1
11/
26/
02
Response
to
Scientific
Advisory
Panel
Comments
on
Applicability
of
the
Up­
and­
Down
Procedure
Methodology
for
Acute
Oral
Toxicity
Testing
Comment
1:
The
Panel
concluded
that
the
Up­
and
Down
Procedure
(
UDP)
assay
can
be
fairly
and
readily
conducted
inasmuch
as
any
laboratory
that
can
conduct
a
traditional
LD50
test
can
conduct
the
UDP.
Laboratory
and
animal
issues
unique
to
the
UDP
for
acute
testing
include:
(
i)
individual
animal
dosing
and
prolonged
study
time
utilizes
laboratory
space
inefficiently
and
adds
cost;
and
(
ii)
maintenance
of
animal
weights
and
age
for
the
duration
of
the
study
is
more
difficult
and
could
necessitate
additional
shipments
of
animals;
and
(
iii)
special
methods
may
be
needed
to
ensure
accurate
dosing
and
dose
dilution
over
time.

Agency
response:
The
Agency
acknowledges
that
these
issues
require
careful
planning
on
the
part
of
laboratories
performing
the
UDP.
However,
the
UDP
guideline
meets
animal
welfare
goals;
it
calls
for
a
reduced
number
of
test
subjects
and
incorporates
OECD
Humane
Practices
for
animal
treatment.

Comment
2:
Since
the
guideline
calls
for
use
of
a
single
sex,
in
order
to
minimize
population
variance,
it
is
important
that
the
most
sensitive
sex
be
selected.
In
the
majority
of
cases,
females
are
more
sensitive
than
males
and
would
be
the
default
choice
in
the
absence
of
information
to
the
contrary
(
e.
g.,
past
experience
with
similar
chemicals,
knowledge
of
detoxification
kinetics,
etc.).

Agency
response:
As
noted
in
the
text
of
the
guideline,
the
emphasis
is
on
utilization
of
all
available
information
on
related
materials
and
mode
of
toxic
information
in
order
to
determine
which
sex
is
likely
to
be
most
sensitive.
Therefore,
toxicologists
should
always
carefully
consider
selection
of
animal
sex
and
selection
of
the
most
sensitive
sex
should
also
be
justified
in
test
reports.

Comment
3:
An
age
range
of
8­
12
weeks
for
test
animals,
while
giving
needed
flexibility
for
the
staggered
schedule
of
this
study,
may
introduce
variability
because
8
week
old
rats
are
still
in
their
rapid
growth
phase
and
may
be
different
in
dose
response
than
ones
of
12
weeks.
Therefore,
the
Panel
recommended
use
of
a
narrower
age
interval
of
9
­
11
weeks.

Agency
response:
Although
a
narrower
age
limit
may
be
desirable,
the
Agency
selected
8
­
12
weeks
to
harmonize
with
OECD
practice
for
acute
testing.

Comment
4:
The
Panel
agreed
that
histopathology
would
add
no
useful
information.
The
suggestion
that
histopathology
be
considered
in
study
design
was
viewed
as
adding
no
useful
information.
More
useful
pathology
will
be
obtained
from
studies
of
longer
duration,
e.
g.
28­
and
90­
day
studies.
2
Agency
response:
Histopathology
is
not
required
in
the
guideline,
but
if
it
is
performed,
a
standard
reporting
format
is
identified.

Comment
5:
The
Panel
agreed
that
selection
of
starting
doses
should
not
be
a
problem
for
informed
toxicologists.
The
Panel
also
recognized
that
reducing
step
size
[
too
far]
might
result
in
too
narrow
a
range
of
doses
and
give
a
poor
LD50
estimate
and
suggested
that
additional
information
could
be
incorporated
in
a
Bayesian
analysis
as
an
alternative
to
the
profile
likelihood
analysis
proposed.
The
Panel
noted
that
if
additional
[
intermediate]
doses
are
added
after
selection
of
initial
sigma
and
dose
progression,
the
software
calculates
the
LD50
and
profile
likelihood.
Calculation
of
the
confidence
interval
remains
valid.

