Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
1
APPENDIX
A.
BENCHMARK
DOSE
MODELING
RESULTS
A.
Introduction
Benchmark
dose
(
BMD)
modeling
was
performed
to
identify
potential
critical
effect
levels
as
alternatives
to
the
study
NOAEL/
LOAELs
for
derivation
of
the
HAs
for
cyanide,
and
by
extension,
cyanogen
chloride.
BMD
modeling
was
not
conducted
for
thiocyanate,
because
thiocyanate
was
not
chosen
as
the
surrogate
for
calculation
of
effect
levels
for
cyanogen
chloride.

Results
of
the
BMD
modeling
are
described
below
and
summarized
in
Table
A­
1.

B.
Methods
Benchmark
Dose
The
cyanide
data
sets
considered
for
dose­
response
modeling
were
all
continuous
endpoints.
The
modeling
was
conducted
according
to
draft
EPA
guidelines
(
U.
S.
EPA,
2000e)

using
Benchmark
Dose
Software
(
BMDS
version
1.3.1),
available
from
the
U.
S.
EPA
(
U.
S.
EPA,

2001b).
The
methods
and
models
applied
to
the
continuous
endpoints
are
presented
here.

The
continuous
endpoints
of
interest
with
respect
to
cyanide
toxicity
were
quantitatively
summarized
by
group
means
and
measures
of
variability
(
standard
errors
or
standard
deviations).
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
2
The
BMDS
software
provides
three
models
(
the
Hill,
power,
and
polynomial
models)
for
modeling
of
continuous
data.
Linear
fits
to
the
data
were
incorporated
into
the
analysis
by
allowing
the
power
and
polynomial
models
to
simplify
to
linear
equations
as
dictated
by
the
data.

The
mathematical
models
fit
to
the
data
are
defined
here.
In
all
cases,
µ
(
d)
indicates
the
mean
of
the
response
variable
following
exposure
to
"
dose"
d.

The
polynomial
model
is
defined
as:

µ
(
d)
=
 
0
+
 
1
d
+
...
+
 
n
dn
where
the
degree
of
the
polynomial,
n,
was
set
less
than
or
equal
to
the
number
of
dose
groups
in
the
experiment
being
analyzed.
Note
that
U.
S.
EPA
(
2000e)
recommends
the
use
of
the
most
parsimonious
model
that
provides
an
adequate
fit
to
the
data.
It
may
appear
that
the
use
of
a
polynomial
model
with
degree
possibly
as
great
as
the
number
of
dose
groups
would
not
yield
the
most
parsimonious
model.
However,
allowing
the
model
to
have
that
degree
is
not
the
same
as
forcing
the
model
to
have
that
degree;
in
the
model
fitting,
if
fewer
parameters
(
e.
g.,
a
lower
degree
polynomial)
is
adequate
and
consistent
with
the
data,
then
the
fitting
will
reflect
that
fact
and
a
more
parsimonious
model
will
be
the
result.
For
these
analyses,
the
values
of
the
 
parameters
allowed
to
be
estimated
were
constrained
to
be
either
all
nonnegative
or
all
nonpositive
(
as
dictated
by
the
data
set
being
modeled,
i.
e.,
nonnegative
if
the
mean
response
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
3
increased
with
increasing
dose
or
nonpositive
if
the
mean
response
decreased
with
increasing
dose).

The
power
model
is
represented
by
the
equation:

µ
(
d)
=
 
+
 d 
where
the
parameter
 
is
restricted
to
be
nonnegative.
[
The
linear
model
is
obtained
when
 
is
fixed
at
a
value
of
1.
The
linear
model
was
not
separately
fit
to
the
data;
if
the
result
of
fitting
the
power
model
does
not
result
in
the
linear
form,
 
=
1,
then
the
linear
model
does
not
fit
as
well
as
the
more
general
power
model,
by
definition.]

The
Hill
model
is
given
by
the
following
equation:

µ
(
d)
=
 
+
(
vdn)
/
(
dn
+
kn))

where
the
parameters
n
and
k
are
restricted
to
be
positive
(
in
fact,
n
>
1).

In
the
case
of
continuous
endpoints,
one
must
assume
something
about
the
distribution
of
individual
observations
around
the
dose­
specific
mean
values
defined
by
the
above
models.
The
assumptions
imposed
by
BMDS
were
used
in
this
analysis:
individual
observations
were
assumed
to
vary
normally
around
the
means
with
variances
given
by
the
following
equation:
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
4
 
i
2
=
 2

[
µ
(
d
i)]
 
where
both
 2
and
 
were
parameters
estimated
by
the
model.

