption
in
the
model
that
Aldicarb
instantaneously
is
converted
to
its
byproducts
would
likely
not
introduce
large
errors
27
of
63
The
set
back
distances
and
travel
times
to
them
should
be
checked.

On
page
4
of
section
II.
D.
7,
it
is
stated
that
in
order
to
implement
the
irrigation
routine
and
obtain
correct
irrigation
rates,
the
depth
of
the
root
zone
in
the
PRZM
model
was
decreased.
Reducing
the
depth
of
the
root
zone
might
have
an
unexpected
effect
on
the
amount
of
evaporation
and
thereby
increase
the
amount
of
recharge
compared
to
the
other
models
that
have
the
correct
root
depth.
See
the
response
to
question
W2
for
further
discussion
of
the
accuracy
of
water
balance.

The
Agency's
working
document
defines
C0
as
the
concentration
of
pesticide
at
the
point
of
application.
Instead
it
should
be
the
concentration
at
the
point
where
pesticide
enters
the
ground
water.

Tables
II.
d.
7.1
through
II.
d.
7.4
use
a
mix
of
English
and
metric
units.
In
some
cases
the
same
numerical
values
appear
with
different
units.
Copy
editing
and
proofreading
is
advised.

WATER
QUESTION
#
2
Comparisons
of
the
Three
Models
The
three
models
used
by
the
Agency
(
PRZM,
RZWQM,
and
LEACHP)
provided
predicted
concentrations
that
were
similar
on
average,
but
short­
term
concentration
differences
among
the
models
varied
considerably.
Differences
in
peak
concentration
estimates
ranged
from
a
factor
of
2
to
5
in
Florida
to
as
much
as
a
factor
of
20
in
North
Carolina;
however,
there
was
no
consistency
with
regard
to
which
model
gave
the
highest
or
lowest
predictions.
Some
of
these
differences
may
due
to
differences
in
the
way
the
models
handle
degradation­
temperature
relationships,
evapotranspiration,
and
weather
generation.

W2.
Given
that
no
model
stands
out
as
superior
when
compared
to
the
monitoring
data
evaluated
so
far,
can
the
SAP
suggest
criteria
for
further
evaluation
of
the
models?

Response
To
facilitate
comparison
of
the
three
models
(
RZWQM,
PRZM,
and
LEACHP),
a
common
weather
file
should
be
used
as
input
for
all
of
them.
Historical
weather
data
is
preferable
to
CLIGEN
(
climate
generator),
which
may
not
simulate
subtropical
weather
accurately.
An
appendix
summarizing
the
major
input
parameters
and
the
rainfall
characteristics
would
help
readers
understand
the
modeling
scenarios
more
fully.

Before
predicting
pesticide
losses,
it
is
important
to
investigate
the
hydrology.
The
most
important
hydrologic
consideration
is
an
accurate
water
balance.
For
that
purpose
28
of
63
the
estimate
of
evapotranspiration
is
crucial.
One
study
on
Long
Island
measured
recharge
over
a
sixth
month
period
(
Steenhuis
et
al.,
1985;
Steenhuis
and
van
der
Molen,
1986).
Information
from
this
study
can
be
used
to
check
the
water
balances
of
the
current
simulations.
Another
hydrologic
consideration
is
the
expected
water
flow
out
of
the
unit
area
(
or
unit
volume)
given
the
lateral
groundwater
velocity
and
aquifer
porosity.
Consider
calibrating
the
three
models
so
that
predicted
flow
from
the
unit
area
is
similar
with
each.
With
RZWQM,
tile
flow
calibration
involves
adjusting
drain
diameter,
lateral
hydraulic
conductivity,
and
"
effective
porosity"
or
porosity
minus
field
capacity
(
Singh
et
al.,
1996).
With
PRZM
the
calibration
parameters
to
equate
predicted
and
observed
flow
from
the
unit
volume
may
be
saturated
hydraulic
conductivity
and/
or
porosity
of
the
aquifer.
By
using
the
calibrated
models,
the
simulations
of
pesticide
fate
can
be
assessed
without
confounding
influences
from
uncertainties
of
water
transport.
In
any
case,
understanding
the
hydrologic
balance
of
the
three
models
will
greatly
help
model
comparison.

As
always
it
is
important
to
consider
how
the
chemicals
are
sampled.
The
sampling
protocol
is
questionable
in
light
of
the
curves
in
Figure
II.
D.
7.17.
Assuming
that
the
bromide
flux
is
the
same
for
the
model
and
the
observations,
then
the
mass
under
the
simulated
and
observed
curves
should
be
the
same.
The
mass
for
the
observed
data
is
considerably
less
(
by
about
half)
than
for
the
simulated
data.
This
discrepancy
points
toward
a
sampling
problem
because
the
model
should
be
able
to
simulate
the
correct
total
mass.
The
difference
between
observed
and
predicted
bromide
flux
may
be
due
to
the
physical
sampling
process.
In
particular,
pumping
of
wells
used
for
irrigation
could
cause
mixing
from
a
depth
increment
that
would
not
be
included
in
the
model
predictions.

The
peak
pesticide
concentrations
in
groundwater
differed
considerably
between
the
three
models.
Peak
exposures
are
important
for
this
risk
assessment.
To
identify
the
source
of
the
model
differences,
including
peak
concentrations
in
groundwater,
a
pesticide
mass
balance
and
a
hydrologic
balance
are
both
important.
Factors
to
consider
with
both
pesticide
and
water
include:
application,
runoff,
percolate
(
or
lateral
flow)
out
of
the
profile,
tile­
drainage,
pesticide
degradation
in
profile,
plant
uptake,
and
evapotranspiration.
Without
completing
a
hydrologic
and
pesticide
mass
balance
on
the
three
models,
it
is
difficult
to
compare
the
models.

The
Panel
suggests
the
following
ways
to
evaluate
the
models
under
consideration.
The
current
simulations
are
primarily
from
areas
of
low
soil
carbon,
sandy
soil,
and
low
sorption
coefficient
(
Koc).
These
conditions
are
consistent
with
" 
drinking
water
that
is
expected
to
be
among
the
most
vulnerable "
(
p.
88).
Vulnerable
sources,
however,
should
include
high
intensity
rainfall
within
a
few
days
of
pesticide
application
where
macropore
flow
occurred.
The
current
version
of
the
NMC
Cumulative
Risk
Assessment
does
not
specify
if
macropore
flow
was
simulated
to
occur
and
does
not
specify
when
intense
rainfall
occurred
in
relation
to
application
and
other
rainfall.
Even
if
macropores
were
parameterized
for
RZWQM,
rainfall
must
be
intense
enough
and
soil
properties
such
that
macropore
flow
actually
occurred
and
reached
groundwater.
On
structured
soils,
low
29
of
63
intensity
rainfall
after
pesticide
application
reduces
pesticide
transport
in
macropores
during
subsequent
rainfall
(
Shipitalo
et
al.,
1990).

Pesticide
leaching
under
some
circumstances
can
be
greater
on
structured
soil
than
on
sandier
soil.
Sadeghi
et
al.
(
2000)
concluded
that
intact
and
repacked
silt
loam
soil
laboratory
columns
leached
more
atrazine
than
a
sandy
loam
soil
that
had
less
carbon
and
less
clay
content
because
of
more
macropore
flow
from
the
silt
loam
soil.

For
the
high
conductivity
soils
in
Florida,
macropore
flow
may
be
less
important
than
on
structured
soils
because
few
storms
will
exceed
the
saturated
conductivity
of
the
soil.
Preferential
flow,
however,
is
still
important
on
high
conductivity
soils
because
of
fingered
and/
or
funnel
flow
(
Kung,
1990;
Glass
et
al.,
1989).
This
is
consistent
with
the
conclusion
of
Jarvis
et
al.
(
1994)
that
it
was
important
to
model
preferential
flow
in
order
to
accurately
predict
herbicide
leaching
on
sandy
textured
soil.
One
method
to
predict
the
portion
of
preferential
flow
within
the
profile
of
sandy
soil
is
to
divide
the
maximum
intensity
of
the
rainfall
or
irrigation
(
on
an
hourly
basis
or
less)
by
the
saturated
conductivity
of
the
soil
(
information
about
this
approach
can
be
found
in
Darnault
et
al.,
2004;
Kim
et
al.,
2005;
Selker
et
al.,
1996).
Another
way
to
estimate
the
preferential
flow
area
of
the
soil
is
to
calibrate
the
model
with
observed
data
using
an
inert
non­
adsorbing
tracer
such
as
bromide.

In
summary,
the
Panel
strongly
favors
comparing
the
hydrologic
and
pesticide
balance
of
the
three
models.
Another
important
step
to
consider,
focusing
on
the
most
vulnerable
groundwater,
is
to
simulate
preferential
flow
shortly
after
pesticide
application.
Many
points
must
be
considered
in
choosing
which
model
to
use
for
the
present
purpose.
Among
these
is
the
question
of
which
model
most
reasonably
represents
the
expected
processes
(
pesticide,
hydrology,
and
preferential
flow).
If
thorough
comparison
doesn't
identify
a
superior
model,
however,
it
may
be
best
to
use
the
simplest
model.

WATER
QUESTION
#
3
Evaluation
of
the
Ground
Water
Model
Estimates
The
Agency
compared
NMC
concentrations
in
ground
water
estimated
with
the
three
models
(
PRZM,
RZWQM,
LEACHP)
to
results
of
available
prospective
ground
water
monitoring
studies
(
oxamyl
in
NC
and
MD
and
methomyl
in
GA),
two
wellmonitoring
studies
along
the
central
ridge
of
FL,
and
published
literature
on
in­
field
monitoring
studies.
Using
the
FL
well
monitoring
data,
known
fate
characteristics
of
the
NMC
pesticides,
and
soil
and
hydrologic
data,
the
Agency
identified
the
conditions
under
which
exposures
similar
to
that
estimated
in
the
NMC
CRA
may
occur:
private
wells
drawing
from
shallow,
acidic
ground
water
with
high
to
very
high
saturated
hydraulic
conductivities
in
the
soil
and
vadose
zone.
This
has
allowed
the
Agency
to
move
toward
a
spatially­
explicit
characterization
of
potential
high
exposure
areas.
30
of
63
W3.
Please
comment
on
the
performance
of
the
models
against
the
available
monitoring
data.
What
additional
considerations
should
be
taken
when
applying
modeled
estimates
to
risk
assessments
for
areas
where
monitoring
data
are
not
available?

Response
Model
Performance
There
are
no
real
surprises
in
performance
of
the
models
relative
to
the
available
monitoring
data.
The
magnitudes
of
the
differences
in
model
predictions
are
expected
given
the
differences
in
the
models
and
the
input
data
they
use.
The
performance
of
the
models
is
reasonable.
Statistical
measures,
such
as
Nash
Sutcliffe
coefficients,
R2,
and
RMSE,
describing
model
performance
relative
to
observed
data
would
be
helpful
in
assessing
and
comparing
model
performance.

It
would
be
useful
to
know
which
of
these
models
is
best
at
estimating
the
hydrology
for
the
cases
being
modeled.
The
model
that
best
estimates
the
hydrology
would
have
the
potential
to
perform
the
best
in
estimating
pesticide
concentrations
in
shallow
ground
water.
If
the
hydrology
is
incorrect,
it
is
difficult
to
estimate
the
pesticide
concentrations
correctly
given
that
the
movement
of
water
is
the
transport
mechanism
for
the
pesticides.
In
such
a
case
pesticide
concentrations
may
at
best
be
correct
for
the
wrong
reasons.

Since
the
models
were
validated
for
only
two
locations,
it
may
be
worthwhile
to
compare
and
validate
the
models
with
additional
observed
data.
Observed
data
from
eastern
Canada
could
be
one
of
these
data
sets
(
http://
www.
hc­
sc.
gc.
ca/
ewhsemt
pubs/
water­
eau/
doc_
sup­
appui/
aldicarb­
aldicarbe/
index_
e.
html).
This
reference
indicates
that
the
following
data
are
available:

"
In
a
survey
of
317
wells
in
eastern
Canada
in
September
1986,
aldicarb
was
detected
in
167
of
782
samples;
concentrations
were
above
10
ppb
in
only
9%
of
the
167
samples.
In
surveys
of
private
and
municipal
drinking
water
supplies
in
five
Canadian
provinces,
conducted
from
1980
to
1986,
aldicarb
was
detected
in
111
of
1017
samples
(
detection
limits
ranged
from
0.01
to
3.0
µ
g/
L);
the
maximum
concentration
was
28
µ
g/
L.
In
Prince
Edward
Island
during
1985
and
1986,
77
of
96
samples
(
80%)
in
two
areas
contained
aldicarb
residues
above
the
detection
limit
of
0.1
µ
g/
L,
with
a
maximum
of
16.4
µ
g/
L.
In
the
same
province,
ground
water
quality
was
monitored
between
1985
and
1988
near
two
potato
fields
to
which
aldicarb
was
applied
at
planting
once
or
twice
between
1983
and
1986.
In
May
1988,
concentrations
of
aldicarb
plus
its
degradation
products
exceeded
9
µ
g/
L
in
12%
of
48
well
samples.
Residues
of
aldicarb
and
its
sulphoxide
and
sulphone
have
also
been
reported
frequently
in
water
samples
in
a
number
of
U.
S.
states;
concentrations
are
typically
in
the
range
1­
50
µ
g/
L,
and
a
31
of
63
maximum
of
400
µ
g/
L
was
recorded
in
one
case
from
Long
Island,
New
York."

Models
should
produce
distributions
similar
to
those
in
monitoring
data.
The
ability
of
PRZM,
RZWQM,
and
LEACHP
to
produce
distributions
of
pesticide
concentrations
in
ground
water
relative
to
monitoring
data
should
be
explored
and
characterized.

From
the
information
provided,
the
Panel
concurs
with
EPA's
statement
in
the
ESTIMATION
OF
CUMULATIVE
RISK
FROM
N­
METHYL
CARBAMATE
PESTICIDES:
Preliminary
Assessment
document
that
"
There
is
no
clear
"
best"
model
to
use
to
assess
pesticide
concentrations
in
ground
water."
However,
this
conclusion
could
change
after
EPA
considers
the
Panel
comments
from
questions
W1,
W2,
and
W3.

