United
States
Environmental
Protection
Agency
Prevention,
Pesticides
and
Toxic
Substances
(
7101)
EPA
712
 
C
 
02
 
190
December
2002
Health
Effects
Test
Guidelines
OPPTS
870.1100
Acute
Oral
Toxicity
i
INTRODUCTION
This
guideline
is
one
of
a
series
of
test
guidelines
that
have
been
developed
by
the
Office
of
Prevention,
Pesticides
and
Toxic
Substances,
United
States
Environmental
Protection
Agency
for
use
in
the
testing
of
pesticides
and
toxic
substances,
and
the
development
of
test
data
that
must
be
submitted
to
the
Agency
for
review
under
Federal
regulations.

The
Office
of
Prevention,
Pesticides
and
Toxic
Substances
(
OPPTS)
has
developed
this
guideline
through
a
process
of
harmonization
that
blended
the
testing
guidance
and
requirements
that
existed
in
the
Office
of
Pollution
Prevention
and
Toxics
(
OPPT)
and
appeared
in
Title
40,
Chapter
I,
Subchapter
R
of
the
Code
of
Federal
Regulations
(
CFR),
the
Office
of
Pesticide
Programs
(
OPP)
which
appeared
in
publications
of
the
National
Technical
Information
Service
(
NTIS)
and
the
guidelines
published
by
the
Organization
for
Economic
Cooperation
and
Development
(
OECD).

The
purpose
of
harmonizing
these
guidelines
into
a
single
set
of
OPPTS
guidelines
is
to
minimize
variations
among
the
testing
procedures
that
must
be
performed
to
meet
the
data
requirements
of
the
U.
S.
Environmental
Protection
Agency
under
the
Toxic
Substances
Control
Act
(
15
U.
S.
C.
2601)
and
the
Federal
Insecticide,
Fungicide
and
Rodenticide
Act
(
7
U.
S.
C.
136,
et
seq.).

Final
Guideline
Release:
This
guideline
is
available
from
the
U.
S.
Government
Printing
Office,
Washington,
DC
20402
on
disks
or
paper
copies:
call
(
202)
512
 
0132.
This
guideline
is
also
available
electronically
in
PDF
(
portable
document
format)
from
EPA's
Internet
Web
site
at
http:/
/
www.
epa.
gov/
opptsfrs/
home/
guidelin.
htm.
Also,
the
Agency
has
developed
and
strongly
recommends
users
to
solely
use,
the
software
program
for
performing
the
Up­
and­
Down
Procedure
and
calculating
the
LD50
and
confidence
interval.
The
software
program
(
AOT425StatPgm)
is
available
on
EPA's
Internet
Web
site
at
http://
www.
epa.
gov/
oppfead1/
harmonized.
1
OPPTS
870.1100
Acute
oral
toxicity.
(
a)
Scope
 
Applicability.
This
guideline
is
intended
to
meet
testing
requirements
of
both
the
Federal
Insecticide,
Fungicide,
and
Rodenticida
Act
(
FIFRA)
(
7
U.
S.
C.
136,
et
seq.)
and
the
Toxic
Substances
Control
Act
(
TSCA)
(
15
U.
S.
C.
2601).

(
2)
Background.
The
source
material
for
this
revised
harmonized
test
guideline
is
OPPTS
870.1100
Acute
Oral
Toxicity,
dated
August
1998
and
OECD
test
Guideline
425
Acute
Oral
Toxicity
 
Up­
and­
Down
Procedure.

(
b)
Purpose.
In
the
assessment
and
evaluation
of
the
toxic
characteristics
of
a
substance,
determination
of
acute
oral
toxicity
is
usually
an
initial
step.
It
provides
information
on
health
and
environmental
hazards
likely
to
arise
from
short­
term
exposure
by
the
oral
route.
Data
from
an
acute
study
may
serve
as
a
basis
for
classification
and
labeling.
It
is
traditionally
a
step
in
establishing
a
dosage
regimen
in
subchronic
and
other
studies
and
may
provide
initial
information
on
the
mode
of
toxic
action
of
a
substance.
An
evaluation
of
acute
toxicity
data
should
include
the
relationship,
if
any,
between
the
exposure
of
animals
to
the
test
substance
and
the
incidence
and
severity
of
all
abnormalities,
including
behavioral
and
clinical
abnormalities,
the
reversibility
of
observed
abnormalities,
gross
lesions,
body
weight
changes,
effects
on
mortality,
and
any
other
toxic
effects.

(
c)
Definitions.
The
definitions
in
Section
3
of
the
Toxic
Substances
Control
Act
(
TSCA)
and
the
definitions
in
40
CFR
Part
792
 
Good
Laboratory
Practice
Standards
apply
to
this
test
guideline.
The
following
definitions
also
apply
to
this
test
guideline.

Acute
oral
toxicity
is
the
adverse
effects
occurring
within
a
short
time
of
oral
administration
of
a
single
dose
of
a
substance
or
multiple
doses
given
within
24
hours.

Confidence
interval
(
CI)
is
an
interval
estimate,
a
range
of
values,
intended
to
include
the
true
LD50
with
a
specified
degree
of
confidence.

Delayed
death
means
that
an
animal
does
not
die
or
appear
moribund
within
48
hours,
but
dies
later
during
the
14­
day
observation
period.

Dose
is
the
amount
of
test
substance
administered.
Dose
is
expressed
as
weight
(
g,
mg
(
grams,
milligrams))
or
as
weight
of
test
substance
per
unit
weight
of
test
animal
(
e.
g.,
mg/
kg
(
milligrams/
kilograms)).

Dose
progression
factor,
sometimes
termed
a
dose
spacing
factor,
refers
to
the
multiple
by
which
a
dose
is
increased
(
i.
e.,
the
dose
progression)
when
an
animal
survives
or
the
divisor
by
which
it
is
decreased
when
an
animal
dies.
The
dose
progression
factor
is
recommended
to
be
the
antilog
of
1/(
the
estimated
slope
of
the
dose­
response
curve).
The
default
2
dose
progression
factor
is
recommended
to
be
3.2
=
antilog
0.5
=
antilog
(
1/
2).

LD50
(
median
lethal
dose),
oral,
is
a
statistically
derived
single
dose
of
a
substance
that
can
be
expected
to
cause
death
in
50
per
cent
of
animals
when
administered
by
the
oral
route.
The
LD50
value
is
expressed
in
terms
of
weight
of
test
substance
per
unit
weight
of
test
animal
(
mg/
kg).

Limit
dose
refers
to
a
dose
at
an
upper
limitation
on
testing
(
2000
 
5000
mg/
kg).

Moribund
status
of
an
animal
refers
to
being
in
a
state
of
dying
or
inability
to
survive,
even
if
treated.

Nominal
sample
size
refers
to
the
total
number
of
tested
animals,
reduced
by
one
less
than
the
number
of
like
responses
at
the
beginning
of
the
series,
or
by
the
number
of
tested
animals
up
to
but
not
including
the
pair
that
creates
the
first
reversal.
For
example,
for
a
series
where
X
and
O
indicate
opposite
animal
outcomes
(
for
instance,
X
could
be
dies
within
48
hours
and
O
survives)
in
a
pattern
as
follows:
OOOXXOXO,
we
have
the
total
number
of
tested
animals
(
or
sample
size
in
the
conventional
sense)
as
8
and
the
nominal
sample
size
as
6.
This
particular
example
shows
4
animals
following
a
reversal.
It
is
important
to
note
whether
a
count
in
a
particular
part
of
the
guideline
refers
to
the
nominal
sample
size
or
to
the
total
number
tested.
For
example,
the
maximum
actual
number
tested
is
15.
When
testing
is
stopped
based
on
that
basis,
the
nominal
sample
size
will
be
less
than
or
equal
to
15.
Members
of
the
nominal
sample
start
with
the
(
r­
1)
st
animal
(
the
animal
before
the
second
in
the
reversal
pair)
(
see
reversal
below).

Probit
is
an
abbreviation
for
the
term
``
probability
integral
transformation''
and
a
probit
dose­
response
model
permits
a
standard
normal
distribution
of
expected
responses
(
i.
e.,
one
centered
to
its
mean
and
scaled
to
its
standard
deviation,
sigma
)
to
doses
(
typically
in
a
logarithmic
scale)
to
be
analyzed
as
if
it
were
a
straight
line
with
slope
the
reciprocal
of
sigma.
A
standard
normal
lethality
distribution
is
symmetric;
hence,
its
mean
is
also
its
true
LD50
or
median
response.

Reversal
is
a
situation
where
nonresponse
is
observed
at
some
dose,
and
a
response
is
observed
at
the
next
dose
tested,
or
vice
versa
(
i.
e.,
response
followed
by
nonresponse).
Thus,
a
reversal
is
created
by
a
pair
of
responses.
The
first
such
pair
occurs
at
animals
numbered
r­
1
and
r.

Sigma
is
the
standard
deviation
of
a
log
normal
curve
describing
the
range
of
tolerances
of
test
subjects
to
the
chemical
(
where
a
subject
is
expected
capable
of
responding
if
the
chemical
dose
exceeds
the
subject's
tolerance).
The
estimated
sigma
provides
an
estimate
of
the
variation
3
among
test
animals
in
response
to
a
full
range
of
doses.
See
slope
and
probit.

Slope
(
of
the
dose­
response
curve)
is
a
value
related
to
the
angle
at
which
the
dose
response
curve
rises
from
the
dose
axis.
In
the
case
of
probit
analysis,
when
responses
are
analyzed
on
a
probit
scale
against
dose
on
a
log
scale
this
curve
will
be
a
straight
line
and
the
slope
is
the
reciprocal
of
sigma,
the
standard
deviation
of
the
underlying
test
subject
tolerances
which
are
assumed
to
be
normally
distributed.
See
probit
and
sigma.

Stopping
rule
is
used
in
this
guideline
synonymously
with
(
1)
a
specific
stopping
criterion
and
(
2)
the
collection
of
all
criteria
determining
when
a
testing
sequence
terminates.
In
particular,
for
the
main
test,
stopping
rule
is
used
in
paragraph
(
e)(
2)(
ii)
of
this
guideline
as
a
shorthand
for
the
criterion
that
relies
on
comparison
of
ratios
to
a
critical
value.

(
d)
Approaches
to
the
determination
of
acute
toxicity.
EPA
recommends
the
Up­
and­
Down
Procedure
(
UDP)
as
detailed
in
this
guideline
and
adopted
by
the
Organization
for
Economic
Cooperation
and
Development
(
OECD)
as
test
Guideline
425
(
see
paragraph
(
n)(
1)
of
this
guideline
to
assess
acute
oral
toxicity.
This
method
provides
a
point
estimate
of
lethality
and
confidence
interval
around
the
LD50.
Acute
oral
toxicity
testing
may
also
be
performed
using
the
Fixed
Dose
Method
of
OECD
Guideline
420
(
see
paragraph
(
n)(
2)
of
this
guideline)
or
the
Acute
Toxic
Class
Method
of
OECD
Guideline
423
(
see
paragraph
(
n)(
3)
of
this
guideline
These
methods
assess
lethality
within
a
dose
range.

(
e)
Introduction
to
the
UDP
 
(
1)
Background.
(
i)
The
concept
of
the
up­
and­
down
testing
approach
was
first
described
by
Dixon
and
Mood
(
see
paragraphs
(
n)(
4)
through
(
n)(
7)
of
this
guideline).
In
1985,
Bruce
proposed
to
use
an
UDP
for
the
determination
of
acute
toxicity
of
chemicals
(
see
paragraph
(
n)(
8)
of
this
guideline).
There
exist
several
variations
of
the
up­
and­
down
experimental
design
for
estimating
an
LD50.
This
guideline
is
derived
from
the
UDP
of
Bruce
as
adopted
by
the
American
Society
for
Testing
and
Materials
(
ASTM)
in
1987
(
see
paragraph
(
n)(
9)
of
this
guideline)
and
revised
in
1990.
A
study
comparing
the
results
obtained
with
the
UDP,
the
conventional
LD50
test
and
the
Fixed
Dose
Procedure
(
FDP,
OECD
Guideline
420)
was
published
in
1995
(
see
paragraph
(
n)(
10)
of
this
guideline).

