U.S. Environmental Protection Agency
Revised OP (Organophosphate)
Cumulative Risk Assessment
June 10, 2002
El. Appendices
B. Hazard / Relative Potency Factor
This document was only published electronically.
Accessed 1/14/04 from:
http://www.epa.gov/pesticides/cumulative
Cover page created by EPA Region 9 Library staff, January 14, 2004.
-------
I III. Appendices
j B. Hazard/RPF
I 1. Technical Aspects of Dose-Response Analysis
CM f
O | Background
*4*<-i —
EPA released a Preliminary Dose-Response Assessment for OPs on July 31, 2001
(USEPA 2001 b) followed by a revised dose-response assessment on December 3,
2001. Both of these analyzes were reviewed by the FIFRA SAP in September 2001
i | and February 2002, respectively (FIFRA SAP 2001 b, 2002). The current approach was
-*--» I supported by the SAP (FIFRA 2002). At the February 5-8, 2002 meeting of the SAP,
£~ | EPA discussed some programming errors found after the December 3, 2001 release of
f the Preliminary Cumulative Risk Assessment. These errors have been corrected; the
[ contents of III.B.4 (R programming code) reflect the corrections.
CO I
CO I Dose-Response Modeling
CD I
(/> [ The goal of the statistical methods was to estimate the dose that would be expected
to result in a 10% reduction in brain AChE activity, the BMD10. The data for this study
were in the form of dose-response studies which measured the effect of different dose
rates of OP pesticides on cholinesterase activities in brain, red blood cells, and plasma.
C/5 1 The mean and standard deviation of cholinesterase activity, and number of animals
examined were available for several dosages in each data set. Females and males
were analyzed separately in each study. For each chemical there were several groups
'11 I of studies labeled by separate MRIDs. Within each major study, one or more studies
> [ were conducted, each with measurements taken for several durations of exposure.
• "*** E
«•*•»' Z
•$ = It is useful to describe the approach to modeling the dose-response data in three
3 I parts:
;~ I • the shape of the dose-response curve to be used;
•^ | • how multiple data sets were modeled at the same time;
1 • the statistical methods used to estimate values for the model parameters.
CL ^
= In this analysis, the dose-response function had to accommodate three important
features of the data. First, since the data came from multiple studies, perhaps carried
_ out in different laboratories and at different times, and even sometimes reporting
0 | cholinesterase activity in different units, activity at a given dosage was expressed as a
fraction of control activity. Second, it was observed that, as dose increased,
cholinesterase activity in quite a few data sets approached a lower non-zero asymptote.
This asymptote varied among chemicals and possibly sexes. Finally, for many of the
chemicals it was apparent that there is a "shoulder" on the dose response curve, such
that the dose-response curve was shallower at lower doses than at higher.
Appendix III.B.1 - Page 1
-------
[ These features of the dose-response were incorporated in the dose-response model
1 in two phases. First, a model was developed relating dose to cholinesterase activity
1 which allowed for a horizontal asympote, and expressed activity at a given dose level as
j a fraction of background, or control, activity. In this document, this first model is called
| the "basic" model. Next, a submodel relating internal dose to administered dose, was
I combined with the first model to make a new model with that could have a shoulder in
I the low-dose region. The next subsections discuss these two models in more detail.
*v= =
^== I Basic Dose-Response Model
| The basic model is described by the equation:
y = A
BAffi)
xDose
Eqn. 1
1 Here, A is the level of cholinesterase activity in the absence of exposure to
| organophosphate, PB is the fraction of cholinesterase activity remaining at a very high
I dose of organophosphate, BMR is the level of inhibition at which to estimate the
| benchmark dose (in this study, BMR is always 0.10), BMD is the benchmark dose, and
I Dose is the dose of organophosphate pesticide, generally in units of mg/kg/day. This
I model is essentially the same as was described in FIFRA SAP (2001 b, 2002), only
I reparametrized so that BMD appears as an explicit parameter, thus simplifying the
„= | calculations. Note that the model is undefined if PB + BMR ^1.
„«. [ Expanded Dose-Response Model
'****>, ~
"^ I A submodel relating internal dose to administered dose was combined with the basic
ff? [ model to make the expanded model which allows for a shoulder in the low-dose region.
V ^3" 5 '
=2 11. Biologically Inspired Mode): Accounting for Potential First-Pass Metabolism
At this time, the appropriate kinetic data needed for the development of a
physiologically based pharmacokinetc model (PBPK model) for all OPs are not
available. EPA has developed a biologically inspired model based on metabolic
pathways for first-pass metabolism which are theorized to influence the shape of the
dose-response curve.
When many chemicals are administered orally, much of the absorbed chemical is
carried to the liver by the portal circulation, where they may be metabolized. In the
presence of saturable metabolism the dose-response curve would be expected to
have a shallower slope at lower doses, and the slope would gradually increase as
metabolism became saturated and more of the active chemical enters the general
circulation. Although a detailed treatment of this process for each chemical is
beyond the scope of this project, this basic idea was used to derive a two-parameter
function of dose that relates administered dose to internal dose. The resulting
function was combined with the basic exponential model giving a model that has a
Appendix III.B.1 - Page 2
-------
CM
O
V
CO
1
c
CD
0)
CO
vt
"35
0
low dose shoulder while retaining the dose-response shape of the basic model for
larger doses.
Consider the simple two-compartment pharmacokinetic model illustrated in Figure
III.B.1-1.
tn
O
0>
Body (C6)
T
Urine (ke)
Ingestion (Dose xBW/24)
1
Liver (C,)
0)'
Metabolism (^ KJ
Figure III.B.1-1: Diagram for two-
compartment PBPK model for the extension to
the basic model
~*
e
"™*<
~-j
o
3»u& ,-(
•*••
O
CO
In this simple model, all the ingested chemical is taken directly to the liver, where it is
metabolized. The residual unmetabolized chemical is then distributed to the rest of
the body through the circulation. Intake of chemical is continuous. In this case, two
differential equations and one algebraic equation describe the concentration in the
liver and the rest of the body:
DosexBW Vr
Qbcb
Here, Cx is the concentration in compartment x, where x is a for arterial blood, b for
the body other than liver, and / for liver. The volume of and blood flow to
Appendix III.B.1 - Page 3
-------
CM
O
CD
C
CD
E
CO
CO
0
CO
CO
compartment x are Vx and Qx, where x is either b or /. Vmax and Km describe
saturable metabolism of the chemical in the liver. The constant ke is a first-order
clearance term. Dose is expressed in milligrams per kilogram per day (hence the
constant "24" to convert to hours), and body weight is expressed in kilograms. Thus,
volumes in this parametrization are expressed in liters and concentrations in
milligrams per liter.
At steady state, the derivatives are both 0: clearance just balances the dose rate. It
can be shown (by solving the system of equations with derivatives set to zero) that
the concentration in the body (Cb) at steady state is:
Q=0.5*
24x ke [{ BW(Q,ke + Qbke + Q,Qb} BW j
*• ' ^\ A S~\ f\ TS" J | " V /
Dose -. "^ , ^DO;C
BW
Here, the odd constants 0.5 and 4 arise because the solution involves finding the
roots of a quadratic polynomial, and 24 arises because dose rates are usually
expressed in terms of "per day", while other coefficients in the model are "per hour".
CO
Equation (2) suggests using the function:
idose = 0.5 * {(Dose -S-D) + ^(Dose -S-D)2 + 4 x DosexS \ Eqn
(3)
i to describe the relationship between administered dose (Dose) and a scaled internal
I dose, where
g
—j jj and
*~ I 24V
13 I £) Q^x. |n this parameterization of the model, Vmax, ke, and total blood flow (= Qb
f \ = BW
I + Q,) should be proportional to body weight, so both S and D are independent of
[ body weight. This is a function of two parameters (S and D), and approaches the
I function idose = Dose - D for larger doses; the slope with respect to dose when
I Dose is close to 0 is S/(S + D). D quantifies the displacement of the relationship
T3 I between Dose and idose from the identity relationship, and S controls the shape of
CD | the relationship at low doses. In the limit as D - 0 or S -°°, Equation (7) converges
.£2 1 to idose = Dose.
> I
(D | In fact, it is reasonable to use Equation (3) to approximate the relationship
CV [ between internal dose and administered dose in the chronic dosing setting, even in
[ the absence of a detailed pharmacokinetic justification. The general properties of
I the equation capture the expected effects of first-pass metabolic clearance of an
1 active compound: a shallow shoulder of the curve at lower doses, with a slope that
I Appendix III.B.1-Page 4
-------
f increases to a limit as the dose increases. As long as S and D are non-negative,
I varying these two parameters should result in a good approximation to virtually any
I low-dose deviation due to metabolic clearance, at least at the resolution available in
I bioassay dose-response data.
CM I 2. Equation for the Expanded Model.
O f
1 The expanded model is just the basic model (Eqn. 1), in which Dose is replaced
I by an expression relating administered dose to internal dose. Note that, in this use
•^ f of the model, the parameter BMD is the internal dose that corresponds to a BMP
CO | level of inhibition. Calculating the benchmark dose that corresponds to that internal
» [ dose requires setting Eqn. 3 equal to BMD, and solving for Dose.
^ [ Incorporating Differences among Datasets in the Modeling and Modeling
| Variability
CO I The data for each chemical were modeled independently of all other chemicals.
GO | However, the data for any one chemical were to some extent from heterogeneous
(!) 1 studies, grouped hierarchically. At the highest level of the hierarchy, the data could
& I come from multiple major studies, indicated by different MRID numbers (A MRID no. is
1 an identification code for a particular study; MRID is used in this discussion to describe
1 the major studies). At that level, it could be expected that analytic methods could differ
| most distinctly, and different major studies might use different units to express
GO I cholinesterase activity. Within a major study were individual dose-response studies,
| often the result of multiple intermediate observations in a sub-chronic or chronic study.
1 Although these were part of the same study, since the data collection was separated by
[ relatively wide time intervals, there is still a reasonable expectation that details of
I method might vary among such data sets. Finally, within each individual dose-response
I study were data for both males and females. In order to combine all the data for a given
[ chemical with a single model, all this variability needed to be incorporated in the model.
I This was done with a combination of allowing fixed effects to take different values in
| different dose-response data sets and sexes, treating a parameter as if it varied
[ randomly across data sets, and treating some parameters as fixed for any given
I chemical. The following describes how each parameter was treated in the modeling.
Parameters f°r the submodel relating administered dose to internal dose (S and
D in Eqn 3) were given a single value for a given chemical, though they could differ
between chemicals.
The parameter governing the horizontal asymptote, PB was allowed to differ between
0 [ sexes, but otherwise to be the same value for all datasets for a given chemical.
C/) ! * The background parameter, A, was estimated as a fixed value for each individual
*£ | data set for each sex.
| • The parameter BMD was treated as a random effect. Specifically,
where JL/IBMD is the log of the geometric mean of the distribution of BMD among data
sets, EMmD and EDataSet are normally distributed random variables with mean 0 and
different standard deviations that reflect variation of \n(BMD) among MRIDs and
Appendix III. B.1 - Page 5
-------
I among datasets within MRID, respectively. The parameter JLIIBMD was allowed to vary
[ between males and females, but for each sex was constant over all data sets for a
I chemical. Some chemicals were represented by only one MRID, and some were
I represented by MRIDs with only a single data set in them. The above formula was
I reduced in the logical way for such chemicals. In particular, when only a single data
I set was available for a chemical, all the random effect terms would drop away,
| leaving only the log geometric mean for each sex.
= • The variation among individual observations from animals of the same sex within a
| data set was assumed to be normal, with mean determined by the above model, and
I variance proportional to the mean cholinesterase activity level. An earlier version of
[ this analysis (FIFRA SAP, 2001 b) had treated the variance to be proportional to the
[ square of the mean, and was based on analyzing the relationship between mean
\ and variance across studies and chemicals. The current model is due to
I reexamining the relationship, .focusing on the relationship within studies. The
I constant of proportionality was allowed to differ among MRIDs for a chemical, to
| allow for differences in units.
I Estimating Parameters
[ It proved to be impossible to jointly estimate all the parameters for either the basic or
[ the expanded model simultaneously. Therefore, parameters in these models were
I estimated using a combination of either nlme, a method for nonlinear mixed effects
I models (when there were multiple MRIDs and/or data sets for a chemical; almost all the
1 chemicals) or gnls, generalized least squares, and profile likelihood. The functions nlme
| and gnls are from the package nlme for the statistical package R (lhaka and Gentleman,
The R package nlme estimates parameters for nonlinear mixed effects models using the
approach described in Lindstrom and Bates (1990). Davidian and Giltinan (1995, pp
164 - 174) give a good description of this model, where they refer to it as being based
on "conditional first-order linearization". This approach involves approximating the
nonlinear function using a Taylor Expansion before carrying out maximum likelihood
estimation. The implementation in nlme allows the fixed and random effects to be
expressed as linear models of other independent variables. In this analysis, for
example, IBMD was allowed to differ between sexes by modeling IBMD ~ sex -1, where
sex is a categorical variable in the data set that takes the values "F" or "M". The term "-
1" indicates that an intercept term should not be fit for this model, so there would be an
estimate of IBMD for each sex. ,
The function gnls in the R package nlme has a similar user interface as does the
function nlme, but is appropriate when there are no random effects terms other than the
error variance. Generalized least squares as a method is well described in Chapter 2 of
Davidian and Giltinan (1995).
Parameters for the basic model were estimated first, and served as the basis for
estimating parameters for the expanded model.
Appendix III.B.1 - Page 6
-------
[ To fit the basic model, the values of PBF and PBM were set to each value on a grid of
[ appropriate values, and the remaining parameters (background parameter for each
i individual data set, mean of In(BMD) for males and females, standard deviations of
I In(BMD) among MRIDs and among datasets within MRID, and parameters for the error
I variance) were estimated by the method appropriate to the dataset (that is, either gnls
C\l I or nlme\ see the discussion of these two methods, below). When all the models for that
O | particular grid of PB values were fit, a new grid was constructed by using the values of
^ I PB on either side of the grid point with the largest loglikelihood as the new extremes,
^..-, I and repeating the process. When no BMD estimate on grid points surrounding the point
-^ [ with the largest loglikelihood differed from the BMD at the maximum by more than 5%,
CD [ the process of iterative refining the grid stopped.
i 1
~*~» i A similar method was used to estimate S and D in the expanded model. First, the
C I values of PBF and PBM were fixed to their best estimates for the basic model, and were
Q^ I not further modified. In the expanded model, the grid being explored and refined was of
^ I values for S and D, but otherwise the process was the same as for the basic model. In
(/) I the expanded model, the criterion for convergence was no difference between the
00 1 maximum on the grid and neighboring points of greater than 10%.
CD j
£# 1 References
(/) | •
| Davidian, M. and Giltinan, D. M. 1995. Nonlinear Models for Repeated Measurement
1 Data. Chapman and Hall. New York.
co !
FIFRA SAP. (2001 b). "Preliminary Cumulative Hazard and Dose-Response
Assessment for Organophosphorus Pesticides: Determination of Relative Potency and
0 | Points of Departure for Cholinesterase Inhibition ." Report from the FIFRA Scientific
> | Advisory Panel Meeting of September 5-6, 2001 (Report dated September 11, 2001).
-£3 . [ FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of
J5J j Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency.
"3 I Washington, DC. SAP Report 2001-OX. Internet:
CI I http://www.epa.gov/scipoly/sap/2001/index.html
mr-^f 5
FIFRA SAP. (2002) "Methods Used to Conduct a Preliminary Cumulative Risk
Assessment for Organophosphate Pesticides". Report from the FIFRA Scientific
Advisory Panel Meeting of February 5-7, 2002. (Report dated March 19, 2002): FIFRA
Scientific Advisory Panel, Office of Science Coordination and Policy, Office of
Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency.
Washington, DC. SAP Report 20020-01. Internet:
http://www.epa. gov/scipolv/sap/index.htm#iune
lhaka, R. and Gentleman, R. 1996. R: A language for data analysis and graphics.
Journal of Computational and Graphical Statistics 5: 299-314.
Pinheiro, J. and Bates, D. M. 2000. Mixed Effects Models in S and S-Plus: Springer.
Berlin.
Appendix III.B.1 - Page 7
CL
"D
0>
C/)
">
CD
-------
I USEPA (2001 b). "Preliminary Cumulative Hazard and Dose-Response Assessment for
i Organophosphorus Pesticides: Determination of Relative Potency and Points of
I Departure for Cholinesterase Inhibition"; (released July 31, 2001) Office of Pesticide
| Programs, US Environmental Protection Agency, Washington DC. Internet:
i http://www.epa.gov/pesticides/cumulative/EPA_approach_methods.htm .
CM j
O USEPA (2001 c). "Preliminary Organophosphorus Pesticide Cumulative Risk
Assessment", (issued in December, 2001), Office of Pesticide Programs, US
Environmental Protection Agency, Washington DC. Internet:
http://www.epa.gov/pesticides/cumulative/pra-op/index.htm
CD
C
CD
E
CD
cn
C/)
D
E
13
O
QL
O
•o
CD
C/)
'>
0
01
Appendix III.B.1 - Page 8
-------
CM
o
"%*""*"
"*•*-».
CO
i
c
CD
fc
C/)
CO
0
00
CO
52
CC
0
"4!3
CO
"r>
D
6
"O
CD
0)
cr
Appendices
B. Hazard/RPF
2. Dose-Response Curves
Appendix III.B.2- Page 1
-------
CM
O
CO
C
0
E
co
(A
0
CO
CO
CO
be
0
O
DL
O
"a
0
CO
">
0
oe
Key to Tables in Appendix III.B.2
Toxicology Profile Tables:
Key to Figures in Appendix III.B.2
a. Dose-response Curve (Basic):
b. Residuals from Basic Model:
c. Profile Likelihood for PR:
For each chemical, the studies reported
in the toxicology profile tables
correspond to the studies listed in the
figures. Specifically, oral studies
containing whole brain rat
cholinesterase data used to determine
potency are reported in the tables. In
addition, dermal and inhalation toxicity
studies are listed only for the chemicals
with residential/ nonoccupational
exposures and for the index chemical
(methamidophos).