Given
those
assumptions
about
variation
around
the
means,
maximum
likelihood
methods
were
applied
to
estimate
all
of
the
parameters,
where
the
log­
likelihood
to
be
maximized
is
(
except
for
an
additive
constant)
given
by
L
=
 
[(
N
i/
2)

ln(
 
i
2)
+
(
N
i
­
1)
s
i
2/
2 
i
2
+
N
i{
m
i
­
µ
(
d
i)}
2/
2 
i
2]

where
N
i
is
the
number
of
individuals
in
group
i
exposed
to
dose
d
i,
and
m
i
and
s
i
are
the
observed
mean
and
standard
deviation
for
that
group.
The
summation
runs
over
i
from
1
to
k
(
the
number
of
dose
groups).

Goodness
of
Fit
Analyses
For
these
continuous
models,
goodness
of
fit
was
determined
based
on
a
likelihood
ratio
statistic.
In
particular,
the
maximized
log­
likelihood
associated
with
the
fitted
model
was
compared
to
the
log­
likelihood
maximized
with
each
dose
group
considered
to
have
a
mean
and
variance
completely
independent
of
the
means
and
variances
of
the
other
dose
groups
(
model
A2
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
11
If
and
when
BMDS
suggested
that
a
homogeneous­
variance
model
was
appropriate,
the
log­
likelihood
of
the
fitted
model
was
compared
to
the
likelihood
maximized
assuming
independent
means
but
a
single,
constant
variance
for
all
dose
groups
(
the
fitted
model
also
assumed
that
to
be
the
case
in
such
cases).
This
was
also
model
A2.

EPA/
OW/
OST/
HECD
Final
draft
A­
5
in
BMDS
output).
1
It
is
always
the
case
that
the
latter
log­
likelihood
will
be
at
least
as
great
as
the
model­
associated
log­
likelihood,
but
if
the
model
does
a
"
reasonable"
job
of
fitting
the
data,

the
difference
between
the
two
log­
likelihoods
will
not
be
too
great.
A
formal
statistical
test
reflecting
this
idea
uses
the
fact
that
twice
the
difference
in
the
log­
likelihoods
is
distributed
as
a
chi­
square
random
variable.
The
degrees
of
freedom
associated
with
that
chi­
squared
test
statistic
are
equal
to
the
difference
between
the
number
of
parameters
fit
by
the
model
(
including
the
parameters
 2
and
 
defining
how
variances
change
as
a
function
of
mean
response
level)
and
twice
the
number
of
dose
groups
(
which
is
equal
to
the
number
of
parameters
estimated
by
the
"
model"
assuming
independence
of
dose
group
means
and
variances).
Parameters
hitting
boundary
values
were
not
included
for
determining
degrees
of
freedom.

Acceptable
fit
was
defined
as
a
goodness­
of­
fit
p­
value
greater
than
or
equal
to
0.1,
or
a
perfect
fit
when
there
were
no
degrees
of
freedom
for
a
statistical
test
of
fit.
Choice
of
0.1
is
consistent
with
current
U.
S.
EPA
guidance
for
BMD
modeling
(
U.
S.
EPA,
2000e).
If
a
model
was
judged
to
provide
a
reasonable
BMDL
estimate,
but
the
p­
value
criterion
of
0.1
was
not
met,

the
rationale
for
waiving
the
p­
value
criterion
is
provided
in
the
discussion
of
the
results.

Goodness­
of­
fit
statistics
are
not
designed
to
compare
different
models,
particularly
if
the
different
models
have
different
numbers
of
parameters.
Within
a
family
of
models,
adding
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
6
parameters
generally
improves
the
fit.
BMDS
reports
the
Akaike
Information
Criterion
(
AIC)
to
aid
in
comparing
the
fit
of
different
models.
The
AIC
is
defined
as
 
2L+
2p,
where
L
is
the
loglikelihood
at
the
maximum
likelihood
estimates
for
the
parameters,
and
p
is
the
number
of
model
parameters
estimated
(
ignoring
parameters
assuming
values
at
the
boundaries
of
their
allowable
ranges).
When
comparing
the
fit
of
two
or
more
models
to
a
single
data
set,
the
model
with
the
lesser
AIC
was
considered
to
provide
a
superior
fit.

Definition
of
the
BMR
and
Corresponding
BMD
and
BMDL
For
the
continuous
models,
BMDs
were
implicitly
defined
as
follows:

 
µ
(
BMD)
­
µ
(
0)
 
=
  
1
where
 
1
is
the
model­
estimated
standard
deviation
in
the
control
group.
In
other
words,
the
BMR
was
defined
as
a
change
in
mean
corresponding
to
some
multiplicative
factor
of
the
control
group
standard
deviation.