Additional
considerations
when
applying
modeled
estimates
to
risk
assessments
for
areas
where
monitoring
data
are
not
available
Monitoring
is
necessarily
limited.
Therefore,
models
are
used
to
extrapolate
beyond
the
monitoring
data.
The
key
factors
that
identify
areas
with
the
greatest
potential
for
pesticides
reaching
ground
water
have
been
captured
in
the
process
that
is
being
used.
However,
some
additional
factors
and
steps
should
potentially
be
considered.

The
models
can
be
readily
extended
to
locate
regions
likely
to
have
high
concentrations
of
carbamates
in
ground
water,
although
some
further
validation
is
desirable.
The
model
simulations,
as
verified
by
experimental
data
from
Florida
and
experiences
in
Long
Island,
show
clearly
that
elevated
levels
of
carbamate
pesticides
(
and
especially
high
aldicarb
concentrations)
are
probable
in
locations
with
sandy
soils
of
high
conductivity,
low
organic
matter,
and
ground
water
with
a
pH<
7.
These
sites
have
travel
times
on
the
same
order
of
magnitude
as
the
abiotic
degradation
half­
life
of
carbamates.

To
predict
oxamyl
concentration
in
ground
water,
the
dissolved
oxygen
in
the
ground
water
also
might
be
an
important
factor.
The
number
of
pesticide
applications
is
important
too.

It
is
unclear
if
the
monitoring
data
to
which
the
model
outputs
were
compared
included
intense
rainfall
shortly
after
pesticide
application.
A
worst­
case
modeling
scenario
should
include
a
high
intensity
rainfall
shortly
after
pesticide
application
where
macropore
flow
occurs.
Research
suggests
that
preferential
flow
occurs
on
sandy
soil.
On
structured
soil,
the
most
significant
preferential
flow
event
is
generally
during
the
first
intense
rainfall
after
application.
Therefore,
the
Agency
should
consider
a
scenario
in
which
intense
rainfall
shortly
after
application
leads
to
macropore
flow
to
shallow
ground
water.
Additional
discussion,
rationale,
and
references
concerning
macropore
and
preferential
flow
are
provided
in
the
Panel
response
to
the
W2
question.
32
of
63
Experts
may
be
able
to
identify
other
regions
in
which
high
pesticide
concentrations
might
reach
ground
water.
It
may
be
desirable
to
consult
with
informed
individuals
in
various
regions
where
carbamates
are
applied,
since
limited
data
are
now
available
on
ground
water
quality
in
these
areas.
In
particular,
the
quality
of
available
spatial
data
does
not
allow
us
to
identify
very
small
areas
in
regions
that
have
just
a
few
fields
with
high
carbamate
use.
Such
areas
may
not
be
important
to
the
national
level
assessments
but
could
present
high
exposure
risk
for
a
small
number
of
people.

Miscellaneous
comments
on
water
questions
Although
it
is
assumed
that
the
private
wells
are
the
most
sensitive,
it
is
not
unlikely
that
the
municipal
wells
can
have
also
carbamates
in
the
drinking
water.
For
example
in
the
city
of
Owen
Sound
in
Ontario
a
sample
taken
in
late
summer
of
2000
had
a
Carbofuran
concentration
of
2
ppb
as
the
only
pesticide
(
http://
city.
owensound
on.
ca/
water/
2000­
thirdquarter.
pdf).

Isolated
cases
of
ground
water
pollution
of
domestic
wells
below
pesticide
treated
fields
will
only
affect
a
few
locations
in
certain
regions
but
possibly
cannot
be
ignored
in
the
aggregate.

FOOD
FOOD
QUESTION
#
1
The
food
portion
of
the
N­
methyl
carbamate
cumulative
risk
assessment
used
similar
data
sources
and
techniques
to
those
used
for
the
organophosphate
pesticide
for
estimating
cumulative
risk
from
food.
This
included
use
of
both
the
USDA's
Continuing
Survey
of
Food
Intakes
by
Individuals
(
CSFII)
as
a
data
source
for
food
consumption
and
Pesticide
Data
Program
Data
(
PDP)
as
a
data
source
for
food
residues.

F1.
Please
comment
on
the
planned
intermediate­
and
longer­
term
activities
associated
with
sensitivity
analyses
identified
in
Section
I
of
the
document.
Does
the
Panel
have
any
suggestions
for
other
or
additional
activities
which
the
Agency
should
consider?

Response
General
Comments
In
a
broad
sense,
the
Panel
agrees
with
the
Agency's
intermediate
and
long­
term
activities,
and
appreciates
the
Agency's
responsiveness
to
the
Panel's
comments
in
the
previous
SAP
meeting
regarding
the
food
component
in
the
N­
methyl
carbamate
cumulative
exposure
and
risk
assessment.
Now
the
Panel
would
like
to
see
the
Agency
33
of
63
prioritize
its
activities
so
that
the
key
outcomes
that
may
help
improve
the
dietary
component
of
the
cumulative
exposure
assessment
will
be
available
in
the
near
future.
A
series
of
recommendations
are
offered
here.

Intermediate
Term
 
Conduct
a
more
detailed
analysis
of
food
exposure
to
identify
major
contributors
to
risk,
identifying
specific
food­
pesticide
combinations.

Identifying
individual
foods
and
food
classes
likely
to
make
major
contributions
to
dietary
exposure
is
an
excellent
next
step,
and
one
that
can
be
accomplished
using
the
databases
described.
It
also
is
of
interest
to
identify
major
contributors
that
are
agespecific
for
example,
the
differential
sensitivity
of
children
to
dietary
intake
of
carbamates.

Of
further
potential
interest
are
NHANES
data
on
biological
markers
of
exposure,
e.
g.,
urinary
1­
naphthol,
that
can
be
correlated
with
certain
intake
data.
EPA
may
wish
to
explore
the
use
of
such
markers
as
part
of
the
overall
risk
calculations.

One
Panel
recommendation
is
that
the
Agency
plan
to
gather
longitudinal
dietary
consumption
patterns
from
individuals
living
in
different
regions
of
the
country.
Autocorrelation
and
"
anti­
autocorrelation"
are
likely
in
dietary
intake
and
cannot
be
obtained
from
cross­
sectional
data.
For
example,
some
individuals
may
tend
to
eat
foods
drawn
from
a
relatively
small
fraction
of
the
total
possible
items
due
to
preferences
that
may
be
purely
personal
(
e.
g.,
a
vegetarian
diet)
or
based
on
culture,
ethnicity,
and
geographical
location.
Such
data
differ
from
the
data
compiled
by
the
CSFII,
which
is
a
cross­
sectional
survey
with
(
one­
time)
repeated
sampling
within
a
10­
day
period.
The
CSFII
lacks
information
on
each
individual's
long­
term
dietary
consumption
pattern.
This
limitation
can
be
problematic
since
certain
food
commodities
are
more
likely
than
others
to
contain
NMC
residues.
Furthermore,
such
differences
may
have
seasonal
components
that
are
overlooked
in
rolling,
cross­
sectional
studies.
A
longitudinal
study
of
consumption
is
a
daunting
task;
however,
recent
improvements
in
survey
tools
should
facilitate
the
process.

 
Conduct
a
series
of
sensitivity
analyses
for
input
parameters
that
are
most
likely
to
impact
the
outcome
of
the
assessment
and
determine
their
effects.
The
effects
of
deleting
earlier
years
and
of
using
PDP
data
translation
protocols
are
worth
study.

Sensitivity
analyses
were
discussed
during
this
meeting
in
other
contexts
and
are
of
particular
interest
here.
Several
kinds
of
sensitivity
analyses
can
be
envisioned
including
various
omissions
from
data
sets
to
assess
the
resulting
change
in
estimated
parameters.
Small
changes
will
increase
confidence
in
the
parameter
estimates,
while
large
changes,
suggesting
that
such
estimates
are
not
robust,
may
focus
attention
on
the
quality
and
comprehensiveness
of
the
data.
34
of
63
Data
omissions
to
consider
in
a
sensitivity
analysis
include
elimination
of
earlier
years
from
the
PDP
data.
This
elimination
may
be
desirable
because
pesticide
registrations
and
use
patterns
have
changed
and
concentrations
measured
in
earlier
years
are
no
longer
appropriate
for
modeling
current
exposures.
Sampling
strategies
for
foods
have
not
changed.
Thus
the
data
from
early
CFSII
collections
can
be
used
and
may
be
indicative
of
long­
term
trends
in
dietary
change.
Examination
of
the
PDP
data
for
carbamates
will
indicate
numerous
foods
that
do
not
contain
measurable
quantities
of
these
materials.
It
may
be
worthwhile
to
test
the
effects
of
eliminating
such
foods.

In
general,
sensitivity
analyses
are
to
be
encouraged
as
they
give
insight
into
the
potential
impact
of
dietary
trends.
The
long­
term
change
that
led
from
high­
fat
to
lowerfat
diets
and
then
to
the
"
low­
carb
craze"
could
be
modeled.
One
could
look
forward
and
test
the
effects
of
possible
dietary
changes
in
the
future.
For
example,
it
would
be
interesting
to
know
if
carbamate
intake
would
rise
or
fall
substantially
if
Americans
began
to
eat
less
fast
food.

With
regard
to
translation
protocols,
the
Panel
finds
it
acceptable
to
use
residue
data
from
one
commodity
as
a
surrogate
for
another
when
both
commodities
are
in
the
same
tolerance
crop
group
published
in
the
CFR.
Moreover,
within
a
given
food
form,
the
processing
factor
for
one
pesticide
should
be
used
for
another
pesticide
when
data
for
all
pesticides
are
not
available
in
the
NMC
CRA
analysis.
This
is
especially
true
when
the
processing
factor
appears
to
reflect
dehydration,
e.
g.,
from
apple
to
dried
apple.

 
Determine
how
the
choice
of
assumed
values
for
"
non­
detects"
affects
the
estimated
exposure.

Choosing
different
values
for
non­
detects
may
well
affect
estimated
exposures
and
risks.
Some
compounds
could
pose
significant
residual
risk
at
or
below
LOD
because
certain
groups
consume
large
amounts
of
material
(
e.
g.,
children
and
milk).
Even
though
a
given
carbamate
may
not
occur
at
levels
above
the
LOD
for
a
given
analytical
technique,
the
total
exposure
to
pesticide,
calculated
as
concentration
in
food
multiplied
by
the
quantity
eaten,
may
nonetheless
be
significant.
Exposures
from
foods
eaten
in
large
quantities
may
be
underestimated
if
concentrations
in
samples
recorded
as
"
below
LOD"
are
arbitrarily
set
at
zero.

The
Agency
has
apparently
conducted
a
preliminary
analysis
that
showed
little
impact
of
LOD
values
on
the
final
dietary
exposure
analysis.
However,
it
is
important
to
qualify
"
little
impact."
We
do
not
advocate
abandoning
the
assumption
of
zero
for
nondetects
and
we
do
not
argue
for
any
other
value
in
particular,
but
we
would
like
better
communication
from
EPA.
We
recognize
that
in
a
cumulative
risk
assessment,
a
zero
residue
for
one
NMC
may
not
necessarily
mean
zero
residues
for
all.
And
we
understand
that
any
assumption
about
the
values
associated
with
"
non­
detects"
will
have
little
effect
on
high­
end
exposures.
Nonetheless,
the
Panel
points
out
that
scoring
non­
detects
with
zero
values
will
most
certainly
distort
the
shape
of
the
exposure
distribution
at
the
lower
35
of
63
end.
That
must
happen
because
the
ultimate
exposure
will
then
be
zero
regardless
of
consumption
rate,
not
the
low
level
that
would
be
computed
from
any
other
finite
value
assumed
for
the
tested
samples.

In
the
long
run
the
best
plan
of
attack
is
to
ensure
that
the
limits
of
detection
(
LOD)
for
the
residue­
monitoring
program
are
toxicologically
relevant
(
i.
e.,
that
a
residue
at
the
LOD
does
not
contribute
significantly
to
overall
risk).
This
plan
can
be
accomplished
with
the
aid
of
better
methods
in
laboratory
analysis.
Alternative
strategies
include
combining
information
from
the
datasets
described
above
and
evaluating
the
most
important
foods
using
both
presumed
residue
data
and
overall
intake
in
a
combined
fashion.
One
can
then
determine
in
a
more
quantitative
manner
how
varying
LOD
values
affect
total
exposure.

 
Evaluate
the
Carbamate
Market
Basket
Residue
Monitoring
Study
and
its
implications
for
cumulative
risk
assessment
(
particularly
with
respect
to
single
item
vs.
composite
samples).

This
proposed
work
fits
in
well
with
the
suggestions
made
above.
Single
item
samples
are
of
interest
since
only
a
few
food
items
may
be
expected
to
have
levels
of
contamination
deemed
"
large"
by
a
given
metric.
Composite
samples
dilute
the
effect
of
the
more
contaminated
samples
and
may
completely
mask
them
by
dropping
concentrations
below
LOD.
An
example
cited
at
the
Panel
meeting
was
a
sample
of
15
apples,
one
with
carbamate
residue
at
15
x
LOD
and
the
rest
at
zero.
A
composited
sample
would
still
be
at
LOD
and
would
score
as
zero.
However,
an
individual
who
eats
the
15
x
LOD
apple
would
receive
a
significant
exposure.
The
most
significant
point
is
that
individual
foods
with
high
levels
must
be
identified
and
not
composited
with
other,
uncontaminated
foods.
At
the
very
least,
the
fraction
of
high­
concentration
items
must
be
assessed.

Market
Basket
Monitoring:
Since
this
survey
is
based
on
a
single
unit
of
each
commodity,
it
is
not
"
just
another
dataset."
The
Agency
is
encouraged
to
evaluate
market
basket
data
for
input
into
the
dietary
exposure
assessment.
Such
data
should
be
compared
to
the
PDP
composite
residue
and
single
unit
residue
data.
The
Panel
does
not
know
if
the
Agency
has
already
conducted
studies
showing
that
replacing
composite
data
with
single
unit
data
does
not
significantly
impact
the
final
outcome
of
the
analysis.
If
that
is
true,
however,
the
effort
and
its
outcome
should
be
communicated
clearly
in
the
document,
in
the
context
of
the
percentile
and
population
basis
that
will
be
used
to
characterize
the
final
risk.