Dose-response curve(s) from the basic
model (low dose linear model). For
chemicals with more than one study, the
studies are plotted separately. Male
data are red and female data are blue.
Plot of residuals from the basic model.
Dotted red line represents 10% brain
cholinesterase inhibition.
Profile likelihood plot for PB (i.e.,
horizontal asymptote). The x-axis gives
the ranges of PB tried for female rat
cholinesterase data (PBF). The y-axis
gives the ranges of PB tried for male rat
cholinesterase data (PBM). As color
moves from red to orange to yellow to
very bright yellow, the likelihood values
increase to a peak. The peak is marked
by an X. Open circles are points that
are not significantly different (P-value >
0.05) from the peak.
Appendix III.B.2 - Page 2
-------
i d.
Profile Likelihood for 0 and S:
CM
CD-
CO
c
0)
= e.
0)
05
CO
GO
0
J29
13
Dose-response Curve (Expanded):
I f.
Residuals from Model
w/Low Dose Curvature:
Profile likelihood plot for D (i.e.,
horizontal displacement along the x-axis
of the dose-response curve) and S (i.e.,
shape). As color moves from red to
orange to yellow to very bright yellow,
the likelihood values increase to a peak
The peak is marked with an X. Closed
circles are points that are not
significantly different (P-value > 0.05)
from the peak. Plot is listed only for
those OPs where the expanded model
fit the cholinesterase data significantly
better than the basic.
Dose-response curve(s) from the
expanded model (low dose flat model).
For chemicals with more than one study,
the studies are plotted separately. Male
data are red and female data are blue.
Plot(s) is/are listed only for those OPs
where the expanded model fit the
cholinesterase data significantly better
than the basic.
Plot of residuals from the expanded
model. Dotted red line represents 10%
brain cholinesterase inhibition. Plot is
listed only for those OPs where the
expanded model fit the cholinesterase
data significantly better than the basic.
DL
O
"O
O
Ml
CD
cc
Appendix III.B.2 - Page 3
-------
Table III.B.2-1. Acephate: Toxicology Profile Table
Acephate
MRID#
40504819
00084017
45134301
44541101
45134302
40504818
40645903
Guideline No.
82-1
(870.3100)
83-5
(870.4300)
82-2
(870-3200)
82-2
(870.3200)
82-4
(870.3465)
82-4
(870.3465)
82-4
(870.3465)
Study Type
Subchronic Oral Toxicity-Rat
(Special ChE inhibition study)
Combined Chronic Oral Toxicity/
Carcinogenicity-Rat
21-Day Dermal Toxicity-Rat
21-Day Dermal Toxicity-Rat
Subchronic Inhalation Toxicity-Rat
4-Week Inhalation Toxicity-Rat
4-Week Inhalation Toxicity-Rat
HED
Doc. No.
006680
012544
14258
004951
012544
14210
41528
13396
14223
41528
12544
12544
Dose
0/0, 0.15/0.12, 0.36/0.28, 0.76/0.58, 11.48/8.90
mg/kg/day (females/males)
0/0, 0.3/0.2,3. 1/ 2.4,47.2/38.2 mg/kg/day
(females/males)
0, 20, 30, 40, 50 mg/kg/day
0, 12, 60, 300 mg/kg/day
0, 0.001064, 0.003123, 0.005550 mg/L
0 (air), 1 .05, 1 0.8, 93.6 mg/m3
0 (air), 0.187, 0.507 mg/m3
Guideline/
Nonguideline
Nonguideline
Guideline
Nonguideline
Guideline
Nonguideline
Guideline
Guideline
Species/
Strain,
Rat/ Sprague Dawley
Rat/ Sprague Dawley
Rat/ Sprague Dawley
Rat/ Sprague Dawley
Rat/ Sprague Dawley
Rat/ Fischer
Rat/ Fischer
Appendix III.B.2 - Page 4
-------
I Figure III.B.2-1. Acephate: Dose-response Curves Using the Basic Model, Plot of the
I Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot For PB
a1 . Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
CM
o
^1
CD
t
_!__»
C
0
E
CO
CO
0
CO
CO
*f 1*
-^
CO
» •muni*
(T
0
1*^
M axmBiiB
-4_~<
_03
D
mmmf
E
ZJ
O
a.
O
0
.£2
^
0
rv*
I «
i
i _£ -
I so
1 1°
| w in -
I °
|
o -
1
1
\
^^^•S'
w
i i i i i
«-.
3 "^
fi:
LU Tj- -
^
O
< ^
o -
^^,
i i i
0 10 20 30 40 024
I Dose (mg/kg/day) Dose
.
6810
(mg/kg/day)
MRID:84017 MRID: 40504819
b. Residuals from Basic Model c. Profile Likelihood for PB
E
W _
i D
! |o -
E ID
1 0
= w
— ^_
cr» _
= I
i
1
s
o o
^o * °o0 *
.^Q % . -o-0d .®-
fl °8 *
o
o
II 1 !
n -j> O O O O
o
to ~
__ CNJ ~
CL
d
rn
C\l -
D
K o o o o
>" o o o o
o o o o
> O O 0 O
> o o M o
> o o o o
> o o o o
> o o o o
t''':'/s-
-------
Table 111. B.2-2. Azinphos-methyl: Toxicology Profile Table
Azinphos-methyl
MRID#
43826601
41119901
Guideline
No.
82-7
(870.6200)
.83-5
(870.4300)
Study Type
Subchronic Neurotoxicity-Rat
Combined Chronic Oral Toxicity/
Carcinogenicity-Rat.
HED
Doc. No.
011898
008300
Dose
0/0, 1.05/0.91, 3.23/2.81, 6.99/7.87 mg/kg/day (females/males)
0/0, 0.31/0.25, 0.96/0.75, 3.1 1/2.33 mg/kg/day (females/males)
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Wistar
Appendix III.B.2 - Page 6
-------
CM
o
c
0)
V)
GO
0
CO
CO
DC
_2
13
E
ZJ
O
Q_
13
0)
£2
*>
CD
o:
I Figure III.B.2-2. Azinphos-methyl: Dose-response Curves Using the Basic and
I Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
[ Profile Likelihood Plots For PB, D, and S
I
I
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
O d
<
D
O
I
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Dose (mg/kg/day)
MRID: 41119901
b. Residuals from Basic Model
LJJ
-C
O
J
0 2 4 B
Dose (mg/kg/day)
MRID: 43826601
c. Profile Likelihood for PB
•
8
n -
CM -
3
-D "- -
JS O -
"S T -
o
c? _
O
o
_ OQ O
060
0
"O" , "0"
8
0
o o
0
i^ l :
0.0 0.2 0.4 0.6 0.8
Fractbn of Inhibition
d. Profile Likelihood for D and S
CM
o ~
d
_
00
S -
o
.
o
d
0
o
o
s w "
>• o
> 0
3 0
t: '•• o
>; ' o
>' 0
) O
5 0
S 1*1
Y
d
0.07
\j
O
O
p
o
o
o
o
o
0
A
o
o
0
o
o
0
0
Q
0
_o_
0.
\J
o
o
o
o
o
o
o
o
0
_i*L
1
oa
o
o
o
o
o
o
o
o
o
rt
0
PBF
0
o
o
o
o
o
o
o
0
09
«J
o
o
o
o
o
o
o:
o
n
0
0
o
o
o
o
o
o
o
f",
I
0.1
1
G
1
O' '•'"
0
Ol;
'dl
1
,
'
1
I
j
0
el. Dose-response Curve (Expanded)
I*
i°
LU
JZ
O
10
'
I
0.002 0.004 0.006 0.008
S
p
D
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Dose (rng/kgAday)
MRID: 41119901
Appendix III.B.2 - Page 7
-------
Figure III.B.2-2. Azinphos-methyl con't: Dose-response Curves Using the Basic and
Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
Profile Likelihood Plots For PB, D, and S
e2. Dose-response Curve (Expanded)
f. Residuals from Model
w/Low Dose Curvature
— 1O _
CD -
LU
O
2 .4 6
Dose (mg/kg/day)
MRID: 43826601
10
t/J O — -
13
O
t/3
o
0
0.0 0.2 0.4 0.6
Fraction of Inhibition
0.8
Appendix III.B.2 - Page 8
-------
Table III.B.2-3. Bensulide: Toxicology Profile Table
Bensulide
MRID#
43919601
44161101
44801101
44809401
Guideline No.
82-1
(870.3100)
83-5
(870.4300)
82-2
(870.3200)
Study Type
Subchronic Oral
Toxicity-Rat
• Combined Chronic Oral Toxicity/
Carcinogenicity-Rat
21 -Day Dermal Toxicity-Rat
HED
Doc. No.
12289
12289
013532
Dose
0/0, 5/5, 15/15, 45/46, or 100/110 mg/kg/day (females/males)
.0/0, 1/1, 15.30/15.10, 61 .30/60.10 mg/kg/day (females/males)
0, 30, 50, 500 mg/kg/day
Guideline/
Nonguideline
Guideline
Guideline
Nonguideline
Species/
Strain
Rat/
Sprague Dawley
Rat/
Sprague Dawley
Rat/CD
Appendix III.B.2 - Page 9
-------
I Figure III.B.2-3. Bensulide: Dose-response Curves Using the Basic and Expanded
[ Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
| Likelihood Plots for PB, D, and S
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
CM I .
d3 i
---«. |
^ I Ig ~
CD I f * "
S
«
c 1 <§ "
CD j B.
(
E i
IT
q^,
T^
. i
m I °
(!n I
0) i
00 I
00 I
< | 2-1
ju^ i q _
.CO j |^
fV ^ JS ° "
LJfc«<* - Lt_
= -a °- -
0= « o
= -a
>= u
1 fl5 •"
"i*"*; s
0
CO i - -
<
s^__^^ (
: "^-^^
i i
20 40
Dose
o
i
""—-— -^,
"^^-i
1
o -
in
I| _
f °
IB
mg -
<
o -
& T
^^^ i^
"^=^^^^^^j
.
iii i i i i i i i
BO 80 100 0 10 20 30 40 50 60
(mg/kg/day) Dose (mg/kg/day)
MRID: 43919601 MRID: 44161 101
b. Residuals
G
&
o
1
1 °
t
l
.
t '»"¥"••
P :
o o o ;
t
>
o
^ : «—
0
E !
13 j
0
1
1
1
from Basic Model c. Profile Likelihood for PB
0
Q f\
o
0 0
a
Q
s .
o
M
B
D
«.
O
C5
J^~~V ™«r— ta.— w—o— -o ~"v~ o -«o v.
3 0 O 0 O O 0 O 0 O C
JOOOOOOOOO^
> O O 0 O O 0 O O O •'<
JOOOOOOOOOC
>O'OOO. OOOOOC
>oooooooooc
>oooooooooc
JOOOO OOOOOC
3OCOOOOOOOC
ill ri ~T l i l s l
0.1 0.2 0.3 0.4 0.5 0.00 0.01 0.02 0.03 0.04 0.05
Fractbn of I nhibition PBF
O j d. Profile Likelihood for D and S e1 . Dose-response Curve (Expanded)
a. I
O I s~
WH3P*** 1
\**r ~
CD I Qg
C/D - ^ ~
^^
> ^
(^^- _ ^
CD I
fV = m -
n
i °
9
»
•
|
ft
r
0.001
B
i
""**™w-1-- ""-
•
*
A
1
0.003
*^ T*"--" ;' ' '-'•••'*
* * «
* » I
• * <
• -» «
fill
=r
5
— 0
^4 P^ —
IS
*c
L^ 0
Q D _
o -
•r
fT_I T
1^'^^r~-^^
" ' ^v"? *w ^^^*^^**^_
"^^^ ^~~i
^"-^J,
1
1 i 1 1 1 1
0.005 0.007 0 20 40 60 80 100
S Dose (mg/kg/day)
MRID: 4391 9601
-------
CO
CD
C/)
CO
CO
0
E
3
O
Q_
O
"O
CD
CO
">
0
cr:
Figure III.B.2-3. Bensulide Con't: Dose-response Curves Using the Basic and
Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
Profile Likelihood Plots for PB, D, and S
CM
O
CO
i
tl
0
E
t/\
Illlllllllllllllllllllllllllllllllll
AChE Activity
e:
o
o
o -
m
r-
O
O
0 -
o
T—
o
o
o ~
m
o -
!. Dose-response Curve (Expanded)
i jt_^
f__|_^^
\
Scaled Residual
: 1 1 1 1 1 i 1
1 0 10 20 30 40 50 60
m
b ""
o
b ~
in
b —
1
o
f. Residuals
w/Low Dose
O
8
9
8
0
8
8
••J 1
0.0 0
0
8
o
0
from Model
Curvature
8
o
o
1 1 1 1
1 0.2 0.3 0.4 0.5
Dose (mg/kg/day)
MRID:44161101
Fractbn of I nhibition
Appendix III.B.2- Page 11
-------
Table III.B.2-4. Chlorethoxyfos: Toxicology Profile Table '
Chlorethoxyfos
MRID#
41290632
42559215
41736837
Guideline No.
82-1
(870.3100)
82-1
(870.3100)
83-5
(870.4300)
Study Type
Six Week Oral Toxicity - Rat
Subchronic Oral Toxicity - Rat
Combined Chronic Oral Toxicity/Carcinogenicity
Study - Rat
HED
Doc. No.
008330
NA
NA
Dose
0/0, 0.014/0.009, 0.132/0.091, 0.66/0:477,
1 .3/0.958 mg/kg/day (females/males)
0, 0.008, 0.080, 0.635, 1.23, 1.63 mg/kg/day
(females only)
0/0, 0.005/0.004, 0.042/ 0.031, 0.208/ 0.154,
0.41 6/ 0.31 1 mg/kg/day (females/males)
Guideline/
Nonguideline
Supplemental
Guideline
Guideline
Species/
Strain
Rat/Crt:CD®BR
Rat/Crl:CD®BR
Rat/Crl:CD®BR
NA=Not available
I.B.2 Page 12
-------
CM
O.
CD
c
CD
CO
CD
If)
CO
0
13
E
13
O
0_
O
CD
00
">
0
I Figure III.B.2-4. Chlorethoxyfos: Dose-response Curves Using the Basic and Expanded
1 Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
's, D, and S
| Likelihood Plots for P,
:
I
a1. Dose-response Curve (Basic)
IO -
= LLJ OJ -
I ° ,_
I
O -
0.0 0.2 0.4 0.6 0.8 1.0
Dose (mg/kg/day)
MRID: 41 290632
1.2
a3. Dose-response Curve (Basic)
in -i
5 CD
•f« H
LU
_C
O
0.0
0.5 1.0
Dose (mg/kg/day)
MRID: 42559215
1.5
c. Profile Likelihood tor PE
8 J
O
8
O
) • • • \r~ ' v ~ \>
O O 0 O OO O . 0 • O
O O O O O O O 0 O
ooooooooo
o o o o o o o o o
ooooooooo
0 O O 000 O O O
0 O O O OO O OO
OOOOQOOOO
ooooooooo
0.000
I
0.005
0.010
PBF
0.015
a2. Dose-response Curve (Basic)
LU
-C
o
f
0.0 0.1 0.2 0.3
Dose (mg/kg/day)
MRID: 41736837
f
0.4
b. Residuals from Basic Model
CO
id
13
«-
r
T
0.0 0.2 0.4 0.6
F reel b not Inhibition
0.8
d. Profile Likelihood tor D and S
0.020
0.002
0.008 0.010
I.B.2 Page 13
-------
I Figure III.B.2-4. Chlorethoxyfos con't: Dose-response Curves Using the Basic and
\ Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
I Profile Likelihood Plots for PB, D, and S
CM
x —
CO
i
Hh-1
c
0
E
CO
CO
0
CO
CO
k
>v
>sv
^^L
i1*^^,^
^*s^~*-twlfll
CO
o
ra
|s -
T) W
W 0 -
"5 1
0
c/?
to
D —
1
0 0
8 o
§c "^^
.
o
8
0
o
o
= 1 1 1 1 1 1 1 I t
\ 0.0 0.5 1.0 ' 1.5 0.0 0.2 0.4 0.6 0.8
1 Dose (mg/kg/day) ' Frastbn of Inhibition
[ MRID: 4255921 5 '
«
S . -
i
1
1 •
| •
s ' •
I.B.2 Page 14
-------
Table III.B.2-5. Chlorpyrifos: Toxicology Profile Table
Chlorpyrifos
MRID#
40952801
42172802
40952802
Guideline
No.
82-1
(870.3100)
83-5
(870.4300)
83-5
(870.4300)
Study Type
Subchronic Oral Toxicity-Rat
Combined Chronic Oral Toxicity/
Carcinogenicity- Rat
Combined Chronic Oral
Toxicity/Carcinogenicity- Rat
HED Doc.
No.
007102
009733
010605
013240
007107
013240
Dose .
0, 0.10, 1.00, 5.00, 15.00 mg/kg/day
0/0, 0.01/0.01, 0.37/0.33, 7.61/6.77 mg/kg/day (females/males)
0, 0.05, 0.10, 1, 10 mg/kg/day
Guideline/
Nonguideline
Guideline
Guideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Fischer
Rat/ Fischer
I.B.2 Page 15
-------
I Figure III.B.2-5 Chlorpyrifos: Dose-response Curves Using the Basic and Expanded
I Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
[ Likelihood Plots For PB, D, and S
CM
O
CO
S
-t— I
c
0
E
V)
60
0
C/)
0
E
13
O
Q_
O
0
CO
0
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
.- ,
= --P
1S
0
5 10
Dose (mg/kg/day)
MRID: 40952801
IT
15
a3. Dose-response Curve (Basic)
GO i s«
I
LU
o
<
I
0246
Dose (mg/kg/day)
MRID: 42172802
c. Profile Likelihood for PB
f
00' OOOOOOO'
ooooooooo
3
b
3
?
i
o
0
:J;.O
O
O
jjLo .'