(
ii)
The
UDP
described
in
this
guideline
is
of
value
in
minimizing
the
number
of
animals
required
to
estimate
the
acute
oral
toxicity
of
a
chemical.
In
addition
to
the
estimation
of
LD50
and
CI,
the
test
procedure
allows
the
observation
of
signs
of
toxicity.
The
UDP
does
not
provide
information
about
the
slope
of
the
dose­
response
curve.

(
iii)
The
guideline
significantly
reduces
the
number
of
animals
used
in
comparison
to
the
traditional
LD50
test,
which
often
required
at
least
30
animals
in
a
test:
(
A)
The
stopping
rule
limits
the
number
of
animals
4
in
a
test;
(
B)
sequential
dosing
introduces
further
efficiencies
in
animal
use;
(
C)
initial
dosing
is
now
set
to
be
below
the
LD50
increasing
the
percentage
of
animals
in
which
dosing
levels
will
be
sublethal
and
thereby
providing
some
reduction
in
pain
and
distress;
and
(
D)
the
use
of
a
single
sex
(
usually
females)
reduces
the
number
of
animals
needed
and
minimizes
the
variability
in
the
test
population.
In
addition,
the
OECD
Guidance
Document
on
Humane
Endpoints
(
see
paragraph
(
n)(
11)
of
this
guideline)
should
be
followed
in
order
to
reduce
the
overall
suffering
of
test
animals
used
in
this
type
of
toxicity
test.

(
2)
Initial
considerations
 
(
i)
Choice
of
starting
dose
and
dose
progression
factor.
All
available
information
on
the
test
substance
should
be
considered
by
the
testing
laboratory
prior
to
conducting
the
study
in
order
to
determine
if
a
preliminary
estimate
of
the
LD50
and
the
slope
of
the
dose­
response
curve
can
be
made.
Because
the
method
has
a
bias
toward
the
starting
dose,
it
is
essential
that
initial
dosing
occur
below
the
LD50.
In
addition,
the
UDP
performs
best
when
the
spacing
between
doses
or
dose
progression
factor
is
based
on
an
accurate
estimate
of
the
slope
of
the
dose­
response
curve.
(
See
paragraphs
(
i)(
3)(
ii)
and
(
m)(
1)
of
this
guideline
for
discussion
of
dose
sequences
and
starting
values.)
Initial
information
may
include
the
identity
and
chemical
structure
of
the
substance;
its
physical
chemical
properties;
the
results
of
any
other
in
vitro
or
in
vivo
toxicity
tests
on
the
substance
or
mixtures;
toxicological
data
on
structurally
related
substances
or
similar
mixtures;
and
the
anticipated
use(
s)
of
the
substance.
For
example,
data
from
an
in
vitro
cytotoxicity
assay
can
also
be
useful
as
one
of
the
tools
in
setting
a
starting
dose
for
the
in
vivo
assessment
of
acute
oral
toxicity
(
see
paragraphs
(
n)(
10)
through
(
n)(
12)
of
this
guideline).
(
A
Guidance
Document
on
Using
In
Vitro
Data
to
Estimate
In
Vivo
Starting
Doses
for
Acute
Toxicity
is
available
(
see
paragraph
(
n)(
11)
of
this
guideline),
and
preliminary
information
suggests
that
the
use
of
this
approach
may
further
reduce
the
number
of
animals
used
for
in
vivo
testing
(
see
paragraph
(
n)(
11)
of
this
guideline).
Preliminary
estimates
of
the
LD50
and
the
dose­
response
slope
will
help
in
selecting
a
dose
progression
factor
and
a
starting
dose
for
testing.

(
ii)
Default
starting
dose
and
dose
progression
factor.
If
no
information
is
available
to
make
a
preliminary
estimate
of
the
LD50
and
the
slope
of
the
dose­
response
curve,
results
of
computer
simulations
have
suggested
that
starting
near
175
mg/
kg
and
using
half­
log
units
(
corresponding
to
a
dose
progression
of
3.2)
between
doses
will
produce
the
best
results.
This
starting
dose
should
be
modified
if
the
substance
is
likely
to
be
highly
toxic.
The
half­
log
spacing
provides
for
a
more
efficient
use
of
animals,
and
increases
accuracy
in
the
prediction
of
the
LD50
value.
However,
for
chemicals
with
large
variability
(
i.
e.,
shallow
dose­
response
slopes),
bias
can
still
be
introduced
in
the
lethality
estimates
and
the
LD50
estimate
will
have
a
large
statistical
error,
similar
to
other
acute
toxicity
methods.
To
correct
for
this,
the
main
test
includes
a
stopping
rule
keyed
5
to
properties
of
the
estimate
rather
than
a
fixed
number
of
test
observations
(
See
paragraph
(
i)(
3)(
iii)
of
this
guideline.)

(
iii)
Delayed
toxicity.
The
method
is
easiest
to
apply
to
materials
that
produce
death
within
one
or
two
days.
The
method
would
not
be
practical
to
use
when
considerably
delayed
death
(
five
days
or
more)
can
be
expected.

(
iv)
Computation.
Computers
are
used
to
facilitate
animal­
by­
animal
calculations
that
establish
testing
sequences
and
provide
final
estimates.
The
users
of
this
protocol
are
strongly
urged
to
solely
use
the
Agencydeveloped
software
package
(
AOT425StatPgm)
for
performing
the
test
and
the
calculation
of
the
LD
50.
The
software
is
available
on
EPA's
Internet
Web
site
at
http://
www.
epa.
gov/
oppfead1/
harmonized.

(
v)
Humane
practices.
Moribund
animals
or
animals
obviously
in
pain
or
showing
signs
of
severe
and
enduring
distress
shall
be
humanely
killed,
and
are
considered
in
the
interpretation
of
the
test
results
in
the
same
way
as
animals
that
died
on
test.
Criteria
for
making
the
decision
to
kill
moribund
or
severely
suffering
animals,
and
guidance
on
the
recognition
of
predictable
or
impending
death
are
the
subject
of
an
OECD
guidance
document
(
see
paragraph
(
n)(
11)
of
this
guideline).

(
vi)
Limit
test.
A
limit
test
can
be
used
efficiently
to
identify
chemicals
that
are
likely
to
have
low
acute
toxicity.

(
f)
Principle
of
the
limit
test.
The
limit
test
is
a
sequential
test
that
uses
a
maximum
of
5
animals
(
see
paragraphs
(
i)(
2)(
i)
through
(
i)(
2)(
iv)
of
this
guideline).
A
test
dose
of
5000
mg/
kg
is
used.
The
selection
of
a
sequential
test
plan
increases
the
statistical
power
and
also
has
been
made
to
intentionally
bias
the
procedure
towards
rejection
of
the
limit
test
for
compounds
with
LD50s
near
the
limit
dose;
i.
e.,
to
err
on
the
side
of
safety.
As
with
any
limit
test
protocol,
the
probability
of
correctly
classifying
a
compound
will
decrease
as
the
actual
LD50
more
nearly
resembles
the
limit
dose.

(
g)
Principle
of
the
Main
Test.
(
1)
The
main
test
consists
of
a
single
ordered
dose
progression
in
which
animals
are
dosed,
one
at
a
time,
at
48­
hour
intervals.
The
first
animal
receives
a
dose
a
step
below
the
level
of
the
best
estimate
of
the
LD50.
If
the
animal
survives,
the
dose
for
the
next
animal
is
increased
to
a
factor
of
one
half
log
times
the
original
dose;
if
it
dies,
the
dose
for
the
next
animal
is
decreased
by
a
similar
dose
progression.
(
Note:
3.2
is
the
default
factor
corresponding
to
a
dose
progression
of
one
half
log
unit
in
the
Agency
developed
software
program
(
AOT425StatPgm).
However,
this
value
may
be
changed.
Paragraphs
(
i)(
3)(
ii)
and
(
m)(
12)
of
this
guideline
provide
further
guidance
for
choice
of
dose
spacing
factor.)
Each
animal
should
be
observed
carefully
for
up
to
48
hours
before
making
a
decision
on
whether
and
how
much
to
dose
the
next
animal.
That
decision
is
based
on
the
48­
hour
survival
pattern
6
of
all
the
animals
up
to
that
time.
(
See
paragraphs
(
i)(
3)(
i)
and
(
i)(
3)(
v)
of
this
guideline
on
choice
of
survival
interval.)
A
combination
of
stopping
criteria
is
used
to
keep
the
number
of
animals
low
while
adjusting
the
dosing
pattern
to
reduce
the
effect
of
a
poor
starting
value
or
low
slope
(
see
paragraph
(
i)(
3)(
iv)
of
this
guideline).
Dosing
is
stopped
when
one
of
these
criteria
is
satisfied
(
see
paragraphs
(
i)(
3)(
iii)
and
(
k)(
2)
of
this
guideline),
at
which
time
an
estimate
of
the
LD50
and
a
CI
are
calculated
for
the
test
based
on
the
status
of
all
the
animals
at
termination.
For
most
applications,
testing
will
be
completed
with
only
4
animals
after
initial
reversal
in
animal
outcome.
The
LD50
is
calculated
using
the
method
of
maximum
likelihood
(
see
paragraphs
(
k)(
2)
and
(
k)(
2)(
iii)
of
this
guideline

(
2)
The
results
of
the
main
test
procedure
serve
as
the
starting
point
for
a
computational
procedure
to
provide
a
CI
estimate
where
feasible.
A
description
of
the
basis
for
this
CI
is
outlined
in
paragraph
(
k)(
3)
of
this
guideline.

(
h)
Preparation
for
testing
 
(
1)
Selection
of
animals
species.
The
preferred
rodent
species
is
the
rat
although
other
rodent
species
may
be
used.

(
2)
Single
sex
selection.
The
test
is
conducted
using
a
single
sex
in
order
to
reduce
variability
and
as
a
means
of
minimizing
the
number
of
animals
used.
Either
sex
may
be
used,
however,
if
there
is
information
available
indicating
differences
in
sensitivity,
the
most
sensitive
sex
(
usually
females)
should
be
tested
(
see
paragraph
(
n)(
11)
of
this
guideline).

(
i)
Literature
surveys
of
conventional
LD50
tests
show
that
usually
there
is
little
difference
in
sensitivity
between
the
sexes
but,
in
those
cases
where
differences
were
observed,
females
were
often
slightly
more
sensitive
(
see
paragraph
(
n)(
10)
of
this
guideline).
For
chemicals
that
are
direct
acting
in
their
toxic
mechanism,
female
rats
may
have
a
lower
detoxification
capacity
than
males,
as
measured
by
specific
activity
of
phase
I
and
II
enzymes.
However,
all
available
information
should
be
evaluated,
for
example
on
chemical
analogues
and
the
results
of
testing
for
other
toxicological
endpoints
on
the
chemical
itself,
as
this
may
indicate
that
males
may
be
more
sensitive
than
females.
Knowledge
that
metabolic
activation
is
required
for
a
chemical's
toxicity
can
also
indicate
that
males
may
be
the
more
sensitive
sex.