Long­
term:

 
Investigate
the
effect
of
seasonal
residues
and
consumption
patterns
on
the
cumulative
assessment.
36
of
63
The
Panel
disagreed
on
this
question.
Most
of
the
Panel
considered
it
is
almost
essential
that
the
Agency
investigate
further
the
effect
of
seasonal
residue
and
consumption
patterns
in
the
cumulative
assessment.
In
fact
they
recommended
that
this
be
done
immediately
and
not
deferred
as
a
long­
term
plan.
In
a
previous
SAP
meeting,
one
Panel
member
raised
the
seasonality
issue
and
asked
why
the
cumulative
exposure
assessment
model
passed
along
the
seasonal
effect
for
water
but
not
for
dietary
consumption.
According
to
the
Agency's
own
assessment,
65%
of
total
NMC
dietary
exposure
comes
from
citrus
fruits,
including
orange,
tangerine,
and
grapefruit,
whose
availability
in
the
US
is
seasonal.
At
least
one
Panel
member
was
concerned
that
the
cumulative
exposure
model
did
not
transmit
this
seasonal
effect
of
dietary
consumption
to
the
higher
end
of
the
total
exposure
profiles.
The
majority
of
the
Panel
had
the
sense
that
the
failure
to
treat
seasonal­
regional
level
effects
reflects
Agency
policy,
not
methodological
constraints
in
the
exposure
model.
The
Agency's
plan
to
conduct
further
investigation
to
identify
specific
food­
pesticide
residues­
consumption
combinations
and
their
contribution
to
risk
is
welcomed
and
should
be
pursued
as
soon
as
possible.

Another
Panel
member
of
the
Panel
suggested
that
investigation
of
seasonal
residues
and
consumption
patterns
would
be
interesting
but
not
an
immediate
priority.
According
to
this
Panel
member,
seasonal
effects
are
likely
to
be
second
order.
Specific
crops
are
likely
to
be
treated
with
a
specific
carbamate,
and
then
stored.
The
likely
exposure
will
be
modified
by
second­
order
effects
like
seasonal
changes
in
food
sources,
e.
g.
from
locally
grown
items
to
imports
from
other
parts
of
the
country
or
abroad.
However,
the
primary
effect
is
still
likely
to
be
food­
item­
specific.
Hence,
a
study
of
seasonal
effects
is
appropriately
deferred
while
attention
focuses
on
identifying
those
foods
likely
to
have
high
contamination.

 
Evaluate
the
tails
of
the
food
exposure
distribution
to
verify
that
unusual
consumption
patterns
are
not
inappropriately
impacting
the
results
of
the
assessment.

The
CSFII
is
designed
to
be
representative
of
the
population
as
a
whole.
Hence
the
"
tails"
of
the
distribution
are
still
part
of
the
distribution
and,
therefore,
cannot
be
said
to
impact
the
results
of
the
assessment
inappropriately.
An
individual
whose
diet
consists
of
nothing
but
the
single
most
contaminated
food
item
may
be
unusual,
even
extreme,
but
is
still
relevant.
Appropriate
statistical
analysis
may
be
all
that
is
needed
to
identify
such
individuals.
The
analysis
of
unusual
but
reasonable
eating
patterns
can
begin
with
the
high­
contributing
commodities
identified
as
recommended
under
the
first
bulleted
item
above.
For
evaluating
the
high­
end
consumption
pattern,
and
with
a
high
contributing
commodity,
the
analysis
should
look
at
"
user­
only"
distribution
and
not
"
per­
capita"
data.
The
joint
probability
of
finding
an
individual
who
eats
large
amounts
of
heavily
contaminated
foods
is
likely
to
be
small,
but
should
also
be
reflected
accurately
in
the
assessment.
In
any
calculation
of
risk,
including
this
type
of
exposure
assessment,
a
significant
fraction
of
the
burden
is
often
carried
by
a
small
number
of
individuals.

The
above
comments
address
the
question
in
the
form
it
was
submitted,
but
37
of
63
probably
not
its
intent.
It
is
an
excellent
idea
to
look
for
unusual
patterns
of
individual
consumption
in
the
tails
of
the
distribution.
As
with
mercury
in
tuna,
individuals
who
eat
large
quantities
of
foods
that
are
known
to
contain
a
specific
carbamate
should
be
made
aware
of
the
risks.
This
is
even
more
appropriate
when
one
examines
the
cumulative
exposure
group.
While
one
particular
food
is
unlikely
to
contain
large
concentrations
of
multiple
carbamates,
several
foods,
each
with
modest
amounts
of
several
carbamates
could
generate
exposure
to
the
whole
class
of
compounds.

RESIDENTIAL
RESIDENTIAL
QUESTION
#
1
Use
of
REJV
Data
and
Professional
Judgment
To
generate
estimates
of
exposure
from
residential
use
of
NMC
pesticides,
the
probabilistic
models
use
a
variety
of
inputs
to
address
potential
exposure
from
multiple
use
scenarios.
Critical
inputs
include
the
percent
of
households
applying
the
various
pesticide
products,
and
the
timing
of
those
applications.
These
two
inputs,
coupled
with
potential
exposure
from
pesticide
residues
in
drinking
water
and
the
diet,
directly
impact
per
capita
estimates
of
cumulative
exposure.
In
its
February
Case
Study,
the
Agency
presented
background
information
on
the
Residential
Exposure
Joint
Venture
(
REJV)
survey.
The
Agency
used
this
database
as
the
primary
source
for
data
on
the
inputs
relating
to
timing
of
applications
and
percent
of
households
using
NMC
products.
Details
regarding
the
empirical
data
of
the
REJV
survey
are
presented
in
Appendix
II.
E.
1.

In
February
2005,
the
SAP
expressed
reservations
regarding
the
REJV
data.
In
response
to
SAP
concerns,
EPA
used
other
non­
survey
information
in
this
preliminary
CRA,
in
addition
to
estimates
from
REJV,
to
develop
use/
usage
inputs
and
seasonal
timelines
of
pesticide
use
which
were
representative
of
the
Southern
region
of
the
U.
S.

As
previously
mentioned,
the
REJV
survey
can
be
used
to
generate
empiricallybased
estimates
of
percent
of
household
use
and
the
frequency
of
product
specific
applications.
But,
because
the
REJV
did
not
collect
information
regarding
the
reason
for
the
reported
pesticide
use
(
pest
treated)
or
how
much
of
the
product
was
used,
the
empirical
timing
and
frequency
information
(
based
on
a
national
survey)
may
not
provide
a
clear
picture
of
regional
use.
Therefore,
to
establish
the
timing
of
pesticide
applications
for
the
scenarios
likely
to
result
in
the
highest
exposure,
the
Agency
made
these
estimates
based
on
a
combination
of
REJV
data,
product
label
information,
professional
judgment,
and
pest
pressure
information
available
from
the
Cooperative
State
Extension
Services.
Specific
examples
of
how
these
sources
were
used
to
determine
timing
and
frequency
of
pesticide
use
for
PNMC
residential
assessment
are
presented
in
Section
E
of
the
preliminary
NMC
CRA
document.

R1.
Please
comment
on
the
use
of
information
sources
other
than
REJV
to
38
of
63
establish
periods
of
pesticide
use
and
other
use/
usage
information.
Does
the
Panel
suggest
an
alternative
method
to
improve
the
use
of
REJV
in
the
NMC
assessment?
Does
the
Panel
know
of
other
data
sources
that
may
be
available?

Response
The
Panel
is
pleased
to
see
more
description
of
the
REJV
data
pertaining
to
residential
exposure
of
NMC,
especially
considering
that
the
Panel
had
been
unable
to
comment
on
the
use
of
REJV
data
at
the
February
2005
SAP
meeting
due
to
lack
of
information
on
this
proprietary
dataset.

The
Agency
should
explore
the
possibility
that
other
proprietary
data
might
be
available.
Nevertheless,
based
on
the
information
in
Appendix
II.
E.
1
of
the
Agency
document
(
August
2,
2005),
the
REJV
dataset
appears
to
contain
much
useful
information
regarding
residential
exposure
to
NMC,
and
in
many
respects
is
more
useful
than
the
National
Home
and
Garden
Pesticide
Use
Survey
(
NHGPUS),
which
is
dated.
The
Agency
also
appears
to
appreciate
the
limitations
of
this
database
and
has
articulated
a
plan
to
consider
the
impact
of
these
limitations
on
assessing
the
cumulative
risk
of
NMC.

It
will
be
almost
unavoidable
to
use
information
sources
other
then
REJV
to
supplement
the
estimation
of
exposure
from
residential
applications
of
NMC,
considering
that
the
REJV
database
contains
insufficient
information
for
cumulative
exposure
assessment.
The
Agency's
general
principles
and
approach
toward
using
other
necessary
data
(
e.
g.,
information
on
pesticide
use
pattern
and
available
formulation,
maximum
application
rate)
are
reasonable
and
represent
the
best
effort
under
the
current
situation.
The
use
of
"
professional
judgment"
is
often
subsumed
in
the
selected
distributional
characteristics,
e.
g.,
analytical
distributions
truncated
at
the
99th
percentile.
Sensitivity
analyses
should
be
performed
to
assess
the
impact
of
such
judgments.

The
Panel
also
supports
the
Agency's
approach
in
using
the
REJV
data
for
empirical
estimates
of
pesticide
use
patterns
for
residential
exposure
scenarios.
The
Panel
provides
the
following
suggestions
and
comments
on
the
residential
exposure
analysis.

1)
Co­
occurring
residential
applications:
The
REJV
data
may
lack
sufficient
information
to
address
co­
occurring
application
events.
One
plausible
scenario
is
a
series
of
NMC
applications
to
trees,
ornamentals,
and
the
home
garden,
all
by
the
same
person
in
one
extended
event.
This
scenario
might
arise
"
for
convenience,"
when
extra
tank
mix
remains
after
an
originally
intended
single
use
on,
for
example,
trees.
Exposures
in
such
a
scenario
may
not
be
adequately
characterized
by
a
probabilistic
approach
based
on
the
REJV.
It
may
be
reasonable
to
conduct
a
separate
deterministic
analysis
to
determine
the
plausible
upper
end
of
exposure.

2)
Exposure
following
a
professional
application:
This
scenario
is
not
specifically
39
of
63
addressed.
The
residential
exposure
analysis
using
REJV
data
deals
only
with
post­
application
exposures
associated
with
homeowner
applications.
Residential
exposure
after
professional
application
is
a
realistic
possibility,
however,
even
though,
as
the
Agency
has
indicated,
the
home
presents
fewer
occasions
for
professionals
to
apply
NMCs
than
it
does
for
the
residents
themselves.
Adding
the
scenario
of
professional
application
in
the
home
would
modestly
increase
the
probability
and
frequency
of
estimated
residential
exposure.

3)
REJV
and
NHGPUS
comparison:
The
REJV
appears
to
be
a
superior
database­­
being
more
recent
than
the
NHGPUS
data­­
and
can
better
address
certain
exposure
scenarios
associated
with
residential
use
of
pesticides.
Nonetheless,
the
proprietary
nature
of
this
database
is
likely
to
limit
its
usefulness
and
presents
some
difficulties
in
achieving
transparency
for
risk
assessment.
Also,
as
a
one­
time
survey,
the
REJV
will
also
be
outdated
in
a
few
years.
Moreover,
while
Calendex
and
CARES
both
use
REJV,
Lifeline
uses
NHGPUS.
The
Panel
advises
the
Agency
to
compare
empirical
use
patterns
generated
from
REJV
and
from
NHGPUS,
especially
in
a
manner
similar
to
that
which
is
used
by
Lifeline.
The
comparison
may
add
to
the
support
for
either
database
and
enhance
the
future
utility
of
the
REJV.

4)
Potential
new
database:
It
is
recognized
that
there
is
no
CSFII­
or
PDP­
like
database
for
residential
exposure
as
yet.
However,
two
5­
years,
multi­
million
dollar
research
projects
funded
by
EPA
NCER
are
now
collecting
longitudinal
data
on
relevant
activities
including
residential
pesticide
uses.
Unless
there
are
policies
that
specifically
bar
it
from
doing
so,
the
Agency
is
encouraged
to
communicate
with
the
grant
recipients
to
ensure
that
the
future
data
will
be
of
a
quality
suitable
for
cumulative
risk
assessment
models.

5)
Further
use
of
REJV
database:
EPA
has
not
explored
any
means
of
using
the
REJV
database
other
than
to
estimate
the
percentage
of
households
that
use
particular
pesticides
and
the
frequency
of
product­
specific
applications.
Current
use
is
limited
to
roughly
1200
complete
records.
All
other
responses
in
the
REJV
go
unused.
A
more
complex
statistical
analysis
might
enable
EPA
to
utilize
all
the
REJV
data.
For
example,
the
methods
of
censored
data
analysis
can
be
employed,
treating
the
incomplete
records
as
time­
censored
data.
This
step
will
require
EPA
to
begin
viewing
the
REJV
data
less
like
the
CSFII
records
and
more
like
its
data
on
water
residues.
To
fully
utilize
the
REJV
data
a
model
of
household
usage
and
frequency
of
use
will
be
needed.
The
data
should
be
sufficient
for
a
three
part
model
that
involves
estimating
i)
the
probability
of
a
productspecific
event
like
lawn
pesticide
treatment
on
a
given
day
or
week,
ii)
the
distribution
of
the
number
of
total
such
events
in
the
household
in
a
year
(
e.
g.,
number
of
lawn
pesticide
treatments)
and
iii)
the
distribution
of
the
times
between
events
(
e.
g.,
time
to
next
lawn
pesticide
treatment).
All
three
of
these
components
would
need
to
be
region­
specific.
Furthermore,
certain
types
of
residents
might
not
perform
certain
types
of
activities
(
condo
and
apartment
dwellers
might
not
do
shrub
and
lawn
applications).
Incorporating
such
considerations
into
the
analysis
should
not
be
too
difficult
and
would
improve
this
aspect
of
the
risk
assessment
by
simplifying
the
process
and
by
helping
describe
the
40
of
63
components
in
terms
of
statistical
distributions.