(v'O
T
0.22
O
o
o
o
o
o
o
0
o
o
o
o
o
o
o
I
23
o
o
o
o
o
o
o
0
o
o
o
o
o
Q
o
I
24
0
o
o
o
o
0
1
0.25
0
o
o
o
o
o
0
0
o
o
o
0
0
0
1
0.26
o
0
0
0
0
0
0
0
c
c
c
c
c
1
2'
00 -
-
LU "fr
2468
Dose (mg/kg/day)
MRID: 40952802
10
b. Residuals from Basic Model
o
o 0|
0 0 fc
o o
it ' OO
|L.°.L. o
. - -S8 °
I« ! °
• ° • «° n
i O O
i
O i
o ! °
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fraction of Inhibition
d. Profile Likelihood for D and S
0.002
0.008 0.010
III.B.2 Page 16
-------
0
= Figure III.B.2-5 Chlorpyrifos con't: Dose-response Curves Using the Basic and
| Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
I Profile Likelihood Plots For PB, D, and S
I e1. Dose-response Curve (Expanded) e2. Dose-response Curve (Expanded)
'oo -
s LU
D
;>
LU
5 10
Dose (mg/kg/day)
MRID: 40952801
15
e3. Dose-response Curve (Expanded)
2468
Dose (mg/kg/day)
MRID: 40952802
f. Residuals from Model
w/LowDose Curvature
10
00 -
= LU
= O
Cvj -
ra
"D
_
ro
O
O
io
(.a..
246
Dose (mg/kg/day)
MRID: 42172802
i i i i i i r
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Frastbn of Inhibition
I.B.2Page17
-------
Table III.B.2-6. Chlorpyrifos-methyl: Toxicology Profile Table
Chlorpyrifos-methyl
MRID#
42269001
44906902
Guideline
No.
83-5
(870.4300)
82-1
(870.3100)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Oral Toxicity-Rat
HED
Doc. No.
009560
014122
Dose
0, 0.05, 0.1, 1, 50 mg/kg/day
0,0.1,1,10, 250 mg/kg/day
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/Fischer
Rat/Fischer
I.B.2 Page 18
-------
5 Figure III.B.2-6. Chlorpyrifos-methyl: Dose-response Curves Using the Basic Model,
1 Plot of the Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot
! ForP8
CM
O
CO
i
c
0
CO
CD
C/)
c/)
r
CD
E
ZJ
O
QL
O
"O
CD
C/}
">
0)
cr
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
o _
= LU
I 5
= <
O _
•> to -
LU
O
<
10 20 30 40
Dose (mg/kg/day)
MRID: 422B9001
50
I
I
50 100 150 200
Dose (mg/kg/day)
MRID: 44906902
250
b. Residuals from Basic Model
c. Profile Likelihood for PB
0.1 0.2 0.3 0.4 0.5
Fractbn of Inhibition
O.B
0.412
0.415 0.41 B
I.B.2Page19
-------
Table III.B.2-7. Diazinon: Toxicology Profile Table
Diazinon
MRID#
40815003
41942002
Guideline No.
82-1
(870.3100)
83-1
(870.4100)
Study Type
Subchronic Oral Toxicity-Rat
Chronic Oral Toxicity-Rat
. HED Doc. No.
007041
007553
012219
010331
012219
Dose
0/0, 0.04/0.03, 0.40/0.30, 19/15, 212/168 mg/kg/day
(females/males)
0, 0.005/0.004, 0.07/0.06, 6/5, or 12/10 mg/kg/day
(males/females)
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/
Sprague Dawley
Rat/
Sprague Dawley
I.B.2 Page 20
-------
C\J
o
CD
I
c
CD
C/)
C/)
0
C/)
Lr
CD
E
ZJ
O
OL
O
CD
CO
">
0
-. Figure III.B.2-7. Diazinon: Dose-response Curves Using the Basic and Expanded
| Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plots for PB, D, and S
a1 . Dose -response Curve (Basic)
a2. Dose-response Curve (Basic)
o
o
o
I :18
; •*• O
I !cS -
— -f
i ^-
:
°
i
3
i
o
o
o
CM
io
ug -
o
o
50 100 150
Dose (mg/kg/day)
MRID: 40815003
200
.
0
b. Residuals from Basic Model
i _ CM -
! &
o
o
S
tv
^
ci
2 4 6 8 10 12
Dose (mg/kg/day)
MRID: 41942002
c. Profile Likelihood for PB
0 O 0 O
ooooooooo
ooooooooo
OOOOQOOOO
000X00000
ooooooooo
O O O O 0 O O 0 O
ooooooooo
o o
0.0 0.1 0.2 0.3 0.4
Fraction of Inhibition
0.5
0.426
0.428
P
0.430
0.432
BF
d. Profile Likelihood for D and S
e1. Dose-response Curve (Expanded)
o
o
o
£•§
LU
0.0010 0.0015 0.0020 0.0025 0.0030
50 100 150
Dose (mg/kg/day)
MRID: 40815003
I
200
I.B.2Page21
-------
CM
O
CD
C
0
E
CO
CO
0
CO
CO
<
J*
CO
o:
0
>
",+_J
_03
D
E
13
O
CL
O
"D
0
CO
">
0
cc
Figure III.B.2-7. Diazinon con't: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S
e2. Dose-response Curve (Expanded)
f. Residuals from Model
w/Low Dose Curvature
U § -
o —
n
T3
0)
-ft
CO
2 4 6 8 10
Dose (mg/kg/day)
MRID: 41942002
t
'12
o
0
8
ft ° o °
S o °
o
| 6 ^
i
8
0
o
o
0
0.0 0.1 0.2 0.3 0.4 0.5
Fractb not Inhibition
I.B.2 Page 22
-------
Table III.B.2-8. Dichlorvos: Toxicology Profile Table
Dichlorvos
MRID#
41004701
00057695
00632569
Guideline No.
82-1
(870.3100)
83-5
(870.4300)
Study Type
Subchronic Oral (Gavage)
Toxicity-Rat
Combined Chronic Inhalation
Toxicity/Carcinogenicity-Rat
HED Doc. No.
007448
001466
006860
Dose
0, 0.1, 1.5, 15 mg/kg/day (gavage)
0, 0.05, 0.5, 5 mg/m3
Guideline/
Nonguideline
Guideline
Supplemental
Species/
Strain
Rat/
Sprague Dawtey
Rat/
Carworth Farm E
(CFE)
I.B.2 Page 23
-------
Figure III.B.2-8. Dichlorvos: Dose-response Curve Using the Basic Model, Plot of the
Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
a. Dose-response Curve (Basic)
b. Residuals from Basic Model
CM
0 g _
x- g"-
'K™"1™ ~~'~ ^3
i 5 g _
"c <
(D
E
CO
"4ZJ
03
D
E
D
o
Q_
o
0
CO
E
1
E
1
E
E
[
i
|
E.
i
E
|
0
s
b
i
Q.
O
to
b
D
CO
(a
b
r*
>
>
}
)
>
*
>
>
)
i
""Xr*™
0
0
0
o
o
0
0
o
0
o
o
o
o
o
0
o
o
0
<-»
0
o
0
o
o
0
0
0
0
jwr^^s.
o
o
o
o
0
o
o
0
0
***£irw'
O
O
0
0
o
o
0
0
o
"•—O™
o
0
0
o
o
o
0
0
o
—tr-
O
O
O
o
6
6
0
6
Q
J""-£JTS— -**(RJ™
0 0
.00
0 O
0 O
0 0
0 0
0 O
o 'o
0 O
""C
c
<
c
c
c
c
c
c
c
0.00 0.01
0.02 0.03
PBF
0.04 0.05
I.B.2 Page 24
-------
Table III.B.2-9. Dicrotophos: Toxicology Profile Table
Dicrotophos
MRID#
44527802
43980201
Guideline No.
83-5 '
(870.4300)
82-7
(870.6200)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Neurotoxicity -Rat
HED
Doc. No.
012994
013048
Dose
01, 0.03/0.02, 0.32/0.25,1.74/1.42 mg/kg/day (females/males)
0/0, 0.04/0.04, 0.45/0.39, 2.38/2.03 mg/kg/day (females/males)
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/Sprague
Dawley
Rat/Sprague
Dawley
I.B.2 Page 25
-------
; Figure III.B.2-9. Dicrotophos: Dose-response Curves Using the Basic Model, Plot of the
1 Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
CM
O
CD
i
C
0
E
to
CO
0
C/)
tO
CO
ir
0
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
*«.. O
i :5«
I I
E 111
[ I
0.0 0.5 1.0 1.5 2.0
Dose (mg/kg/day)
MRID: 43980201
b. Residuals from Basic Model
UJ •
0.0 0.5 1.0 1.5
Dose (mg/kg/day)
MRID: 44527802
c. Profile Likelihood for PB
^~
8
°Q
O
o
0
0
-e-
uo
CD
00
o
0.0 0.2 0.4 0.6 0.8
Frastbn of Inhibition
T
0.106
>••
J
kl ...
o
o
o
•_o
0
'&'':
6
o
o
o
o
o
o
o
o
o
o
o
0
o
o
o
o
0
o
o
o
...e\.
O
o
o
o
o
o
o
o
,o ,
o
o
M
o
o
o
o
o
o
__.&,...
o
0
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
0
0
o
o
o
0
o
o
o
o
o
o
o
•O''
o
o
o
'.'e
c
c
c
c
c
c
c
0.108
P
0.110 0.112
BF
O
CL
O
"O
0
GO
0
.B.2 Page 26
-------
Table III.B.2-10. Dimethoate: Toxicology Profile Table
Dimethoate
MRID#
164177
Guideline No.
83-5
(870.4300)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
HED Doc. No.
006398
008457
Dose
0/0, 0.06/0.04, 0.30/0.23, 1.48/1.16, 6.29/4.82 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Species/
Strain
Rat/ Wistar
I.B.2 Page 27
-------
O
CD
i
C
0
E
0)
0
C/)
C/)
E
13
O
DL
O
"O
0
0
s Figure III.B.2-10. Dimethoate: Dose-response Curve Using the Basic Model, Plot of the
I Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
a. Dose-response Curve (Basic)
b. Residuals from Basic Model
o
o -
: rs D
I <° H
:
TJ
0> -^
13 I
O
(f)
1 23 4 5
Dose (mg/kg/day)
MRID: 164177
c. Profile Likelihood for PB
^»ft
^
C/5
MMMM
•v^y
r
JL
0
4— '
fR
E f? -
D
I i^
s a. n -
D
1 ™
i d
s
>
5
p
>
>
)
>
>
O
0
0
:'O
0
'l:o
o
0
o
0
o
o
o
o
o
o
J-v
i
o
0
o
o
o
o
o
o
a
0
o
o
o
o
o
o
o
A
1
o
0
o
o
0
X
o
o
B
t
o
o
o
o
0
o
o
o
^
o
o
o
0
o
o
o
o
T
o
0
o
o
0
0
o
o
a
6
o.
o
6
0
o
o
" ffl
.. /"ii
c
•d
c
d
d
3
o
d
., rt
0.350
0.3BO
0.370
o
-o--
0
-eo-
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fraction of I nhibition
I.B.2Page28
-------
Table III.B.2-11. Disulfoton: Toxicology Profile Table
Disulfoton
MRID#
42977401
43058401
146873/
41850002
44758404
00162338
45239601
41224301
Guideline No.
82-7
(870.6200)
Non-guideline study
83-5
(870.4300)
82-1
(870.3100)
82-2
(870.3200)
82-2
(870.3200)
82-4
(870.3465)
Study Type
Subchronic Neurotoxicity-Rat
Special 6-month Cholinesterase-Rat
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
28-Day Dietary Study - Rat
21 -Day Dermal Toxicity-Rabbit
21 -Day Dermal Toxicity-Rabbit
Subchronic Inhalation Toxicity-Rat
HED Doc.
No.
011456
011249
005029
NA
005556
014448
011242
Dose
0/0, 0.07/0.06, 0.31/0.27, 1.30/1.08 mg/kg/day
(females/males)
0/0, 0.02/0.02, 0.03/0.03, 0.07/0.06 mg/kg/day
(females/males)
0/0, 0.08/0.06, 0.26/0.22, 1.25/0.92 mg/kg/day
(females/males)
Prep 1: 0/0, 0.18/0.17, 1.11/1.04 mg/kg/day
Prep 2: 0/0, 0.16/0.14, 1.29/1.16 mg/kg/day
0, 0.4, 1 .6, 6.5 mg/kg/day
0,0.8, 1, 3 mg/kg/day
Air and PEG-400:50% ethanol vehicle controls,
0.016/0.018, 0.16/0.16, 1.4/1.4 mg/m3 (females/males)
Guideline/
Nonguideline
Guideline
Nonguideline
Guideline
NA
Guideline
Guideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Fischer
Rat/ Fischer
Rat/Fischer
Rabbit/
New Zealand
Rabbit/
New Zealand
Rat/ Fischer
I.B.2 Page 29
-------
Figure III.B.2-11. Disulfoton: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S
C\J
*'**•-
'T~~
x —
CO
5
+_!
c.
0
E
V)
t/)
0
5 1
s
1
i 0"
I ^
1 £"
! i
s LU "0 -
1 °
1
D -
E
B
"*
\\
»
%
Vs.
^V
^x^^-^^^
^^^-^Z^^~-« x
I I 1 1 I 1 1
m ^
""
o
3.0 .
*•'-
1
LU in -
o
*"•
o «*
v^
—f
1 il 1 1 1 1 1
0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
| Dose (mg/kg/day) Dose (mg/kg/day)
MRID: 146873 MRID: 42977401
E
5
(/) i a3. Dose-response Curve (Basic) a4. Dose-response Curve (Basic)
C/>
^B
0)
b<
0
15
_2
E
13
O
Q_
O
-a
0
C/)
^•5*
0
a:
:
| _« _
i CD
z 2*
[ Jo _
I |
1 LU !Q ™.
-C
I "
D -
— _
yai'macaii..! MaW*'*'" « j gi
*~" — ~- — — . i
i f i i i i i i
« -
5"""
1
i
LU m -
-C
O
D -
1
III!
0.00 0.02 0.04 0.06 -1.0 -0.5 0.0 0.5 1
| Dose (mg/kg/day) Dose (mg/kg/day)
MRID: 43058401 MRID: 44758404
z •
i
1 b. Residuals from Basic Model c. Profile Likelihood for PB
i
i
D
| "5
i ~o
E CC
1 Wo
z "B ' -.
E O ^
= W
i o
CNi >
0
io^ ix
OV ^3
rw O
g
0
o
o
1 1 1 1 1
S _
C3
-
^ O
S • t«
o_ ^
0
-
to
to
lK!OOOOOOOQO
JOOOOOOOOO
JOOOOOOOOO
JOOOOOOOOO
>OOOOOQOOO
)OOOOOOOOO
>OOOO)8[OOOO
> O O 0 O Q O 0 0 O
JOOOOOOOOO
.0
\,
(.
c
c
c
c
c
c
c
O k A A A A J
1 1 1 i 1 1
0.0 0.2 0.4 0.6 0.8 0.128 0.132 0.136
Fract on of Inhibition PBF
i
I.B.2 Page 30
-------
Figure III.1.1-11. Disulfoton con't: Dose-response Curves Using the Basic and
Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
Revised OP Cumulative Risk Assessment - 6/1 1/02
=
i
I
|
E
=
=
E
:
=
E
=
E
i
=
E
E
E
=
E
E
=
E
E
E
=
E
=
E
E
E
i
E
E
E
E
=
E
=
E
E
E
E
E
=
i
E
i
E
E
i
E
«,
o
m
s
o
o
D
O
n
AChE Activity (U/G) AChE Activity (U/G)
0 5 10 15 0 5 10 15 a 0.03
I I II? i i i i SI a
d. Profile Likelihood tor D and S
* '* ' * * i
» • » * «
* • * • i
AChE Activity (U/G)
0 5 10 15
• . , . •.
. Dose-response Curve (Expanded)
V^
i i ^r i i i i i i i
010 0.0014 0.0018 0.0 0.2 0.4 O.B 0.8 1.0 1.2
S Dose (mg/kg/day)
MRID: 146873
!. Dose-response Curve (Expanded) e3. Dose-response Curve (Expanded)
\^
AChE Activity (U/G)
0 5 10 15
_
.
i i i i i i i i i t i i i i i
0.0 0.2 0.4 O.B 0.8 1.0 1.2 0.00 0.02 0.04 O.OB
Dose (mg/kg/day) Dose (mg/kg/day)
MRID: 42977401 MRID: 43058401
Innse resnanse Curve f Fvnanrlern '• Residuals from Model
• UOSl B uurve (txpanoecy w/Low Dose Curvature
1
Scaled Residual
-2-101 2
o
o
o
o
-1.0 -0.5 0.0 0.5
Dose (mg/kg&Jay)
MRID: 44758404
1.0
0.0 0.2 0.4 O.B
Fraction of Inhibition
0.8
I.B.2 Page 31
-------
Table III.B.2-12. Ethoprop: Toxicology Profile Table
MRID#
75239
40291801
138636
42530201
Guideline No.
82-1
(870.3100)
83-5
(870.4300)
83-5
(870.4300)
83-5
(870.4300)
Study Type
Subchronic Oral Toxicity-Rat
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
HED
Doc. No.