Agency
response:
Simulations
show
that
the
UDP
works
best
when
all
available
information
is
brought
to
bear
when
selecting
starting
dose
and
dose
progression.
If
initial
choice
of
sigma
and
dose
progression
appears
to
be
too
wide,
the
software
can
accept
intermediate
doses
and
will
calculate
an
acceptable
LD50s,
but
the
confidence
interval
provided
will
not
be
accurate,
posing
significant
problems
in
interpretation
of
test
results.

Flexible
stopping
rules
tend
to
compensate
if
choice
of
step
size
is
too
narrow.
Sequential
frequentist
and
Bayesian
analyses
are
a
new
direction
for
acute
testing
analyses
in
general,
and
the
statistical
community
is
being
encouraged
to
explore
how
the
design
and
estimation
may
be
improved
by
their
consideration.
Incorporation
of
additional
information,
as
through
use
of
Bayesian
statistical
analysis,
may
be
feasible
in
the
future
as
the
science
develops.

Comment
6:
The
Panel
agreed
with
the
48
hour
dosing
interval.

Agency
response:
The
Agency
takes
note
of
this
comment.
Of
course,
the
guideline
specifies
48
hours
as
normative
and
for
certain
chemicals
this
dosing
time
interval
can
be
changed.

Comment
7:
The
Panel
agreed
with
the
Agency
that
LD50
estimates
generated
from
the
UDP
will
be
sufficient
for
hazard
classification
and
labeling
of
products
and
noted
that
the
point
estimates
will
be
especially
useful
in
the
classification
of
mixtures.
However,
the
Panel
cautioned
that,
due
to
the
small
number
of
animals
used,
the
confidence
interval
must
be
considered
inexact
and,
in
many
instances,
it
will
be
very
wide.
Even
for
hazard
classification,
the
Panel
noted,
there
is
the
possibility
that
a
small
error
in
estimation
could
put
a
chemical
into
a
different
hazard
class.
The
Panel
cited
the
fact
that
the
Agency
suggested
that
whenever
there
was
uncertainty
at
a
boundary,
the
chemical
or
mixture
would
be
assigned
to
the
more
toxic
class.

Agency
response:
The
Agency
agrees
that
a
primary
application
of
the
UDP
guideline
is
for
hazard
classification
and
labeling
of
products.
As
it
interprets
acute
testing
results,
the
Agency
will
carefully
consider
the
caveats
identified
in
the
Scientific
Advisory
Panel
report
in
order
to
maintain
protection
of
human
health.

Comment
8:
The
Panel
pointed
out
that
the
UDP
will
not
be
sufficient
for
risk
assessment
3
procedures
that
require
information
on
the
slope
or
shape
of
the
dose­
response
curve
especially
for
tiered
ecological
risk
assessments.
In
cases
where
the
confidence
bounds
on
a
probabilistic
risk
assessment
are
too
wide,
the
Agency
will
have
to
add
additional
animals
to
the
UDP
or
perform
a
LD50
test
with
standard
techniques
traditionally
used.
Limitations
of
the
UDP
for
ecological
risk
assessments
are
as
follows
and
should
be
clearly
recognized.
(
1)
UDP
test
results
do
not
lend
themselves
to
the
generation
of
a
NOAEL
or
an
estimate
of
the
NOAEL
by
means
of
a
point
estimate
such
as
the
LDx;
(
2)
The
proposed
UDP
test
does
not
provide
information
about
the
slope
of
the
dose­
response
curve,
especially
when
a
minimal
number
of
animals
are
used
and
no
partial
kills
are
obtained;
(
3)
Reduction
in
the
number
of
animals,
while
beneficial
from
one
perspective,
makes
it
more
likely
that
the
confidence
interval
will
be
larger,
increasing
the
uncertainty
in
the
LD50
value,
raising
issues
of
what
is
an
acceptable
confidence
interval
and
when
should
additional
animals
be
tested.
In
consequence,
LD50
values
obtained
via
this
methodology
will
often
have
limited
utility
for
risk
assessment
procedures
where
exposure
and
effects
curves
are
compared
since
a
reliable
dose­
response
curve
will
not
be
available
from
the
acute
rat
test
without
additional
testing.