(
ii)
Occasionally,
the
results
of
subsequent
testing,
for
example
a
subchronic
test,
may
raise
concerns
that
the
more
sensitive
sex
had
not
been
used.
In
such
cases,
and
only
when
considerable
differences
between
the
sexes
are
suspected,
it
may
be
necessary
to
conduct
another
full
acute
oral
toxicity
study
in
the
second
sex.
This
is
preferable
to
conducting
confirmatory
testing
in
a
small
group
of
animals
of
the
second
sex
as
a
late
satellite
to
the
original
test
because
there
is
a
strong
possibility
that
this
7
would
produce
results
that
are
difficult
to
interpret.
The
impact
of
conducting
a
second
full
test
on
the
overall
number
of
animals
used
in
acute
toxicity
testing
should
be
small
because
re­
testing
is
anticipated
to
be
infrequent
and
the
results
of
the
test
in
one
sex,
together
with
data
from
any
subsequent
studies,
will
greatly
assist
in
the
selection
of
starting
doses
closer
to
the
LD50
in
the
second
test.

(
3)
Age
and
weight
ranges.
Healthy
young
adult
animals
of
commonly
used
laboratory
strains
should
be
employed.
Females
should
be
nulliparous
and
non­
pregnant.
At
the
commencement
of
its
dosing,
each
animal
should
be
between
8
weeks
and
12
weeks
old.
In
order
to
minimize
the
contribution
of
developmental
variability
to
study
outcome,
10
weeks,
with
a
range
of
±
1
week
is
recommended
if
practical.
The
weight
of
each
animal
should
fall
in
an
interval
±
20%
of
the
mean
initial
weight
of
all
previously
dosed
animals.

(
4)
Housing
and
feeding
conditions.
The
temperature
in
the
experimental
animal
room
should
be
22
°
C
(
±
3
°
C).
The
relative
humidity
should
be
at
least
30%
and
preferably
not
exceed
70%
other
than
during
room
cleaning.
Lighting
should
be
artificial,
the
sequence
being
12
hours
light
and
12
hours
dark.
The
animals
are
housed
individually.
For
feeding,
conventional
rodent
laboratory
diets
may
be
used
with
an
unlimited
supply
of
drinking
water.

(
5)
Preparation
of
animals.
The
animals
are
randomly
selected,
marked
to
permit
individual
identification,
and
kept
in
their
cages
for
at
least
5
days
prior
to
dosing
to
allow
for
acclimatization
to
the
laboratory
conditions.
As
with
other
sequential
test
designs,
care
must
be
taken
to
ensure
that
animals
are
available
in
the
appropriate
size
and
age
range
for
the
entire
study.

(
6)
Preparation
of
doses.
(
i)
When
necessary,
the
test
substance
is
dissolved
or
suspended
in
a
suitable
vehicle.
The
use
of
an
aqueous
solution
suspension/
emulsion
is
recommended
wherever
possible,
followed
in
order
of
preference
by
a
solution/
suspension/
emulsion
in
oil
(
e.
g.
corn
oil)
and
then
possibly
solution
in
other
vehicles.
For
vehicles
other
than
water
the
toxicological
characteristics
of
the
vehicle
should
be
known.
Dosing
preparations
must
be
prepared
shortly
prior
to
administration
unless
the
stability
of
the
preparation
over
the
period
during
which
it
will
be
used
is
known.
Where
preparation
shortly
before
administration
is
not
practicable
and
the
stability
of
the
preparation
is
not
known,
this
will
need
to
be
demonstrated
analytically.

(
ii)
Constant
concentration
should
be
used
in
dosing
unless
there
is
clear
scientific
or
regulatory
justification
for
not
doing
so.
The
maximum
dose
volume
for
administration
must
not
be
exceeded.
The
maximum
volume
of
liquid
that
can
be
administered
at
one
time
depends
on
the
size
of
the
test
animal.
In
rodents,
the
volume
should
not
normally
exceed
8
1
ml/
100g
of
body
weight;
however,
in
the
case
of
aqueous
solutions,
2
ml/
100g
body
weight
can
be
considered.

(
7)
Administration
of
doses.
(
i)
The
test
substance
is
administered
in
a
single
dose
by
gavage
using
a
stomach
tube
or
a
suitable
intubation
cannula.
In
the
unusual
circumstance
that
a
single
dose
is
not
possible,
the
dose
may
be
given
in
smaller
fractions
over
a
period
not
exceeding
24
hours.

(
ii)
Animals
should
be
fasted
prior
to
dosing
(
e.
g.,
with
the
rat,
food
but
not
water
should
be
withheld
overnight;
with
the
mouse,
food
but
not
water
should
be
withheld
for
3
 
4
hours).
Following
the
period
of
fasting,
the
animals
should
be
weighed
and
the
test
substance
administered.
The
fasted
body
weight
of
each
animal
is
determined
and
the
dose
is
calculated
according
to
the
body
weight.
After
the
substance
has
been
administered,
food
may
be
withheld
for
a
further
3
 
4
hours
in
rats
or
1
 
2
hours
in
mice.
Where
a
dose
is
administered
in
fractions
over
a
period
of
time,
it
may
be
necessary
to
provide
the
animals
with
food
and
water
depending
on
the
length
of
the
period.

(
i)
The
up­
and­
down
testing
procedure
 
(
1)
Choice
of
limit
test
and
main
test.
The
limit
test
is
primarily
used
in
situations
where
the
experimenter
has
information
indicating
that
the
test
material
is
likely
to
be
nontoxic,
i.
e.,
having
toxicity
below
regulatory
limit
doses.
Information
about
the
toxicity
of
the
test
material
can
be
gained
from
knowledge
about
similar
tested
compounds
or
similar
tested
mixtures
or
products,
taking
into
consideration
the
identity
and
percentage
of
components
known
to
be
of
toxicological
significance.
In
those
situations
where
there
is
little
or
no
information
about
its
toxicity,
or
in
which
the
test
material
is
expected
to
be
toxic,
the
main
test
should
be
performed.

(
2)
Implementation
of
the
limit
test.
(
i)
The
Agency
has
developed
dedicated
software
for
performing
the
test
and
calculation
of
test
results
(
see
paragraph
(
e)
(
2)(
iv)
of
this
guideline).

(
ii)
Dose
one
animal
at
5000
mg/
kg.
If
the
animal
dies,
conduct
the
main
test
starting
at
175
mg/
kg
to
determine
the
LD50.
If
the
animal
survives
dose
two
additional
animals.
If
both
animals
survive,
the
LD50
is
greater
than
the
limit
dose
and
the
test
is
terminated
(
i.
e.
carried
to
full
14­
day
observation
without
dosing
of
further
animals).
If
one
or
both
animals
die,
then
dose
an
additional
two
animals,
one
at
a
time.
If
an
animal
unexpectedly
dies
late
in
the
study,
and
there
are
other
survivors,
it
is
appropriate
to
stop
dosing
and
observe
all
animals
to
see
if
other
animals
will
also
die
during
a
similar
observation
period
(
see
paragraph
(
g)(
1)
of
this
guideline
for
initial
observation
period).
Late
deaths
should
be
counted
the
same
as
other
deaths.
The
results
are
evaluated
as
follows
(
O=
survival
and
X=
death).
9
(
iii)
The
LD50
is
less
than
the
test
dose
(
5000
mg/
kg)
when
three
or
more
animals
die.
If
a
third
animal
dies,
conduct
the
main
test.

O
XO
XX
O
OX
XX
O
XX
OX
O
XX
X
(
iv)
The
LD50
is
greater
than
the
test
dose
(
5000
mg/
kg)
when
three
or
more
animals
survive.

O
OO
O
XO
XO
O
XO
O
O
OX
XO
O
OX
O
O
XX
OO
(
v)
If
a
limit
test
is
performed
at
2000
mg/
kg,
animals
should
be
dosed
sequentially
and
testing
should
be
performed
on
all
five
animals.

(
3)
Implementation
of
the
main
test.
(
i)
The
Agency
has
developed
dedicated
software
for
performing
the
test
and
calculation
of
test
results
(
see
paragraph
(
e)
(
2)(
iv)
of
this
guideline).

(
ii)
Performing
the
UDP.
Single
animals
are
dosed
in
sequence
usually
at
48­
hour
intervals.
However,
the
time
interval
between
dosing
is
determined
by
the
onset,
duration,
and
severity
of
toxic
signs.
Treatment
of
an
animal
at
the
next
dose
should
be
delayed
until
one
is
confident
of
survival
of
the
previously
dosed
animal.
The
time
interval
may
be
adjusted
as
appropriate,
e.
g.,
in
case
of
inconclusive
response.
The
test
is
simpler
to
implement
when
a
single
time
interval
is
used
for
making
sequential
dosing
decisions.
Nevertheless,
it
is
not
necessary
to
recalculate
dosing
or
likelihood­
ratios
if
the
time
interval
changes
midtest.
For
selecting
the
starting
dose,
all
available
information,
including
information
on
structurally
related
substances
and
results
of
any
other
toxicity
tests
on
the
test
material,
should
be
used
to
approximate
the
LD50
as
well
as
the
slope
of
the
dose­
response
curve.

(
iii)
Choice
of
starting
dose
and
dose
progression.
The
first
animal
is
dosed
a
step
below
the
toxicologist's
best
estimate
of
the
LD50.
If
the
animal
survives,
the
second
animal
receives
a
higher
dose.
If
the
first
animal
dies
or
appears
moribund,
the
second
animal
receives
a
lower
dose.
The
same
dosing
decision
pattern
is
followed
for
each
subsequent
animal.
10
The
dose
progression
factor
should
be
chosen
to
be
the
antilog
of
1/(
the
estimated
slope
of
the
dose­
response
curve)
(
a
progression
of
3.2
corresponds
to
a
slope
of
2)
and
should
remain
constant
throughout
testing.
Thus,
when
there
is
no
information
on
the
slope
of
the
substance
to
be
tested,
a
default
dose
progression
factor
of
3.2
is
used.
Using
the
default
progression
factor,
doses
would
be
selected
from
the
sequence
1.75,
5.5,
17.5,
55,
175,
550,
1750,
5000.
If
no
estimate
of
the
substance's
lethality
is
available,
dosing
should
be
initiated
at
175
mg/
kg.
In
most
cases,
this
dose
is
sublethal
and
therefore
serves
to
reduce
the
level
of
pain
and
suffering
If
animal
tolerances
to
the
chemical
are
expected
to
be
highly
variable
(
i.
e.,
slopes
are
expected
to
be
less
than
2.0),
consideration
should
be
given
to
increasing
the
dose
progression
factor
beyond
the
default
0.5
on
a
log
dose
scale
(
i.
e.,
3.2
progression
factor)
prior
to
starting
the
test.
Similarly,
for
test
substances
known
to
have
very
steep
slopes,
dose
progression
factors
smaller
than
the
default
should
be
chosen.
(
Paragraph
(
m)(
3)
of
this
guideline
relates
choice
of
dose
progression
to
assumed
slope
and
sigma
and
discusses
test
performance.
Paragraph
(
m)(
1)
of
this
guideline
includes
a
table
of
dose
progressions
for
whole
number
slopes
ranging
from
1
to
8
with
starting
dose
175
mg/
kg.)

(
iv)
Stopping
rules.
Dosing
continues
depending
on
the
fixed­
time
interval
(
e.
g.,
48­
hours)
outcomes
of
all
the
animals
up
to
that
time.
The
testing
stops
when
one
of
the
following
stopping
criteria
first
is
met:

(
A)
3
consecutive
animals
survive
at
the
upper
bound;

(
B)
5
reversals
occur
in
any
6
consecutive
animals
tested;

(
C)
At
least
4
animals
have
followed
the
first
reversal
and
the
specified
likelihood­
ratios
exceed
the
critical
value.
(
See
paragraphs
(
k)(
2)(
iv)
and
(
m)(
2)
of
this
guideline).
Calculations
are
made
at
each
dosing,
following
the
fourth
animal
after
the
first
reversal.).