001789
001795
002775
012589
006006
005741
012589
012589
010775
Ethoprop
Dose
0, 0.015, 0.05, 5 mg/kg/day
0/0, 0.052/0.041, 0.51/0.4, 5.12/4.19 mg/kg/day (females/males)
0/0, 13.1/10.28/ 21,59.3/44.8 mg/kg/day (females/males)
0/0, 0.06/0.04, 3.27/2.62, 23.98/18.55 mg/kg/day (females/males)
Guideline/
Nonguideline
Supplementary
Supplementary
Supplementary
Guideline
Species/
Strain
Rat/Charles River
Rat/Fischer
Rat/Fischer
Rat/Crl:CD
III.B.2 Page 32
-------
, Figure III.B.2-12. Ethoprop: Dose-response Curves Using the Basic Model, Plot of the
I Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
a1 . Dose-response Curve (Basic)
1234
Dose (mg/kg/day)
MRID: 75239
a2. Dose-response Curve (Basic)
_
°
is-
111
0
D
i
10 20 30 40 50
Dose (mg/kg/day)
MRID: 138636
60
a3. Dose-response Curve (Basic)
J 1 2 3 4 5
Dose (mg/kg/day)
MRID: 40291801
b. Residuals from Basic Model
o
°0
<*> O °
-> . .9. .0. - .
00
a4. Dose-response Curve (Basic)
So
:tS
SI
LU
< EM
o -
5 10 15 20
Dose (mg/kg/day)
MRID: 42530201
c. Profile Likelihood for PB
m
?j
o
'B
<*?
o
in
S
o
•f
3
J
>'"
»
>
>
>
•v
u
O
O
O
o
o
0
0
0
0
•^-
* I/'
o
o
o
o
o
0
o
O"
0
— n: ,
* 'U'"
0
.0
•:b
o
o
o
o
o
o
_£) —
•\*
o
o
o
0
o
0
o
o
0
-A,
0
o
o
o
o
o
o
o
o
— e> —
-XJ~
O
O
0
o
o
o
o
o
o
.Cl
— \r~
O
o
o
o
o
M
o
o
o
£V_
0
0
0
0
o
o
0
0
o
,«s,,
'XJ •"
o
o
o
o
o
o
o
0
0
_jQ_
^_
c
c
c
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Fraction of I inhibition
0.295
0.305
P
0.315
BF
I.B.2 Page 33
-------
Table III.B.2-1.3. Fenamiphos: Toxicology Profile Table
Fenamiphos
MRID#
00161361
44051401
00161360
00154497
40774809
Guideline No.
83-5
(870.4300)
82-7
(870.6200)
82-1
(870.3100)
82-2
(870.3200)
82-4
(870.3465)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Neurotoxicity-Rat
90-Day Cholinesterase Study-Rat
21 -Day Dermal Toxicity-Rabbit
21-Day Inhalation Toxicity-Rat
(nose only)
HED
Doc. No.
003331
003606
005722
012019
003606
004531
005722
004531
010301
011035
Dose
0/0, 0.12/0.10, 0.60/0.46, 3.36/2.45 mg/kg/day (females/males)
0/0, 0.08/0.06, 0.80/0.61, 3.98/3.13 mg/kg/day (females/males)
0, 0.018, 0.03, or 0.05 mg/kg/day
0, 0.5, 2.5, 10 mg/kg/day
0, 0.03, 0.25, 3.5 ug/L
Guideline/
Nonguideline
Guideline
Guideline
Minimum
Guideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Wistar
. Rat/Fischer
Rabbit/
New Zealand White
Rat/ Wistar
I.B.2 Page 34
-------
. Figure III.B.2-13. Fenamiphos: Dose-response Curves Using the Basic Model, Plot of
| the Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
CM
O
T"~
v
CO
I
_j_j
id
0
E
a1 . Dose-response Curve (Basic)
E
£ -
| JD
3
E - — o
- .&>'r~
E .i
: -fj
I UJ « -
E -c
E O
1 "*•
o -
& .
9 9
1 i 1 1 1 1
0.00 0.01 0.02 0.03 0.04 O.OE
1 Dose (mg/kg/day)
0
CO
o:
0
03
0
">
0
C£
a2. Dose-response Curve (Basic)
LU
O
m
I a
i LU
E
13
O
CL
O
•Q
E
i
s
E
r
E
E
|
E
E
E
E
B
o
B o
MRID: 161 360
a3. Dose-response Curve (Basic)
0.0 0.5 1.0 1.5 2.0 2.5
Dose (mg/kg/day)
MRID: 161361
3.0
b. Residuals from Basic Model
o
< w H
—-8
in
T>
is «
iq
o
..«- -«„.
o o
1 2 3
Dose (mg/kg/day)
MRID: 44 051401
c. Profile Likelihood for PB
0.00
0.05 0.10 0.15 0.20
Fraction of Inhibition
S
o
>
>
>
i
i
0
o
0
0
jo
b
0
o
o
o
0
o
o
o
o
o
0
o
o
0
o
o
0
-------
Table III.B.2-14. Fenthion: Toxicology Profile Table
Fenthion
MRID#
41743101
44339401
40329501
Guideline No.
83-5
(870.4300)
82-7
(870.6200)
82-2
(870.3200)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Neurotoxicity-Rat
21 -Day Dermal Toxicity- Rabbit
HED
Doc. No.
011804
009870
012511
011765
Dose
0/0, 0.3/0.2, 1.3/0.8, 7.3/5.2 mg/kg/day
0, 0.17/0.13, 2.19/1.63,12.62/8.5 mg/kg/day (females/males)
0, 5, 50, 100, 200, 400 mg/kg/day
Guideline/
Nonguideline
Guideline
Guideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Wistar
Rabbit/ New
Zealand
I.B.2 Page 36
-------
-. Figure III.B.2-14. Fenthion: Dose-response Curves Using the Basic Model, Plot of the
| Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
CM
O
CD
C
0
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
O
<
o -
>;;; 0
5 0
?,:.*.
• w
O
O
O
o
o
o
o
o
u
0
o
0
o
o
o
o
o
p
™"W
o
o
o
0
o
o
o
o
0
o
0
K
" o
o
o
o
o
.0
1
\J
o
o
o
o
o
o
0
o
o
0.0 0.2 0.4 0.6
Fractbn of Inhibition
0.8
0.190
0.195
0.200 0.205
0.210
CD
o:
I.B.2 Page 37
-------
Table III.B.2-15. Fosthiazate: Toxicology Profile Table
Fosthiazate*
MRID#
44269905
41347632
43559703
Guideline
No.
82-1
(870.3100)
82-1
(870.3100)
83-5
(870.4300)
Study Type
Subchronic Oral Toxicity-Rat
Subchronic Oral Toxicity-Rat
Combined Chronic Oral Toxicity/Carcinogenicity-Rat
HED
Doc. No.
In review
008039
008039
Dose
0/0, 0.05/0.05, 0.1/0.1, 0.5/0.48, 1/0.97,
10.67/9.69, 43.52/40.87 mg/kg/day
(females/males)
0/0, 0.09/0.08, 0.89/0.77, 4.74/4.12, 41.03/36.37
mg/kg/day (females/males)
0/0, 0.055/0.042, 0.54/0.41 , 2.63/2.08,
12:53/8.94 mg/kg/day (females/males)
Guideline/
Nonguideline
In review
Guideline
Guideline
Species/
Strain
Rat/
Charles River CD
(remote SD origin)
Rat/ CD
Rat/
Charles River CD
*Not yet registered
I.B.2 Page 38
-------
Figure III.B.2-15. Fosthiazate: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S
CM
O
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
C
0
E
CO
CO
0
CO
CO
CO
0
13
13
E
^
O
CL
O
"O
0
CO
">
0
o:
i .5 to -
UJ
_c
O
<
10 20 3D
Dose (mg/kg/day)
MRID: 41347632
40
a3. Dose-response Curve (Basic)
< CM -
o -
10 20 30
Dose (mg/kg/day)
MRID: 442B9905
40
c. Profile Likelihood for PE
i a.
a
c>
O
O
O
o
0
0
o
o
o
o
••o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
0
0
o
o
o
o
o
o
o
o
M
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
0
o
o
o
o
o
o
o
o
o
o
o
0.090
0.094
0.093
0.102
0
3
CO -
LU **
I-
o -
2 4 6 6 10
Dose (mg/kg/day)
MRID: 43559703
12
b. Residuals from Basic Model
'at CM
i
13 D
O
CO
O
o
o g°
lip0
mr
o
1 1
0.0 0.2
O
0 o °
8 I
9 0
1 II
0.4 0.6 0.8
Fraction of Inhibition
d. Profile Likelihood for D and S
o
ES _
CVj
o
Q o
1 -
D
O
S
C\J
°.oo
r ~—wr~~ _—
»
:
» • •
» M *
T
0135 0.000145
S
0.0001E
I.B.2 Page 39
-------
CD
CO
CD
o:
Figure III.B.2-15. Fosthiazate con't: Dose-response Curves Using the Basic and
Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
Profile Likelihood Plots for PB, D, and S
CN
o
CD
C
CD
E
CD
(f)
CO
C/)
CD
IJZj
&
13
E
13
O
CL
O
e1. Dose-response Curve (Expanded) e2. Dose-response Curve (Expanded)
D
UJ
oo -
CD -
,J- _
O -
,00 -
LU
D -
ffi '
10 20 30
Dose (mg/kg/day)
MRID: 41347632
40
at Hneo raennnea run/A /Pvnanrtarn
63. uose-response curve (Expanded)
2^03 -
'>
10 20 30
Dose (mg/kg/day)
MRID: 44269905
i
40
ra
3
12 w -
U?
ra o - •
O
CO
o
0.0
2 4 6 8 10
Dose (mg/kg/day)
MRID: 43559703
TrOITI Model
w/Low Dose Curvature
0
8
12
8
0°
0.2 0.4 0.6 0.8
Fractb not Inhibition
I.B.2Page40
-------
Table III.B.2-16. Malathion: Toxicology Profile Table
Malathion
MRID#
43942901
41054201
43266601
Guideline No.
83-5
(870.4300)
82-2
(870.3200)
82-4
(870.3465)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
21 -Day Dermal Toxicity-Rabbit
13-Week Inhalation Toxicity-Rat
HED
Doc. No.
013822
014120
014121
008714
009385
012433
012433
011516
Dose
0/0, 5/4 , 35/29, 415/359, 868/739
mg/kg/day (females/males )
0, 50, 300, 1000 mg/kg/day
O(air), 0.1,0.45, 2.01 mg/L
Guideline/
Nonguideline
Guideline
Guideline
Nonguideline
Species/
Strain
Rat/ Fischer
Rabbit/
New Zealand Albino
Rat/
Sprague Dawley
I.B.2Page41
-------
1 Figure III.B.2-16. Malathion: Dose-Response Curves Using the Basic and Expanded
I Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plots for PB, D, and S
04
a. Dose-response Curve (Basic)
b. Residuals from Basic Model
v_x
^
^
CD
1
"c
0
E
C/)
V)
0)
C/)
C/)
tf^T
jy
MMNMMM
ff\
.*
ft.
0
» MXH^
MM*
"5
E
Z3
o
Q_
MMMMM
O
"O
0
C/)
*>
0
tr
:
9
D _
E CD
Zj^co -
j :i« -
1* -
E -C
1 <
0 -
i
^•W
lK*^^*!s=«r—
^ss&Biitlr^
^^=sllr" I^~~^~~w=t5_
A^::^^"-^I
^'^-^1
I
i i t i i
r> -
"id
3 CM -
1- -
"S
-m ° -
o
1
o
0
IxLpo&v.^
6*0 Q
o
o
( 1 1
| 0 200 400 600 800 0.0 0.1 0.2 0.3 0.4 0.5 0.6
I Dose (mg/kg/day) Fraction of Inhibition
MRID: 43942901
s
c. Profile Likelihood for PB d. Profile Likelihood for D and 5
t\l
I
00 —
D
I
1 £so "
i L | -
i
E
«
»
15
1 °
* ' . ik« '\f *j \j -\,f-- -\,f ^f • w . : ; -.-ur- : i^ „:
>oooooooooci . ~
> OOO O O 0 0 O OC
> O O O O O O O 0 O C
>oooooooooc
>OOOOOOOOO(
0
QJ? -
(oooooooooa
joooooooood
00
>oooooooood PJ.
CD
>oooooooooc W
. :;.-
a T n t *~1
V ^ 1
i • • • <
1 • * * 4
i
.
1 * • * 4
. •
ft - A ,-- A.. A *
i r i i
0.000 0.002 0.004 0.006 0.008 0.0034 0.0036 0.0038 0.0040 0.004J
PBF S
I
> Rnn raennne.a r^nrua /CvnanHaH^ ^- ReSldllBlS fTOm Model
e. Dose-response curve (bxpanoea) w/Low Dose Curvature
s
I
1 ,~° -
i 3.10 -
S j>»
E '> to —
i '-p
I m'* ~
s [—
E
-------
Table IM.B.2-17. Methamidophos: Toxicology Profile Table
Methamidophos
MRID#
41867201
43197901
00148452
44525301
00147935
41402401
Guideline No.
82-1
(870.3100)
82-7
(870-6200)
83-5
(870.4300)
82-2
(870.3200)
82-2
(870.3200)
82-3
(870.3465)
Study Type
Subchronic Oral Toxicity-Rat
(Special ChE study)
Subchronic Neurotoxicity-Rat
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
21 -Day Dermal Toxicity-Rat
21 -Day Dermal Toxicity-Rabbit
Subchronic Inhalation
Toxicity-Rat
HED Doc. No.
008846
012826
011530
012826
005313
007124
012514
13394
• 11779
011550
012826
Dose
0/0, 0.06/0.03, 0.06/0.07, 0.17/0.13, 0.28/0.24 mg/kg/day
(females/males)
0/0, 0.07/0.07, 0.90/0.79, 4.94/4.26 mg/kg/day
(females/males)
0/0, 0.116/0.095, 0.351/0.288, 1.056/0.848, 3.49/2.847
mg/kg/day (females/males)
0, 0.75, 1 1.2, 36.5 mg/kg/day
0, 0.5, 5 mg/kg/day
Air and vehicle [PEG E400:ethanol] controls,
0.0011, 0.0054, 0.0231 mg/L
Guideline/
Nonguideline
Guideline
Guideline
Guideline
Guideline
Nonguideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Fischer
Rat/ Fischer
Rat/
Sprague Dawley
Rabbit/
NZW
Rat/ Wistar
III.B.2Page43
-------
CM
O
CD
C
0
I Figure III.B.2-17A. Methamidophos: Dose-response Curves Using the Basic Model for
I the Oral Route, Plot of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plot for PB
\
a1. Dose-response Curve (Basic)
i £2
i 3i
I i-1
m m
JC
o
0 -
a2. Dose-response Curve (Basic)
I
LU IT)
-C
O
CO
CO
0
CO
CO
CO
o:
0
E
Z3
o
£L
o
0
to
0
a:
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Dose (mg/kg/day)
MRID: 148452
3.5
a3. Dose-response Curve (Basic)
> -r-
= r.
1 2 3
Dose (mg/kg/day)
MRID: 43197901
4
c. Profile Likelihood for PB
S8
CL cy
g
CJ
d
0.200
0.204
I
0.208
5
3 • O
3 0
0
• o
0
o
3 0
3 O
•j J*J
O
O
o
o
o
o
o
o
o
o
0
0
o
0
o
o
0
o
o
o
o
o
o
o
o
0
o
o
o
K
o
o
o
0
o
o
0
o
o
0
0
o
o
o
o
0
o
o
o
o
Q
0
0
o
o
0
0
o
o
o
0
o
o
o
o
o
0.:
0
o
(<
(.
<
c
••<;
c
<
<
c
0.00 0.05 0.10 0.15 0.20
Dose (mg/kg/day)
MRID:418B7201
0.25
b. Residuals from Basic Model
I
13
-------
Figure III.B.2-17B. Methamidophos: Dose-response Curves Using the Basic Model for
the Dermal and Inhalation Routes.
CM
O
CO
Dermal
c
0
E
CO
CO
0
CO
to
CO
or
CD
CD
O
CL
O
"O
0
to
">
0
15 - T
10 -
5 -
0 -
1.5 -
>- 1.0
5
p
u
I
0.5 -
0.0 -
10 20 30
Methamidophos Dose
Inhalation
MelhainicJophos Dose
.B.2 Page 45
-------
Table III.B.2-18. Methidathion: Toxicology Profile Table
MRID#
00160260
Guideline
No.
83-5
(870.4300)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
HED
Doc. No.
005743
006587
Methidathion
Dose
0/0, 0.22/0.16, 2.2/1.72, 6.93/4:91 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Species/
Strain
Rat/
Sprague Dawley
III.B.2Page46
-------
i
I Figure III.B.2-18. Methidathion: Dose-response Curve Using the Basic Model, Plot of
j the Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
CN
O
a. Dose-response Curve (Basic)
b. Residuals from Basic Model
CD
C
0
E
C/)
0
CO
CO
. *— !_*
^O
fil
di
2345
Dose (mg/kg/day)
MRID: 160260
I
7
T3
32
CO '
"If
c. Profile Likelihood for PB
Lr
0
"4^1
CO
3
E
13
O
Q_
O
"O
0
GO
0
o:
e
Z
Z
I
Z
Z
E
E
E
E
E
z
E
z
z
z
E
z
E
E
E
z
i
i
z
E
E
i
i
E
z
:
z
z
i
—p O O O O, O ,Q . O,, , C ;-Q . <
jn j£> O 0 0 0 -f- O O 00 c
n \ ~ > o O O O o o6o"O:C
_p o o o o X °^^^^^H
S i o OOOOOOOOC
^_>O O O O O.O O 0 O
cipOOOOOOOOO
-booooooooo
10 p O O O O O O O O O
CO '"TV".- f^ ... r^' . (+L ^*v A -A. A ft -f^ ' " f
a lH
0.280 0.2S5 0.290 0.295
o oo o
r~.fs, *. ». » «. _ O .
»* O
o o
o o
0.0 0.1 0.2 0.3 0.4 0.5 O.B
Fraction of Inhibition
0.7
I.B.2 Page 47
-------
Table III.B.2-19. Methyl Parathion: Toxicology Profile Table
Methyl Parathion
MRID#
00074299
41853801
Guideline No.