Agency
Response:
The
Agency
recognizes
these
limitations
of
the
UDP
for
tiered
ecological
risk
assessment
purposes
and
certain
health
effects
purposes.
However,
in
preparation
for
the
transition
to
alternative
testing
approaches,
Agency
scientists
and
statisticians
performed
simulations
of
performance
of
the
acute
oral
study
utilizing
the
minimum
of
three
groups
of
five
animals.
These
simulations
showed
confidence
intervals
for
slope
in
such
traditional
tests
could
be
fairly
wide
and
not
necessarily
suited
to
risk
assessment
purposes.

Comment
9:
Current
Agency
policy
requires
a
tiered
approach
for
product
evaluation
under
FIFRA
in
order
to
protect
wildlife
from
adverse
effects
that
may
be
manifested
following
pesticide
exposure.
Current
ecological
risk
assessment
procedures
are
thought
to
provide
some
measure
of
protection
for
wildlife.
However,
in
the
future,
for
ecological
assessments
which
rely
upon
rat
studies
to
estimate
risk
to
other
mammals,
the
Agency
must
be
cautious
that
free­
ranging
mammals
are
not
placed
at
substantial
risk.

Agency
response:
The
Agency
agrees
and
is
developing
additional
guidance
consistent
with
Part
158
for
higher
tier
testing
to
reduce
uncertainty
in
ecological
risk
assessment.

Comment
10:
At
the
first
tier
of
the
ecological
risk
assessment
process,
an
LD50
is
required
for
deterministic
assessment.
This
estimate
must
be
accurate
enough
to
prevent
erroneous
conclusions
progressing
to
higher
tiers.
Substantial
concern
exists
that
uncertainties
at
the
screening
level
are
already
too
great
to
provide
useful
decisions.
Known
biases
in
the
UDP
may
impact
the
validity
of
assumptions
made
in
current
early
tier/
level
ecological
risk
assessments.
If
tests
which
provide
data
for
screening
level
assessments
are
altered,
the
alteration
must
not
result
in
wider
confidence
limits
and
a
reduced
level
of
certainty
in
screening
or
higher­
level
assessments.

Agency
response:
In
determining
the
need
for
the
wild
mammal
test
for
terrestrial
ecological
effects,
Part
158
calls
for
the
Agency
to
consider
results
of
acute
and
subacute
testing
at
Tier
1,
4
along
with
use
pattern
and
environmental
fate
characteristics.
Results
of
simulations
show
that
if
the
initial
dose
and
dose
progression
for
the
UDP
are
chosen
in
the
optimum
fashion
described
in
the
guideline,
the
confidence
interval
which
describes
uncertainty
in
the
LD50
may
be
relatively
narrow.
However,
until
the
results
of
laboratory
testing
of
pesticides
are
available,
the
degree
to
which
the
UDP
will
be
useful
in
Tier
1
ecological
risk
assessment
is
uncertain.
The
Agency
recognizes
that
the
confidence
interval
from
UDP
tests
submitted
for
registration
or
reregistration
may
not
always
be
sufficiently
narrow
for
use
in
Tier
1
ecological
risk
assessments
and
is
developing
a
policy,
including
consideration
of
higher
tier
testing
such
as
wild
mammal
testing,
to
address
this.

Comment
11:
The
Panel
noted
that
multiple
LD
50
values
for
different
(
mammalian)
species
are
utilized
to
identify
the
most
sensitive
species.
Full
dose
response
data
for
most
of
these
(
focal)
species
are
not
typically
available
and
are
not
needed
to
develop
the
acute
5th
percentile
species.
Utilization
of
LD
50
values
generated
with
the
up­
down
procedure
are
not
expected
to
significantly
affect
the
outcome
of
this
procedure.
However,
lack
of
dose­
response
data
from
an
up­
down
test
for
a
key
focal
species
or
for
the
5th
percentile
species
would
prevent
accurate
probabalistic
risk
assessment
(
PRA)
for
that
species,
i.
e.,
it
would
not
be
possible
to
accurately
assess
effects
on
that
species
due
to
exposures
above
or
below
the
LD
50
exposure
level.
Without
slope
information,
the
dose
response
for
a
given
species
will
be
largely
unknown
and
the
confidence
bound
placed
on
modeled
probabilistic
risk
assessments
may
be
unacceptably
wide.

Agency
response:
The
Agency
will
take
note
of
these
points
as
it
continues
to
develop
approaches
for
probabilistic
risk
assessment
for
terrestrial
species.