The
value
of
 
used
in
this
analysis
was
1.0.
This
value
was
chosen
based
on
EPA
draft
guidelines
for
BMD
analyses
(
U.
S.
EPA,
2000e),
in
the
absence
of
a
clear
biological
rationale
for
selecting
an
alternative
response
level.
It
is
roughly
consistent
with
(
though
slightly
more
conservative
than)
a
choice
of
1.1,
which
according
to
Crump
(
1995)
corresponds
to
an
additional
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
7
risk
of
10%
when
the
background
response
rate
was
assumed
to
be
1%,
with
normal
variation
around
the
mean
(
and
constant
standard
deviation).

Choice
of
BMDL
The
following
guidance
was
followed
with
regard
to
the
choice
of
the
BMDL
to
use
as
a
point
of
departure
for
calculation
of
a
health
advisory.
This
guidance
is
consistent
with
recommendations
in
U.
S.
EPA
(
2000e).
For
each
endpoint,
the
following
procedure
is
recommended:

1.
Models
with
an
unacceptable
fit
are
excluded.

2.
If
the
BMDL
values
for
the
remaining
models
for
a
given
endpoint
are
within
a
factor
of
3,

no
model
dependence
is
assumed,
and
the
models
are
considered
indistinguishable
in
the
context
of
the
precision
of
the
methods.
The
models
are
then
ranked
according
to
the
AIC,
and
the
model
with
the
lowest
AIC
is
chosen
as
the
basis
for
the
BMDL.

3.
If
the
BMDL
values
are
not
within
a
factor
of
3,
some
model
dependence
is
assumed,
and
the
lowest
BMDL
is
selected
as
a
reasonable
conservative
estimate,
unless
it
is
an
outlier
compared
to
the
results
from
all
of
the
other
models.
Note
that
when
outliers
are
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
8
removed,
the
remaining
BMDLs
may
then
be
within
a
factor
of
3,
and
so
the
criteria
given
in
item
2.
would
be
applied.

4.
The
BMDL
values
from
all
modeled
endpoints
are
compared,
along
with
any
NOAELs
or
LOAELs
from
data
sets
that
were
not
amenable
to
modeling,
and
the
lowest
NOAEL
or
BMDL
is
chosen.

C.
Modeling
Results
Short­
term
Studies
Several
of
the
short­
term
studies
on
cyanide
toxicity
summarized
in
Table
VIII­
2
could
not
be
modeled.
The
human
studies
typically
reported
a
single
dose
level,
no
control
group,
and
often
small
sample
sizes,
and
so
could
not
be
modeled.
LD
50
and
mortality
studies
were
not
modeled,
because
they
did
not
evaluate
sensitive
endpoints.
The
drinking
water
and
feed
studies
of
Palmer
and
Olson
(
1979)
were
not
modeled,
because
no
measure
of
variability
was
reported
for
the
continuous
endpoints
evaluated.

The
only
short­
term
study
that
could
be
modeled
was
the
study
of
Kreutler
et
al.
(
1977),

in
which
rats
were
exposed
to
cyanide
in
the
diet,
and
thyroid
effects
were
evaluated.
In
this
study,
there
were
only
two
dose
groups,
so
only
a
linear
model
could
be
fit
to
the
endpoints
from
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
2These
values
doses
differ
from
those
shown
in
the
BMDS
output
by
a
factor
of
87/
40,
because
of
a
correction
to
the
dose
conversions
made
after
the
BMD
modeling.
Since
dose
enters
the
modeling
in
a
linear
form,
such
posthoc
corrections
can
be
done.

EPA/
OW/
OST/
HECD
Final
draft
A­
9
that
study.
The
BMDS
output
for
the
thyroid
weight
endpoint
suggested
that
a
model
with
constant
variance
(
as
opposed
to
a
model
with
dose
group­
specific
variances
depending
on
the
group
means)
would
be
the
preferred
approach.
When
such
a
model
was
fit
to
the
data,
the
goodness
of
fit
was
satisfactory
(
p
=
0.34)
and
the
BMD
and
BMDL
estimates
were
17.4
and
14.6
mg
CN/
kg­
day,
respectively2.
For
the
endpoint
of
increased
plasma
TSH,
the
variances
did
not
appear
to
be
constant;
the
linear
model
with
group­
dependent
variances
fit
the
data
as
well
as
possible
(
i.
e.,
the
fit
was
as
good
as
the
model
maximizing
the
likelihood,
model
A2
in
BMDS
parlance).
The
BMD
and
BMDL
for
plasma
TSH
were
2.6
and
2.1
mg
CN/
kg­
day,
respectively.