82-1
(870.3100)
83-1
(870.4100)
Study Type
Subchronic Oral Toxicity-Rat
Chronic Oral Toxicity with Special
Focus on Sciatic Nerve Effects
HED Doc.
No.
001882
010333
Dose
0/0, 0.20/0.16, 2.10/1.64, 6.90/5.90 mg/kg/day .
(females/males)
0, 0.03/0.02, 0.14/0.11, 0.70/0.53, 3.09/2.21 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Nonguideline
Species/
Strain
Rat/
Sprague Dawley
Rat/
Sprague Dawley
I.B.2 Page 48
-------
I Figure III.B.2-19. Methyl-parathion: Dose-response Curves Using the Basic and
| Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
I Profile Likelihood Plots for PB, D, and S
CM
O
C
0
E
0
C/)
V)
0
Z5
E
13
O
CL
O
0
c/2
">
0
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
1O _!
I t
'
LU
I
O -
CD
LU
CM -
O -J
2345
Dose (mg/kg/day)
MRID: 74299
i
7
0.0 0.5 1.0 1.5 2.0 2.5
Dose (ing/kg/day)
MRID: 41853801
3.0
b. Residuals from Basic Model
c. Profile Likelihood for PB
O
_ .o.
p
D
OD
o
o
CL
o
o
o
o
D
•VI
O
O
0
o
0
o
O
o
o
O
O
0
o
0
o
O
o
o
O
O O
00
o o
00
O
O
0
o
O
o
o
o
o
o o o
o o o
o o o
o o o
o o
o o
o o
O 0
O 0
O 0
o o
o
o
o
o
o
o
o
0.0 0.2 0.4 0.6
Fraction of Inhibition
d. Profile Likelihood tor D and 5
0.000
0.005
0.010
0.015
0.020
e1. Dose-response Curve (Expanded)
LLJ
0
0.002 0.004 0.006 0.006 0.010
S
2345
Dose (mg/kg/day)
MRID: 74299
I.B.2 Page 49
-------
CC
0
13
E
D
o
Q,
O
"O
0
CO
">
0
Figure III.B.2-19. Methyl-parathion con't: Dose-response Curves Using the Basic and
Expanded Models, Plots of the'Scaled Residuals Versus Predicted Inhibition, and the
Profile Likelihood Plots for PB, D, and S ,
CsJ
C
CD
E
(/)
CO
0
(f>
V)
e2. Dose-response Curve (Expanded)
,00 -i
(3
LU
O CM
P.O 0.5 1.0 1.5 2.0' 2.5
Dose (mg/kg/day)
MRID: 41853801
3.0
-------
Table III.B.2-20. Mevinphos: Toxicology Profile Table
Mevinphos
MRID#
42588501
Guideline No.
82-1
(870.3100)
Study Type
Subchronic Oral
Toxicity-Rat
RED
Doc. No.
015801
Dose
0/0, 0.011/0.056, 0.056/0.56, 0.56/1.12, 0.84/1.67 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Species/
Strain
Rat/
Sprague Dawley
I.B.2 Page 51
-------
C\J
CO
•
C
0
E
CO
(D
CO
CO
CO
0)
J3
Z5
E
13
O
Q,,
O
"O
CD
0)
t£
Figure III.B.2-20. Mevinphos: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S
a. Dose-response Curve (Basic)
D
D
O -i
CM
- ^T'"
1 -s
E .fro
s •; 03
[ I
o
111 B -
o -
0.0 0.2
0.4 O.B 0.8
Dose (mg/kg/day)
MRID: 42588501
1.0
c. Profile Likelihood for PB
T
0.32 0.33 0.34 0.35 0.36
0.37
e. Dose-response Curve (Expanded)
B _
CM
I 3"
| 3
| J
i i
i w
o
s <
o
D
O "
-*
O -
0.0 0.2
0.4 0.6 O.S
Dose (mg/kg/day)
MRID: 42588501
1.0
b. Residuals from Basic Model
3 q
!5 ••-
'in
C °.
TB o
I
p
T
oo
o
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fraztion of Inhibition
d. Profile Likelihood for D and 5
o
10
o
8
o
S J
0.002 0.004
O.OOB
S
0.008 0.010
f. Residuals from Model
w/Low Dose Curvature
to
ci
.y c\i
Ic
"S CXJ
"S O
% '
o
0.0 0.1
I [
0.2 0.3 0.4 0.5 0.6
Fractbnof Inhibition
I.B.2 Page 52
-------
Table III.B.2-21. Naled: Toxicology Profile Table
Naled
MRID#
00088871
00141784
45222001
00160750
00164224
40087201
Guideline
No.
82-1
(870.3100)
83-5
(870.4300)
82-2
(870.3200)
82-2
(870.3200)
82-4
(870.3465)
82-4
(870.3465)
Study Type
Four-Week Subchronic Oral (Gavage)
Toxicity-Rat
Combined Chronic Oral (Gavage)
Toxicity/Carcinogenicity-Rat
28-Day Dermal Toxicity-Rat
28-Day Dermal Toxicity-Rat
Subchronic Inhalation Toxicity-Rat
21 -Day Inhalation Toxicity-Rat
HED
Doc. No.
1460
002997
004128
004521
0144336
5774
5784
004580
006709
Dose
0, 0.25, 1 , 1 0, 1 00 mg/kg/day (gavage)
0, 0.2, 2, 10 mg/kg/day (gavage)
0, 5, 1 0, 40 mg/kg/day
0, 1 , 20, 80 mg/kg/day
0, 0.2, 1.2,or6ug/L
0 (air), 4, 8, 16 pg/L (nominal) actual chamber concentration:
0,3.4, 7.2, 12.1 ug/L
Guideline/
Nonguideline
Supplementary
Guideline
Guideline
Guideline
Guideline
Supplementary
Species/
Strain
Rat/Sprague
Dawley
Rat/Sprague
Dawley
Rat/Sprague
Dawley
Rat/Sprague
Dawley
Rat/
Fischer
Rat/
Fischer
I.B.2 Page 53
-------
I Figure III.B.2-21. Naled: Dose-response Curves Using the Basic Model, Plot of the
I Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
at. Dose-response Curve (Basic)
32. Dose-response Curve (Basic)
CM
O
CD
C
0
CO
CO
0
CO
CO
CO
be
0
D
E
D
O
CL
O
13
0
CO
0
oe
D _
w -
o -
0 CO
3
*•
LU
O
20 40 60 SO
Dose (mg/kg/dayj
MRID: 88871
100
b. Residuals from Basic Model
2468
Dose (mg/kg/day!
MRID: 141784
c. Profile Likelihood for PE
10
0
13 "~
$ o
BE
~p ° -
03 0
u
OT
q
I
O
o Q
80 *
T" "t" :**"
o
o
0 0
III!
0.0 0.2 0.4 0.6
D
CD
04
O
' in
S.
D
O
O
>
>
t;
}
>'
>
>
:
^
0
o
0
o
0
0
o
o
0
o
o
o
o
o
0
o
o
0
o
0
o
o
o
o
o
o
o
o
o
o
0
o
o
o
0
o
o
o
o
0
o
0
o
o
0
o
o
o
o
0
0
o
o
o
0
Q
K
o
o
0
o
0
o
0
o
o
o
0
0
o
o
o
o
0
o
o
o
o
o
c
<:
c
c
c
c
c
<
c
0.253
Fraction of Inhibition
0.262 0.266
PBF
0.270
I.B.2 Page 54
-------
Table III.B.2-22. Omethoate: Toxicity Profile Table
Omethoate
MRID#
ACP28Day
(MRID Not
assigned)*
Guideline
No.
NA
Study Type
28-Day Feeding Study - Rat
HED
Doc. No.
NA
Dose
0, 0.01 , 0.02, 0.04, 0.08, 0.4 mg/kg/day
Guideline/
Nonguideline
NA
Species/
Strain
Rat/Nelson
*Fax and email communications from D. Allemang, Cheminova, Inc. to A. Lowit, EPA, 3/18/02, 3/20/02, 3/27/02
NA=Not applicable
III.B.2 Page 55
-------
I Figure III.B.2-22. Omethoate: Dose:response Curve Using the Basic Model, Plot of the
1 Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
a. Dose-response Curve (Basic)
b. Residuals from Basic Model
CM
O
T~""
^~
CD
i
-4-1
C
0
E
CO
CO
0
CO
CO
10
i |6 "
2 **.
E LJJ
E x: o
E O o "
$
^nJhvffli
^""^^Sasfc- '
^!5;^=;;,;^
^:^a:=&=^^
o
~ra "*""• ~
S o
-a o
") o
Q) Q ^. , jQ
"go °
ra
W "7 " °
w „ o
E < 1 1 1 I 1 ' 1 1 1 1 1 1
1 0.0 0.2 0.4 0.6 0.8 . 0.0 0.1 0.2 0.3 0.4. 0.5
| Dose (mg/kg/day) Frastbn of Inhibition
i MRID: ACP28DAY
E
1 c. Profile Likelihood for PB
E
: to *"
- o
E I -
1 & .
| zm -
E | -
E V
I
= o
: o
E V
> o o o c
> o o o c
> o o o c
, • A rt „ v
i i i i
£ O
| 0.4125 0.4130 0.4135 0.4140
I PBP :
|
~
E
1
1
i
I.B.2 Page 56
-------
Table III.B.2-23. Oxydemeton-methyl: Toxicology Profile Table
Oxydemeton-methyl
MRID#
00151806
00143351
41834002
44141301
Guideline
No.
83-5
870.4300
82-1
870.3100
Non-
guideline
82-1
870.3100
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Oral Toxicity-Rat
Special NTP Study
Subchronic Oral Toxicity (13-week
Cholinesterase Study)-Rat
HED
Doc. No.
005174
005752
009544
005752
012221
012216
Dose
0/0, 0.06/0.05, 0.62/0.49, 6.92/5.84 mg/kg/day
(females/males)
0/0, 0.09/0.08, 0.93/0.75, 13.22/8.25 mg/kg/day
(females/males)
0, 0.15, 0.45 or 2.5 mg/kg/day
(males only)
0/0, 0.0073/0.006,0.0224/0.0184, 0.074/0.0616,
0.7475/0.6201,
6.5697/5.3925 mg/kg/day (females/males)
Guideline/
Nonguideline
Guideline
Supplementary
Nonguideline
Nonguideline
Species/
Strain
Rat/
Fischer
Rat/
SPF
Rat/
Sprague Dawley
Rat/
Sprague Dawley
I.B.2 Page 57
-------
CM
O
CD
C
0
E
CO
CO
CD
CO
CO
0
D
E
13
O
a ......
O
"O
0
0
I Figure III.B.2-23. Oxydemeton-methyl: Dose-response Curves Using the Basic Model,
I Plot of the Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot
I ForP«
CO I a
i
E
E
II 1 Illl
|
E
E
E
=
i
E
E
:
E
E
i
E
E
E
ChE Activity (U/G)
0.5 1.0 1.5 2.0
_>
>
_J
•>
->
o
o
o
X
o
•
-------
Table III.B.2-24. Phorate: Toxicology Profile Table
Phorate
MRID#
44895301
44895302
Guideline.
No.
82-1
(870.3100)
82-7
(870.6200)
Study Type
21-Day Rangefinding-Rat
Subchronic Neurotoxicity-Rat
HED
Doc. No.
137767
13767
Dose
0/0, 0.10/0.09, 0.20/0.19, 0.52/0.69 mg/kg/day (females/males)
0/0, 0.04/0.04, 0.08/0.07, 0.33/0.54 mg/kg/day (females/males)
Guideline/
Nonguideline
Supplementary
Guideline
Species/
Strain
Rat/
Sprague
Dawley
Rat/
Sprague
Dawley
I.B.2 Page 59
-------
I Figure III.B.2-24. Phorate: Dose-response Curves Using the Basic and Expanded
1 Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plots for PB, D, and S
CM
O
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
C
0
E
V)
CO
0
to
tO
CO
CD
13
E
13
O
CL
O
"O
0
*£
0
o:
E
LU
<
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Dose (mg/kgAJay)
MRID: 44895301
b. Residuals from Basic Model
0.0 0.2 0.4 0.6
Fractbn of Inhibition
d. Profile Likelihood for D and S
01
O
0.001 0.002 0.003 0.004 0.005
E £ _
is
S
LU
6
1O -
0.0 0.1 0.2 0.3 0.4
Dose (mg/kg/day)
MRID: 44895302
0.5
c. Profile Likelihood for PB
i
TJ
'u> "*"
1
"P 0
5 °
•a
"T "
o
00
o o
w °
«*' = °
o o
g
0 0
I 1 1 I
S
P
D
^d
O
0
o
o
>
>
>
3
)
>
J
>
>
f
0
O
O
O
o
0
0
o
o
^
0
o
o
o
0
o
o
o
o
/•*
t
o
o
0
o
o
o
o
0
o
rt
l
o
o
o
o
o
o
o
o
o
A
0
o
o
o
o
o
0
0
o
J^
T
O
0
o
o
o
o
o
0
o
y^
O
o
o
o
o
o
o
o
o
A
I
0
0
0
0
0
0
0
0
0
A
1
0
o
o
0
0
0
o
o
0
A
c
<
c
t
<
c
<
£
c
.,
0.000 0.002 0.004
PBC
0.006
e1. Dose-response Curve (Expanded)
o
CM
E 10 _
^
O
i
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Dose (mg/kg/day)
MRID: 44895301
I.B.2 Page 60
-------
C\!
O
CO
c
0)
E
CO
(/)
0)
CO
CO
CO
cr
CD
03
Figure III.B.2-24. Phorate con't: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S
e2. Dose-response Curve (Expanded)
f. Residuals from Model
w/Low Dose Curvature
E « -
:fs
LLJ
6
in -
o —
0.0 0.1 0.2 0.3 0.4
Dose (mg/kg/day)
MRID: 44^5302
0.5
133 -
ia
u
10
d
0
o
o
0.0 0.2 0.4 O.B
Fraction of Inhibition
0.8
E
13
O
DL
O
"O
0
CO
">
0)
I.B.2 Page 61
-------
Table III.B.2-25. Phosalone: Toxicology Profile Table
Phosalone
MRID#
44801002
45317902
Guideline No.
83-5
(870.4300)
82-7
(870.6200)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Neurotoxicity-Rat
HED
Doc. No.
13753
13753
Dose
0/0, 0.28/0.23, 2.87/2.19, 46.54/31.82 mg/kg/day
(females/males)
0/0, 5/4.6, 14.70/13.80, 61.90/55.80 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/
Sprague Dawley
Rat/ Crl:CD BR
I.B.2 Page 62
-------
I Figure III.B.2-25. Phosalone: Dose-response Curves Using the Basic and Expanded
I Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
1 Likelihood Plots for Pg, D, and S
CM
""---
v'
T—
CO
1
~£
0
E
V)
0)
V)
^c
J*
(/)
>Mfe. jd»
u:
0
>
"rirt
vU
'
E
Z3
O
n
LL.
O
•Q
0
\Jj
"~
Z3p
0
rv
:
|
r- -i
S
^T" t£) —
i 52
1 2.w -
E -^ >* —
z .5
i 5 t*5 ~
E ^
UJ -.
1 0
Z «»"-__
E
o -
E
s
z
=
i
z
=
z
=
E
[
5.
Z
z
1
1 "a"1"
E =J
s '^
W3 *-K __
i D?
j |T _
i ^
^ -
i
E
:
E
I
E
s
E
E
:
1
E
:
z
= ~"
i pi
o -
D'
E
1 Q
5 CQ
E q .
D
E
^.
1 3 -
E
I
ai. wwa< =apwii i ««uii c ^L->«-J^I<_;
Ij
1 IT^*s^
f^-i^^
^*~^--~i>->^_
^^**^**™^^>
^
i
i i i i i
0 10 20 30 40
Dose (mg/kg/day)
MRID: 44801 002
b. Residuals from Basic Model
o
o °
o ^ ° o o
© 0 rt ° ° 0
o q ° o
Q 0
O
0
0 0
1 i I 1
0.0 0.2 0.4 O.B 0.8
Fraction of Inhibition
d. Profile Likelihood for D and S
f *j, «PW*W|j|«««S» •*™B^g(™™~ «W^||
• • M • i
,
» * * * • 4
» • • * 1
n i
0.015 0.020 0.025
a2. Dose-response Curve (Basic)
00
§
•~r tO ~
*«
LU
Q N -
10 20 30 40 50 BO
Dose (mg/kg/day)
MRID: 45317902
c. Profile Likelihood for PB
5 _
ci
-
-r-_ _
O
-
CO
o
d
j-
>
>
>
>
>
K
>
3
>
"™w~
O
O
0
o
o
o
0
o
o
~"er~
O
O
O
o
0
o
o
o
o
"TST"
o
o
0
o
o
o
o
o
o
—ty
O
o
o
o
o
o
o
o
o
"\f--
o
o
o
0
o
o
0
0
o
-XT'
o
o
o
o
o
o
0
o
o
—tar-
O
o
o
o
o
o
0
o
o
,., u-
0
0
o
0
o
o
0
o
o
-o—
0
o
o
o
o
o
o
o
o
~*
c
<
<
It
c
£
c
<
<
f
0.000
0.002
0.004
PBF
O.OOB
0.008
el . Dose-response Curve (Expanded)
. m -
f'
o
<
10 20 30
Dose (mg/kg/day)
MRID: 44801 002
40
I.B.2Page63
-------
I Figure III.B.2-25. Phosalone con't: Dose-response Curves Using the Basic and
I Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
1 Profile Likelihood Plots for PB, D, and S
C\l
o
CD
c
CD
E
CO
CO
CD
CO
CO
CO
CD
>
E
13
O
CL
O
"O
0
CO
'>
CD
e2. Dose-response Curve (Expanded)
f. Residuals from Model
w/Low Dose Curvature
CO
1 I
LU
O
o -
10 20 30 40 50
Dose (mg/kg/day)
MRID: 45317902
60
ra
•
I
-o1
JU
ra
o
T- ~
O_
*T—
1
CM _
1
O
0
0
o
QD o
'O
0° 0
0 O
-o
o
o
1 1
O.Q 0.2
o
0
0
o
%
o
1 1 1
0.4 0.6 0.8
Frastbn of Inhibition
I.B.2Page64.