(
v)
Total
number
of
doses.
For
a
wide
variety
of
combinations
of
LD50
and
slopes,
stopping
rule
in
paragraph
(
i)(
3)(
iii)(
C)
of
this
guideline
will
be
satisfied
with
4
to
6
animals
after
the
test
reversal.
In
some
cases
for
chemicals
with
shallow
slope
dose­
response
curves,
additional
animals
(
up
to
a
total
of
fifteen
tested)
may
be
needed.

(
vi)
Calculation.
When
the
stopping
criteria
have
been
attained,
the
estimated
LD50
should
be
calculated
from
the
animal
outcomes
at
test
termination
using
the
method
described
in
paragraphs
(
k)(
1)(
i)
and
(
k)(
2)(
i)
of
this
guideline.

(
vii)
Humane
practices.
Moribund
animals
killed
for
humane
reasons
are
considered
in
the
same
way
as
animals
that
died
on
test.
If
an
animal
unexpectedly
dies
late
in
the
study
and
there
are
other
survivors
at
that
dose
or
above,
it
is
appropriate
to
stop
dosing
and
observe
all
animals
to
see
if
other
animals
will
also
die
during
a
similar
observation
period.
11
If
subsequent
survivors
also
die,
and
it
appears
that
all
dose
levels
exceed
the
LD50
it
would
be
most
appropriate
to
start
the
study
again
beginning
at
least
two
steps
below
the
lowest
dose
with
deaths
(
and
increasing
the
observation
period)
since
the
technique
is
most
accurate
when
the
starting
dose
is
below
the
LD50.
If
subsequent
animals
survive
at
or
above
the
dose
of
the
animal
that
dies,
it
is
not
necessary
to
change
the
dose
progression
since
the
information
from
the
animal
that
has
now
died
will
be
included
into
the
calculations
as
a
death
at
a
lower
dose
than
subsequent
survivors,
pulling
the
LD50
down.

(
j)
Observations.
Animals
are
observed
individually
at
least
once
during
the
first
30
minutes
after
dosing,
periodically
during
the
first
24
hours
(
with
special
attention
given
during
the
first
4
hours),
and
daily
thereafter,
for
a
total
of
14
days,
except
where
they
need
to
be
removed
from
the
study
and
humanely
killed
for
animal
welfare
reasons
or
are
found
dead.
However,
the
duration
of
observation
should
not
be
fixed
rigidly
It
should
be
determined
by
the
toxic
reactions
and
time
of
onset
and
length
of
recovery
period,
and
may
thus
be
extended
when
considered
necessary
The
times
at
which
signs
of
toxicity
appear
and
disappear
are
important
especially
if
there
is
a
tendency
for
toxic
signs
to
be
delayed
(
see
paragraph
(
n)(
15)
of
this
guideline).
All
observations
of
toxic
signs
are
systematically
recorded
with
individual
records
being
maintained
for
each
animal.
Additional
observations
will
be
necessary
if
the
animals
continue
to
display
signs
of
toxicity.

(
1)
Toxic
signs.
Observations
should
include
changes
in
skin
and
fur,
eyes
and
mucous
membranes,
and
also
respiratory,
circulatory,
autonomic
and
central
nervous
systems,
and
somatomotor
activity
and
behavior
pattern
Attention
should
be
directed
to
observations
of
tremors,
convulsions,
salivation,
diarrhea,
lethargy,
sleep
and
coma.
The
principles
and
criteria
summarized
in
the
Humane
Endpoints
Guidance
Document
(
see
paragraph
(
n)(
11)
of
this
guideline)
should
be
taken
into
consideration.
Animals
found
in
a
moribund
condition
and
animals
showing
severe
pain
and
enduring
signs
of
severe
distress
should
be
humanely
killed.
When
animals
are
killed
for
humane
reasons
or
found
dead,
the
time
of
death
should
be
recorded
as
precisely
as
possible.

(
2)
Body
weight.
Individual
weights
of
animals
should
be
determined
shortly
before
the
test
substance
is
administered
and
at
least
weekly
thereafter
Weight
changes
should
be
calculated
and
recorded.
At
the
end
of
the
test
surviving
animals
are
weighed
and
then
humanely
killed.

(
3)
Pathology.
All
animals
(
including
those
which
die
during
the
test
or
are
removed
from
the
study
for
animal
welfare
reasons)
should
be
subjected
to
gross
necropsy.
All
gross
pathological
changes
should
be
recorded
for
each
animal.
Microscopic
examination
of
organs
showing
evidence
of
gross
pathology
in
animals
surviving
24
or
more
hours
after
the
12
initial
dosing
may
also
be
considered
because
it
may
yield
useful
information

(
k)
Data
and
reporting
 
(
1)
Data.
Individual
animal
data
should
be
provided.
Additionally,
all
data
should
be
summarized
in
tabular
form,
showing
for
each
test
dose
the
number
of
animals
used,
the
number
of
animals
displaying
signs
of
toxicity
(
see
paragraph
(
n)(
15)
of
this
guideline
the
number
of
animals
found
dead
during
the
test
or
killed
for
humane
reasons,
time
of
death
of
individual
animals,
a
description
and
the
time
course
of
toxic
effects
and
reversibility,
and
necropsy
findings.
A
rationale
for
the
starting
dose
and
the
dose
progression
and
any
data
used
to
support
this
choice
should
be
provided.

(
2)
Calculation
of
LD50
for
the
main
test
 
(
i)
Maximum
likelihood
The
LD50
is
calculated
using
the
maximum
likelihood
method,
except
in
the
exceptional
cases
described
in
paragraphs
(
k)(
2)(
ii)
and
(
m)(
3)
of
this
guideline.
The
Agency­
developed
software
program
(
AOT425StatPgm)
available
on
EPA's
Internet
Web
site
at
http://
www.
epa.
gov/
oppfead1/
harmonized
should
be
used
to
perform
this
calculation
The
following
statistical
details
may
be
helpful
in
implementing
the
maximum
likelihood
calculations
suggested
(
with
an
assumed
sigma).
All
deaths,
whether
immediate
or
delayed
or
humane
kills,
are
incorporated
for
the
purpose
of
the
maximum
likelihood
analysis.
Following
Dixon
(
see
paragraph
(
n)(
5)
of
this
guideline),
the
likelihood
function
is
written
as
follows:

L
=
L1
L2
....
Ln
,

where
L
is
the
likelihood
of
the
experimental
outcome,
given
µ
and
sigma,
and
n
the
total
number
of
animals
tested.

Li
=
1
­
F(
Zi)
if
the
ith
animal
survived,
or
Li
=
F(
Zi)
if
the
ith
animal
died,

where
F
=
cumulative
standard
normal
distribution,

Zi
=
[
log(
di)
­
µ
]
/
sigma
di
=
dose
given
to
the
ith
animal,
and
sigma
=
standard
deviation
in
log
units
of
dose
(
which
is
not
the
log
standard
deviation).

An
estimate
of
the
log
of
the
true
LD50
is
given
by
the
value
of
µ
that
maximizes
the
likelihood
L
(
see
paragraph
(
k)(
2)(
iii)
of
this
guideline
13
An
estimate
of
sigma
of
0.5
is
used
unless
a
better
generic
or
casespecific
value
is
available.

(
ii)
Special
circumstances.
Under
some
circumstances,
statistical
computation
will
not
be
possible
or
will
likely
give
erroneous
results.
Special
means
to
determine/
report
an
estimated
LD50
are
available
for
these
circumstances
as
described
in
the
following
paragraphs
(
k)(
2)(
ii)(
A),
(
k)(
2)(
ii)(
B),
and
(
k)(
2)(
ii)(
C).
If
none
of
these
situations
occurs,
then
the
LD50
is
calculated
using
the
maximum
likelihood
method.

(
A)
If
testing
stopped
based
on
the
criterion
in
paragraph
(
i)(
3)(
iii)(
C)
of
this
guideline
(
i.
e.,
a
boundary
dose
was
tested
repeatedly),
or
if
the
upper
bound
dose
ended
testing,
then
the
LD50
is
reported
to
be
above
the
upper
bound.

(
B)
If
all
the
dead
animals
have
higher
doses
than
all
the
live
animals
(
or
if
all
live
animals
have
higher
doses
than
all
the
dead
animals,
although
this
is
practically
unlikely),
then
the
LD50
is
between
the
doses
for
the
live
and
the
dead
animals.
These
observations
give
no
further
information
on
the
exact
value
of
the
LD50.
Still,
a
maximum
likelihood
LD50
estimate
can
be
made
provided
there
is
a
prior
value
for
sigma.
The
LD50
estimate
is
only
as
good
as
the
validity
of
the
assumed
signa.
However,
Case
3
as
described
in
paragraph
(
m)(
3)(
iii)
of
this
guideline
and
here
is
most
likely
to
occur
because
the
dose
progression
(
based
on
the
assumed
signma)
is
too
wide.
The
stopping
criterion
in
paragraph
(
i)(
3)(
iii)(
C)
describes
one
such
circumstance.

(
C)
If
the
live
and
dead
animals
have
only
one
dose
in
common
and
all
the
other
dead
animals
have
higher
doses
and
all
the
other
live
animals
lower
doses,
or
vice
versa,
then
the
LD50
equals
their
common
dose.
If
a
closely
related
substance
is
tested,
testing
should
proceed
with
a
smaller
dose
progression.

(
iii)
Maximum
likelihood
calculation.
Maximum
likelihood
calculation
should
be
performed
using
a
dedicated
program
developed
by
and
available
from
EPA
(
see
paragraph
(
n)(
16)
of
this
guideline).
If
other
computer
programs
are
used,
the
laboratory
should
take
care
in
handling
special
cases
described
in
this
guideline
and
the
documentation
of
test
performance
available
on
EPA's
Internet
Web
site
at
http://
www.
epa.
gov/
oppfead1/
harmonized.
Typical
instructions
for
these
packages
are
given
in
appendices
to
the
ASTM
Standard
E
1163­
87
(
see
paragraph
(
n)(
9)
of
this
guideline).
(
The
sigma
used
in
the
BASIC
program
in
(
see
paragraph
(
n)(
9)
of
this
guideline)
will
need
to
be
edited
to
reflect
the
parameters
of
the
UDP.)
The
program's
output
is
an
estimate
of
log
(
LD50)
and
its
standard
error.

(
iv)
Stopping
rule.
The
likelihood­
ratio
stopping
rule
in
paragraph
(
i)(
3)(
iii)(
C)
of
this
guideline
is
based
on
three
measures
of
test
progress,
that
are
of
the
form
of
the
likelihood
in
paragraph
(
k)(
2)
of
this
guideline,
14
with
different
values
for
µ
.
Comparisons
are
made
after
each
animal
tested
after
the
sixth
that
does
not
already
satisfy
the
criteria
in
paragraph
(
i)(
3)(
iii)(
A)
or
paragraph
(
i)(
3)(
iii)(
B)
guideline.
The
equations
for
the
likelihood­
ratio
criteria
are
provided
by
following
the
steps
in
paragraph
(
m)(
2)(
vii)
of
this
guideline.
These
comparisons
are
most
readily
performed
in
an
automated
manner
and
can
be
executed
repeatedly,
for
instance
by
a
spreadsheet
routine
such
as
that
also
provided
in
paragraph
(
m)(
2)(
vii)
of
this
guideline.
If
the
criterion
is
met,
testing
stops
and
the
LD50
can
be
calculated
by
the
maximum
likelihood
method.