Because
the
slope
of
the
dose­
response
for
this
endpoint
was
much
greater
than
that
for
thyroid
weight,
lower
values
were
calculated
for
the
BMD
and
BMDL.
Plasma
TSH
was
the
more
sensitive
of
the
two
endpoints
modeled.
Although
the
BMDL
calculated
for
plasma
TSH
was
one
of
the
lower
BMDLs
calculated
for
cyanide,
confidence
in
this
value
is
limited,
based
on
the
absence
of
information
on
the
shape
of
the
dose­
response
curve,
and
due
to
the
significant
degree
of
extrapolation
below
the
single
data
point
of
87
mg
CN/
kg­
day.
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
10
Longer­
term
Studies
As
for
the
short­
term
studies,
several
of
the
longer­
term
studies
for
cyanide
(
summarized
in
Table
VIII­
2)
did
not
report
data
in
a
form
that
could
be
modeled.
Data
from
the
study
by
Kamalu
(
1993)
were
not
modeled
due
to
concerns
about
the
relevance
of
the
results
in
dogs
(
see
Chapter
V),
and
because
no
incidence
data
were
reported
for
the
observed
histopathology.
The
results
of
Jackson
(
1988)
were
not
modeled
because
the
quantitative
results
were
not
reliable.

The
data
were
reported
only
as
the
average
across
males
and
females,
but
the
number
of
males/
females
per
group
varied,
and
the
endpoints
of
interest
(
T3,
T4,
fighting,
victimization)

have
different
baselines
for
males
and
females.
Howard
and
Hanzal
(
1955)
did
not
report
any
adverse
effects,
and
so
that
study
could
not
be
modeled.

One
of
the
longer­
term
studies
(
Philbrick
et
al.,
1979)
tested
only
a
single
positive
dose,

although
the
study
included
two
parallel
experiments.
In
one
experiment
(
designated
here
as
the
(+)
control
experiment),
the
rats
received
a
normal
diet,
while
in
the
second
experiment
(
the
(­)

control
experiment),
the
rats
received
a
diet
deficient
in
methionine,
iodine,
and
vitamin
B
12.

Linear
models
were
fit
to
the
data
for
both
thyroid
weight
and
thyroxine
(
T4)
secretion
for
each
of
these
experiments.
In
all
four
cases,
BMDS
suggested
that
a
constant
variance
model
was
the
preferred
option.
In
all
four
cases,
the
fits
of
the
constant­
variance,
linear
model
was
satisfactory
(
p­
values
all
greater
than
0.14).
Consistent
with
the
larger
response
noted
at
the
single
dose
tested,
the
BMDs
and
BMDLs
calculated
for
the
data
sets
for
the
animals
given
deficient
diet
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
3These
values
doses
differ
from
those
shown
in
the
BMDS
output
by
a
factor
of
44/
30,
because
of
a
correction
to
the
dose
conversions
made
after
the
BMD
modeling.

EPA/
OW/
OST/
HECD
Final
draft
A­
11
were
smaller
than
the
corresponding
values
for
the
animals
given
the
normal
diet;
within
a
given
diet,
similar
values
were
calculated
for
BMDLs
for
both
endpoints.
For
the
animals
given
the
normal
diet,
the
BMDL
was
19
mg/
kg­
day,
for
both
the
thyroid
weight
and
T4
secretion
endpoints.
For
the
animals
given
the
deficient
diet,
the
BMDL
for
T4
secretion
was
12
mg/

kgday
slightly
lower
than
the
BMDL
of
13
mg/
kg­
day
for
thyroid
weight.
3
Thus,
the
BMDL
for
this
study
was
19
mg
CN/
kg­
day
in
rats
fed
a
normal
diet
and
12
in
rats
fed
a
deficient
diet,
both
based
on
decreased
T4
secretion.
Interestingly,
these
values
are
comparable
to
the
BMDL
for
increased
thyroid
weight
in
Kreutler
et
al.
(
1977),
even
though
Philbrick
et
al.
(
1979)
was
an
11.5­
month
study,
while
exposure
was
for
2
weeks
in
Kreutler
et
al.
(
1977).
This
supports
the
conclusion
in
Chapter
VIII
that
this
endpoint
does
not
progress
with
increasing
exposure
duration,
although
again
the
conclusion
is
limited
due
to
the
use
of
only
one
dose
level
in
both
studies.
In
addition,
as
noted
in
Chapter
V,
the
observed
thyroid
effects
may
be
adaptive,
rather
than
adverse.
Philbrick
et
al.
(
1979)
also
reported
histopathological
evidence
of
neurological
effects,
but
this
endpoint
could
not
be
modeled,
because
no
quantitative
results
were
reported.