-------
Table III.B.2-26. Phosmet: Toxicology Profile Table
MRID#
41916401
44811801
Guideline No.
83-5
(870.4300)
82-7
(870.6200)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Neurotoxicity-Rat
HED
Doc. No.
9828
10756
13522
Phosmet
Dose
0/0, 1.1/1.1, 2.1/1.8, 10.9/9.4, 27.1/22.7 mg/kg/day
(females/males)
0/0, 1.9/1.7, 3.9/3.4, 12.1/10.4 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/
Sprague Dawley
Rat/
Sprague Dawley
I.B.2 Page 65
-------
CM
O
CD
i
C
CD
E
c/>
09
CD
GO
C/)
GO
a:
CD
I Figure III.B.2-26. Phosmet: Dose-response Curves Using the Basic and Expanded
| Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plots for PB, D, and S
a1. Dose-response Curve (Basic)
OJ
= 2,2
i i^
: £ O
i o o
i °
i o
25
5 10 15 20
Dose (mg/kg/day)
MRID:4191B401
b. Residuals from Basic Model
Residual
J 1 2 3
1 *
= "~
1 ° °
•M. %-,- T -CU. . A , „->„ ., ....
/J CO T TQf
§
i
0 0
I I
0.0 0.1 0.2 0.3 0.4 0.5 O.B
Fraction of Inhibition
a2. Dose-response Curve (Basic)
CM
UJ
.c
O 10
0
S
o
1
O.
D
O
4 B 8
Dose (mg/kgAday)
MRID: 44811801
10 12
c. Profile Likelihood for PB
r~
>
>
>
J
)
>
>
>
>
O
0
0
0
o
0
0
o
o
A
o
o
o
o
0
o
o
o
o
o
n
o
o
o
o
o
o
o
0
o
f.
I
o
o
o
o
o
o
o
o
o
n
a
o
o
o
o
o
• o
o
o
„
1
o
o
o
o
o
o
o
o
o
A
—o—
o
o
o
o
0
o
0
o
o
^
0
0
0
0
0
0
0
o
o
„
o
o
o
o
0
o
0
o
o
f±
<
••<
<
c
<
e
c
c
c
r
i
D.OOO
0.002
0.004
0.006
0.008
O
DL
O
13
CD
CO
">
0
= Q
d. Profile Likelihood for D and S
o
to
:!-::•!,-•:.
*
0.002 0.004 O.OOB 0.008 0.010
S
e1. Dose-response Curve (Expanded)
c\i
CD
O o
I.B.2 Page 66
10 15 20
Dose (mg/kg/day)
MRID:4191B401
25
-------
CO
be
0
^
1
D
O
CL
O
0
CO
0
Figure III.B.2-26. Phosmet con't: Dose-response Curves Using the Basic and
Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
Profile Likelihood Plots for PB, D, and S
CM
O
CD
i
"c
0
E
CO
CO
0
CO
CO
e2. Dose-response Curve (Expanded)
f. Residuals from Model
w/Low Dose Curvature
S -
|S -
*£ -
_>
5 °
LU
JC
O to -
o -
5 1-
w
I
~o °
IU
O ,-
W I
CM
|
o o
o
o
o
"6"
o
o
o
o
468
Dose (mg/kg/day)
MRID: 44811801
10 12
0.0 0.1 0.2 0.3 0.4 0.5
Fraction of Inhibition
0.6
I.B.2Page67
-------
Table III.B.2-27. Phostebupirim: Toxicology Profile Table
Phostebupirim
MRID#
43656302
42005451
42005447
Guideline No.
82-7
(870.6200)
83-5
(870.4300)
82-1
(870.3100)
Study Type
Subchronic Dietary Neurotoxicity - Rat
Combined Chronic Oral Toxicity/Oncogenicity - Rat
Subchronic Oral Toxicity - Rat
HED
Doc. No.
013283
009954
009954
Dose
0, 0.30/0.26, 0.96/1.2, and 3.6/4.4 mg/kg/day
(females/males)
0/0, 0.08/0.06, 0.42/0.30, 2.37/1.71 mg/kg/day
(females/males)
0/0, 0.2/0.2, 0.4/0.3, 1.2/1.0, 4.9/3.6 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Minimum
Guideline
Species/
Strain
Rat/Fischer
Rat/Wistar
Rat/Wistar
I.B.2 Page 68
-------
= Figure III.B.2-27. Phostebupirim: Dose-response Curves Using the Basic and
[ Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
| Profile Likelihood Plots for PB, D, and S
CM
O
CD
C
0
E
(ft
(ft
0
(ft
(ft
(ft
0
JS
Z5
E
13
O
CL
O
0
(ft
">
0
a1. Dose-response Curve (Basic)
in
i 5
! g
1O
o
o
0
I
234
Dose (mg/kg/day)
MRID: 42005447
i
5
a3. Dose-response Curve (Basic)
II
,o -
= LU
| £
= 4!
: ^
|
=
!
I
|
i
I
:
1234
Dose (mg/kg/day)
MRID: 43B5 6302
c. Prof ile Likelihood for PE
o
OJ
p
o
o
o
1
>
>
>
>
>
>
)
>
u
0
0
o
0
0
o
0
0
o
. 1^
"
0
o
o
o
o
o
o
o
o
0
o
o
o
o
o
0
o
0
f\
o
0
o
o
o
o
o
o
0
J*fc
n
0
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
t*}
o
o
o
o
o
o
o
o
o
^
T
0
0
0
o
o
0
0
o
o
r*i
o
o
0
o
o
0
o
0
o
c
c
c
c
(
c
c
c
c
/•
0.03
0.04
0.05
0.06
BF
a2. Dose-response Curve (Basic)
n
^ O
LLI
JZ
O
o
O
1
0.0 0.5 1.0 1.5 2.0
Dose (mg/kg/day)
MRID: 42005451
b. Residuals from Basic Model
CM -
IS
O
03
-------
CM
O
= Figure III.B.2-27. Phostebupirim con't: Dose-response Curves Using the Basic and
I Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
1 Profile Likelihood Plots for PB, D, and S
el. Dose-response Curve (Expanded) e2. Dose-response Curve (Expanded)
in
CD
i
C
0)
CO
0
to
CO
= o
R
O
p
CO
S
LU '
JC
O
01 2345
Dose (mg/kg/day)
MRID: 42005447
e3. Dose-response Curve (Expanded)
0.0
0.5 1.0 1.5
Dose (mg/kgfcfsy)
MRID: 42 005451
2.0
f. Residuals from Model
w/Low Dose Curvature
CO
CD
>
v° _
10 -
o
CO
CO -
(J °
TS
O
*f 4
O
o o
123
Dose (mg/kg/day)
MRID: 43E56302
0:0 0.2 0.4 0.6
Fractbn of Inhibition
O.S
O
a.
O
"O
CD
00
">
0
cr
I.B.2 Page 70
-------
Table III.B.2-28. Pirimiphos-methyl: Toxicology Profile Table
Pirimiphos-methyl
MRID#
00129343
92147035.
Guideline No.
82-1
(870.3100)
83-5
(870.4300)
Study Type
Subchronic Oral Toxicity-Rat
Combined Chronic Oral
Toxitity/Carcinogenicity-Rat
HED
Doc. No.
014067
3582
14067
3582
5105
8819
Dose
0, 0.25, 0.40, 0.50, 2.50 mg/kg/day
0/0,0.4/0.4, 2.1/2.1, 12.6/12.6 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Guideline
Species/
Strain
Rat/
Wistar
Rat/
Wistar
I.B.2 Page 71
-------
I Figure III.B.2-28. Pirimiphos-methyl: Dose-response Curves Using the Basic Model,
I Plot of the Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot
I forPn
CM
O
^_
CD
i
<4— >
C
0
E
CO
CO
0
CO
CO
y
CO
f^jf
[JV
0
>
'j~i
CO
MMMMMM
^5
E
13
o
CL
o
"O
0
CO
m mmmm
^s#
0
a:
i
•
z
j
5
i
z
z
z
z
z
E
I
I
a1 . Dose-response Curve (Basic) a2. Dose-response
1
i "
Q.
I'M "
i
LU ° -
-C
O
o -
fcifcr- — - —
"^"p- — — - —TIJ
!^ O
I
Q_
*^ O
LU ,- -
r/~
0
o -
Curve (Basic)
i^S&_J
T8-— j
1 ! 1 1 1 1 1 1 1 1
0.0 0.5 1.0 1.5 2.0 2.5 0246
a 10 12
Dose (mg/kg/day) Dose (mg/kg/day)
MRID: 129343 MRID: 92147035
E
E
E
=
|
5
i
—
i
z
z
z
1
z
z
E
z
i
z
:
z
z
z
z
z
z
z
z
z
1
z
E
i
i
E
z
z
z
z
z
b. Residuals from Basic Model c. Profile Likelihood for PB
iq
"Hi
^ ,,
« *°. .
,¥ °
£
"8,0 "
C/3 '
IO
t
o
0
0|
°° o o °°
/ny rt
8*-*
L-
^3 O j»
off g o
-. Q O
8
O O 0
0 0
II 1 1 1 1
fS .
0
f;
•
~
f^ tO
h- -
D
in
> o o o o o
> o o o o o
> 0 O 0 O O
> o o o o o
J O O O O O
J 0 O O O O'
5 O O 0 O 0
> 0 O 0 O O
D a o o o o o
- .
T I
o o o o c
O 0 0 0 C
0000 <
s: o o o c
H '::
0 O O O C
o •;.] o :: i
O O 0 0 ',
o o o o c
O 0 O O C
. .- j
1"'
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.55 0.60 0.65
F rait bn of Inhibition PBF
I.B.2 Page 72
-------
Table III.B.2-29. Profenofos: Toxicology Profile Table
Profenofos
MRID#
92148022
43213303
92148031
Guideline No.
82-1
(870.3100)
82-7
(870.6200)
83-5
(870.4300)
Study Type
Subchronic Oral Toxicity - Rat
Subchronic Dietary Neurotoxicity - Rat
Combined Chronic Oral Toxicity/Oncogenicity - Rat
HED
Doc. No.
NA
011795
011916
Dose
0/0, 0.001/0.001,0.003/0.003,
0.01/0.009,0.03/0.02, 0.09/0.09,0.25/0.21 ,
0.96/0.87, 2.6/2.1, 9.2/8.4, 24.8/21.1, 96.8/85.9
(females/males)
0/0, 1.84/1.7, 8.4/7.7, 37.9/36 mg/kg/day
(females/males)
0/0, 0.02/0.017, 0.694/0.559, 6.951/5.685
mg/kg/day (females/males)
Guideline/
Nonguideline
NA
Acceptable
Acceptable
Species/
Strain
Rat/Fischer
Rat/Sprague Dawley
Rat/Fischer
NA=Not available
III.B.2 Page 73
-------
I Figure III.B.2-29. Profenofos: Dose-response Curves Using the Basic Model, Plot of the
| Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot for PB
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
Csj
O
X —
S
-1— '
c
0
.ssessm
<
*
r?
LL.
0
IS
i
E
r
E
E
E
-
\
E
=
E
=
E
=
E
E
E
E
E
E
-
=
i
E
I
a
E
E
=
E
E
E
E
E
E
ACHE Activity (U/G)
10 0.5 1.0 i.5 2.0
,*-*.
.— ' . M*
EL
•— ' D
1,
UJ D
O
o
<*=>==*,
1 1 1 1
0 10 20 30
Dose (mg/kg/day)
MRID: 4321 3303
a3. Dose-response Curve (Basic)
i I
JL,., »r J™ J
I i I t 1 1 i I
01234567
*ChE Activity (PHCHANGEj
3.0 0.5 1.0 1.5 2.0
CM -
10 T- _
1
(
N^
i i t i i i
0 20 40 60 80 10
Dose (mg/kg/day)
MRID: 92148 022
b. Residuals from Basic Model
o
o
Oo o
l°o° 8 ° o °
o
0
II III)
J.OO 0.10 0.20 0.30
13
O
CL
O
T3
0
C/)
">
0
XtX
E
i
E
E
|
E
E
=
E
E
E
E
E
E
E
E
Dose (mg/kg/day)
MRID: 92148 031
c. Profile Likelihood for PB
Frast ion of Inhibition
S
SCM
O
= 0-
O
O
>
0
O
0
o
0
0
o
0
o
o
o
o
0
o
o
o
o
o
0
o
o
o
0
0
0
o
-^
o
0
o
o.
o
o
o
o
o
o
o
o
o
0
o
o
o
0
0
o
o
o
o
o
o
0
0
o
o
o
o
o
o
o
0
o
o
0
0
o
0
o
o
o
o
o
0
0
o
o
o
o
0
1
c
t
e,
c
c
c
c
c
c
0.40
0.45
0.50
0.55
I.B.2 Page 74
-------
Table III.B.2-30. Terbufos: Toxicology Profile Table
Terbufos
MRID#
00109446
40089602
00049236
44842302
Guideline No.
82-1
(870.3100)
83-1
(870.4100)
83-5
(870.4300)
82-7
(870.6200)
Study Type
Subchronic Oral Toxicity-Rat
Chronic Oral Toxicity-Rat
Combined Chronic Oral
Toxicity/Carcinogenicity-Rat
Subchronic Neurotoxicity-Rat
HED
Doc. No.
002377
005612
006352
004898
003847
001514
005612
006352
013572
Dose
0/0, 0.01/0.01, 0.02/0.02, 0.05/0.04, 0.095/0.08 mg/kg/day
(females/males)
0/0, 0.009/0.007, 0.04/0.03, 0.07/0.06 mg/kg/day
(females/males)
0/0, 0.01/0.01, 0.05/0.04, 0.22/0.33 mg/kg/day
(females/males)
0/0, 0.04/0.04, 0.06/0,06, 0.25/0.37 mg/kg/day
(females/males)
Guideline/
Nonguideline
Guideline
Guideline
Guideline
Guideline
Species/
Strain
Rat/
Sprague Dawley
Rat/
Sprague Dawtey
Rat/
Long Evans
Rat/
Sprague Dawley
I.B.2 Page 75
-------
I Figure III.B.2-30. Terbufos: Dose-response Curves Using the Basic and Expanded
| Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plots for PB, D, and S
CM
O
CO
C
0
E
CO
CO
0
CO
CO
CO
0
JD
15
E
13
O
CL
O
"O
0
CO
0
a:
a1. Dose-response Curve (Basic)
a2. Dose-response Curve
"P
s >-
E =E
i i
I u
I o
IT)
b
p
c\i
o
o
O
0.00 0.10 0.20
Dose (mg/kg/day)
MRID: 49236
0.30
a3. Dose-response Curve (Basic)
0.00 0.02 0.04 O.OB 0.08
Dose (mg/kg/day)
MRID: 109446
a4. Dose-response Curve (Basic)
to
E LU
E _c
i y
! T
0.00 0.02 0.04 O.OB
Dose (mg/kg/day)
MRID: 40089 602
b. Residuals from Basic Model
E in
T-
•&D
.a -- "
^ 10 H
O
0.0 0.1 0.2 0.3
Dose (mg/kg/day)
MRID: 44842302
c. Profile Likelihood for PB
I "B
5 II
I £
D
C)
as
O- d
p
b
D
O
1 tJ""'""W"'""W"
>
>
)
)
>
}
J
>
y
O
o
o
o
o
o
0
o
o
o
o
©
o
o
0
o
o
0
A
o
o
O'
o
o
o
0
0
o
n
' \J'"
o
o
o
o
o
0
0
o
o
••••LT-
o
o
o
o
o
0
o
o
o
f\
T
-~u~*
o
o
o
o
o
o
o
0
o
iy
•»«/••••
o
o
'. O...
o
o
o
o
o
o
rt 1
•tf~
0
0
o
o
o
o
O;
o
o
f"l
— cr—
o
o
.'O
o
o
o
o
o
o
-"'.
(
<
c
c
(
c
<
<
c
1*
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Fraction of I nhibition
0.000
0.001
0.002
0.003
0.004
III.B.2 Page 76
-------
I Figure III.B.2-30. Terbufos con't: Dose-response Curves Using the Basic and
1 Expanded Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the
[ Profile Likelihood Plots for PB, D, and S
CsJ
o
^1
CD
i
^^™*
C
0
E
AMM
t f\
CO
CO
0
CO
CO
^
**wmm
CO
Lr
0
•^^
„;£_
-i— •
CO
~zz
Z3
E
d
CL
0
"O
0
CO
" xT"
0)
X.