Comment
12:
Overall,
the
guideline,
additional
guidance,
software
and
software
documentation
provide
sufficient
guidance,
supplemented
with
the
Interagency
Coordinating
Committee
on
the
Validation
of
Alternative
Methods
(
ICCVAM)
report
and
special
simulation
results.
Specific
suggestions
are
as
follows:
(
i)
the
quality
of
the
written
guideline
is
adequate,
but
could
be
improved;
although
the
panel
recognizes
that
EPA
must
follow
certain
formatting
rules.
In
addition,
the
Panel
recommended
that
the
following
topics
be
moved
from
"
Additional
Guidance"
to
an
Appendix:
the
linear
probit
dose
response
model;
the
likelihood
function;
the
assumption
of
constant
variance;
the
maximum
likelihood
estimation
of
LD50;
the
stopping
rule;
and
the
profile
likelihood
confidence
bounds.
and
(
ii)
a
better
illustration
should
be
provided
of
procedural
sequence
in
the
course
of
testing
and
calculation
for
test
scenarios.

Agency
response:

For
point
(
i),
the
EPA
guideline
format
has
been
adjusted
in
so
far
as
possible
to
provide
clear
headings
for
many
of
the
topics
identified
by
the
Panel.
Additional
definitions
also
provide
clarification.
Because
the
EPA
870.1100
guideline
must
track
the
OECD
425
guideline,
the
order
of
sections
has
not
been
changed.
In
addition,
the
Agency
has
developed
a
web
site,
to
be
used
with
the
guideline,
containing
the
AOT425StatPgm
software
and
additional
background
material
and
guidance
addressing
these
topics.
5
For
point
(
ii),
as
an
illustration
of
procedural
sequence
and
calculation,
the
Agency
has
added
a
diagram
of
the
computer
screen
to
illustrate
each
software
image.

Comment
13:
The
Panel
noted
that
documentation
of
AOT425StatPgm
software
for
implementing
the
UDP
dosing
and
LD50
calculations
is
well
written
and
provides
sufficient
guidance
to
the
user.
The
Panel
suggests
that
the
Agency
simplify
installation
procedures
to
make
it
efficient
and
errorfree
from
user
perspective.

Agency
response:
The
software
has
been
fixed
to
eliminate
the
need
for
the
user
to
change
installation
file
names
or
to
`
ignore'
error
messages.

Comment
14:
The
Panel
recommended
that
the
Agency
include
an
initial
schematic
that
outlines
the
relation
of
the
program's
Data
Edit
and
Report,
the
window
task
bars,
and
menu
choices
to
provide
the
user
an
overview
and
facilitate
migration
between
windows,
tasks
and
options.

Agency
response:
These
recommendations
will
be
addressed
in
future
revision
of
the
AOT425StatPgm
software.

Comment
15:
The
Panel
recommended
establishment
of
a
data
set
and
test
protocol
for
verification
of
installation
and
software
operation
of
the
AOT425StatPgm
on
the
users'
computers.

Agency
response:
The
Agency
agreed
with
this
comment
and
developed
a
test
data
collection
consisting
of
a
set
of
15
sample
experiments
which
can
be
used
to
verify
the
proper
installation
and
operation
of
the
software
on
the
user's
computer.
The
test
data
sets,
instructions,
and
solutions
to
the
sample
runs
are
included
on
the
EPA
web
site
at
www.
epa.
gov/
oppfead1/
harmonization,
along
with
the
AOT425StatPgm
program
,
Users'
Manual,
and
other
related
documents.
The
Users'
Manual
also
includes
instructions
on
how
comments
and
questions
pertaining
to
the
software
program
can
be
directed
to
a
specific
AOT425
email
address
at
EPA.

Comment
16:
The
Panel
recommended
that
the
software
output
show
explicit
calculations
for:
the
linear
probit
dose­
response
model;
the
likelihood
function;
the
assumption
of
constant
variance;
the
maximum
likelihood
estimate
of
LD50;
the
stopping
rule;
and
the
profile
likelihood
confidence
bounds.

Agency
response:
The
Agency
agrees
that
it
would
be
valuable
for
the
output
of
the
software
to
provide
explicit
calculations
and
will
include
such
calculations
and
other
changes
in
the
next
generation
of
the
software,
after
more
experience
has
been
gained
in
its
use.