In
a
developmental
toxicity
study,
Tewe
and
Maner
(
1981)
exposed
rats
to
two
dose
levels
of
cyanide
in
the
diet
throughout
mating,
gestation,
and
lactation;
decreased
daily
body
weight
gain
was
observed
in
the
weanlings
and
this
endpoint
was
modeled.
Even
though
no
control
group
was
included
(
the
low­
exposure
group
did
have
some
CN
exposure),
a
linear
model
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
4These
values
doses
differ
from
those
shown
in
the
BMDS
output
by
a
factor
of
34.3/
32.8,
because
of
a
correction
to
the
dose
conversions
made
after
the
BMD
modeling.

EPA/
OW/
OST/
HECD
Final
draft
A­
12
could
be
fit
to
the
data
and
BMDs
could
be
estimated
based
on
imputed
(
model­
predicted)

background
means
and
standard
deviations.
There
was
no
indication
that
the
variance
depended
on
dose
group
(
through
group
mean
weight
gains),
so
a
constant
variance
model
was
fit
to
the
data.
The
fit
was
perfect,
and
the
BMD
and
BMDL
estimates
were
23
and
15
mg
CN/
kg­
day,

respectively.
4
The
NTP
(
1993)
study
included
several
endpoints
potentially
related
to
cyanide
exposure.

For
the
female
rats
and
mice,
the
only
endpoint
of
interest
was
relative
liver
weight.
The
doseresponse
for
that
endpoint
was
non­
monotonic
for
both
species.
For
both
species,
BMDS
suggested
that
the
preferred
approach
would
be
to
treat
the
variances
as
if
they
were
constant.

For
female
rats,
despite
the
non­
monotonicity,
the
polynomial
model
fit
with
borderline
adequacy
(
p
=
0.09),
whereas
the
tests
of
fit
for
the
power
and
Hill
models
suggested
a
poorer
fit
(
p­
values
of
0.05
and
0.02,
respectively),
but
that
is
an
artifact
of
the
fact
that
the
latter
two
models
have
more
parameters
(
as
also
reflected
in
the
greater
AICs
for
the
latter
two
models).
The
fit
of
the
data
was
good
in
the
region
of
the
BMDL,
and
fit
to
the
control
standard
deviation
was
good.
In
addition,
visual
fit
to
the
control
and
lower
doses
was
generally
reasonable,
except
for
the
outlier
datapoint
(
second
lowest
dose).
In
any
case,
with
BMD
estimates
very
similar
across
models,
the
polynomial
model
would
be
the
model
of
choice
(
with
the
smallest
AIC),
so
the
BMD
and
BMDL
selected
as
the
best
estimates
for
increased
relative
liver
weight
in
female
mice
endpoint
are
11
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
5
Although,
it
should
be
noted
that
the
BMDS
output
has
an
error
somewhere
because
the
likelihood
reported
for
model
A1
is
different
for
the
polynomial
run
and
the
power
model
run.
Model
A1
does
not
depend
on
the
model,
and
so
should
be
the
same
for
both
runs.
In
any
case,
the
fit
is
unacceptably
poor
for
either
model.

EPA/
OW/
OST/
HECD
Final
draft
A­
13
and
8.7
mg
CN/
kg­
day,
respectively.
However,
confidence
in
the
BMDL
is
limited
because
a
clear
increase
was
seen
only
at
the
high
dose.

For
relative
liver
weight
in
female
mice,
the
non­
monotonicities
in
the
data
were
such
that
the
polynomial
and
power
models
defaulted
to
linearity,
and
still
could
not
achieve
an
adequate
fit
(
p­
values
of
0.008
and
0.005,
respectively5).
Even
for
the
Hill
model,
which
accommodates
more
"
downward"
curvature
than
do
the
power
or
polynomial
models,
the
fit
was
poor
(
p
=
0.02).

Although
the
Hill
model
might
otherwise
be
the
preferred
model
for
this
data
set,
it
might
be
best
not
to
use
BMD
or
BMDL
estimates
from
any
model
for
relative
liver
weight
changes
in
female
mice.
The
steep
slope
of
the
Hill
model
in
the
low­
dose
region
also
resulted
in
a
very
low
BMDL
that
does
not
appear
to
be
otherwise
consistent
with
the
data
for
this
endpoint.
In
addition,
the
flat
slope
of
the
dose
response
curve
at
all
cyanide
doses
suggests
that
the
control
relative
liver
weight
may
have
been
abnormally
low,
leading
to
a
low
BMDL.