(
3)
Computation
of
CI.
(
i)
Following
the
main
test
and
estimated
LD50
calculation,
it
may
be
possible
to
compute
interval
estimates
for
the
LD50.
The
Agency­
developed
software
program
AOT425StatPgm
will
perform
the
calculations.
Any
of
these
CIs
provides
valuable
information
on
the
reliability
and
utility
of
the
main
test
that
was
conducted.
A
wide
CI
indicates
that
there
is
more
uncertainty
associated
with
the
estimated
LD50.
In
this
case,
the
reliability
of
the
estimated
LD50
is
low
and
the
usefulness
of
the
estimated
LD50
may
be
marginal.
A
narrow
interval
indicates
that
there
is
relatively
little
uncertainty
associated
with
the
estimated
LD50.
In
this
case,
the
reliability
of
the
estimated
LD50
is
high
and
the
usefulness
of
the
estimated
LD50
is
good.
This
means
that
if
the
main
test
were
to
be
repeated,
the
new
estimated
LD50
is
expected
to
be
close
to
the
original
estimated
LD50
and
both
of
these
estimates
are
expected
to
be
close
to
the
true
LD50.

(
ii)
Depending
on
the
outcome
of
the
main
test,
one
of
two
different
types
of
interval
estimates
of
the
true
LD50
is
calculated:

(
A)
When
at
least
three
different
doses
have
been
tested
and
the
middle
dose
has
at
least
one
animal
that
survived
and
one
animal
that
died,
a
profile­
likelihood­
based
computational
procedure
is
used
to
obtain
a
CI
that
is
expected
to
contain
the
true
LD50
95%
of
the
time.
However,
because
small
numbers
of
animals
are
expected
to
be
used,
the
actual
level
of
confidence
is
generally
not
exact
(
see
paragraph
(
n)(
19)
of
this
guideline
The
random
stopping
rule
improves
the
ability
of
the
test
overall
to
respond
to
varying
underlying
conditions,
but
also
causes
the
reported
level
of
confidence
and
the
actual
level
of
confidence
to
differ
somewhat
(
see
paragraph
(
n)(
18)
of
this
guideline).

(
B)
If
all
animals
survive
at
or
below
a
given
dose
level
and
all
animals
die
when
dosed
at
the
next
higher
dose
level,
an
interval
is
calculated
that
has
as
its
lower
limit
the
highest
dose
tested
where
all
the
animals
survive
and
has
as
its
upper
limit
the
dose
level
where
all
the
animals
died.
This
interval
is
labeled
as
``
approximate.''
The
exact
confidence
level
associated
with
this
interval
cannot
be
specifically
determined.
However
because
this
type
of
response
would
only
occur
when
the
dose­
response
is
steep,
in
most
cases,
the
true
LD50
is
expected
to
be
contained
15
within
the
calculated
interval
or
be
very
close
to
it.
This
interval
will
be
relatively
narrow
and
sufficiently
accurate
for
most
practical
use.

(
iii)
In
some
instances,
CIs
are
reported
as
infinite,
through
including
either
zero
at
the
lower
end
or
infinity
at
the
upper
end,
or
both.
Such
intervals
may
occur,
for
example,
when
the
response
profile
is
relatively
flat
or
relatively
uncertain.

(
iv)
Implementing
this
set
of
procedures
requires
specialized
computation
which
is
either
by
use
of
a
dedicated
program
to
be
available
through
the
Environmental
Protection
Agency
(
EPA)
or
OECD
or
developed
following
technical
details
available
from
the
EPA
or
OECD.
Achieved
coverage
of
these
intervals
and
properties
of
the
dedicated
program
are
described
in
a
report
(
see
paragraph
(
n)(
16)
of
this
guideline)
also
available
through
the
EPA.
Paragraph
(
m)(
3)
of
this
guideline
provides
information
on
choice
of
dose
progression
and
initial
dose
level
for
the
UDP
and
describes
test
performance
under
a
variety
of
circumstances.

(
l)
Test
reporting.
The
test
report
must
include
the
following
information

(
1)
Test
substance:

(
i)
Physical
nature,
purity
and
physicochemical
properties
(
including
isomerization);

(
ii)
Identification
data.

(
2)
Vehicle
(
if
appropriate):
Justification
for
choice
of
vehicle,
if
other
than
water.

(
3)
Test
animals:

(
i)
Species/
strain
used;

(
ii)
Microbiological
status
of
the
animals,
when
known;

(
iii)
Number,
age
and
sex
of
animals;

(
iv)
Rationale
for
use
of
males
instead
of
females;

(
v)
Source,
housing
conditions,
diet,
etc.;

(
vi)
Individual
weights
of
animals
at
the
start
of
the
test,
at
day
7,
and
at
day
14.

(
4)
Test
conditions:

(
i)
Rationale
for
initial
dose
level
selection,
dose
progression
factor
and
for
follow­
up
dose
levels;

(
ii)
Details
of
test
substance
formulation;
16
(
iii)
Details
of
the
administration
of
the
test
substance;

(
iv)
Details
of
food
and
water
quality
(
including
diet
type/
source,
water
source).

(
5)
Results:

(
i)
Body
weight/
body
weight
changes;

(
ii)
Tabulation
of
response
data
by
sex
(
if
both
sexes
are
used)
and
dose
level
for
each
animal
(
i.
e.,
animals
showing
signs
of
toxicity
including
nature,
severity,
duration
of
effects,
and
mortality);

(
iii)
Time
course
of
onset
of
signs
of
toxicity
and
whether
these
were
reversible
for
each
animal;

(
iv)
Necropsy
findings
and
any
histopathological
findings
for
each
animal,
if
available;

(
v)
LD50
and
CI
(
which
the
AOT425StatPgm
software
package
uses);

(
vi)
Statistical
treatment
of
results
(
description
of
computer
routine
used
and
spreadsheet
tabulation
of
calculations).
If
other
than
Agency­
supplied
software
is
used,
give
explanation
of
now
the
program
was
verified
against
Agency
software.

(
6)
Discussion
and
interpretation
of
results.

(
7)
Conclusions.

(
m)
Additional
guidance
for
toxicologists
 
(
1)
Dosing
procedure
 
dose
sequence
for
main
test.
(
i)
Up­
and­
down
dosing
procedure.
For
each
run,
animals
are
dosed,
one
at
a
time,
usually
at
48­
hour
intervals.
The
first
animal
receives
a
dose
a
step
below
the
level
of
the
best
estimate
of
the
LD50.
This
selection
reflects
an
adjustment
for
a
tendency
to
bias
away
from
the
LD50
in
the
direction
of
the
initial
starting
dose
in
the
final
estimate
(
see
paragraph
(
e)(
2)(
ii)
of
the
guideline).
The
overall
pattern
of
outcomes
is
expected
to
stabilize
as
dosing
is
adjusted
for
each
subsequent
animal.
Paragraph
(
m)(
1)(
iii)
of
this
guideline
provides
further
guidance
for
choice
of
dose
spacing
factor.

(
ii)
Default
dose
progression.
Once
the
starting
dose
and
dose
spacing
are
decided,
the
toxicologist
should
list
all
possible
doses
including
the
upper
bound
(
usually
2000
or
5000
mg/
kg).
Doses
that
are
close
to
the
upper
bound
should
be
removed
from
the
progression.
The
stepped
nature
of
the
UDP
design
provides
for
the
first
few
doses
to
function
as
a
selfadjusting
sequence.
Because
of
the
tendency
for
positive
bias,
in
the
event
that
nothing
is
known
about
the
substance,
a
starting
dose
of
175
mg/
kg
is
recommended.
If
the
default
procedure
is
to
be
used
for
the
main
test,
dosing
will
be
initiated
at
175
mg/
kg
and
doses
will
be
spaced
by
a
factor
of
0.5
on
a
log
dose
scale.
The
doses
to
be
used
include
1.75,
17
5.5,
17.5,
55,
175,
550,
2000
or,
for
specific
regulatory
needs,
1.75,
5.5,
17.5,
55,
175,
550,
1750,
5000.
For
certain
highly
toxic
substances,
the
dosing
sequence
may
need
to
be
extended
to
lower
values.

(
iii)
In
the
event
a
dose
progression
factor
other
than
the
default
is
deemed
suitable,
the
following
Table
1
provides
dose
progressions
for
whole
number
multiples
of
slope,
from
1
to
8.
(
See
paragraph
(
m)(
3)
of
this
guideline
for
discussion
of
influence
of
dose
progression
on
test
performance
18
Table
1.
 
Dose
Progressions
for
UDP
(
Choose
a
Slope
and
Read
Down
the
Column.
All
doses
in
mg/
kg
body
weight)

Slope
=
1
2
3
4
5
6
7
8
0.175*
0.175*
0.175*
0.175*
0.175*
0.175*
0.175*
0.175*
......................
......................
......................
......................
......................
......................
0.243*
0.233*
......................
......................
......................
......................
0.28
0.26
......................
......................
......................
......................
......................
0.31
......................
......................
0.34
0.31
......................
......................
0.38
......................
......................
0.38
......................
......................
......................
......................
......................
......................
......................
......................
......................
0.41
......................
......................
......................
......................
0.44
......................
0.47
......................
......................
0.55
......................
.55
......................
0.55
......................
0.55
......................
......................
......................
0.70
......................
0.65
......................
......................
......................
......................
......................
......................
......................
0.74
......................
......................
.81
......................
......................
.81
......................
......................
......................
......................
......................
0.98
......................
......................
0.91
0.98
......................
......................
......................
......................
110
1.19
......................
......................
......................
......................
......................
......................
......................
......................
1.26
1.31
1.75
1.75
1.75
1.75
1.75
1.75
1.75
1.75
......................
......................
......................
......................
......................
......................
2.43
2.33
......................
......................
......................
......................
2.8
2.6
......................
......................
......................
......................
......................
3.1
......................
......................
3.4
3.1
......................
......................
3.8
......................
......................
3.8
......................
......................
......................
......................
......................
......................
4.4
......................
......................
4.1
......................
......................
......................
......................
......................
......................
4.7
......................
......................
5.5
......................
5.5
5.5
......................
5.5
......................
......................
......................
......................
7.0
......................
6.5
......................
......................
......................
......................
......................
......................
......................
......................
7.4
......................
......................
8.1
......................
......................
8.1
......................
......................
......................
......................
......................
9.8
......................
......................
9.1
9.8
......................
......................
......................
......................
11.0
11.9
......................
......................
......................
......................
......................
......................
......................
......................
12.6
13.1
17.5
17.5
17.5
17.5
17.5
17.5
17.5
17.5
......................
......................
......................
......................
......................
......................
24.3
23.3
......................
......................
......................
......................
28
26
......................
......................
......................
......................
......................
31
......................
......................
34
31
......................
......................
38
......................
......................
38
......................
......................
......................
......................
......................
......................
44
......................
......................
41
......................
......................
......................
......................
......................
......................
47
......................
......................
55
......................
55
......................
55
......................
55
......................
......................
......................
......................
......................
......................
65
......................
......................
......................
......................
......................
70
......................
......................
74
......................
......................
81
......................
......................
81
......................
......................
......................
......................
......................
98
......................
......................
91
98
......................
......................
......................
......................
110
119
......................
......................
......................
......................
......................
......................
......................
......................
126
131
175
175
175
175
175
175
175
175
......................
......................
......................
......................
......................
......................
243
233
......................
......................
......................
......................
280
260
......................
......................
......................
......................
......................
310
......................
......................
340
310
......................
......................
380
......................
......................
380
......................
......................
......................
......................
......................
......................
440
......................
......................
410
......................
......................
......................
......................
......................
......................
470
......................
......................
550
......................
550
......................
550
......................
550
......................
......................
......................
......................
......................
......................
650
......................
......................
......................
......................
......................
700
......................
......................
740
......................
......................
810
......................
......................
810
......................
......................
......................
......................
......................
980
......................
......................
910
980
......................
......................
......................
......................
1100
1190
......................
......................
......................
......................
......................
......................
......................
......................
1260
1310
1750
1750
1750
1750
1750
1750
1750
1750
......................
......................
......................
......................
......................
......................
2430
2330
......................
......................
......................
......................
2800
2600
......................
......................
......................
......................
......................
3100
......................
......................
......................
3100
......................
......................
......................
......................
......................
3800
3400
......................
......................
......................
......................
......................
......................
......................
......................
4100
5000
5000
5000
5000
5000
5000
5000
5000
*
If
lower
doses
are
needed,
continue
progressions
to
a
lower
dose
(
2)
Computations
for
the
likelihood­
ratio
stopping
rules.
(
i)
As
described
in
paragraph
(
i)(
3)(
iii)
of
this
guideline,
the
main
test
may
be
completed
on
the
basis
of
the
first
of
three
stopping
criteria
to
occur.
In
any
case,
even
if
none
of
the
stopping
criteria
is
satisfied,
dosing
would
stop
when
15
animals
are
dosed.
Tables
2,
4,
and
6
in
paragraphs
(
m)(
2)(
ii),
(
m)(
2)(
iii),
and
(
m)(
2)(
iv),
respectively,
of
this
guideline
illustrate
examples
where
testing
has
started
with
no
information,
so
the
rec­
19
ommended
default
starting
value,
175
mg/
kg,
and
the
recommended
default
dose
progression
factor,
3.2
or
one
half
log,
have
been
used.
Tables
3,
5,
and
7
in
paragraphs
(
m)(
2)(
ii),
(
m)(
2)(
iii),
and
(
m)(
2)(
iv),
respectively
illustrate
how
Tables
2,
4,
and
6,
respectively,
would
appear
in
the
dedicated
program
referenced
in
paragraph
(
k)(
3)(
iv)
(
see
also
paragraph
(
n)(
16)).