RESIDENTIAL
QUESTION
#
2
Uncertainties
Associated
with
the
Hand­
To­
Mouth
Assessment
To
assess
non­
dietary
ingestion
(
mg/
day),
the
following
four
key
factors
are
used
in
the
models:


Residue
Concentration
(
turf
residues,
pet
fur
residues,
and
residues
from
hard
indoor
surfaces)


Hand
to
mouth
frequency
(
number
of
events
per
hour)


Surface
area
of
the
inserted
hand
parts
(
cm2)


Exposure
time
(
hours/
day)

Other
factors
include
both
saliva
extraction
efficiency
and
wet
hand
adjustment
factor.
This
exposure
estimate
is
then
used
along
with
the
Relative
Potency
Factor
(
RPF)
and
Benchmark
Dose
to
estimate
risk.
In
the
Preliminary
N­
methyl
carbamate
assessment,
risk
estimates
for
non­
dietary
oral
exposure
result
in
the
lowest
Margins
of
Exposure
(
MOEs),
and
would
therefore
be
of
greatest
concern
to
the
Agency;
however,
these
low
MOEs
appear
to
be
due
in
part
to
the
incorporation
of
micro­
activity
data
into
our
macro
activity
models.
As
a
result,
the
non­
dietary
ingestion
scenarios
in
the
Preliminary
Nmethyl
carbamate
cumulative
risk
assessment
are
the
least
refined.

The
residue
concentration
values
are
derived
from
individual
residue
dissipation
or
deposition
studies
which
are
discussed
in
the
Residential
Chapter
(
Section
E)
of
the
Cumulative
Risk
Assessment
document.
The
exposure
durations
are
taken
from
the
Agency's
Exposure
Factors
Handbook.
The
hand
to
mouth
frequencies
and
hand
surface
areas
come
from
behavior
studies
relying
either
on
observational
data
of
young
children
using
video
tape
analysis,
trained
observers,
or
parental
observers.
However,
study
data
that
evaluated
hand­
to­
mouth
frequency
and
surface
area
mouthed
is
difficult
to
interpret.
Specifically,
comparison
of
study
results
can
be
difficult
due
to
differences
in
study
practices
and
methodologies.
For
example,
there
are
no
standard
definitions
of
mouthing
(
superficial
contact,
licking,
biting,
fraction
of
hand
inserted)
and
thus
the
data
for
these
behaviors
likely
differs
among
studies
as
a
result
of
the
investigators
definitions.
In
addition,
the
degree
to
which
ancillary
data
(
such
as
surface
area
of
hand
contacted
or
inserted,
the
duration
of
contact,
and
the
length
of
videotaping)
are
collected
and
reported
differ
among
studies.
This
makes
broad­
based
and
generally­
applicable
interpretation
difficult.
Nevertheless,
Drs.
Zartarian
and
Xue
allowed
us
to
use
their
preliminary
distributional
analyses
of
these
children's
video
data
in
this
assessment.
The
studies
used
in
the
hand
to
mouth
frequency
analysis
performed
by
Zartarian
and
Xue
are
briefly
41
of
63
summarized
in
a
table
provided
in
a
memorandum
dated
August
8,
2005
and
provided
to
the
Panel
under
separate
cover.

The
distributions
of
hand­
to­
mouth
frequencies
and
surface
area
mouthed
used
in
the
Preliminary
NMC
CRA
were
based
on
the
analysis
performed
by
Zartarian
and
Xue
(
as
detailed
above).
In
the
aggregate
models
used
in
the
NMC
cumulative
assessment,
each
separate
iteration
selects
a
single
value
for
the
hand
to
mouth
events
variable
from
a
distribution
of
hand
to
mouth
frequency
values.
Also,
each
separate
iteration
of
the
model
selects
a
single
surface
area
from
a
distribution
of
the
fraction
of
hand
mouthed.
These
values
are
multiplied
by
the
residues
and
exposure
durations
which
are
similarly
selected
from
a
distribution
of
residue
and
exposure
durations
as
described
above.
This
relatively
simple
selection
process,
however,
ignores
the
numerous
complexities
and
interrelationships
involved
in
this
critical
behavior
pattern.
For
example,
the
area
of
hand
that
is
mouthed
during
a
given
event
may
correlate
inversely
with
the
frequency
of
mouthing
events.
Specifically,
more
frequent
hand­
to­
mouth
events
may
be
associated
with
mouthing
smaller
fractions
of
the
hand.
The
algorithms
used
in
the
NMC
CRA
however,
(
as
established
by
the
OPP
Residential
Standard
Operating
Procedures
(
SOPs)
assume
independence
between
these
two
parameters.
This
assumption
likely
leads
to
overestimates
of
exposures
when
upper
percentiles
of
the
hand­
to­
mouth
frequency
and
area
of
hand
mouthed
distributions
are
combined.
In
addition,
the
macroactivity
approach
used
in
the
NMC
CRA
aggregate
models
is
based
on
the
following
assumptions:


The
mouthing
frequency
(
events
per
hour),
as
recorded
during
the
course
of
observational
studies,
continue
at
the
same
rate
for
the
entire
exposure
duration
selected;
in
reality,
a
high­
end
mouthing
frequency
recorded
over
a
short
time
interval
(
e.
g.,
one
hour)
may
not
be
likely
to
continue
at
the
same
intensity
over
a
longer
time
period
(
e.
g.,
6
or
8
hours)


The
hand
is
fully
replenished
with
residues
from
a
contaminated
surface
(
e.
g.,
the
lawn,
pet
or
hard
flooring)
between
each
hand
to
mouth
event

The
contact
frequency
and
surface
area
data
used
in
this
assessment
are
taken
from
observational
studies
in
which
all
hand
contacts
were
recorded
as
hand­
to­
mouth
events,
regardless
of
the
fraction
of
hand
mouthed.
Additionally,
no
adjustment
was
made
for
the
duration
of
time
the
hand
remained
in
the
mouth.

R2a.
The
methodology
used
in
the
NMC
CRA
in
which
micro­
activity
data
are
used
in
macro­
activity
approach
likely
leads
to
systematic
overestimates
of
exposure
when
upper
percentiles
of
mouthing
frequency
and
surface
area
of
hand
mouthed
are
combined.
Does
the
Panel
agree
that
this
methodology
does
indeed
overestimate
exposure?
Can
the
Panel
suggest
improvements
to
this
methodology
to
further
refine
exposure
estimates?

R2b.
Does
the
Panel
have
suggestions
for
an
alternative
approach
than
the
one
42
of
63
used
to
estimate
the
non­
dietary
oral
exposure
pathway
in
the
Preliminary
NMC
CRA?
For
example,
would
the
use
of
a
time
weighted
frequency
value
based
on
random
hourly
draws
of
hand
frequency
distributions
more
accurately
estimate
hand­
to­
mouth
exposures?

Response
This
Panel
appreciates
the
Agency's
effort
in
responding
the
comments
made
by
the
previous
SAP
meeting
in
February
2005
regarding
the
issue
of
incorporating
the
nondietary
oral
ingestion
in
the
cumulative
residential
exposure
assessment.
The
Panel
is
pleased
to
see
the
Agency's
effort
in
this
development.

The
Panel
recognized
that
non­
dietary
oral
ingestion
is
a
highly
variable
element
of
the
residential
exposure
assessment,
and
perhaps
in
the
overall
cumulative
exposure
assessment.
It
is
therefore
understandable
that
the
non­
dietary
oral
ingestion
is
the
least
refined
component
in
the
present
version
of
the
NMC
cumulative
risk
assessment.
As
the
Agency
indicated
in
the
PNMC
documentation,
an
accurate
assessment
of
non­
dietary
oral
ingestion
requires
a
substantial
amount
of
information
for
four
key
factors.
These
are
residue
concentration,
hand­
to­
mouth
frequency,
surface
area
of
the
mouthing
part,
and
duration
of
exposure.
At
present,
complete
information
on
these
four
factors
is
lacking
or
conflicting.

Even
more
problematic
is
the
analysis
of
hand­
to­
mouth
activity
data
in
a
macroactivity
approach
using
the
default
assumptions
set
out
in
the
documentation.
The
Panel
agreed
that
such
an
approach
will
overestimate
exposure,
as
expressed
by
the
MOE.
However,
different
Panel
members
have
different
suggestions
to
mitigate
this
problem.

As
pointed
out
by
one
Panel
member,
problems
can
arise
when
simulation
procedures
use
independent
draws
from
two
distributions
for
variables
that
are
related
in
reality.
This
may
result
in
overestimates
or
underestimates
of
exposure,
depending
on
the
nature
of
the
association
between
the
two
variables.
The
variance
of
the
product
of
two
random
variables,
say
p=
x
·
y
where
x=
mouthing
frequency
and
y=
surface
area
of
mouthed
hand,
is
var(
p) 
y2var(
x)+
x2var(
y)+
2xy
·
cov(
x,
y).
If
the
input
values
for
x
and
y
are
drawn
independently
in
the
model
run,
the
covariance
term
in
this
variance
expression
is
treated
as
zero
and
the
variance
of
the
distribution
for
the
resulting
product
values
is
simply
equal
to
the
first
two
terms
of
this
expression.
On
the
other
hand,
if
x
and
y
are
positively
correlated
in
reality,
the
variance
of
the
true
p
(
the
target
value)
includes
a
positive
covariance
term.
Therefore,
the
simulated
values
of
p
will
have
less
variability
than
the
target
distribution
and
exposures
will
be
underestimated
(
assuming
that
greater
variability
in
the
target
distribution
corresponds
to
greater
extremes
in
the
tails
of
the
simulated
distribution).
The
reverse
will
be
true
if
x
and
y
are
negatively
correlated
(
e.
g.,
more
frequent
mouthing
is
associated
with
smaller
areas
of
hand
mouthed).
In
that
case,
modeling
the
product
of
two
jointly
distributed
random
variables
as
the
product
of
independent
draws
from
the
x
and
y
distributions
will
overestimate
the
variances
of
the
43
of
63
target
distribution
for
the
product.
In
that
way,
draws
from
the
simulated
distribution
would
yield
more
extreme
values
than
might
be
encountered
in
the
real­
world
process
being
modeled,
or,
in
other
words,
overestimate
exposure
at
the
high
end.

The
rules
that
govern
variances
of
functions
of
random
variables
also
inform
us
about
the
potential
impact
of
employing
a
single
value
for
an
input
variable
over
a
protracted
time
as
opposed
to
refreshing
the
value
through
independent
draws
throughout
the
exposure
window.
Without
presenting
the
formal
statistical
argument
here,
using
a
single
draw
of
an
input
variable
for
a
protracted
period
of
exposure
will
result
in
greater
variability
in
exposures
(
hence
greater
extremes
in
values)
than
a
modeling
procedure
that
periodically
returns
to
refresh
the
value
of
the
input
during
the
window
of
exposure.
For
example,
fixing
the
residue
on
a
child's
hands
for
a
two
hour
play
period,
while
simpler
to
implement,
will
yield
greater
variability
in
the
modeled
distribution
of
exposures
than
a
run
that
updates
the
residue
concentration
hourly
during
the
exposure.
Almost
certainly,
the
composite
simulated
distribution
will
contain
more
extreme
values
than
the
target
distribution,
which
arises
from
fluctuating
exposures.
The
degree
to
which
variability
in
the
target
distribution
is
overestimated
will
be
inversely
related
to
the
autocorrelation
of
the
input
variable
over
time.
A
high
autocorrelation
will
lead
to
a
small
overestimation.
Note:
fixing
the
value
of
an
input
variable
for
a
time
period
is
equivalent
to
assuming
perfect
auto­
correlation
of
its
values
for
subintervals
of
the
larger
time
window.
A
sensitivity
analysis
is
the
best
way
to
determine
whether
or
not
these
effects
have
practical
importance
for
the
interpretation
and
use
of
the
final
exposure
simulation.

Moving
away
from
statistics,
several
Panel
members
suggested
collecting
more
data
and
qualitative
information
from
the
videotapes
and
transcripts
of
studies
on
hand­
tomouth
exposure.
A
key
question
is
whether
the
surface
area
of
the
hand
(
or
fingers
in
mouth)
co­
varies
with
the
frequency
of
mouthing.
Panel
members
recognized
that
a
major
undertaking
might
be
needed
to
answer
this
question.
However,
some
preliminary
work
is
justified
to
see
how
much
additional
information
can
be
culled
from
the
existing
investigations
and
data.

One
Panel
member
concluded
that
the
Agency's
idea
of
time
weighting
might
lead
to
a
more
realistic
assessment
of
exposure.
This
individual
encouraged
the
Agency
to
pursue
this
approach
by
segmenting
the
time
of
contact
into
multiple
short
periods
with
separately
determined
frequency
of
behaviors.
Using
relatively
large
blocks
of
time
would
simplify
an
initial
analysis
along
these
lines.
The
impact
of
the
time­
weighted
approach
could
then
be
evaluated
before
deciding
whether
to
increase
temporal
resolution
and
use
shorter
time
segments.
The
Agency
followed
a
similar
approach
with
its
CCA
exposure
assessment.
However,
other
Panel
members
noted
that
the
paucity
of
available
data
would
make
it
hard
to
tell
which
outcome
is
"
better"
if
time
weighting
affects
the
analysis.

Another
Panel
member
recommended
excluding
hand­
to­
mouth
activity
from
the
residential
exposure
assessment
model
until
more
data
becomes
available.
Meanwhile
the
Agency
should
run
simulations
with
a
deterministic
model
to
learn
whether
hand­
to­
mouth
44
of
63
activity
contributes
appreciably
to
the
overall
residential
exposure
assessment.
If
the
estimated
exposure
from
hand­
to­
mouth
activity
does
represent
a
substantial
portion
of
the
total
residential
exposure,
the
Agency
is
recommended
to
make
the
macro­
activity
assumptions
of
the
model
more
realistic.
In
contrast,
if
hand­
to­
mouth
exposure
indeed
represents
only
a
small
fraction
of
the
total
residential
exposure,
excluding
that
component
from
the
assessment
is
justified
for
the
following
reasons:

1.
The
dermal
exposure
component
takes
into
account
the
fraction
of
pesticide
residue
that
would
be
ingested
if
hand­
to­
mouth
activity
does
occur.
2.
Oral
exposures
from
dietary
ingestion,
and
water
consumption
in
certain
regions,
are
a
much
more
important
component
of
the
cumulative
assessment
and
deserve
more
resources
and
attention
as
assessment
methods
are
refined.