Both
male
rats
and
male
mice
in
the
NTP
(
1993)
study
exhibited
responses
related
to
reproductive
toxicity.
For
these
endpoints,
however,
the
study
authors
only
reported
the
results
in
the
control
group
and
at
the
top
three
dose
levels.
All
significant
endpoints
were
modeled.
In
rats,
the
endpoints
modeled
were
left
epididymis
weight,
left
cauda
epididymis
weight,
left
testis
weight,
spermatid
head
(
per
testis),
spermatid
count,
and
percent
motility.
With
the
exception
of
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
14
the
motility
endpoint,
BMDS
suggested
that
constant
variance
models
were
preferred,
and
these
models
are
therefore
the
basis
of
the
BMD
estimates
for
the
other
male
rat
endpoints
in
the
NTP
(
1993)
study.

For
left
epididymis
weight,
all
of
the
models
fit
adequately
(
p­
values
all
greater
than
or
equal
to
0.22).
The
polynomial
and
power
models
defaulted
to
the
same
linear
model,
while
the
Hill
model
captured
more
of
the
curvature,
resulting
in
a
BMD
estimate
that
was
a
little
more
than
3
times
smaller
than
that
for
the
linear
model
(
2.7
as
opposed
to
8.2).
Moreover,
the
BMDL
estimate
from
the
Hill
model
(
0.79)
was
about
7
times
less
than
the
BMDL
from
the
linear
model.

Therefore,
according
to
the
draft
EPA
guidelines
(
U.
S.
EPA,
2000e),
the
results
from
the
Hill
model
would
be
the
preferred
ones
to
use.

For
left
cauda
epididymis
weight,
the
polynomial
and
power
models
defaulted
to
the
same
linear
model.
The
Hill
model
captured
more
of
the
curvature
and
fit
the
data
better,
based
on
the
AIC
values.
The
BMD
for
the
Hill
model
was
1.5
(
compared
to
8.4
for
the
linear
model)
but,

unfortunately
the
Hill
model
failed
to
estimate
a
BMDL.
It
is
difficult
to
recommend
use
of
the
BMD
or
BMDL
from
the
linear
model,
given
that
the
Hill
model
did
describe
the
data
better
and
provided
a
BMD
estimate
less
than
the
BMDL
estimate
from
the
linear
model.
Therefore,
no
BMDL
was
identified
for
this
endpoint.
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
15
For
left
testis
weight,
the
polynomial
and
power
models
defaulted
to
the
same
linear
model.
The
Hill
model
did
capture
more
of
the
curvature
and
resulting
curve
passed
through
all
of
the
observed
means,
but
the
additional
parameters
necessary
to
do
so
result
in
an
AIC
that
is
greater
than
the
AIC
for
the
linear
model.
The
BMD
for
the
linear
model
was
7.0
(
compared
to
5.0
for
the
Hill
model)
and
the
BMDL
for
the
linear
model
was
4.9.
The
linear
model
estimates
(
shown
in
Table
A­
1
as
polynomial/
power
model)
are
the
preferred
values
for
this
data
set,
since
the
BMDLs
are
within
a
factor
of
three
and
the
AIC
is
smaller
for
the
linear
model.

For
spermatid
heads
per
testis,
the
polynomial
and
power
models
defaulted
to
the
same
linear
model.
The
Hill
model
captured
more
of
the
curvature
in
the
data
but,
like
the
previous
two
data
sets,
the
improvement
was
not
enough
to
offset
the
need
to
estimate
an
additional
parameter
(
i.
e.,
the
AIC
for
the
Hill
model
was
larger
than
that
for
the
linear
model).
However,

given
that
the
BMDL
estimates
differ
by
more
than
a
factor
of
three
and
both
models
fit
adequately
(
p­
values
all
greater
than
or
equal
to
0.58)
the
model
with
the
smaller
BMDL
would
be
the
recommended
one.
The
Hill
model
is
therefore
preferred
for
this
data
set
and
it
yields
a
BMD
estimate
of
8.3
with
a
lower
bound
of
1.3.

Very
similar
results
were
observed
for
spermatid
count
as
for
spermatid
heads
per
testis.

In
this
case
also,
the
Hill
model
is
preferred
over
the
linear
model
resulting
from
fitting
either
the
power
or
polynomial
models
(
the
former
has
a
BMDL
more
than
three
times
smaller
than
the
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
16
latter),
and
it
gives
the
same
BMD
and
BMDL
estimates
as
for
spermatid
heads
per
testis
(
8.3
and
1.3,
respectively).

For
the
sperm
motility
endpoint,
the
polynomial
and
power
models
both
defaulted
to
linearity,
but
they
did
not
converge
to
the
same
result.
The
linear
model
to
which
the
power
model
defaulted
fit
the
data
slightly
better
than
did
the
polynomial
model
(
which
did
not
default
to
a
linear
model),
based
on
the
maximized
log­
likelihood
and
the
AIC,
but
neither
fit
describes
the
data
well
at
all.
The
non­
monotonicities
in
the
data
make
it
difficult
to
fit
these
monotonic
models
and,
in
addition,
the
variances
are
not
well
represented
by
the
models
(
also
due
to
the
fact
that
the
models
impose
monotonicity
in
variance
but
such
monotonicity
is
absent
in
the
data).
On
the
other
hand,
the
Hill
model
gives
only
a
slightly
better
maximized
likelihood,
but
the
added
number
of
parameters
in
that
model
make
the
AIC
larger
than
that
for
the
linear
models
and
the
fit
to
the
data
is
still
poor.
No
BMD
results
should
be
used
for
the
motility
endpoint.