\ IS
= m ~
E °
E
E O ID ~
E b
E
z _
E
= IO
E 10
i M •
i • •
i • •
: b
i 0.0120 0.0124
1 S
E
i e2. Dose-response
~ /n
E Q o
1 2.™
I t "
01 *~
x:
O
o
5 $ lj I
0 1 1 1
0.00 0.02 0.04
h*WWI 1^1 ^ tlllVd ^ ^ 1 • ^^^^^1 «r)«|^««l I^W ^^««l w %^ ^^B«%|rfM*l l^mw*^
- ..... ^. ... - ...^
• <
• t
• 4
tp
to
[>
•§ 10
LU b
-C
O
o
r" i
fe=fc—
r^^^^
\\ N^v
\$ ^\|
\_ •*
%
1 1 1 1 1 1 1 1 1 1
0.0128 0.00 0.10 0.20 0.30
Dose (mg/kg/day)
MRID: 49236
Curve (Expanded) e3. Dose-response Curve (Expanded)
•*• •^~
T—
CJ 00 *"
D
J5-CD -
LU
JC
0 CM -
O •«
-„ ,
i
i i i i i i i i i i
0.06 0.08 0.00 0.02 0.04 O.OB
Dose (mg/kg/day) Dose (mg/kg/day)
MRID: 1
e4. Dose-response
?5 -
3,
•^° -
tj
^
&"-
0 -
09446 MRID: 40089602
Curve f FvnandedA f- Residuals from Model
curve (Expanded) w/Low Dose Curvature
""^N
1 1
p
^ b ~
ti °
TJ ^
-=
cH "
0
TT ~
0
o
0
oe o
& o
^o o
Co °
0
^OQ
•^ o °
i i i i i i i i i i
0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Dose (mg/kg/day) Fraction of I nhibition
I.B.2 Page 77
-------
Table III.B.2-31. Tetrachlorvinphos: Toxicology Profile Table
Tetrachlorvinphos
MRID#
43371201
00112525
42980901
45570601
41342001
Guideline No.
82-1
(870.3100)
83-2
(870.4200)
83-2
(870.4200)
Nonguideline
82-2
(870.3200)
Study Type
Subchronic Oral Toxicity-Rat
Chronic Oral Toxicity-Rat
Chronic Oral Toxicity-Rat
21 -Day Cholinesterase
Study-Rat
21 -Day Dermal Toxicity-Rat
HED
Doc. No.
11295
002607
,007181
010884
010884
011295
TXR No.
0050614
7844
Dose
0, 5, 100, 250 mg/kg/day
0, 0.25, 1 .25, 6.25, 1 00 mg/kg/day
0/0, 5.93/4.23, 62.7/43.2, 125.3/88.5 mg/kg/day (females/males)
0, 8, 1 2, 50 mg/kg/day
0,10, 100, 1000 mg/kg/day
Guideline/
Nonguideline
Guideline
Guideline
Guideline
Acceptable
Guideline
Species/
Strain
Rat/Sprague Dawley
Rat/
Porton strain derived
from Tumstall Lab
Rat/Sprague Dawley
Rat/Crl:CD®(SD)IGS BR
Rat/Sprague Dawley
I.B.2 Page 78
-------
Figure III.B.2-31. Tetrachlorvinphos: Dose-response Curves Using the Basic Model, .
Plot of the Scaled Residuals Versus Predicted Inhibition, and the Profile Likelihood Plot
for PB
ai. uose- response wurve uaasicj a^. uwse-ie
CNJ
o ^ -
T— i «> _
. • .S-'-'
**««»-_ -&»
CD |2 -
1 uw
+-j go
c. <
0 S -
9k i
*"* — " — — — — - - — .. I
In^ -
» ° .
~» _,
2. D _
.-^b
< D _
LU d
o
"^ 0 _
af- — -\
i •
$• >— - j
T<
A
spun
i — i
>
jjgjjjj— -.™
£ffi UUI
.
L_j
^" " ii _
-(
P- —
fZ. 1 t i i i i <=> i i i i i
Li 0 20 40 60 80 100 0 20 40 60 80
ve ^Dasiiu
i ^
~ — 1
, t
ii ,
/
i
i 1
100 120
CO Dose (mg/kg/day) Dose (mg/kg/day)
CO MRID: 112525 MRID: 42980901
0
CO a3. Dose-response Curve (Basic) a4. Dose-response Curve (Basic)
co
^ IT 10
,00 JL
0 1° -
>s» ' LU o
.:_. o
-4— • < o
t _! j
f T- — — |
— '
1 ° "
"6 w _
< <=>
LU
o
< D
fi
^-_,
vU o 1 i 1 1 t II
13 0 50 100 150 200 250 0 10
20
30
I i
40 50
EDose (mg/kg/day) Dose (mg/kg/day)
MRID: 43371 201 MRID: 45570601
D
(J b. Residuals from Basic Model c. Profile Likelihood for PB
CL
o g°:
•O lev,
0 (8°
CO "S^ "
13 W
^> $ °
0
rV "*•
O
o o
0%
<& o 0
"O*
o
'Jo Q ^
r
<6 °
o
§
°- s
d
o
^^ I I I 1 I o
/' \J U-V? \y
> o
> o
> 0
f
\ 1
0.0 0.1 0.2 0.3 0.00 0.01
Frarf bn of Inhibition
III. B.2 Page 79
o
o
o
1
0.02
P
•n
I
0.03
BF
o
o
o
1
c
c
c
0.04 0.05
-------
Table III.B.2-32. Tribufos: Toxicology Profile Table
Tribufos
MRID#
42335101
45369101
Guideline No.
83-5
(870.4300)
82-7
(870.6200)
Study Type
Combined Chronic Oral
Toxicity/Carcinogenicity/Neurotoxicity-Rat
Subchronic Neurotoxicity-Rat
HED
Doc. No.
010119
NA
Dose
0/0, 0.2/0.2, 2.3/1.8, 21.1/16.8 mg/kg/day
. (females/males)
0/0, 0.17/0.14, 3.54/2.89, 46.2/36.8
mg/kg/day (females/males)
Guideline/
Nonguideline
Guideline
NA
Species/
Strain
Rat/Fischer
Rat/ Wistar
NA=Not available
I.B.2 Page 80
-------
I Figure III.B.2-32. Tribufos: Dose-response Curves Using the Basic and Expanded
| Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
I Likelihood Plots for PB, D, and S
Csl
O
CD
a1. Dose-response Curve (Basic)
a2. Dose-response Curve (Basic)
C
0
E
CO
co
0
CO
CO
CO
o:
0
....
J25
13
E
Z5
o
CL
o
"a
0
CO
">
0
rr
i q
1 2-
S **•
~ XI
I o
J
LU rf
JZ
< CVJ -
o ~J
0 5 10 15 20
Dose (mg/kg/day)
MRID: 42335101
b. Residuals from Basic Model
10 20 30
Dose (mg/kg/day)
MRID: 45369101
40
c. Profile Likelihood for PB
js:
!«,
(J O ""
in
1
'0
a
0
° °o o
.0 _rj.
0 0
0
O
1 1 1 I
0.0 0.2 0.4 0.6
00
o
o
q
D
o
o
o
/ • "VJ •
0
O
0
O
O
) O
> 0
) O
o
o
o
o
o
0
o
o
o
o
o
0
o
o
o
o
0
o
0
o
•'o
o
o
0
o
o
o
0
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
0
o
o
o
o
o
o
o
o
o
" 'U'"
o
o
o
o
0
0
o
o
o
71
o
.0 <
o <
o
o
o c
o c
0 <
Fraction of Inhibition
d. Profile Likelihood tor D and S
000 0.002 0.004 0.006 0.008 0.010
el. Dose-response Curve (Expanded)
o
E
o
o
•*
o
n
o
0.002 0.004
O.OOB
S
0.008 0.010
5 10 15
Dose (mg/kg/day)
MRID: 42335101
20
I.B.2 Page 81
-------
CD
s
C
CO
Figure III.B.2-32. Tribufos con't: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S >
e2. Dose-response Curve (Expanded)
f. Residuals from Model
w/Low Dose Curvature
-Sr-co -
o
< CM
O -I
10 20 30
Dose (mg/kg/day)
MRID: 45369101
i
40
ro
13
o
0.0 0.2 0.4 O.B
Fraction erf I nhibition
I.B.2 Page 82
-------
Table III. B.2-33. Trichlorfon: Toxicology Profile Table
Trichlorfon
MRID#
43871701
41056201
41973001
40306901
00152137
Guideline No.
82-7
(870.6200)
83-5
(870.4300)
83-5
(870.4300)
82-2
(870.3200)
82-4
(870.3465)
Study Type
Subchronic Neurotoxicity-Rat
Combined Chronic Oral
Toxicity/ Carcinogenicity-Rat
Combined Chronic Oral
Toxicity/ Carcinogenicity-Rat
21-Day Dermal Toxicity-Rabbit
21 -Day Inhalation Toxicity-Rat
HED
Doc. No.
13967
9626
013703
6476
004509
004915
Dose
0/0, 6.9/6.1, 35.4/31.2, 188.7/164.7 mg/kg/day (females/males)
0/0, 5.8/4.5, 17.4/13.3, 109.2/85.7 mg/kg/day (females/males)
0/0, 159/129 mg/kg/day (females/males)
0, 100, 300, 1000 mg/kg/day
0 (EtON/PEG), 12.7, 35.4, 103.5 mg/m3
Guideline/
Nonguideline
Guideline
Guideline
NA
Guideline
Guideline
Species/
Strain
Rat/ Fischer
Rat/ Fischer
Rat/ Fischer
Rabbit/New
Zealand
Rat/ Wistar
NA=Not available
I.B.2 Page 83
-------
Figure III.B.2-33. Trichlorfon: Dose-response Curves Using the Basic and Expanded
Models, Plots of the Scaled Residuals Versus Predicted Inhibition, and the Profile
Likelihood Plots for PB, D, and S
CM
O
CO
c
CD
E
t/)
C/5
CD
C/>
CD
.—.
JS
1
D
O
a.
O
CD
c/)
CD
a1. Dose-response Curve (Basic)
"O _
I
I I
UJ 10
-C
O
20 40 60 SO
Dose (mg/kg/day)
MRID: 41 056201
100
a3. Dose-response Curve (Basic)
10 -
I 5
I
E
E
:
=
E
E
E
:
E
i
0
S
P
d
50 100 150
Dose (mg/kg/day)
MRID: 43S71701
c. Profile Likelihood for PB
0
o
>
>
>
5
>
)
J
0
0
o
0
o
0
0
0
0
o
o
0
o
0
o
o
o
o
o
o
o
o
o
o
o
o
o
f
o
o
o
o
o
o
o
o
0
o
o
o
Q
o
o
o
o
0
I
\J
o
o
o
0
o
0
o
o
0
o.
o
o
0
o
0
o
o
o
r
0
0
o
o
o
o
0
o
o
o
0
o
o
o
o
o
o
o
d
4
H
d
c
ej
d
d
t
1
32. Dose-response Curve (Basic)
ir>
CD
So
:>
UJ
-------
CM
O
CD
i
C
-------
I III. Appendices
I B. Hazard/RPF
I 3. Response to SAP Comments from September 2001 and March 2002
(N I Reports
O f
x— =
T— i a. Response to SAP Comments from September 2001
I OPP in collaboration with ORD presented its July 31st, 2001 document
i 1 entitled, "Determination of Relative Potency and Points of Departure for
+-• j Cholinesterase Inhibition" to the FIFRA SAP on September 5-6, 2001. The
*••• [ key recommendations from the September 2001 report
I (http://www.epa.gov/scipoly/sap/index.htm) and OPP's responses are given
I below:
C/) f
GO i i. Derivation of the Adjustment Factor "B" and Modification of
Q) I Decision Tree for use of "B"
CO i
< f The SAP Report noted that a plot of the "scaled residuals" against
! "predicted % inhibition" indicates that the weighting strategy used for
Jsd | calculating the adjustment factor "B" does not adequately reflect how the
C/) | variance changes with response. The SAP was specifically concerned
| EPA "focused the modeling effort on achieving fidelity with observations at
I the high end of the range of doses tested, to the likely detriment of fitting
0 I points at the low end of the dose response relationship."
I In the current analysis, all available Cholinesterase datasets for the
[ brain compartment were analyzed using a fixed horizontal y-asymptote for
13 i each chemical. The weight function was changed from one in which the
C | variance was presumed proportional to the square of the mean to one in
-~j 1 which the variance is proportional to the mean. The revised methodology
Ofor the determination of the horizontal y- asymptote is described in I.B and
III.B.1.
CL
O
13
CD
00
">
0
ii. Conduct a Formal Analysis of Residuals as a Function of Dose
Residual plots for the basic and expanded models for each chemical
for the brain compartment are given in III.B.2.
I.B.3 Page 1
-------
I iii. Accuracy of the "Chi Square Approximation" for the "Goodness
[ of Fit" Statistic
I In the July 31st document, a Chi-Square Approximation was calculated
I for each cholinesterase dataset. This statistic was used as a measure of
CM | the goodness-of-fit for the exponential function. The concern expressed
CD 1 by the SAP does not apply to the current methodology. Although the
^ | OPCumRisk program was not used to determine potency of OPs in the
current analysis, the program was revised to deliver a warning message to
the program user indicating possible calculation inaccuracy for this
statistic. The revised version of the OPCumRisk is available for download
i I at http://www.epa.gov/scipolv/sap/index.htm and
•+- ' [ http://www.epa.gov/pesticides/cumulative/
\*mm E
P~ | iv. Confidence Interval Calculations
C/) I The SAP report suggested that HED "reconsider the confidence
00 1 interval calculations"and "perhaps try bootstrapping or some other more
CD | robust method . . . ." In the current analysis, HED has revised the
^ | calculation of the confidence intervals (See III. B.1). Bootstrapping is a
* I very time and resource-intensive procedure. Although bootstrapping may
I be the preferred approach for calculating confidence intervals, due to
Jad | limited availability of resources, the Agency has not conducted any
CO | bootstrapping procedures. At this time, the current method for calculating
confidence intervals is adequate and satisfactory. Because it is important
to evaluate the range of uncertainty around any potency or benchmark
0 | dose values used to extrapolate to human risk, the Agency will consider
!> I bootstrapping procedures in future assessments;
v. Deleting p- and t- values
'
-0
.
CZ I The SAP Report recommended deleting the p- and t- values that are
=Z [ produced by the Agency's OPCumRisk program. As stated previously,
Of the OPCumRisk program was not used in the current analysis to calculate
| potency or benchmark dose estimates. The requested deletions have
n I been incorporated; the revised version of the OPCumRisk is available for
Oi download at http://www.epa.gov/scipoly/sap/index.htm and
i http://www.epa.gov/pesticides/cumulative/
•o i
(D i vi. Estimates of Relative Potency
CO f
">> I The SAP Report included considerable discussion regarding whether
0 | relative potency factors should be based on ratios of the "Benchmark
I Dose 10's" (BMD10) or on ratios of the dose-scaling factors. OPP has
= derived potency in the present analysis on BMD10 (See I.B).
I.B.3 Page 2
-------
I vii. Inhalation Dose
| The SAP Report recommended that inhalation exposure be expressed
| in the same units as the oral doses and that the doses be adjusted for
| actual treatment durations. HED has calculated the inhalation doses as
CM | mg/kg/day using conversion factors that account for respiratory volume
O I and body weight for the strain of rat used, as well as the duration of
I exposure in terms of hours exposed per day.
i viii. Use of Individual Animal Data
CD [
i | The SAP Report from the September 2000 SAP meeting
| recommended that study data on individual animals be used in calculating
| relative potencies. Due to the fact that all the data on organophosphates
f are not in an electronic format, HED has not taken this step. However, the
I September, 2001 Report recognizes that "individual data would not be
(/) | likely to change the results using current methods." In addition, by
C/) I switching from RBC to the brain compartment, some of the concern about
CD | not using individual animal data should be reduced, since the
W I experimental designs for the brain measurements do not include a
[ repeated measures component, unlike the RBC data.
= iv. Use of NOAEL's and LOAEL's for Inhalation and Dermal Routes
c/> f
/y I Several Panel members objected to EPA's use of No Observed
"•• ! Adverse Effect Levels ("NOAEL's") and Lowest Observed Adverse Effect
0 1 Levels ("LOAEL's") for cholinesterase inhibition data by the dermal and
> i inhalation routes of exposure instead of actual dose-response models as
are used for the oral data set. HED does not intend to use dose-response
modeling to determine relative potency estimates for dermal and
inhalation exposure because the data are not sufficiently robust to justify
the resources required.
However, it is to be noted that the current analysis uses Comparative
Effect Levels (CEL's) for cholinesterase inhibition data for these two
routes of exposure. The dermal and inhalation database was not suitable
for dose-response analysis. Cholinesterase determinations in these
studies were typically made at only one time point and several of the
studies had no cholinesterase inhibition at the highest dose. For the
current assessment, potencies by the dermal and inhalation routes were
compared using brain cholinesterase inhibition at a dose causing a
O
Q_
O
CD
en
maximum of 15% brain cholinesterase inhibition.
ff~-
0
o:
I.B.3 Page 3
-------
I v. Derivation of Doses from the Actual Dietary Intake Rates
I The SAP Report recommends that "the doses used for evaluation of
[ potencies at various ages within specific data sets should be derived from
[ the actual dietary intake rates observed in the study for those ages where
CM 1 the consumption data are available."
O I
T_ i In feeding toxicity studies, laboratory rats are exposed to the test
T— I compound via the diet. Generally, the test compound is mixed in the
I animal feed which the laboratory animals eat. Over the course of a
I toxicity study, as the animals age, they will not only gain weight and but
I they will naturally change their rate of food consumption. The data
! collected for the oral route and used in both the July and December 2001
I preliminary cumulative risk assessments include average compound
| intake (mg of active ingredient per kg per day). HED has conducted a
I pilot analysis in response to this recommendation to evaluate the effect of
(/) I age and food consumption rate on the potency estimates. In this pilot
(/) | compound intake analysis, OP potency was determined for a subset of
CD I studies [=10% of total studies in the dose-response assessment] using
I compound intake measured at or around the time of cholinesterase
I measurements [duration-specific compound intake].
st
I Seventy-nine oral toxicity studies were included in the dose-response
£/) I assessment for the December, 2001 Cumulative Risk Assessment for
1 OP's. Of these 79 studies, the test article was administered via the diet
1 for 73. For each of the seven OPs selected for this analysis, the
0 I calculated compound intake (mg/kg/day) given in the study report for a
> I weekly, biweekly, or monthly time interval closest to the time of
*-* I cholinesterase measurement was extracted from the feeding toxicity
CO | studies [duration-specific compound intakes]. For example, if brain
35 | cholinesterase was measured at a one-year interim sacrifice, the
| ' compound intake for the 50-52 week reported interval was collected. The
| potency values obtained were compared to those in the July, 2001
I analysis, which utilized average compound intake values. Potency
| estimates given below (Table III. B. 3-4) were calculated using the
pi [ OPCumRisk program with the methodology described in the July 31
O[ document prior to the completion of the current methodology for the joint
I analysis. The pilot analysis was performed in three stages : 1) impact of
I age on relative potency for chronic studies only; 2) impact of age on
| relative potency for complete database of subchronic and chronic studies;
(/) I and 3) impact of age on the points of departure on the index chemical.