The
Panel
member
who
raised
these
points
argued
that
current
data
for
assessing
non­
dietary
oral
ingestion
are
insufficient
both
in
quantity
and
in
quality,
and
are
unlikely
to
be
sufficient
in
the
foreseeable
future.
This
individual
concluded
that,
without
good
quality
data
to
facilitate
model
development,
inclusion
of
the
hand­
to­
mouth
component
actually
carries
additional
and
unnecessary
error
and
uncertainty
forward
to
the
cumulative
risk.

RESIDENTIAL
QUESTION
#
3
Distributional
Analysis
Assessing
residential
exposure
to
pesticides
is
a
complex
process
that
must
consider
exposure
from
a
variety
of
sources
via
multiple
routes.
To
account
for
exposure
from
different
sources,
the
PNMC
residential
exposure
assessment
identifies
scenarios
where
significant
exposure
may
occur.
Each
of
these
scenarios
is
defined
by
a
specific
type
of
activity
or
set
of
activities
that
may
result
in
exposure.
Generally
the
relationships
between
these
activities
and
the
resulting
exposures
are
well­
defined
in
that
algorithms,
equations,
and
standard
operating
procedures
exist
for
calculating
exposure
based
on
the
activity
being
performed.
However
the
supporting
data
sets
used
to
estimate
exposure
for
various
residential
scenarios
range
from
robust
(
e.
g.,
unit
exposure
values)
to
limited
or
sparse
(
e.
g.,
lawn
sizes,
area
treated,
duration
of
exposure,
and
saliva
extraction
factors).
Additionally,
information
characterizing
the
extent
to
which
each
activity
contributes
to
exposure
for
a
particular
scenario
does
not
always
exist
(
e.
g.,
the
amount
of
time
spent
in
home
gardens
performing
activities
such
as
hand
weeding
versus
staking
tomatoes
or
harvesting
sweet
corn).

In
general,
the
Agency
has
attempted
to
fit
distributions
(
as
described
in
Appendix
II.
E.
2
of
the
NMC
CRA)
to
the
exposure
measurements
for
residential
activities
when
supporting
information
exists
to
characterize
the
extent
to
which
the
activity
contributes
to
exposure
for
the
residential
scenario
of
interest.
However,
the
Agency
has
employed
uniform
distributions
to
the
data
sets
for
which
such
supporting
information
does
not
45
of
63
exist,
(
e.
g.
lawn
sizes,
area
treated,
duration
of
exposure,
and
saliva
extraction
factors).
The
Agency
has
elected
to
create
such
distributions
when
the
available
data
are
limited
to
such
an
extent
that
it
is
uncertain
how
well
they
represent
national
variability.
The
Agency
believes
use
of
uniform
distributions
to
be
conservative
in
estimating
potential
exposure
since
uniform
distributions
tend
to
overestimate
exposure.

R3a.
Please
comment
specifically
on
the
Agency's
use
of
lognormal
distributions
to
estimate
residential
exposure
and
the
statistical
methods
and
procedures
by
which
the
Agency
has
selected
particular
distributions
(
e.
g.,
probability
plots
and
goodness­
of­
fit
statistics).

Response
Probability
plots
and
Shapiro­
Wilk
tests
are
as
good
a
method
as
any
for
assessing
a
distribution.
The
lognormal
has
a
moderately
long
tail
and
the
results
presented
here,
even
though
based
on
a
few
small
samples,
suggest
that
the
lognormal
is
good
enough.
A
lognormal
distribution
arises
naturally
when
an
observed
variable
is
the
product
of
many
arbitrarily
distributed
variables,
or
a
sum
of
variables
on
the
log
scale.
The
lognormal
distribution
is
a
common
choice
for
environmental
measurements
and
these
results
come
as
no
surprise.

The
question
asks
for
a
more
rigorous
evaluation
of
the
goodness­
of­
fit
methodology,
however,
and
the
following
points
need
to
be
made.

 
The
power
of
any
goodness­
of­
fit
test
will
be
too
low
in
small
samples
and
too
high
in
large
samples,
so
that
a
goodness­
of­
fit
test
never
answers
the
relevant
question:
whether
or
not
the
distribution
is
good
enough
for
the
model
to
give
reasonably
accurate
predictions.

 
When
the
null
hypothesis
is
true
and
many
tests
are
done,
the
p­
values
for
the
tests
should
follow
a
uniform
distribution
(
5%
less
than
.05,
1%
less
than
.01,
etc.).
One
does
not
want
all
values
to
be
"
close
to
1"
as
that
would
indicate
that
the
data
are
closer
to
lognormal
than
they
should
be.
There
are
a
few
too
many
very
small
p­
values
among
the
tests
shown
in
the
current
document
but,
on
closer
inspection,
the
low
p­
values
are
mostly
from
inhalation
data
where
several
points
are
tied
for
minimum
and
lie
on
a
horizontal
line
at
the
bottom
of
the
plot:
that
is,
points
that
are
below
the
limit
of
detection
and
reported
as
1/
2
LOD.
These
can
be
ignored
in
a
visual
evaluation
of
the
probability
plot
but
will
invalidate
the
Shapiro­
Wilk
test.
The
maximum
likelihood
estimation
of
the
parameters,
allowing
for
censoring,
is
correct,
but
the
goodness­
of­
fit
tests
are
not
correct
as
shown.

 
Censored
data
must
be
compared
to
a
censored
lognormal,
with
the
tail
below
the
detection
level
removed
and
replaced
by
a
point
mass
at
the
LOD.
A
grouped­
data
chi­
square
test
could
be
used
but
will
be
less
powerful
than
Shapiro­
Wilk.
A
46
of
63
reasonable
and
quick
adaptation
of
Shapiro­
Wilk
for
left
censored
data
is
accomplished
by
fitting
a
linear
regression
line
to
the
QQ
normal
plot
(
with
the
normal
quantiles
on
the
X­
axis
and
the
log
concentration
quantiles
on
the
Y­
axis)
and
omitting
censored
values
while
performing
the
regression.
This
regression
line
provides
estimates
for
the
distribution
mean
and
variance
that
are
comparable
to
the
censored­
data
MLE
estimates
and
the
R2
value
is
the
Shapiro­
Francia
statistic.
Testing
for
normality
(
or
log
normality
here,
since
the
data
have
been
logtransformed
is
accomplished
by
determining
whether
the
R2
term
is
close
enough
to
1.
The
critical
value
depends
on
the
number
of
censored
values
and
the
total
number
of
observations
but
in
general,
if
the
R2
is
not
greater
than
about
0.96,
there
is
evidence
that
the
data
are
not
normal
(
or,
in
this
case,
log
normal).
Visual
inspection
of
the
straight
line
is
often
good
enough
and
outlier
values
are
usually
very
visible.

R3b.
Does
the
Panel
agree
that
the
Agency's
approach
to
creating
and
using
of
uniform
distributions
(
i.
e.,
ranges
of
values)
for
residential
scenarios
lacking
adequate
supporting
information
tends
to
overestimate
exposure?
Is
the
Panel
aware
of
other
data
sources
that
may
be
better
suited
for
assessing
residential
exposure
scenarios
of
interest?
Does
the
Panel
have
any
suggestions
regarding
alternative
distributions
to
use
for
scenarios
where
supporting
exposure
information
is
inadequate?
To
what
extent
should
sensitivity
analyses
be
used
to
assess
the
appropriateness
of
alternative
distributions?

Response
Uniform
distributions
should
never
be
used
as
they
have
no
tails
and
will
never
generate
extreme
cases.
In
these
applications,
the
true
distributions
are
generally
skewed
to
the
right.
In
consequence,
models
using
uniform
distributions
will
understate
the
upper
tails
of
the
exposure
distribution
and
lead
to
underestimates,
not
overestimates,
of
exposure.
At
a
previous
SAP
meeting
the
Panel
commented
extensively
on
the
use
of
uniform
distributions
in
the
context
of
the
SHEDS
analysis
(
Minutes
of
the
meeting
of
August
30,
2002,
No.
2002­
06).
Some
pertinent
quotations
from
that
document
are
reproduced
below:

"
The
Panel
felt
that
the
extensive
use
of
uniform
distributions
to
represent
either
uncertainty
or
variability
should
be
discouraged
in
favor
of
parametric
distributions
that
do
not
have
such
strictly
defined
limits.
Distributions
with
defined
limits
should
generally
be
used
only
in
cases
where
the
limits
can
be
firmly
based
on
physical
principles.
The
model
should
also
allow
use
of
Beta,
Gamma
and
Weibull
distributions,
mixtures
of
any
of
the
available
distributions,
and
the
ability
to
establish
a
distribution
with
a
spike
of
probability
at
0.
The
Beta
distribution
includes
the
Uniform
as
a
special
case
and
is
more
general
as
the
distribution
of
a
proportion.
In
the
technical
documentation
the
user
should
be
cautioned
to
avoid
47
of
63
the
Normal
distribution
for
values
that
are
known
to
be
non­
negative
and
positively
skewed,
particularly
where
the
standard
deviation
is
over
half
of
the
mean.

One
Panel
member
expressed
reservations
about
the
use
of
a
normal
distribution
for
both
the
variability
and
the
uncertainty
about
the
mean
of
the
surface­
to­
hand
transfer
coefficient;
i.
e.
the
surface­
to­
hand
transfer
coefficient
among
children
is
assumed
to
follow
a
Normal
distribution
and
the
uncertainty
in
the
mean
of
that
distribution
is
also
described
by
a
normal
distribution.
The
Panelist
expressed
the
belief
that
this
Normal­
Normal
assumption
for
the
surface­
to­
hand
transfer
coefficient
could
lead
to
substantial
understatement
of
the
uncertainty
in
this
factor.
In
particular,
the
model
as
implemented
had
the
variance
of
the
mean
surface­
to­
hand
transfer
coefficient
less
than
the
variance
among
children
in
surface­
to­
hand
transfer
factor.
Given
that
the
variance
in
surface­
to­
hand
transfer
coefficients
is
limited
by
the
variability
in
hand
surface
area
among
children,
this
was
considered
highly
implausible."

During
previous
SAP
reviews
of
other
probabilistic
modeling
efforts
(
e.
g.,
CARES,
Lifeline),
Panel
members
have
commented
on
the
use
of
uniform
distributions.
Synopses
from
these
past
comments
follow:

°
Analysts
often
give
the
perceived
simplicity
of
the
uniform
distribution
as
an
important
attraction
for
cases
where
there
are
limited
empirical
data.
The
uniform
distribution,
with
its
defined
absolute
upper
and
lower
limits,
unfortunately
provides
an
opportunity
for
analysis
to
fall
into
a
trap
that
a
particular
parameter
has
zero
chance
of
having
values
outside
the
range
of
a
limited
available
data
set.
It
is
completely
incorrect
in
general
to
assume
that
the
largest
and
smallest
values
in
a
group
of
9­
30
data
points
or
fewer
represents
the
true
minimum
and
maximum
values
that
the
variable
can
assume.

°
Moreover
there
are
few
cases
where
the
mechanisms
that
cause
measurements
or
estimates
of
exposure­
related
parameters
to
vary
among
people
create
situations
where
there
is
no
greater
chance
of
producing
a
case
near
the
center
of
a
distribution
than
at
its
extreme
end
(
as
required
for
the
uniform
distribution
to
be
correct).
Factors
that
cause
exposure
to
differ
from
one
individual
to
another
tend
to
interact
multiplicatively
 
leading,
when
these
factors
are
numerous,
to
expectations
of
a
lognormal
distribution.
When
one
or
more
categorical
factors
are
likely
to
have
a
strong
influence
on
exposure
(
e.
g.,
wearing
short­
sleeved
vs.
long­
sleeved
shirts)
it
is
desirable
to
create
mixtures
of
lognormal
distributions,
weighted
by
their
expected
frequency,
to
represent
the
influence
of
those
different
known
cases.

°
The
uniform
distribution
is
appropriate
in
cases
where
(
1)
it
is
physically
impossible
for
the
parameter
to
take
on
values
outside
the
limits
and
(
2)
there
48
of
63
really
is
no
greater
likelihood
for
values
close
to
the
center
of
the
range
rather
than
at
either
end.
For
example,
there
would
be
no
problem
in
using
a
uniform
distribution
to
represent
the
day
of
the
week
that
a
meteor
might
land.
However,
as
many
of
the
applications
in
the
current
model
for
both
variability
and
uncertainty,
the
uniform
distribution
is
often
selected
in
cases
where
there
can
be
no
solid
assurance
that
the
parameter
cannot
take
on
values
outside
the
stated
range.
In
attempting
to
select
a
defined
absolute
range,
the
analyst
is
very
vulnerable
to
the
psychic
trap
of
"
overconfidence".
"
Overconfidence"
 
the
general
underestimation
of
uncertainty
(
assigning
confidence
limits
that
are
too
narrow)
is
one
of
the
best
documented
phenomena
in
risk
analysis.
This
applies
to
both
subjective
evaluations
by
experts
and
non­
experts
(
Tversky
and
Kahneman,
1974;
Alpert
and
Raiffa,
1982;
Lichstenstein
and
Fischoff,
1977),
and
to
supposedly
"
objective"
numerical
calculations
by
physicists
(
Shlyakhter
and
Kammen,
1992).