For
male
mice
in
the
NTP
(
1993)
study,
the
endpoints
in
which
effects
were
seen
were
the
male
reproductive
endpoints
of
decreased
left
epididymis
weight
and
decreased
left
cauda
epididymis
weight;
both
endpoints
were
modeled.
In
the
case
of
epididymis
weight,
the
polynomial
and
the
power
models
both
defaulted
to
a
linear
model,
but
because
of
some
bug
in
BMDS,
not
to
the
same
linear
model.
The
linear
model
parameter
estimates
for
the
power
model
yielded
a
slightly
better
(
larger)
log­
likelihood
and
therefore
a
slightly
lower
AIC.
Either
of
these
linear
models
is
better
than
the
Hill
model
(
which
has
a
larger
AIC
because
of
the
extra
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
17
parameters
needed
to
be
estimated).
Hence
the
power
(
linear)
model
is
the
preferred
model
for
this
endpoint;
it
estimates
a
BMD
of
25,
with
a
lower
bound
of
18.

For
left
cauda
epididymis
weight,
BMDS
had
several
problems.
As
in
the
previous
case,

the
polynomial
and
power
models
defaulted
to
linear
models
but
not
the
same
linear
models.
The
fit
for
the
power
model
was
slightly
better,
on
the
basis
of
the
AICs,
even
though
the
chi­
squared
goodness­
of­
fit
p­
values
could
not
be
estimated
because
the
model
designated
as
A3
in
BMDS
(
to
which
the
fitted
model
is
normally
compared)
yielded
a
log­
likelihood
that
was
so
much
worse
than
all
the
other
models
that
it
was
obviously
in
error.
This
error
was
also
indicated
by
the
fact
that
the
likelihood
for
that
model
was
different
for
each
of
the
three
model
runs,
even
though
it
should
be
the
same
no
matter
which
model
is
fit.
The
fit
of
the
Hill
model
was
so
bad
(
corresponding
observed
and
predicted
means
not
even
close
to
one
another)
that
it
was
considered
a
failed
run.
Nevertheless,
the
fit
of
the
polynomial­
linear
model
was
visually
satisfactory
(
similar
to
the
fit
of
the
linear
model
fit
to
the
left
epididymis
weight
data
set
above)

and
close
enough
to
the
power­
linear
model
in
terms
of
log­
likelihood
that
the
BMD
and
BMDL
from
the
polynomial
model
output
(
20
and
12,
respectively)
can
be
considered
adequate
for
an
initial
analysis.

Overall,
results
of
the
BMD
modeling
for
the
NTP
(
1993)
study
supported
the
conclusion
from
comparison
of
NOAELs/
LOAELs
that
the
rats
were
more
sensitive
than
mice
to
the
male
reproductive
effects
of
cyanide.
Similarly,
the
BMD
modeling
data
also
found
that
male
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
18
reproductive
endpoints
tended
to
be
more
sensitive
than
increased
liver
weight
in
female
rats
or
mice.
Overall,
the
most
sensitive
endpoint
in
this
study
was
decreased
left
epididymis
weight
in
male
rats,
with
a
BMDL
of
0.79
mg/
kg­
day.
This
BMDL
is
supported
by
a
BMD
of
1.5
mg/