(
ii)
The
following
Tables
2
and
3
show
how
the
main
test
would
stop
if
3
animals
have
survived
at
the
limit
dose
of
5000
mg/
kg.
(
This
example
illustrates
situations
where
a
limit
test
was
not
thought
appropriate
a
priori).
Table
2.
Example
of
Stopping
Criterion
in
Paragraph
(
i)(
3)(
iii)(
A)
using
5000
mg/
kg.

1
2
3
4
5
6
7
8
9
10
11
12
Step
(
I)
nclude;
Dose
(
X)
response
Included
log10
LD50
=
#
DIV/
0!
LD50
=
#
DIV/
0!
LD50
=
#
DIV/
0!

(
E)
xclude
(
O)
non­
resp.
in
nominal
Dose
Prob.
of
likelihood
Prob.
of
likelihood
Prob.
of
likelihood
n
response
contribn.
response
contribn.
response
contribn.

OK
(
ln
Li
)
(
ln
Li
)
(
ln
Li
)

1
I
175
O
no
2.2430
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!

2
I
550
O
no
2.7404
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!

3
I
1750
O
no
3.2430
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!

4
I
5000
O
no
3.6990
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!

5
I
5000
O
no
3.6990
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!

6
I
5000
O
no
3.6990
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!
#
DIV/
0!

7
E
­
­
­
­
­
­
­

8
E
­
­
­
­
­
­
­

9
E
­
­
­
­
­
­
­

10
E
­
­
­
­
­
­
­

11
E
­
­
­
­
­
­
­

12
E
­
­
­
­
­
­
­

13
E
­
­
­
­
­
­
­

14
E
­
­
­
­
­
­
­

15
E
­
­
­
­
­
­
­

Nominal
Sample
size
=
0
Actual
number
tested
=
6
Calculated
maximum
likelihood
estimate
of
LD50
=
none
Stop
after
animal
#
6
because
3
animals
survive
at
limit
of
5000
mg/
kg
(#
4­#
6).
Ignore
all
calculation
cells.
No
reversal
in
direction
of
response.

Maximum
Likelihood
Calculations
cannot
be
completed.
LD50
is
greater
than
5000
mg/
kg.
Table
3.
Example
of
Stopping
Criterion
in
Paragraph
(
i)(
3)(
iii)(
A)
of
this
Guideline
Using
5000
mg/
kg
22
(
iii)
The
following
Tables
4
and
5
show
how
a
particular
sequence
of
5
reversals
in
6
tested
animals
could
occur
and
allow
test
completion.
Table
4.
Example
of
Stopping
Criterion
in
Paragraph
(
i)(
3)(
iii)(
B).

1
2
3
4
5
6
7
8
9
10
11
12
Step
(
I)
nclude;
Dose
(
X)
response
Included
log10
LD50
=
31.0
LD50
=
12.4
LD50
=
77.6
(
E)
xclude
(
O)
non­
resp.
in
nominal
Dose
Prob.
of
likelihood
Prob.
of
likelihood
Prob.
of
likelihood
n
response
contribn.
response
contribn.
response
contribn.

OK
(
ln
Li
)
(
ln
Li
)
(
ln
Li
)

1
I
175
X
no
2.2430
0.9335
­
0.0688
0.9892
­
0.0108
0.7602
­
0.2742
2
I
55
X
yes
1.7404
0.6905
­
0.3703
0.9020
­
0.1031
0.3826
­
0.9607
3
I
17.5
O
yes
1.2430
0.3095
­
0.3703
0.6174
­
0.9607
0.0980
­
0.1031
4
I
55
X
yes
1.7404
0.6905
­
0.3703
0.9020
­
0.1031
0.3826
­
0.9607
5
I
17.5
O
yes
1.2430
0.3095
­
0.3703
0.6174
­
0.9607
0.0980
­
0.1031
6
I
55
X
yes
1.7404
0.6905
­
0.3703
0.9020
­
0.1031
0.3826
­
0.9607
7
I
17.5
O
yes
1.2430
0.3095
­
0.3703
0.6174
­
0.9607
0.0980
­
0.1031
8
E
­
­
­
­
­
­
­

9
E
­
­
­
­
­
­
­

10
E
­
­
­
­
­
­
­

11
E
­
­
­
­
­
­
­

12
E
­
­
­
­
­
­
­

13
E
­
­
­
­
­
­
­

14
E
­
­
­
­
­
­
­

15
E
­
­
­
­
­
­
­

Nominal
Sample
size
=
6
Actual
number
tested
=
7
Dose­
averaging
estimator
31.02
log10
=
1.492
log­
likelihood
sums:
­
2.2906
­
3.2021
­
3.4655
likelihoods:
0.1012
0.0407
0.0313
likelihood
ratios:
2.4880
3.2378
Individual
ratios
exceed
critical
value?
critical=
2.5
FALSE
TRUE
Both
ratios
exceed
critical
value?
FALSE
Calculated
maximum
likelihood
estimate
of
LD50
=
29.6
Stop
after
animal
#
7
because
5
reversals
in
6
consecutive
animals
tested
(#
2­#
7).
Automated
calculation;
not
relevant
to
this
case.

Final
estimate
obtained
from
Maximum
Likelihood
Calculations
Table
5.
Example
of
Stopping
Criterion
in
Paragraph
(
i)(
3)(
iii)(
B)
of
this
Guideline.
25
(
iv)
Finally,
the
following
Tables
6
and
7
illustrate
a
situation
several
animals
into
a
test,
where
neither
the
criterion
in
paragraph
(
i)(
3)(
iii)(
A)
nor
the
criterion
in
paragraph
(
i)(
3)(
iii)(
B)
of
this
guideline
has
been
met,
a
reversal
of
response
has
occurred
followed
by
4
tested
animals,
and,
consequently,
the
criterion
in
paragraph
(
i)(
3)(
iii)(
C)
of
this
guideline
must
be
evaluated
as
well.
Table
6.
Example
of
Stopping
Criterion
in
Paragraph
(
i)(
3)(
iii)(
C).

Assumed
slope
2
sigma
=
0.5
Parameters
of
convergence
criterion
critical
LR
2.5
Result:
The
LR
criterion
is
met
factor
of
LD50
2.5
1
2
3
4
5
6
7
8
9
10
11
12
Step
(
I)
nclude;
Dose
(
X)
response
Included
log10
Contrib.
to
LD50
=
1292.8
LD50
=
517.1
LD50
=
3232.0
(
E)
xclude
(
O)
non­
resp.
in
nominal
Dose
DAE
Prob.
of
likelihood
Prob.
of
likelihood
Prob.
of
likelihood
n
response
contribn.
response
contribn.
response
contribn.

OK
(
ln
Li
)
(
ln
Li
)
(
ln
Li
)

1
I
175
O
no
2.2430
0.0000
0.0412
­
0.0421
0.1733
­
0.1903
0.0057
­
0.0057
2
I
550
O
yes
2.7404
2.7404
0.2289
­
0.2600
0.5214
­
0.7368
0.0620
­
0.0640
3
I
1750
X
yes
3.2430
3.2430
0.6037
­
0.5046
0.8552
­
0.1564
0.2971
­
1.2138
4
I
550
O
yes
2.7404
2.7404
0.2289
­
0.2600
0.5214
­
0.7368
0.0620
­
0.0640
5
I
1750
X
yes
3.2430
3.2430
0.6037
­
0.5046
0.8552
­
0.1564
0.2971
­
1.2138
6
I
550
O
yes
2.7404
2.7404
0.2289
­
0.2600
0.5214
­
0.7368
0.0620
­
0.0640
7
I
1750
O
yes
3.2430
3.2430
0.6037
­
0.9257
0.8552
­
1.9323
0.2971
­
0.3525
8
I
5000
X
yes
3.6990
3.6990
0.8800
­
0.1279
0.9756
­
0.0247
0.6477
­
0.4344
9
I
1750
X
yes
3.2430
3.2430
0.6037
­
0.5046
0.8552
­
0.1564
0.2971
­
1.2138
10
E
­
0.0000
­
­
­
­
­
­

11
E
­
0.0000
­
­
­
­
­
­

12
E
­
0.0000
­
­
­
­
­
­

13
E
­
0.0000
­
­
­
­
­
­

14
E
­
0.0000
­
­
­
­
­
­

15
E
­
0.0000
­
­
­
­
­
­

Nominal
Sample
size
=
8
Actual
number
tested
=
9
Dose­
averaging
estimator
1292.78
log10
=
3.112
log­
likelihood
sums:
­
3.3894
­
4.8270
­
4.6260
likelihoods:
0.0337
0.0080
0.0098
likelihood
ratios:
4.2104
3.4436
Individual
ratios
exceed
critical
value?
critical=
2.5
TRUE
TRUE
Both
ratios
exceed
critical
value?
TRUE
Calculated
maximum
likelihood
estimate
of
LD50
=
1329.6
Stop
when
LR
criterion
is
first
met,
here
at
animal
#
9.

Check
LR
criterion
starting
at
animal
#
6.
Final
estimate
obtained
from
Maximum
Likelihood
Calculations
Table
7.
Example
of
Stopping
Criterion
in
Paragraph
(
i)(
3)(
iii)(
C)
of
this
Guideline.
28
(
v)
Criterion
in
paragraph
(
i)(
3)(
iii)(
C)
of
this
guideline
calls
for
a
likelihood­
ratio
stopping
rule
to
be
evaluated
after
testing
each
animal,
starting
with
the
fourth
tested
following
the
reversal.
Three
``
measures
of
test
progress''
are
calculated.
Technically,
these
measures
of
progress
are
likelihoods,
as
recommended
for
the
maximum­
likelihood
estimation
of
the
LD50.
The
procedure
is
closely
related
to
calculation
of
a
CI
by
a
likelihood­
based
procedure.

(
vi)
The
basis
of
the
procedure
is
that
when
enough
data
have
been
collected,
a
point
estimate
of
the
LD50
should
be
more
strongly
supported
than
values
above
and
below
the
point
estimate,
where
statistical
support
is
quantified
using
likelihood.
Therefore
three
likelihood
values
are
calculated
A
likelihood
at
an
LD50
point
estimate
(
called
the
rough
estimate
or
dose­
averaging
estimate
in
the
example),
a
likelihood
at
a
value
below
the
point
estimate,
and
a
likelihood
at
a
value
above
the
point
estimate.
Specifically,
the
low
value
is
taken
to
be
the
point
estimate
divided
by
2.5
and
the
high
value
is
taken
to
be
the
point
estimate
multiplied
by
2.5.