°
Hattis
and
Burmaster
(
1994)
gave
a
series
of
rules
and
examples
of
mechanisms
that
give
rise
to
different
distributional
forms.
Experience
and
the
basic
idea
that
variability
is
often
the
result
of
many
factors
acting
multiplicatively
indicates
that
the
lognormal
form
is
most
often
the
best
choice
for
exposurerelated
data
where
there
is
limited
information.
Both
normal
and
lognormal
distributions
have
just
two
parameters,
and
are
thus
no
more
"
complex"
statistically
than
a
uniform
distribution
(
and
in
that
sense,
less
complex
than
the
three­
parameter
triangular
distribution).
Derivation
of
the
parameters
of
lognormal
distributions
can
be
done
if
a
simple
range
is
given
together
with
the
number
of
independent
observations
that
gave
rise
to
that
range.
Means
and
other
measures
of
dispersion,
such
as
a
standard
deviation,
can
also
be
used
to
estimate
the
parameters
of
lognormal
distributions.

One
example
of
the
use
of
uniform
distributions
in
the
NMC
CRA
is
the
breathing
rate
distribution.
In
describing
the
breathing
rate
data,
the
document
says
(
p.
128)

"
Breathing
Rates:
The
breathing
rates
used
for
this
assessment
are
represented
by
a
uniform
distribution
from
1
to
2
m3/
hour
for
light
to
moderate
activity.
This
assumption
is
based
on
information
from
the
EFH
(
USEPA,
1997).
This
distribution
was
used
to
assess
exposure
for
all
age
groups."

In
general,
use
of
uniform
distributions
for
describing
inter­
individual
variability
should
be
discouraged.
Here,
for
comparison,
are
some
breathing
rate
distributions
collected
in
another
context,
direct
breathing
rate
measurements
from
activity
survey
data
of
coal
miners
(
Figures
1­
3)
and
a
general
population
of
adults
and
children
(
Figure
4).
It
can
be
seen
in
the
probability
plots
that
these
distributions
are
reasonably
described
using
normal
or
lognormal
distributions.
A
better
distribution
for
the
breathing
rates
in
the
NMC
CRA
might
be
to
combine
the
assumed
mean
breathing
rate
with
the
dispersion
from
the
observational
studies
depicted
in
these
figures.
49
of
63
Fig.
1
Distribution
of
Measured
Breathing
Rates
in
Working
Coal
Miners
Data
Source:
Jones,
C.
O.,
Gauld,
S.,
Hurley,
J.
F.,
and
Rickmann,
A.
M.
(
1981).
Personal
differences
in
the
breathing
patterns
and
volumes
and
dust
intakes
of
working
miners.
Report
to
the
Commission
of
the
European
Communities,
Report
No.
TM/
81/
11,
Environmental
Branch,
Institute
of
Occupational
Medicine,
Roxburgh
Place,
Edinburgh,
Scotland.
0
2
4
6
8
10
12
14
16
18
20
12
14
16
18
20
22
24
26
28
30
Mean,
Z
=
0
1
Std.
Dev.
1
Std.
Dev.
Z
=
+
1
Z
=
­
1
50
of
63
Figure
2
3
2
1
0
­
1
­
2
­
3
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
Fit
of
a
Lognormal
Distribution
to
the
Minute
Volumes
of
62
British
Coal
Miners
(
Data
of
Jones
et
al.,
1981)

Z­
Score
log(
Min.
Vol.)
y
=
1.3055
+
0.10273x
R^
2
=
0.962
Figure
3
3
2
1
0
­
1
­
2
­
3
10
20
30
40
Fit
of
a
Normal
Distribution
to
the
Minute
Volumes
Observed
in
62
British
Coal
Miners
(
Data
of
Jones
et
al.,
1981)

Z­
Score
Minute
Vol.
(
L)
y
=
20.760
+
4.6980x
R^
2
=
0.983
Source:
Hattis,
D.,
and
Silver,
K.,
"
Human
Interindividual
Variability­­
a
Major
Source
of
Uncertainty
in
Assessing
Risks
for
Non­
Cancer
Health
Effects,"
Risk
Analysis,
Vol
14,
pp.
421­
431,
1994.
51
of
63
Figure
4
3
2
1
0
­
1
­
2
­
3
2.0
2.2
2.4
2.6
2.8
3.0
Children
S
12
Yrs
Adolescents
and
Adults
Lognormal
Distributions
of
Cal­
EPA
Estimated
1­
Day
Activity­
Based
Breathing
Rates
Z­
Score
Log(
L/
kg
Body
Weight)
Estimated
Activity­
Based
Ventilation
Rates
y
=
2.66
+
.0618x
R^
2
=
0.983
y
=
2.36
+
0.117x
R^
2
=
0.940
In
summary,
the
Panel
agreed
that
it
is
probably
best
to
standardize
on
lognormal
distributions;
if
you
have
enough
information
to
pick
the
2
parameters
of
a
uniform,
you
should
be
able
to
pick
the
2
parameters
of
a
lognormal.
If
you
restrict
attention
to
2­
parameter
positively
skewed
distributions
on
the
positive
axis,
the
best­
known
distributions
(
lognormal,
gamma,
logistic
and
Weibull)
are
quite
similar
to
each
other
and
the
choice
will
not
affect
the
model
output
significantly.
A
small
sensitivity
analysis
comparing
uniform
with
lognormal
would
be
interesting.

Defaulting
to
the
lognormal
assumes
that
the
value
being
modeled
is
positivevalued
if
a
distribution
is
symmetric
and
long­
tailed,
a
shifted
and
scaled
t
on
low
degrees
of
freedom
(
3
parameters)
could
be
tried,
if
you
want
something
more
general
than
the
uniform
(
2
parameters)
you
could
use
the
beta
(
4
parameters).
If
the
uniform
distribution
is
used,
it
will
usually
be
a
good
idea
to
set
the
limits
a
bit
wider
than
the
observed
range
of
values.

Use
of
uniform
distributions
as
uncertainty
distributions
on
unknown
distributional
parameters
in
2­
D
Monte
Carlo
simulations
is
slightly
more
acceptable
but
still
to
be
discouraged.

The
Panel
was
not
aware
of
other
data
sources
better
suited
for
assessing
52
of
63
residential
exposure
scenarios.

R3c.
When
the
Agency
fits
distributions
to
various
exposure
values,
the
maximum
value
entered
into
the
probabilistic
models
for
a
particular
distribution
is
usually
defined
to
be
an
upper
percentile
value
such
as
the
99th
percentile
in
order
to
ensure
realistic
input
parameters.
Recognizing
that
the
Agency
intends
to
perform
sensitivity
analyses
to
evaluate
the
effects
of
this
truncation,
please
comment
on
the
Agency's
approach
of
truncating
distributions
that
are
input
to
the
probabilistic
models.
Please
comment
on
any
other
approaches
that
the
Agency
might
use
to
evaluate
uncertainties
associated
with
choices
about
whether
and
where
to
truncate
distributions.

Response
The
Panel
agreed
that
distributions
should
not
be
truncated
unless
there
is
a
strong
physical
or
biological
reason
to
set
an
upper
or
lower
limit.
Truncation
may
eliminate
only
1%,
say,
of
the
population,
but
it
may
be
the
most
interesting
1%.
Under
truncation,
the
means
and
standard
deviations
of
the
distributions
will
be
less
than
the
nominal
values
and
the
assumptions
of
the
simulation
will
not
be
quite
the
same
as
advertised.
Truncation
rules
need
to
be
set
out
in
the
documentation.
A
sensitivity
analysis,
comparing
output
with
and
without
truncation,
would
help
answer
this
question.

INTEGRATION
INTEGRATION
QUESTION
#
1
The
cumulative
risk
assessment
guidance
describes
key
principles
for
conducting
these
risk
assessments.
One
such
principle
is
the
need
to
consider
the
time
frame
of
both
the
exposure
(
e.
g.,
When
does
exposure
occur?
What
is
the
exposure
duration?)
and
of
the
toxic
effect
(
e.
g.,
What
are
the
time
to
peak
effects
and
the
time
to
recovery?
How
quickly
is
the
effect
reversed?).
EPA's
Preliminary
Cumulative
Risk
Assessment
for
the
N­
methyl
carbamates
describes
the
current
limitations
in
data
and
software
to
fully
characterize
the
dynamic
nature
of
exposure,
effect,
and
recovery
for
this
common
mechanism
group.
In
order
to
address
these
limitations,
OPP
performed
an
examination
of
the
exposure
patterns
for
records
from
the
high
end
of
exposure
distribution
and
found
that
that
a
large
fraction
(~
70%)
of
daily
records
contributing
to
the
upper
tail
of
the
food
exposure
distribution
represent
single
eating
occasions.
Regarding
drinking
water
and
residential/
non­
occupational
exposure,
EPA's
preliminary
assessment
provided
a
characterization
of
the
current
availability
regarding
datasets
and
models
and
a
description
of
the
impact
of
these
limitations
on
the
risk
estimates
from
specific
exposure
pathways
(
i.
e.,
drinking
water,
residential).

I1a.
Please
comment
on
clarity
and
adequacy
of
the
risk
characterization
provided
in
the
preliminary
cumulative
risk
assessment.
Are
there
important
53
of
63
aspects
with
respect
to
the
strengths
and
weaknesses
of
the
risk
characterization
other
than
the
ones
we
identified?

Response
The
consensus
of
the
Panel
was
that
the
cumulative
risk
analysis
document,
including
the
risk
characterization,
shows
enormous
progress
since
the
last
SAP
review
in
February
of
this
year.
Even
so,
the
next
round
would
benefit
from
some
additional
background
material
to
improve
its
accessibility
to
readers
less
familiar
with
the
basic
assumptions
and
features
of
the
analysis.
Some
Panelists
found
the
oral
presentation
of
the
risk
characterization
clearer
and
more
helpful
than
the
written
version
in
the
existing
document.
Finally,
one
Panelist
suggested
that,
in
light
of
the
specific
geographic
focus
of
the
high­
end
risk
analysis
involving
significant
ground
water
exposures,
it
might
be
more
appropriate
to
present
this
material
as
a
descriptive
scenario
rather
than
emphasizing
the
limited
geographic
derivation
of
the
underlying
data.

The
next
report
also
would
benefit
from
further
thought
about
the
application
of
the
relative
potency
factor
paradigm
to
a
group
of
AChE
inhibitors
that
inhibit
cholinesterase
on
time
scales
that
are
short
but
somewhat
different.
The
relatively
rapid
recovery
of
the
inhibition
and
the
consequently
short
time
unit
for
analysis
pose
an
enormous
challenge
for
risk
characterization.
With
daily
possibilities
for
exposure
and
inhibition
there
are
365
(
or
366)
opportunities
per
year
for
an
adverse
event
to
occur.
Therefore
a
risk
characterization
document
needs
to
discuss
and
explain
why
readers
should
focus
on
such
alternatives
as
(
A)
the
worst
day
experienced
by
any
individual
in
a
one­
year
period
(
B)
the
entire
life
stage
(
0­
2
years?
0­
20
years?)
when
there
might
be
unusual
developmental
susceptibility
or
(
C)
through
(
Z)
other
plausibly
relevant
exposure
and
risk
descriptors.
These
are
to
some
extent
risk­
management
judgments,
but
the
risk
characterization
should
be
designed
to
frame
and
clarify
the
information
that
the
risk
manager
and
the
public
might
reasonably
consider,
using
the
best
technical
insights
we
have
into
the
likely
dynamics
of
causation
and
the
relevant
dosimeters
for
adverse
effects.

Other
than
AChE
inhibition
itself,
the
acute
and
readily
apparent
health
effects
from
exposure
to
cholinesterase
inhibitors
go
under
the
acronym
SLUD.
For
these
acute
effects,
experience
indicates
that
a
BMD10
for
brain
cholinesterase
inhibition
is
a
conservative
(
health
protective)
value.
Moreover,
because
of
the
short
time
between
cholinesterase
inhibition
and
the
manifestation
of
these
signs,
it
is
reasonable
to
consider
that
peak
cholinesterase
inhibition
levels
are
the
causally
relevant
measure
of
internal
dose
for
modeling
the
risk
of
adverse
responses
in
individuals.

However,
readily
observable
high­
dose
effects
are
not
the
only,
or
even
the
most
important
responses
of
concern
for
population
exposures
to
cholinesterase
inhibitors.
As
recently
reviewed
by
Slotkin
(
2004),
cholinergic
signaling
plays
a
vital
role
in
several
phases
of
neurodevelopment
including
the
migration
of
the
cells
that
will
become
mature
neurons
to
the
locations
in
the
brain
(
and,
likely,
elsewhere)
where
they
are
needed;
the
formation
of
connections
with
other
neurons,
and
the
survival
of
the
connections
54
of
63
(
synapses)
through
phases
where
unused
connections
are
pared
back
and
lost.
It
is
possible
that
even
subtle
strengthening
of
the
signaling
via
some
cholinergic
synapses
will
lead
to
survival
of
some
connections
in
preference
to
other
non­
cholinergic
pathways,
and
therefore
have
subtle
long
term
consequences
for
function.
This
theory
is
based
on
current
understanding
of
fundamental
processes
of
neurodevelopment.
For
representative
research
articles
and
reviews
on
this
large
topic,
see
Lauder
(
1985),
Whitaker­
Azmitia
(
1991),
Hohmann
and
Berger­
Sweeney
(
1998),
Lauder
and
Schambra
(
1999),
Weiss
et
al.
(
1998).
We
now
know
that
"
neurodevelopment"
of
this
type
is
not
restricted
to
fetal
life,
but
continues
well
after
birth.
In
fact,
there
is
evidence
that
synaptic
rearrangement,
as
well
as
the
proliferation
and
planned
death
of
neurons,
continues
well
into
adolescence
in
rats
(
Bayer
et
al.,
1982;
Bayer,
1983)
and
also
humans
(
Huttenlocher,
1990).