kgday
for
the
Hill
model
for
decreased
left
cauda
epididymis
weight
(
for
which
BMDS
could
not
estimate
a
BMDL),
and
by
BMDLs
of
1.3
mg/
kg­
day
for
decreased
spermatid
heads/
testis
and
decreased
spermatid
count.
Only
preliminary
results
could
be
obtained
for
BMD
modeling
of
the
mouse
data
in
the
NTP
(
1993)
study,
but
the
lowest
BMDL
of
12
mg/
kg­
day
(
for
decreased
left
caudal
epididymis
weight)
is
higher
than
the
rat
BMDL
in
the
study.
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
EPA/
OW/
OST/
HECD
Final
draft
A­
19
Table
A­
1
Benchmark
Dose
Modeling
Results
for
Cyanide
Study
Endpoint
Model
AIC
G­
o­
F
pvalue
BMDa
BMDL
Kreutler
et
al.,
1977
Thyroid
Weight
Linear
192.844b
.34c
17.4
14.6
Plasma
TSH
Linear
309.486
1.0d
2.6
2.1
Philbrick
et
al.,
1979;
(+)
control
Thyroid
Weight
Linear
24.838b
.14c
38
19
Thyroxine
Secretion
Linear
­
23.120b
.14c
37
19
Philbrick
et
al.,
1979;
(­)
control
Thyroid
Weight
Linear
15.674b
1.0d
21
13
Thyroxine
Secretion
Linear
­
23.120b
.14c
19
12
Tewe
and
Maner,
1981
Daily
BW
Gain
Linear
­
24.266b
1.0d
23
15
NTP,
1993;
female
rats
Relative
Liver
Weight
Polynomial
133.910b
.09c
11
8.7
Power
135.881
.05
12
8.6
Hill
137.881
.02
12
Failed
NTP,
1993;
female
mice
Relative
Liver
Weight
Polynomial
188.390b
.008b
14
11
Power
188.390b
.005b
14
11
Hill
186.149
.02b
0.64
0.18
NTP,
1993;
male
rats
Left
Epididymis
Weight
Polynomial
­
274.730b
.22b
8.2
5.6
Power
­
274.730b
.22b
8.2
5.6
Hill
­
273.765b
.69c
2.7
0.79
Left
Cauda
Epididymis
Weight
Polynomial
­
312.753b
.08b
8.4
5.6
Power
­
312.753b
.08b
8.4
5.6
Hill
­
314.828
.35b
1.5
Failed
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
Study
Endpoint
Model
AIC
G­
o­
F
pvalue
BMDa
BMDL
EPA/
OW/
OST/
HECD
Final
draft
A­
20
Left
Testis
Weight
Polynomial
­
171.981b
.81b
7.0
4.9
Power
­
171.981b
.81b
7.0
4.9
Hill
­
168.410
.43c
5.0
1.8
Spermatid
Heads
per
Testis
Polynomial
98.363b
.70b
11
6.9
Power
98.363b
.70b
11
6.9
Hill
99.962b
.58b
8.3
1.3
Spermatid
Count
Polynomial
227.040b
.70b
11
6.9
Power
227.040b
.70b
11
6.9
Hill
228.642
.58b
8.3
1.3
Motility
Polynomial
147.508
.001b
19
8.5
Power
144.243b
.002b
15
6.7
Hill
147.583
.007c
Failed
Failed
NTP,
1993;
male
mice
Left
Epididymis
Weight
Polynomial
­
378.489
.12c
22
Failed
Power
­
378.910b
.14c
25
18
Hill
­
374.586
.12c
21
Failed
Left
Cauda
Epididymis
Weight
Polynomial
­
409.479
Failed
20
12
Power
­
409.568a
Failed
20
Failed
Hill
Failed
aBMD
and
BMDL
are
based
on
benchmark
response
of
1SD.
Results
are
presented
in
units
of
mg/
kg­
day.
BMD
and
BMDL
estimates
in
bold
type
are
the
estimates
judged
to
be
the
best
estimates
to
use
for
the
quantitative
dose­
response
assessment
for
each
chemical.
For
some
data
sets,
none
are
bolded
because
of
poor
model
fits
or
inability
of
BMDS
to
obtain
BMDL
estimates.
"
Failed"
indicates
that
BMDS
was
unable
to
produce
the
estimate
or
the
information
required
to
be
able
to
present
a
value.

bCorrected
from
erroneous
BMDS
output.
Errors
were
identified
in
the
degrees
of
freedom
(
DF)
provided
in
the
output
for
the
fitted
model
in
several
cases.
For
these
cases,
the
AIC
was
calculated
independently
using
the
log
likelihoods
provided
in
the
output
and
the
correct
number
of
DF.
Similarly,
the
goodness­
of­
fit
pvalues
were
corrected
by
calculating
manually
the
chi
square
p­
value
using
the
appropriate
number
of
DF.

cBased
on
a
comparison
of
the
fitted
model
to
the
model
maximizing
the
likelihood
(
i.
e.
model
with
independent
means
and
variances
for
each
dose
group,
model
A2
from
BMDS).
Drinking
Water
Criteria
Document
for
Cyanogen
Chloride
and
Metabolites
Study
Endpoint
Model
AIC
G­
o­
F
pvalue
BMDa
BMDL
EPA/
OW/
OST/
HECD
Final
draft
A­
21
dA
fit
that
maximizes
the
likelihood
(
i.
e.,
gives
a
likelihood
as
great
as
that
for
model
A2)
is
assigned
a
pvalue
of
1.0,
even
if
there
were
no
degrees
of
freedom
for
a
formal
statistical
test.