(
vii)
The
likelihood
values
are
compared
by
calculating
ratios
of
likelihoods,
and
then
determining
whether
these
likelihood­
ratios
(
LR)
exceed
a
critical
value.
Testing
stops
when
the
ratio
of
the
likelihood
for
the
point
estimate
exceeds
each
of
the
other
likelihoods
by
a
factor
of
2.5,
which
is
taken
to
indicate
relatively
strong
statistical
support
for
the
point
estimate.
Therefore
two
likelihood­
ratios
(
LRs)
are
calculated,
a
ratio
of
likelihoods
for
the
point
estimate
and
the
point
estimate
divided
by
2.5,
and
a
ratio
for
the
point
estimate
and
the
estimate
times
2.5.

(
viii)
The
calculations
are
easily
performed
in
any
spreadsheet
with
normal
probability
functions.
The
calculations
are
illustrated
in
Tables
6
and
7
in
paragraph
(
m)(
2)(
iv)
of
this
guideline,
which
is
structured
to
promote
spreadsheet
implementation.
The
computation
steps
are
illustrated
using
an
example
where
the
upper
limit
dose
is
5000
mg/
kg.

(
A)
Hypothetical
example
(
Tables
6
and
7
in
paragraph
(
m)(
2)(
iv)
of
this
guideline).
In
the
hypothetical
example
utilizing
an
upper
boundary
dose
of
5000
mg/
kg,
the
LR
stopping
criterion
was
met
after
nine
animals
had
been
tested.
The
first
``
reversal''
occurred
with
the
3rd
animal
tested.
The
LR
stopping
criterion
is
checked
when
four
animals
have
been
tested
following
the
reversal.
In
this
example,
the
fourth
animal
tested
following
the
reversal
is
the
seventh
animal
actually
tested.
Therefore,
for
this
example
the
spreadsheet
calculations
are
only
needed
after
the
seventh
animal
had
been
tested
and
the
data
could
be
entered
at
that
time.
Subsequently,
the
LR
stopping
criterion
would
have
been
checked
after
testing
the
seventh
animal,
the
eighth
animal,
and
the
ninth.
The
LR
stopping
criterion
is
first
satisfied
after
the
ninth
animal
is
tested
in
this
example.

(
1)
Enter
the
dose­
response
information
animal
by
animal.
29
(
i)
Column
1.
Steps
are
numbered
1
 
15.
No
more
than
15
animals
may
be
tested.

(
ii)
Column
2.
Place
an
I
in
this
column
as
each
animal
is
tested.

(
iii)
Column
3.
Enter
the
dose
received
by
the
ith
animal.

(
iv)
Column
4.
Indicate
whether
the
animal
responded
(
shown
by
an
X)
or
did
not
respond
(
shown
by
an
O).

(
2)
The
nominal
and
actual
sample
sizes.
The
nominal
sample
consists
of
the
two
animals
that
represent
the
first
reversal
(
here
the
second
and
third
animals),
plus
all
animals
tested
subsequently.
Here,
Column
5
indicates
whether
or
not
a
given
animal
is
included
in
the
nominal
sample.

(
i)
The
nominal
sample
size
(
nominal
n)
appears
in
Row
16.
This
is
the
number
of
animals
in
the
nominal
sample.
In
the
example,
nominal
n
is
8.

(
ii)
The
actual
number
tested
appears
in
Row
17.

(
3)
Rough
estimate
of
the
LD50.
The
geometric
mean
of
doses
for
the
animals
in
the
current
nominal
sample
is
used
as
a
rough
estimate
of
the
LD50
from
which
to
gauge
progress.
In
the
table,
this
is
called
the
``
dose­
averaging
estimator.''
It
is
updated
with
each
animal
tested.
This
average
is
restricted
to
the
nominal
sample
in
order
to
allow
for
a
poor
choice
of
initial
test
dose,
which
could
generate
either
an
initial
string
of
responses
or
an
initial
string
of
nonresponses.
(
However,
the
results
for
all
animals
are
used
in
the
likelihood
calculations
for
final
LD50
calculation
below.)
Recall
that
the
geometric
mean
of
n
numbers
is
the
product
of
the
n
numbers,
raised
to
a
power
of
1/
n.

(
i)
The
dose­
averaging
estimate
appears
in
Row
18
(
e.
g.,
(
175
*
550
*
...
*
1750)
1/
8
=
1292.78).

(
ii)
Row
19
shows
the
logarithm
(
base
10)
of
the
value
in
Row
18
(
e.
g.,
log10
1292.8
=
3.112).

(
4)
Likelihood
for
the
rough
LD50
estimate.

(
i)
``
Likelihood''
is
a
statistical
measure
of
how
strongly
the
data
support
an
estimate
of
the
LD50
or
other
parameter.
Ratios
of
likelihood
values
can
be
used
to
compare
how
well
the
data
support
different
estimates
of
the
LD50.

(
ii)
In
Column
8
calculate
the
likelihood
for
Step
C's
rough
LD50
estimate.
The
likelihood
(
Row
21)
is
the
product
of
likelihood
contributions
for
individual
animals
(
see
paragraph
(
k)(
2)
of
this
guideline).
The
likelihood
contribution
for
the
ith
animal
is
denoted
Li.
30
(
iii)
Column
7.
Enter
the
estimate
of
the
probability
of
response
at
dose
di,
denoted
Pi.
Pi
is
calculated
from
a
dose­
response
curve.
Note
that
the
parameters
of
a
probit
dose­
response
curve
are
the
slope
and
the
LD50,
so
values
are
needed
for
each
of
those
parameters.
For
the
LD50
the
dose­
averaging
estimate
from
Row
18
is
used.
For
the
slope
in
this
example
the
default
value
of
2
is
used.
The
following
steps
may
be
used
to
calculate
the
response
probability
Pi.

1.
Calculate
the
base­
10
log
of
dose
di
(
Column
6).

2.
For
each
animal
calculate
the
z­
score,
denoted
Zi
(
not
shown
in
the
table),
using
the
formulae
sigma
=
1
/
slope,

Zi
=
(
log10(
di)
­
log10(
LD50))
/
sigma
For
example,
for
the
first
animal
(
Row
1),

sigma
=
1
/
2
Z1
=
(
2.243
­
3.112)
/
0.500
=
­
1.738
3.
For
the
ith
dose
the
estimated
response
probability
is
Pi
=
F(
Zi)

where
F
denotes
the
cumulative
distribution
function
for
the
standard
normal
distribution
(
i.
e.,
the
normal
distribution
with
mean
0
and
variance
1).

For
example
(
Row
1),

P1
=
F(­
1.738)
=
0.0412
The
function
F
(
or
something
very
close)
is
ordinarily
what
is
given
for
the
normal
distribution
in
statistical
tables,
but
the
function
is
also
widely
available
as
a
spreadsheet
function.
It
is
available
under
different
names,
for
example
the
@
NORMAL
function
of
Lotus
1­
2­
3
(
see
paragraph
(
n)(
19)
of
this
guideline)
and
the
@
NORMDIST
function
in
Excel
(
see
paragraph
(
n)(
20)
of
this
guideline).
To
confirm
that
you
have
used
correctly
the
function
available
in
your
software,
you
may
wish
to
verify
familiar
values
such
as
F(
1.96)
 
0.975
or
F(
1.64)
 
0.95.

(
iv)
Column
8.
Calculate
the
natural
log
of
the
likelihood
contribution
(
ln(
Li)).
Li
is
simply
the
probability
of
the
response
that
actually
was
observed
for
the
ith
animal:

Responding
animals:
ln(
Li)
=
ln(
Pi)

Non­
responding
animals:
ln(
Li)
=
ln(
1
­
Pi)
31
Note
that
here
the
natural
logarithm
(
ln)
is
used,
whereas
elsewhere
the
base­
10
(
common)
logarithm
was
used.
These
choices
are
what
are
ordinarily
expected
in
a
given
context.

The
steps
above
are
performed
for
each
animal.
Finally:

Row
20:
Sum
the
log­
likelihood
contributions
in
Column
8.

Row
21:
Calculate
the
likelihood
by
applying
the
exp
function
applied
to
the
log­
likelihood
value
in
Row
20
(
e.
g.,
exp(­
3.389)
=
e­
3.389
=
0.0337).

(
5)
Calculate
likelihoods
for
two
dose
values
above
and
below
the
rough
estimate.
If
the
data
permit
a
precise
estimate,
then
one
expects
the
likelihood
should
be
high
if
the
estimate
is
a
reasonable
estimate
of
the
LD50,
relative
to
likelihoods
for
values
distant
from
this
estimate.
Compare
the
likelihood
for
the
dose­
averaging
estimate
(
1292.8,
Row
18)
to
values
differing
by
a
factor
of
2.5
from
that
value
(
i.
e.,
to
1292.8*
2.5
and
1292.8/
2.5).
The
calculations
(
displayed
in
Columns
9
 
12)
are
carried
out
in
a
fashion
similar
to
those
described
above,
except
that
the
values
517.1
(=
1292.8/
2.5)
and
3232.0
(=
1292.8*
2.5)
have
been
used
for
the
LD50,
instead
of
1292.8.
The
likelihoods
and
log­
likelihoods
are
displayed
in
Rows
20
 
21.

(
6)
Calculate
likelihood­
ratios.
The
three
likelihood
values
(
Row
21)
are
used
to
calculate
two
likelihood­
ratios
(
Row
22).
A
likelihood­
ratio
is
used
to
compare
the
statistical
support
for
the
estimate
of
1292.8
to
the
support
for
each
of
the
other
values,
517.1
and
3232.0.
The
two
likelihood
ratios
are
therefore:

LR1
=
[
likelihood
of
1292.8]
/
[
likelihood
of
517.1]

=
0.0337
/
0.0080
=
4.21
and
LR2
=
[
likelihood
of
1292.8]
/
[
likelihood
of
3232.0]

=
0.0337
/
0.0098
=
3.44
(
7)
Determine
if
the
likelihood­
ratios
exceed
the
critical
value.
High
likelihood­
ratios
are
taken
to
indicate
relatively
high
support
for
the
point
estimate
of
the
LD50.
Both
of
the
likelihood­
ratios
calculated
in
paragraph
(
m)(
2)(
viii)(
A)(
6)
of
this
guideline
(
4.21
and
3.44)
exceed
the
critical
likelihood
ratio,
which
is
2.5.
Therefore
the
LR
stopping
criterion
is
satisfied
and
testing
stops.
This
is
indicated
by
a
TRUE
in
Row
24
and
a
note
at
the
top
of
the
example
spreadsheet
that
the
LR
criterion
is
met.
Determination
of
the
point
estimate
and
CI
is
carried
out
separately.
32
(
B)
[
Reserved]

(
3)
Performance
of
the
UDP.
This
section
addresses
choice
of
dose
progression
and
initial
dose
level
for
the
UDP
and
describes
the
performance
of
the
test
under
a
variety
of
circumstances.
A
companion
document
titled
``
Toxicology
Summary:
Performance
of
the
Up­
and­
Down
Procedure
provides
assistance
to
the
user
in
interpretation
of
the
test
results
and
is
available
on
the
ICCVAM
web
site
at
http://
iccvam.
niehs.
nih.
gov/
methods/
udpdocs/
udprpt/
udp
 
ciprop.
htm.
The
statistical
methods
applied
will
depend
upon
the
case
into
which
the
test
response
patterns
fall
(
see
Table
8
in
paragraph
(
m)(
3)(
iii)
of
this
guideline.