Long­
term
health
effects
may
also
result
from
the
adaptation
of
cholinergic
signaling
systems
to
cholinesterase
inhibition.
There
are
several
recent
reports
of
unexpected
long­
term
effects
from
military
and
agricultural
anticholinesterase
agents,
in
some
cases
when
exposures
were
insufficient
to
induce
acute
cholinergic
signs
(
Baker
and
Sedgwick,
1996;
Kelly
et
al.,
1997;
Kassa
et
al.
2004;
Tochigi
et
al.
2002;
Abu­
Qare
and
Abou­
Donia,
2002;
Yokoyama
et
al.
1998;
Sanchez­
Santed
et
al.
2004;
Jamal
et
al.
2002).
Long­
term
animal
studies
with
military
nerve
agents
have
recently
led
to
suggestions
of
a
need
to
revise
LOAELs
determined
on
the
basis
of
short­
term
experiments
(
VanHelden
et
al.
2003;
2004),
even
though
the
short­
term
experiments
utilized
quite
a
mild
effect
(
changes
in
pupil
size)
as
the
measure
of
response.
Neuroscientists
on
the
Panel
considered
that
longer­
term
adaptive
responses
are
unlikely
unless
cholinesterase
inhibition
reaches
levels
that
alter
synaptic
physiology
by
overriding
the
margin
of
safety
for
cholinergic
transmission.
The
Panel
finds,
however
that
the
concern
for
subtle
developmental
and
adaptive
effects
does
warrant
further
discussion
in
the
risk
characterization,
to
help
decision­
makers
and
the
public
put
the
Agency's
choice
of
benchmark
doses
in
perspective.

Suspected
developmental
and
adaptive
effects
might
be
more
directly
dependent
on
a
time­
weighted
integral
of
cholinesterase
inhibition
than
on
peak
inhibition
levels
on
specific
days.
There
is
a
relatively
straightforward
way
that
EPA
can
use
information
it
has
already
developed
to
perform
an
alternative
set
of
exposure
assessments
based
on
this
"
Area
Under
the
Curve"
of
cholinesterase
inhibition
dosimeter.
That
is,
the
Agency
can
do
a
parallel
set
of
exposure
analyses
using
an
alternative
set
of
Relative
Potency
Factors
that
incorporate
both
the
potency
of
each
carbamate
for
producing
peak
inhibition,
and
the
rate
at
which
that
inhibition
is
reversed
(
at
the
lowest
available
doses
in
the
experiments
already
analyzed
by
EPA).

In
rats,
the
estimated
reversal
half­
lives
for
inhibition
by
carbamates
vary
widely
and
the
associated
95%
confidence
limits
suggest
statistical
significance
among
some
pairs
(
Table
1.
B.
6).
In
particular,
the
reversal
half­
life
of
the
proposed
index
chemical,
oxamyl
(
0.75h
with
tight
confidence
limits
of
0.66h­
0.88h)
is
much
shorter
than
that
of
formetanate
(
4.05h,
with
confidence
limits
of
3.02h­
5.44h)
and
methiocarb
(
2.77h,
with
55
of
63
confidence
limits
of
1.91h­
4.01h).

Table
1
below
illustrates
a
simple
approach
to
construct
an
alternative
set
of
Relative
Potency
Factors
(
these
might
be
designated
"
RPF*")
using
a
preliminary
index
of
time­
integrated
relative
potencies.
The
index
would
simply
consist
of
the
product
of
the
BMD­
based
RPFs
and
the
individual
chemical
half­
lives
for
cholinesterase
inhibition
reversal
 
renormalized
so
that,
as
before,
the
value
for
the
index
chemical
is
set
at
1.
Because
oxamyl
has
the
shortest
half­
life
for
inhibition
reversal,
this
approach
tends
to
increase
the
relative
potency
factors
for
all
the
other
compounds
 
by
over
7­
fold
in
the
case
of
formetanate.
Table
1
Preliminary
Calculation
of
Time­
Integrated
Relative
Potency
Factors
(
RPF*)
Peak
BMDBased
RPF
Brain
Inhib
Reversal
T1/
2
(
hr)
BMD*
T1/
2
Renormalized
RPF*
Relative
to
Oxamyl
Aldicarb
3.32
1.52
5.05
6.73
Carbaryl
(
0­
10
dose)
0.12
1.83
0.22
0.29
Carbofuran
1.19
2.49
2.96
3.95
Formetanate
1.89
5.4
10.21
13.6
Methiocarb
0.14
2.77
0.39
0.52
Methomyl
0.38
0.8
0.30
0.41
Oxamyl
1
0.75
0.75
1.00
Primicarb
0.02
1.90a
0.038a
0.051
Propoxur
0.09
2.69
0.24
0.32
Thiodicarb
0.7
1.90a
1.33a
1.77
aFor
compounds
with
missing
values
for
the
inhibition
reversal
half­
life,
the
geometric
mean
of
the
half
lives
for
other
compounds
(
1.90)
has
been
substituted
in
the
calculation.

Some
mechanistic
reasoning
and
modeling
could
lead
to
modest
further
modifications
to
the
RPF*
calculations
when
translated
into
human
equivalents.
All
of
the
N­
methyl
carbamates
are
expected
to
leave
behind
the
same
chemical
moiety
on
AChE.
Purely
spontaneous
chemical
reversal
of
the
inhibition
therefore
is
expected
to
occur
at
an
identical
rate
across
different
carbamates
in
the
common
mechanism
group,
and
should
at
least
be
similar
across
species
from
rats
to
humans.
Data
to
confirm
this
expectation
would
not
be
difficult
to
obtain.

Given
the
similar
behavior
of
the
N­
methyl
carbamates
at
the
level
of
the
target
enzyme,
the
differences
in
their
inhibition
reversal
half­
lives
are
likely
to
reflect
differences
in
their
rates
of
clearance
from
the
body
by
excretion
and
metabolism.
To
the
degree
that
a
particular
chemical's
brain
inhibition
reversal
does
depend
on
this
kind
of
slow
clearance,
then
the
ordinary
scaling
principles
for
metabolism­
based
pharmacokinetics
(
Travis
et
al.
1990;
Boxenbaum,
1980
and
1982;
Reese
and
Hattis,
1994)
suggest
that
the
clearance
will
be
slower
in
people
 
according
to
convention,
roughly
in
proportion
to
Body
Weight
raised
to
the
0.25
power.
Thus,
the
baseline
expectation
is
that
this
kind
of
56
of
63
pharmacokinetic
clearance
should
be
about
4
times
slower
in
a
70
kg
person
than
in
a
0.3
kg
rat:
(
70/
0.3)
0.25
=
4.
To
translate
the
rat
brain
inhibition
reversal
half
lives
to
human
equivalents,
therefore,
a
factor
of
4
might
be
applied
to
the
chemicals
whose
brain
inhibition
reversal
rate
is
markedly
slower
than
the
in
vitro
reactivation
half­
lives
that
may
be
observed
in
experiments
now
under
way.
In
cases
where
the
estimated
brain
AChE
inhibition
reversal
half­
lives
are
similar
to
those
measured
at
the
appropriate
temperature
in
vitro,
however,
no
such
interspecies
projection
factor
should
be
applied.
If
modeling
studies
indicate
that
the
processes
limiting
the
rate
of
inhibition
reversal
can
be
apportioned
between
the
spontaneous
chemical
processes
and
clearance
from
the
body
reservoir(
s),
then
of
course
the
interspecies
slowing
factor
should
only
be
applied
to
the
active
metabolic
clearance
process.
This
circumstance
could
therefore
eventually
lead
to
some
differential
adjustment
of
the
RPF*
across
species
[
and,
by
extension,
across
age
groups
for
the
very
young
and
the
very
old
whose
clearance
rates
tend
to
be
less
than
those
for
young
adults
(
Ginsberg
et
al.
2002
and
2005).
Meanwhile,
these
considerations
suggest
that
the
in
vivo
inhibition
half­
lives
for
some
N­
methyl
carbamates
might
be
long
enough
to
call
into
question
a
basic
assumption
of
the
proposed
cumulative
risk
assessment
for
this
common
mechanism
group.
In
particular,
if
one
applies
a
4.1­
fold
inter­
species
scaling
factor
to
the
5.4
hr
half­
time
for
reversal
of
brain
AChE
inhibition
in
rats,
one
obtains
a
predicted
half­
time
of
22
hr
in
the
70
kg
human
adult.
Such
a
long
half­
time
would
force
the
risk
assessment
model
to
address
carryover
of
inhibition
from
one
day
to
the
next.
In
considering
this
issue,
the
Agency
should
take
into
account
cases
where
there
is
a
dose
dependency
for
inhibition
reversal
half­
lives.
In
these
cases,
projections
should
utilize
the
estimates
from
the
lowest
feasible
dose
rate
because
it
may
be
most
relevant
to
the
expectations
for
exposure
at
BMD10.

I1b.
Is
the
Panel
aware
of
additional
data
which
would
aid
the
Agency
in
its
cumulative
risk
characterization
for
the
N­
methyl
carbamate
pesticides?
For
example,
is
the
Panel
aware
of
any
available
data
on
the
timing
of
water
consumption
events
or
can
the
Panel
make
any
recommendations
regarding
reasonable
assumptions
that
could
be
made
to
help
characterize
the
estimated
risk?
Are
there
other
sensitivity
analyses
and
further
investigations
that
would
be
equally
or
more
important
than
the
ones
we
identified?

Response
In
response
to
this
question,
and
because
of
the
clear
importance
of
the
local
drinking
water
pathway
to
the
analysis,
the
Panel
has
undertaken
some
very
simple
pharmacokinetic
modeling
on
this
subject.
There
is
good
reason
to
suspect
that
the
current
method
of
analysis
 
lumping
all
daily
exposure
into
a
single
event
 
introduces
a
systematic
distortion
in
the
expected
effects
on
peak
inhibition
levels
between
dietary
exposure
(
which,
for
upper
percentiles,
seem
to
be
mostly
traceable
to
single
eating
events
per
day)
vs.
drinking
water,
which
clearly
occurs
in
several
different
events
distributed
throughout
the
day.
The
Panel
has
asked
itself,
how
might
the
RPFs
for
peak
brain
AChE
57
of
63
inhibition
be
adapted
to
accommodate
a
likely
scenario
for
time­
dependent
water
consumption?

The
following
analysis
is
based
on
what
the
Panel
understands
to
be
one
plausible
pattern
of
diurnal
drinking
water
consumption
that
the
EPA
uses
for
pharmacokinetic
analyses.
This
pattern
consists
of
three
water
consumption
events
(
each
delivering
25%
of
daily
consumption)
at
meal
times
separated
by
5­
hour
intervals,
and
two
between­
meal
drinking
events,
each
delivering
12.5%
of
daily
consumption.
Figure
5
compares
the
results
of
this
consumption
scenario
for
expected
peak
inhibition
levels
in
rats
following
water
intake
of
BMD10
amounts
of
the
NMC
with
the
longest
inhibition
reversal
half
life
(
formetanate)
and
the
NMC
with
the
shortest
half
life
(
oxamyl).
(
The
Panel
has
not
attempted
to
incorporate
further
adjustments
to
translate
results
into
half­
lives
for
general
and
special
human
subpopulations).

Figure
5
shows
that
this
hypothesized
pattern
of
drinking
water
exposure
does
indeed
lead
to
quite
different
expectations
for
peak
daily
cholinesterase
inhibition
for
the
daily
drinking
water
doses
of
the
two
NMCs.
The
5.4
hour
half
life
for
formetanate
leads
to
appreciable
buildup
during
the
day
to
about
5.9%
inhibition,
whereas
the
predicted
peak
inhibition
for
oxamyl
is
only
a
little
more
than
the
2.5%
expected
from
each
mealtime
drinking
event
considered
separately.
If
inhibition
did
not
reverse
between
events,
of
course,
or
if
the
total
daily
dose
were
delivered
in
one
bolus,
the
expected
peak
inhibition
would
be
10%.
Table
2
illustrates
how
this
finding
can
be
translated
into
a
simple
numerical
adjustment
to
the
Relative
Potency
Factors.

Figure
5
58
of
63
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0
1
2
3
4
5
6
Expected
Time
Pattern
of
Brain
Acetylcholinesterase
Inhibition
for
a
Drinking
Water
Pattern
of
Exposure
to
Formetanate
(
Inhibition
T1/
2
=
5.4
Hours)
vs
Oxamyl
(
Inhibition
T1/
2
=
0.75
Hours)

Hours
After
Start
of
Exposure
(
Breakfast)
%
Brain
ACHE
Inhibition
Maximum
=
5.89%

Maximum
=
2.65%
Formetanate
Oxamyl
Table
2
Indicated
Relative
Potencies
for
Formetanate
vs.
Oxamyl
for
Maximal
Daily
Inhibition
for
a
Drinking
Water
Pattern
of
Exposure
Half
Life
(
hr)
Maximum
%
Inhibition
for
BMD10
Exposure
Formetanate
5.4
5.89
Oxamyl
0.75
2.65
Ratio
7.2
2.23
In
Table
2,
the
2.2­
fold
upward
adjustment
of
the
formetanate
RPF
relative
to
oxamyl
is
significant,
but
less
than
the
full
ratio
of
the
two
half
lives
that
would
be
indicated
if
AUCs
were
the
desired
causal
dosimeter
for
a
particular
type
of
toxic
response.
Alternatively,
the
overall
RPF
for
oxamyl
itself
in
drinking
water
might
be
adjusted
downward
to
.265
of
the
RPF
for
oxamyl
in
the
single
oral
doses
expected
for
most
high­
percentile
dietary
exposures.
59
of
63
In
further
development
of
this
approach,
EPA
should
make
use
of
any
reliable
source
of
relevant
empirical
data
on
daily
patterns
of
drinking
water
consumption;
ideally
adapted
to
the
likely
consumption
behavior
in
specific
regions
or
smaller
areas
of
the
country.

Other
comments
on
the
analysis:

The
discontinuity
that
is
apparent
for
some
exposure
routes
between
the
end
of
the
year
vs.
the
beginning
of
the
year
needs
to
be
resolved.
It
cannot
be
true
that
all
pet
collars
are
really
applied
on
Jan
1;
a
more
random
day
needs
to
be
chosen
for
this
and
similar
modes
of
exposure,
with
allowance
for
cross­
year
exposures
as
needed
to
fully
represent
realistic
patterns
for
the
start
of
an
exposure
event
vs.
the
day
of
actually
delivered
exposures.

Institutional
(
e.
g.,
school)
and
occupational
exposures
should
be
incorporated
into
the
assessment.

It
will
become
important
to
supplement
whole­
and
half­
brain
data
with
measurements
of
causally
relevant
amounts
and
durations
of
AChE
inhibition
in
brain
regions
that
are
mechanistically
connected
to
specific
developmental
and
other
effects.
Brain
regions
differ
in
basal
levels
of
AChE,
and
may
therefore
differ
somewhat
in
sensitivity
to
inhibition.
Whole­
brain
and
half­
brain
measurements
are
considered
by
some
in
the
field
to
be
the
"
wave
of
the
past".
On
the
other
hand,
data
on
region­
specific
inhibition
of
AChE
are
currently
quite
limited,
and
it
is
recognized
that
such
measurements
are
associated
with
much
higher
variability
than
those
from
whole
brain.
For
these
reasons
there
was
no
Panel
consensus
in
favor
of
incorporating
regional
studies
at
this
time.

One
Panelist
considered
that
the
non­
dietary
oral
exposures
were
likely
to
be
much
more
uncertain
than
other
sources
of
exposure.
If
the
EPA
agrees,
the
next
document
might
discuss
the
relative
confidence
of
the
analysis
in
the
quantification
of
exposures
by
various
routes,
and
consequent
implications
for
research
and
risk
management
priorities.

Another
topic
for
discussion
in
the
next
iteration
of
the
document
is
the
fact
that
areas
of
the
country
were
chosen
where
carbamate
residential
exposure
and
use
is
likely
to
be
higher
than
that
in
the
rest
of
the
country
because
of
greater
pest
pressures,
among
other
circumstances.
This
may
suggest
a
scenario
presentation
rather
than
the
implication
of
a
full
US­
South
regional
analysis.
Other
suggestions
by
individual
Panelists
included
subpopulation
analyses
by
ethnic
groups,
groups
with
different
dietary
habits
(
e.
g.,
vegetarians)
and
other
categorizations
of
people
that
might
be
associated
with
differences
in
exposures.
Such
categorizations
might
ultimately
be
helpful
in
formulating
options
for
information
programs
and
other
risk
management
efforts.
60
of
63
REFERENCES
HAZARD
Hammond,
P.,
T.
Jelacic,
S.
Padilla,
and
S.
Brimijoin.
1996.
Quantitative,
video­
based
histochemistry
to
measure
regional
effects
of
anticholinesterase
pesticides
in
rat
brain.
Anal.
Biochem.
241:
82­
92.

WATER
(
In
response
to
question
W1)

Liu
G.,
S.
Dai,
Y.
Qian,
and
Q.
Gan.
2003.
Experimental
Study
on
Effect
of
Anion
Surfactant
on
Degradation
Rate
of
Aldicarb
in
Soil.
Journal
of
Environmental
Science
and
Health,
Part
B:
Pesticides,
Food
Contaminants,
and
Agricultural
Wastes
38:
405­
416.

Ou,
L.
T.,
P.
S.
C.
Rao,
K.
S.
V.
Edvardsson,
P.
E.
Jessup,
and
A.
G.
Hornsby.
1988.
Aldicarb
degradation
in
sandy
soils
from
different
depths.
Pesticide
Science
23
(
1):
1­
12.

Steenhuis,
T.
S.,
S.
Pacenka,
and
K.
S.
Porter.
1987.
MOUSE:
A
Management
Model
for
Evaluating
Groundwater
Contamination
from
Diffuse
Surface
Sources
Aided
by
Computer
Graphics.
Appl.
Agric.
Res.
2:
277­
289.

(
In
response
to
question
W2)

Darnault,
C.
J.
D.,
T.
S.
Steenhuis,
P.
Garnier,
Y.­
J.
Kim,
M.
B.
Jenkins,
W.
C.
Chiorse,
P.
C.
Baveye,
and
J.
Y.
Parlange.
2004.
Preferential
flow
and
transport
of
Cryptosporidium
parvum
oocysts
through
the
vadose
zone:
Experiments
and
modeling.
Vadose
Zone
J.
3:
262­
270.

Glass,
R.
J.,
G.
H.
Oosting,
and
T.
S.
Steenhuis.
1989.
Preferential
Solute
Transport
in
Layered
Homogeneous
Sands
as
a
Consequence
of
Wetting
Front
Instability.
J.
Hydrol.
110:
87­
105.

Jarvis,
N.
J.,
M.
Stahli,
L.
Bergstrom,
and
H.
Johnsson.
1994.
Simulation
of
dichlorprop
and
bentazon
leaching
in
soils
of
contrasting
texture
using
the
MACRO
model.
J.
Environ.
Sci.
Health,
A
29(
6):
1255­
1277.

Kim,
Y.­
J.,
C.
J.
G.
Darnault,
N.
O.
Bailey,
J.­
Y.
Parlange,
and
T.
S.
Steenhuis.
2005.
Equation
for
Describing
Solute
Transport
in
Field
Soils
with
Preferential
Flow
Paths.
Soil
Science
Society
of
America
J.
69:
291­
300.

Kung,
K­
J.
S.
1990.
Preferential
flow
in
a
sandy
vadose
zone:
1.
field
observation.
61
of
63
Geoderma
46:
51­
58.

Sadeghi,
A.
M.,
A.
R.
Isensee,
and
A.
Shirmohammadi.
2000.
Influence
of
soil
texture
and
tillage
on
herbicide
transport.
Chemosphere
41:
1327­
1332.

Selker,
J.
S.,
T.
S.
Steenhuis,
and
J.­
Y.
Parlange.
1996.
An
engineering
approach
to
fingered
vadose
pollutant
transport.
Geoderma
70:
197­
206.

Shipitalo,
M.
J.,
W.
M.
Edwards,
and
W.
A.
Dick.
1990.
Initial
storm
effects
on
macropore
transport
of
surface­
applied
chemicals
in
no­
till
soil.
Soil
Science
Society
of
America
Journal
54(
6):
1530­
1536.

Singh,
P.,
R.
Kanwar,
K.
E.
Johnsen,
and
L.
R.
Ahuja.
1996.
Calibration
and
evaluation
of
subsurface
drainage
component
of
RZWQM
V.
2.5.
Journal
of
Environmental
Quality
25(
1):
56­
63.

Steenhuis,
T.
S.,
C.
Jackson,
K.­
J.
S.
Kung,
and
W.
H.
Brutsaert.
1985.
Measurement
of
Groundwater
Recharge
on
Eastern
Long
Island.
J.
Hydrol.
79:
145­
169.

Steenhuis,
T.
S.
and
W.
H.
van
der
Molen.
1986.
The
Thornthwaite­
Mather
Procedure
as
a
Simple
Engineering
Method
to
Predict
Recharge.
J.
Hydrol.
84:
221­
229.

RESIDENTIAL
(
In
response
to
question
R3b)

Alpert,
M.
and
H.
Raiffa.
1982.
A
progress
report
on
the
training
of
probability
assessors.
in
Judgment
Under
Uncertainty,
Heuristics
and
Biases,
D.
Kahneman,
P.
Slovic,
and
A.
Tversky,
eds.,
Cambridge
University
Press.
N.
Y.
pp.
294­
305.

Hattis,
D.
and
D.
E.
Burmaster.
1994.
Assessment
of
variability
and
uncertainty
distributions
for
practical
risk
analyses.
Risk
Analysis
14:
713­
730.

Lichtenstein
S.
and
B.
Fischoff.
1977.
Do
those
who
know
more
also
know
more
about
how
much
they
know?
Organizational
Behavior
and
Human
Performance
20:
159­
183.

Shlyakhter,
A.
I.
and
D.
M.
Kammen.
1992.
Sea­
level
rise
or
fall?"
Nature
253:
25.

Tversky
A.
and
D.
Kahneman.
1974.
Judgment
under
uncertainty:
Heuristics
and
biases.
Science
185,
1124­
1131,
In:
Judgment
Under
Uncertainty:
Heuristics
and
Biases,
Edited
by:
D.
Kahneman,
P.
Slovic
and
A.
Tversky.
Cambridge
University
Press.
N.
Y.
1982
pp.
3­
20.

INTEGRATION
62
of
63
Abou­
Donia,
M.
B.
2003.
Organophosphorus
ester­
induced
chronic
neurotoxicity.
Arch.
Environ.
Health.
58(
8):
484­
497.

Abu­
Qare,
A.
W.
and
M.
B.
Abou­
Donia.
2002.
Sarin:
health
effects,
metabolism,
and
methods
of
analysis.
Food
Chem.
Toxicol.
40(
10):
1327­
1333.

Baker,
D.
J.,
and
E.
M.
Sedgwick.
1996.
Single
fibre
electromyographic
changes
in
man
after
organophosphate
exposure.
Human
&
Experimental
Toxicol.
15:
369­
375.

Bayer,
S.
A.,
1983.
[
3H]
Thymidine­
radiographic
studies
of
neurogenesis
in
the
rat
olfactory
bulb.
Exp.
Brain
Res.
50:
329­
340.

Bayer,
S.
A,
J.
W.
Yackel,
and
P.
S.
Puri.
1982.
Neurons
in
the
rat
dentate
gyrus
granular
layer
substantially
increase
during
juvenile
and
adult
life.
Science
216:
890­
892.

Boxenbaum
H.
1980.
Interspecies
variation
in
liver
weight,
hepatic
blood
flow,
and
antipyrine
intrinsic
clearance:
extrapolation
of
data
to
benzodiazepines
and
phenytoin.
J.
Pharmacokinet.
Biopharm.
8(
2):
165­
176.

Boxenbaum,
H.
1982.
Interspecies
scaling,
allometry,
physiological
time,
and
the
ground
plan
of
pharmacokinetics.
J.
Pharmacokinet.
Biopharm.
10:
201­
227.

Ginsberg,
G.,
D.
Hattis,
B.
Sonawane,
A.
Russ,
P.
Banati,
M.
Kozlak,
S.
Smolenski,
and
R.
Goble.
2002.
Evaluation
of
Child/
Adult
Pharmacokinetic
Differences
from
a
Database
derived
from
the
Therapeutic
Drug
Literature.
Toxicological
Sciences
66:
185­
200.

Ginsberg,
G.,
D.
Hattis,
A.
Russ,
and
B.
Sonawane.
"
Pharmacokinetic
and
Pharmacodynamic
Factors
that
can
Affect
Sensitivity
to
Neurotoxic
Sequelae
in
the
Elderly."
Environmental
Health
Perspectives
doi:
10.1289/
ehp.
7568
(
available
at
http://
dx.
doi.
org/)
Online
26
May
2005,
in
press.

Hohmann,
C.
F.
and
J.
Berger­
Sweeney.
1998.
Cholinergic
regulation
of
cortical
development
and
plasticity:
new
twists
to
an
old
story.
Perspect.
Dev.
Neurobiol.
5:
401­
425.

Huttenlocher,
P.
R.
1900.
Morphometric
study
of
human
cerebral
cortex
development.
Neuropsychologia
28:
517­
527.

Jamal
G.
A.,
S.
Hansen,
and
P.
O.
Julu.
2002.
Low
level
exposures
to
organophosphorus
esters
may
cause
neurotoxicity.
Toxicology
181­
182:
23­
33.

Kassa
J.,
G.
Krejcová,
F.
Skopec,
J.
Herink,
J.
Bajgar,
L.
Sevelová,
M.
Tichý,
and
M.
Pecka.
2004.
The
influence
of
sarin
on
various
physiological
functions
in
rats
following
single
or
repeated
low­
level
inhalation
exposure.
Inhal.
Toxicol.
16(
8):
517­
30.
63
of
63
Kelly,
S.
S.,
G.
E.
de
Blaquière,
F.
M.
Williams,
and
P.
G.
Blain.
1997.
Effects
of
multiple
doses
of
organophosphates
on
evoked
potentials
in
mouse
diaphragm.
Hum.
Exp.
Toxicol.
16(
2):
72­
78.

Lauder,
J.
M.
and
U.
B.
Schambra.
1999.
Morphogenetic
roles
of
acetylcholine.
Environ.
Health
Perspect.
107
Suppl.
1:
65­
69.

Rees,
D.
C.
and
D.
Hattis.
"
Developing
Quantitative
Strategies
for
Animal
to
Human
Extrapolation"
Chapter
8
in
Principles
and
Methods
of
Toxicology,
3rd
Edition,
A.
W.
Hayes,
ed.,
Raven
Press,
New
York,
1994,
pp.
275­
315.

Sánchez­
Santed
F.,
F.
Cañadas,
P.
Flores,
M.
López­
Grancha,
and
D.
Cardona.
2004.
Long­
term
functional
neurotoxicity
of
paraoxon
and
chlorpyrifos:
behavioural
and
pharmacological
evidence.
Neurotoxicol.
Teratol.
26(
2):
305­
17.

Tochigi
M.,
T.
Umekage,
T.
Otani,
T.
Kato,
A.
Iwanami,
N.
Asukai,
T.
Sasaki,
and
N.
Kato.
2002.
Serum
cholesterol,
uric
acid
and
cholinesterase
in
victims
of
the
Tokyo
subway
sarin
poisoning:
a
relation
with
post­
traumatic
stress
disorder.
Neurosci.
Res.
44(
3):
267­
272.

Travis,
D.
C.,
R.
K.
White,
and
R.
C.
Ward.
1990.
Interspecies
extrapolation
of
pharmacokinetics.
J.
Theor.
Biol.
142:
285­
304.

Weiss,
E.
R.,
P.
Maness,
and
J.
M.
Lauder.
1998.
Why
do
neurotransmitters
act
like
growth
factors?
Perspect.
Dev.
Neuobiol.
5:
323­
335.

Whitaker­
Azmitia,
P.
M.
1991.
Role
of
serotonin
and
other
neurotransmitter
receptors
in
brain
development:
basis
for
developmental
pharmacology.
Pharmacol.
Rev.
43:
553­
561.

Yokoyama
K.,
S.
Araki,
K.
Murata,
M.
Nishikitani,
T.
Okumura,
S.
Ishimatsu,
and
N.
Takasu.
1998.
A
preliminary
study
on
delayed
vestibulo­
cerebellar
effects
of
Tokyo
Subway
Sarin
Poisoning
in
relation
to
gender
difference:
frequency
analysis
of
postural
sway.
J.
Occup.
Environ.
Med.
40(
1):
17­
21.