(
i)
Adjusting
the
dose
progression
and
initial
dose.
For
optimum
performance
of
the
UDP,
the
dose
progression
used
should
be
based
on
an
accurate
prior
estimate
of
sigma.
The
following
two
cases
describe
the
outcome
when
an
accurate
estimate
of
sigma
is
not
available.
In
addition,
to
account
conservatively
for
any
bias
in
the
LD50
estimate,
it
is
essential
that
dosing
be
initiated
below
the
actual
LD50.

(
A)
Assumed
sigma
<<
true
sigma:
When
the
assumed
sigma
(
i.
e.,
the
sigma
on
which
the
dose
progression
is
based)
is
much
smaller
than
the
true
sigma
of
the
actual
test
population,
the
estimated
LD50
may
be
``
biased''
in
the
direction
of
starting
dose.
For
example,
if
the
starting
dose
is
less
than
the
true
LD50
of
the
test
population,
the
estimated
LD50
will
generally
be
below
the
true
LD50.
Also,
if
the
starting
dose
is
greater
than
the
true
LD50
of
the
test
population,
the
estimated
LD50
will
tend
to
be
greater
than
the
true
LD50.
To
minimize
the
chance
of
overestimating
the
LD50
due
to
this
bias,
the
UDP
guideline
recommends
a
choice
of
starting
dose
just
below
the
assumed
LD50.

(
B)
Assumed
sigma
>>
true
sigma:
If
the
assumed
sigma
on
which
the
dose
progression
is
based
is
much
larger
than
the
true
sigma
of
the
test
population,
the
median
estimated
LD50
can
be
much
larger
or
much
smaller
than
the
true
LD50
depending
on
the
starting
dose.
In
this
case,
the
LD50
can
be
estimated
only
within
a
range.
(
This
is
Case
3
described
below.)

(
ii)
CI.
Coverage
of
the
CI
is
the
probability
that
a
calculated
CI
encloses
the
true
LD50
for
an
experimental
sample.
Because
the
profile
likelihood
method
is
approximate,
coverage
of
the
CI
does
not
always
correspond
to
its
nominal
value.
For
example,
coverage
falls
below
95%
for
populations
with
shallow
slopes
and
is
better
than
95%
for
populations
with
steep
slopes.
In
addition,
the
width
of
the
CI
is
limited
by
the
dose
progression
chosen.
Generally,
no
type
of
CI
would
be
more
narrow
than
the
dose
progression.

(
iii)
Response
Patterns.
Data
gathered
under
the
UDP
fall
into
one
of
five
animal
response
patterns.
The
five
types
of
animal
response
patterns
referred
to
as
Case
1
through
Case
5
in
the
following
Table
8,
can
33
be
distinguished
for
the
purpose
of
describing
the
performance
of
the
UDP.
These
cases
can
be
distinguished
by
looking
at
the
experimental
outcome
(
survival
or
death)
as
reflected
in
the
AOT425StatPgm
Data
Grid
or
Report
windows
(
see
paragraph
(
n)(
18)
of
this
guideline).
In
considering
these
cases,
note
that
doses
can
be
repeated
more
than
once
in
the
course
of
sequential
dosing.

Table
8.
 
Outcomes
of
the
UDP:
Cases
and
Confidence
Intervals
Case
#
Definition
of
Case
Approach
Proposed
Possible
Findings
1
.......................
No
positive
dose­
response
association.
(
1a)
All
animals
tested
in
the
study
responded
or
(
1b)
none
responded,
or
(
1c)
the
geometric
mean
dose
is
lower
for
animals
that
responded
than
for
animals
that
did
not
respond.
LD50
cannot
be
calculated.
CI
not
applicable
Possible
inferences:
(
1a)
LD50
<
lowest
dose;
(
1b)
LD50
>
highest
dose;
(
1c)
reverse
dose­
response
curve;
unlikely
test
outcome.
In
case
1b,
the
highest
dose
tested
is
equivalent
to
a
limit
dose.
2
.......................
Multiple
partial
responses.
One
or
more
animals
responded
at
a
dose
below
some
other
dose
where
one
or
more
did
not
respond.
The
conditions
defining
Case
1
do
not
hold.
(
The
definition
of
Case
2
holds
if
there
are
2
doses
with
partial
responses,
but
holds
in
some
other
cases
as
well.)
Maximum
likelihood
estimate
and
profile
likelihood
computations
of
CI
are
straightforward.
The
LD50
can
be
estimated
and
its
CI
calculated.

3
.......................
No
intermediate
response
fractions.
One
or
more
test
doses
is
associated
with
0%
response
and
one
or
more
is
associated
with
100%
response
(
all
of
the
latter
being
greater
than
all
of
the
former),
and
no
test
doses
are
associated
with
a
partial
response.
Lower
bound
=
highest
test
dose
with
0%
response.
Upper
bound
=
lowest
test
dose
with
100%
response.
High
confidence
that
the
true
LD50
falls
between
the
two
bounding
doses.
Any
value
of
LD50
between
highest
dose
with
0%
response
and
lowest
dose
with
100%
response
is
equally
plausible

4
.......................
One
partial
response
fraction,
first
subcase.
An
intermediate
partial
response
is
observed
at
a
single
test
dose.
That
dose
is
greater
than
doses
associated
with
0%
response
and
lower
than
doses
associated
with
100%
response.
The
LD50
is
set
at
the
single
dose
showing
partial
response
and
its
CI
is
calculated
using
profile
likelihood
method.
The
LD50
can
be
estimated
and
its
CI
calculated.

5
.......................
One
partial
response
fraction,
second
subcase.
There
is
a
single
dose
associated
with
partial
response,
which
is
either
the
highest
test
dose
(
with
no
responses
at
all
other
test
doses)
or
the
lowest
test
dose
(
with
100%
response
at
all
other
test
doses).
The
LD50
is
set
at
the
dose
with
the
partial
response.
A
profile
likelihood
CI
is
calculated
and
may
be
finite
or
infinite.
The
true
LD50
could
be
at
the
boundary
of
the
testing
range
with
more
or
less
confidence.

(
n)
References.
The
following
references
should
be
consulted
for
additional
background
material
on
this
test
guideline.

(
1)
Organization
for
Economic
Cooperation
and
Development.
OECD
Guidelines
for
the
Testing
of
Chemicals.
Guideline
425:
Acute
Oral
Toxicity
 
Up­
and­
Down
Procedure.
Adopted:
December
2001.

(
2)
Organization
for
Economic
Cooperation
and
Development.
OECD
Guidelines
for
the
Testing
of
Chemicals.
Guideline
420:
Acute
Oral
Toxicity
 
Fixed
Dose
Method.
Adopted:
December
2001.

(
3)
Organization
for
Economic
Cooperation
and
Development.
OECD
Guidelines
for
the
Testing
of
Chemicals.
Guideline
423:
Acute
Oral
Toxcity
 
Acute
Toxic
Class
Method.
Adopted:
December
2001.

(
4)
Dixon,
W.
J.
and
A.
M.
Mood.
(
1948).
A
Method
for
Obtaining
and
Analyzing
Sensitivity
Data.
J.
Amer.
Statist.
Assoc.,
43,
109
 
126.
34
(
5)
Dixon,
W.
J.
(
1965).
The
Up­
and­
Down
Method
for
Small
Samples
J.
Amer.
Statist.
Assoc.
60,
967
 
978.

(
6)
Dixon,
W.
J.
(
1991).
Staircase
Bioassay:
The
Up­
and­
Down
Method
Neurosci.
Biobehav.
Rev.,
15,
47
 
50.

(
7)
Dixon,
W.
J.
(
1991).
Design
and
Analysis
of
Quantal
Dose­
Response
Experiments
(
with
Emphasis
on
Staircase
Designs).
Dixon
Statistical
Associates,
Los
Angeles
CA,
USA.

(
8)
Bruce,
R.
D.
(
1985).
An
Up­
and­
Down
Procedure
for
Acute
Toxicity
Testing.
Fundam.
Appl.
Tox.,
5,
151
 
157.

(
9)
ASTM
(
1987).
E
1163
 
87,
Standard
Test
Method
for
Estimating
Acute
Oral
Toxicity
in
Rats.
American
Society
for
Testing
and
Materials,
Philadelphia
Pa,
USA.

(
10)
Lipnick,
R.
L.,
J.
A.
Cotruvo,
R.
N.
Hill,
R.
D.
Bruce,
K.
A.
Stitzel,
A.
P.
Walker,
I.
Chu,
M.
Goddard,
L.
Segal,
J.
A.
Springer,
and
R.
C.
Myers.
(
1995).
Comparison
of
the
Up­
and­
Down,
Conventional
LD50
and
Fixed
Dose
Acute
Toxicity
Procedures.
Fd.
Chem.
Toxicol.,
33,
223
 
231.

(
11)
Organization
for
Economic
Cooperation
and
Development.
(
2000).
Guidance
Document
on
the
Recognition,
Assessment
and
Use
of
Clinical
Signs
as
Humane
Endpoints
for
Experimental
Animals
Used
in
Safety
Evaluation.
Environmental
Health
and
Safety
Monograph
Series
on
Testing
and
Assessment
No.
19.

(
12)
Report
of
the
International
Workshop
on
In
Vitro
Methods
for
Assessing
Acute
Systemic
Toxicity.
NIH
Publication
No.
01­
4499.
Research
Triangle
Park,
NC:
National
Institute
of
Environmental
Health
Sciences,
August
2001.

(
13)
Guidance
Document
on
Using
In
Vitro
Data
to
Estimate
In
Vivo
Starting
Doses
for
Acute
Toxicity.
NIH
Publication
No.
01­
4500.
Research
Triangle
Park,
NC:
National
Institute
of
Environmental
Health
Sciences,
August
2001.

(
14)
Spielmann,
H.
E.,
M.
Genschow,
M.
Leibsch,
and
W.
Halle.
(
1999).
Determination
of
the
Starting
Dose
for
Acute
Oral
Toxicity
(
LD50)
Testing
in
the
Up­
and­
Down
Procedure
(
UDP)
from
Cytotoxicity
Data
ATLA
27:
957
 
966.

(
15)
Chan,
P.
K.
and
A.
W.
Hayes.
(
1994).
Chap.
16.
Acute
Toxicity
and
Eye
Irritancy.
Principles
and
Methods
of
Toxicology.
Third
Edition.
A.
W.
Hayes,
Editor.
Raven
Press,
Ltd.,
New
York,
USA.

(
16)
Westat.
(
2001).
Acute
Oral
Toxicity
Software
Program;
AOT
425StatPgm;
AOT425StatPgm
Program
User's
Manual;
and
Simulation
Results
for
the
AOT425StatPgm
Program.
Reports
prepared
for
U.
S.
35
E.
P.
A.
under
Contract
68
 
W7
 
0025,
Task
Order
5­
03.
Currently
available
at
web
site:
http://
iccvam.
niehs.
nih.
gov/
methods/
udpdocs/
udprpt/
udp
 
ciprop.
htm
(
17)
Rosenberger,
W.
F.,
Flournoy,
N.
and
Durham,
S.
D.
(
1997).
Asymptotic
Normality
of
Maximum
Likelihood
Estimators
from
Multiparameter
Response­
driven
Designs.
Journal
of
Statistical
Planning
and
Inference
60,
69
 
76.

(
18)
Jennison,
C.
and
B.
W.
Turnbull.
(
2000).
Group
Sequential
Methods
with
Applications
to
Clinical
Trials.
Chapman
&
Hall/
CRC:
Boca
Raton,
FL.

(
19)
Lotus
Development
Corporation.
(
1999).
Lotus
 
1­
2­
3.
Version
9.5,
Millennium
edition.
Cambridge,
MA,
USA.

(
20)
Microsoft
Corporation.
(
1985
 
1997).
Microsoft
 
Excel.
Version
5.0
or
later.
Seattle,
WA,
USA.