HiiimiiL —
> I
0 1 Stage 1 : The purpose of this pilot analysis was to investigate the impact
1 of age on food consumption and body weight, and ultimately OP
| potency. In order to maximize the age-related differences in
I body weight and food consumption, chronic studies were
| analyzed first. Seven chronic feeding studies were selected
I III.B.3Page4
-------
I randomly and analyzed as described above. Relative potency
I of each was calculated using the methamidophos chronic study.
[ Results given in Table III. B. 3-1.
| In the chronic study analysis (Table III. B. 3-1) comparing the
CNI | RPFs calculated using the slope scale factor (m) and also the
O 1 BMD10sfor ChE data using the average and duration-specific
^H: I compound intakes, the RBC and brain data for both sexes
^~ 1 display comparable potency values. For tribufos a 5-fold
---» i difference between the average and duration-specific intake
CD | assessments for male brain CHel was observed. This
i 1 difference is an artifact of the decision tree for the determination
<•+-' 1 B (horizontal asymptote) and not from differences in potency
£^ | between the average and duration specific intakes. Two
| timepoints (364 and 721 days) are available for the male brain
I ChE data in MRID 42335101 . In the duration specific analysis,
(/) | the 364 day time point did not converge and was therefore not
CO I included in the potency estimates.
0 I
CO I
CO I
< I
o:
CD
Z5
E
13
O
CL
"O
0
CO
">
0)
I.B.3 Page 5
-------
Table III.B.3-1 a. Results of Dietary Intake Comparison [actual vs average] Using Chronic Studies
CHEMICAL
BENSULIDE
DIAZINON
DICROTOPHOS
METHAMIDOPHOS
PHOSALONE
PHOSMET
TRIBUFOS
BENSULIDE
DIAZINON
DICROTOPHOS
METHAMIDOPHOS
PHOSALONE
PHOSMET
TRIBUFOS
MR1D
44161101
41942002
44527802
00148452
44801002
41916401
42335101
44161101
41942002
44527802
00148452
44801002
41916401
42335101
COMPARTMENT
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
SEX
F
F
F
F'
F
F
F
M
M
M
M
M
M
M
Dietary Intake
Calculation
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
•average
biweekly
average
biweekly
Relative
Potency using
'm'
0.005
0.004
0.034
0.031
1.77
1.89
1.00
1.00
0.015
0.024
0.023
0.021
0.018
0.017
0.002
0.002
0.011
0.011
2.06
2.32
1.00
1.00
0.021
' 0.038
0.011
0.013
0.020
0.004
Lower 95%
CL
0.004
0.004
0.031
0.028
1.41
1.51
1.00
1.00
0.013
-OT020- ~
0.010
0.016
0.007
0.007
0.002
0.001
0.003
0.003
1.70
2.03
1.00
1.00
0.018
0.033
0.008
0.009
0.017
0.001
Upper 95%
CL
0.006
0.005
0.038
0.035
2.22
2.38
1.00
1.00
0.018
"OT029~
0.053
0.027
0.048
0.045
0.003
0.003
0.041
0.035
2.38
2.67
1.00
1.00
0.025
0.044
0.015
0.018
0.022
0.020
BMD10
14.11
14.04
1.85
1.85
0.041
0.035 .
0.071
0.063
4.13
" 2.40
4.41
2.76
3.26
3.14
24.69
24.93
3.38
3.31
0.028
0.022
0.062
0.055
2.58
1.29
5.35
3.71
4.22
15.64
BMDL
12.40
12.17
1.78
1.80
0.035
0.030
0.063
0.058
3.70
~ 2.T4 "
3.74
2.33
1.88 ,
1.83
19.37
19.54
1.83
1.83
0.026
0.020
0.057
0.049
2.37
1.18
4.33
2.98
2.51
6.19
Relative Potency
using BMD10
0.005
0.004
0.038
0.034
1.74
1.79
1.00
1.00
0.017
0.026
0.016
0.023
0.022
0.020
0.003
0.002
0.018
0.016
2.23
2.45
1.00
1.00
0.024
0.042
0.012
0.015
0.015
0.003
I.B.3Page6
-------
Table III. B. 3-1 b. Results of Dietar
*»-»•>' JL.V, * n«
-------
Stage 2: Out of the seven OPs analyzed in Stage 1 , the entire oral
databases; i.e., both chronic and subchronic studies, of three
randomly selected OPs were analyzed as in Stage 1. Relative
potency was calculated using all available methamidophos
I studies (Table III.B.3-2).
CM i
CD i In the pilot analysis of the complete oral database for three OPs
^H | (diazinon, dimethoate, and phosalohe; Table III.B.3-2)
T— | comparing the RPFs calculated with slope scale factors and
BMD10s for ChE data using the average and duration-specific
compound intakes, the RBC and brain data for both sexes
display comparable potency values. For phosalone RBC male
only, a 7-fold difference between the average and duration-
specific intake assessments was observed.
Graphs of potency vs. time are shown in Figures III. B. 3-1 ,2 for
(/)
o:
0
>
13
O
CL
O
"D
CD
•GO
">
0)
01
the analyzes of average chemical intake and for duration
CO 1 specific chemical intake. The patterns observed in the graphs
(D | for the average intake analyzes are similar to those of the
[ duration specific intakes.
I.B.3 Page8
-------
Table III.B.3-2. Results of Dietary Intake [actual vs average] Using All Available Studies
DIAZINON
DIMETHOATE
METHAMIDOPHOS
PHOSALONE
DIAZINON
DIMETHOATE
METHAMIDOPHOS
PHOSALONE
DIAZINON
DIMETHOATE
SP
43543901
43543902
40815003
41942002.
43128201
164177
41867201
00148452
43197901
44852504
44801002
43543901
43543902
40815003
41942002
43128201
164177
41867201
148452
43197901
44852504
44801002
43543901
43543902
40815003
41942002
43128201
164177
'i'L '}&;&&££•: *
'COMPARTMENT^
;: _ %$ft£;t'.
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
BRAIN
RBC
RBC
t^sSc '. 3
.-> -»» **
>»?
F
F
F
F
M
M
M
M
F
F
^.Dietary
* Intake ^
Calculation '
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
^Relative Potency
•,','- using *m'
0.031
0.033
0.531
0.58
1.00
1.00
0.019
0.021
0.005
0.005
0.71
0.83
1.00
1.00
0.019
0.028
0.38
0.41
0.32
0.27
Lower¥5%.
•cu\." .
0.018
0.019
0.41
0.45
1.00
1.00
0.014
0.010
0.002
0.002
0.53
0.60
1.00
1.00
0.011
0.012
0.22
0.27
0.14
0.14
itfpper 95%
' CL
0.053
0.058
0.69
0.75
1.00
1.00
0.025
0.040
0.012
0.010
0.94
1.15
1.00
1.00
0.032
0.063
0.65
0.62
0.73
0.53
-v-v-v
- liJB £
BMD;* •
2.48
2.08
0.25
0.20
0.09
0.08
5.05
3.37
24.77
18.28
0.10
0.08
0.08
0.07
3.49
1.96
0.24
0.18
0.29
0.33
*„ t »*.
"BMDL
1.78
1.51
0.23
0.18
0.08
0.07
3.83
2.24
24.15
17.83
0.08
0.06
0.07
0.06
2.49
1.22
0.22
0.17
0.14
0.16
'#fia '>j?-9.
Rela'tive Potency
'using BMD10
0.036
0.038
0.36
0.40
1.00
1.00
0.018
0.024
0.003
0.004
0.80
0.88
1.00
1.00
0.023
0.036
0.38
0.44
0.31
0.24
I.B.3 Page 9
-------
CHEMICAL
METHAMIDOPHOS
PHOSALONE
DIAZINON
DIMETHOATE
METHAMIDOPHOS
PHOSALONE
MRID
41867201
148452
43197901.,
44852504
44801002
43543901
43543902
40815003
41942002
43128201
164177
41867201
148452
43197901
44852504
44801002
COMPARTMENT
RBC
RBC
RBC
RBC
RBC
RBC
SEX
F
F
M
M
M
M
Dietary
; Intake -
Calculation
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
average
biweekly
Relative Potency
using 'm'
1.00
1.00
0.044
0.048
0.12
0.14
0.27
0.25
1.00
1.00
0.054
0.072
Lower 95%
CL
1.00
1.00
0.015
0.017
0.024
0.027
'0.15
0.13
1.00
1.00
0.022
0.032
Upper 95%
-•-• CL ••/•;
1.00
1.00
0.13
0.14
0.63
0.68
0.48
0.47
1.00
1.00
0.13
0.16
BMD10
0.09
0.08 .
1.45
1.31
0.40
0.34
0.36
0.40
0.07
0.06
18.07
2.72
BMDL:
0.07
0.06
0.77
0.68
0.22
0.18
0.20
0.22
0.05
0.05
9.81
1.40
Relative Potency
using BMD10
1.00
1.00 .
0.062
0.061
0.18
0.18
0.19
0.15
1.00
1.00
0.004
0.023
I.B.3Page10
-------
Figure III.B.3-1a. Plots of potency versus time for brain cholinesterase measured
in rats exposed to diazinon
CM
O
CO
C
0
ay
(A
0
co
CO
0
JS
13
E
D
O
Q.
O
"O
0
0
Average Dose
Female
Male
e .
a
0
g -
0
f
1 1 1 1 1 I 1
103 20D 300 403 800 OH 703
100 200 300 4
-------
Figure III.B.3-1 b. Plots of potency versus time for brain cholinesterase measured
in rats exposed to dimethoate
CM
O
Average Dose
Female
Male
CO
c
0)
E
CO
CO
0
CO
CO
0
0
K:
n
"4 .
a ^
6 o
I
100 200 330 403 SBCi 800 TO)
K*
a ^
I I I I I I
iaa mi ma «a eoa eaa ma
Duration Specific Dose
Female
Male
MNM/
O
a.
O
13
0
CO
E
;•
|
§
=
=
:
z
ep
"> -
e ^
a ~
- - y
J
1
II 1 1 1 1 1
CB 2CB 3Xi -2Q3 SCO SM 7CB
0
2Xi 3X1 4X1 KB 8M TOO
I.B.3Page12
-------
CM
O
Figure III.B.3-1 c. Plots of potency versus time for brain cholinesterase measured
in rats exposed to methamidophos
Female
Average Dose
Male
CD
C
0
E
V)
c/)
0
C/)
q
H
-
P .
I I I I I I I
100 330 600 TOO
v 1 1
I I I
ten 3x1
TOO
GO
CC
0
Female
Duration Specific Dose
Male
O
CL
O
"D
0
GO
">
0
01
I I I I I I I
103 3QQ fiOQ 700
TBrt»{ )
i i i i I I
103 3X1 KB .700
I.B.3Page 13
-------
Figure III.B.3-1 d. Plots of potency versus time for brain cholinesterase measured
in rats exposed to phosalone
Average Dose
CM
O
Female
Male
CO
C
0
0
CO
CO
CO
0
15
D
E
D
O
CL
O
"O
0
_co
0
a
5 -
a
Female
9 -
9 -
Q -100
sea ern
T»*sa( )
a
8 -
a
-*
Q -
a
^ .
g-
T
f
i
a iaa aoa sn raa
a 100 aaa saa ma
Duration Specific Dose
TOO
g H
Male
1
i i i I I l i
a ten s» SM raa
I.B.3Page 14
-------
Figure III.B.3-2a. Plots of potency versus time for RBC cholinesterase measured
in rats exposed to diazinon
CM
O
Average Dose
Female
Male
CD
C
CD
E
CO
CO
0)
CO
CO
CO
cr
CD
P
^ -
"* .
E 2 -I
103 200 300 400 603 600 TOO
1QQ 200 333 401 503 630 TOD
Duration Specific Dose
Female
Male
O
CL
O
"D
CD
CO
'>
0
«
P .
P .
P
H
100 2QQ 300 400 600 GOO 703
100 200 300 400 fiQQ 600 TOO
I.B.3Page15
-------
Figure III.B.3-2b. Plots of potency versus time for RBC cholinesterase measured
in rats exposed to dimethoate
(N
O
CD
C
CD
E
0
CO
CO
to
•0
Average Dose
Female
Male
R -
CM
"3 -
P-I
Duration Specific Dose
Female
Male
JO
ZS
E
13
O
QL
O
"O
0)
_co
0
a:
I:}:::::::;:::::
0 1CC 3QQ SCO (tG
Q 10Q 330 SB TOO
I.B.3Page16
-------
Figure III.B.3-2c. Plots of potency versus time for RBC cholinesterase measured
in rats exposed to methamidophos
Average Dose
CNI | Female Male
o i
T~ |
CD I
-A,,* i n -
"f p "l" -
c i
0i e N-
E 1
% f
F f 4
\ i f f
<1
i _ __.
ij
* 1
ep> -
* '
w -
a -
:-f----t-f
J«
CO | ' i i . . i i i
d) 1 0100330600703 0103300930703
jl H TinV6| } TB«6( )
^ jj Duration Specific Dose
CO !
ir i
0 1 Female Male
> I
1 I
m^f* Z -
M««I» •:
O 1
CL ! e --•
O I --:
T^ i
r
•}-;-!-'T--f
U) -
-^ .
to -
e
N -
a -
-} i-rr-f
it CT
^ -
CD 1 i i . . i i i
fA | 0100303600703 01003X1600703
0 !
cr: i
I.B.3Page17
-------
CM
Figure III.B.3-2d. Plots of potency versus time for RBC cholinesterase measured
in rats exposed to phosalone
Average Dose
Female
Male
T—
c
0
CO
CO
0
CO
CO
Qi
•*
rd "
r?
a
a ~
ri ~
q
a
i
I" <
•
1
*
1 1
<
'
i , t 4 1
a 100 30Q W TOO
Duration Specific Dose
Female
a KM acc aoa
Male
J-S
"3
E
13
o
CL
o
"a
0
JO
0
$tr.
0 100 33Q
a too aaa au TOO
)
I.B.3Page18
-------
CM
O
CO
I
•4— *
C
0
E
t/)
(/)
0
GO
60
Stage 3: Compare the BMD10 's and BMDL's of the index chemical
calculated from the average compound intakes and the
duration-specific compound intakes (Table III.B.3-3).
As shown in Table III.B.3-3, BMD10 and BMDL calculated using
the average compound intake from July analysis are similar to
but slightly smaller those calculated with the July methods with
duration-specific compound intakes. BMD10 and BMDL
calculated using the average compound intake from July
analysis are similar those calculated with the December
methods with duration-specific compound intakes.
Table III.B.3-3. Comparison of Average Intake vs Duration-Specific Intake BMD10s
and BMDLs
'Compartment!
yf*
• - \
BMD
BMDUjp
FEMALE RBC
0.09
0.07
0.08
0.06
FEMALE brain
0.09
0.08
0.08
0.07
FEMALE brain
0.08
0.07
oo
MALE RBC
0.07
0.05
0.06
0.05
MALE brain
0.08
0.07
0.07
0.06
MA E brain
0.07
0.06
0
15
E
D
O
Q.
O
"O
0
GO
">
0
tr
Conclusions: The pilot analysis of compound intakes using duration
specific values showed that relate potency estimates
calculated from slope-scaling fac;:ors and BMD10s are
similar to those calculated using the average study
compound intake. Based on this •. nalysis, it is
reasonable for OPP to continue uying the average
compound intake for its potency estimates. Concerning
the PODs for the index chemical, although the values are
very similar, the PODs calculated from duration-specific
intake values result in slightly smaller BMD10s.
b. Response to SAP Comments from March 2002
The following analyses were performed following disci's.sion and
recommendations from the February 5-8. 2002 meeting the FIFRA SAP
meeting on the "Methods Used to Conduct a Preliminary Cumulative Risk
Assessment for Organophosphate Pesticides":
i. Selecting the Benchmark Response Level
At the February 5-8, 2002 meeting of the FIFRA SAP! some panel
members and some Public Commenters discussed the Arency's selection
I.B.3 Page 19
-------
I of the BMD10 as the benchmark response level. In response to this
I discussion, the Agency analyzed the detection limits of the studies
I assessing female brain cholinesterase levels used in the Preliminary
I Cumulative Risk Assessment of the OPs. This analysis has shown that
1 generally these studies can reliably detect around 10% cholinesterase
CM | inhibition and that such .levels were generally achieved in the studies.
CD | Therefore, the Agency's use of the BMD10 as the benchmark response is
| appropriate.
"~--~ I According the Agency's draft benchmark dose guidance (USEPA,
CD | 2000a), generally, the response level selected to calculate the benchmark
i | dose should lie in the low end of the range of the responses but within
| assay detectability. Figure III.B.3-3 shows a plot of the range of mean
| brain cholinesterase inhibition observed in all treatment groups (i.e.,
I controls were not included). That figure shows that all chemicals include
| at least one dose level that yields approximately 10% inhibition. Thus, it
CO i is possible to directly assess the fit of the model to data in this critical
00 I region.
0 f
Vb | The ability of a study to detect a given amount of change is measured
~5 I by the power of the study. In general, the power of a study depends on
i the sample size and the variability of the observations, measured as the
| standard deviation among individual measurements. Both of these
00 I factors vary among datasets in this risk assessment. The power for each
[ study to detect a difference between control and a single treatment group
| of mean brain cholinesterase activity by 1%, 5%, 7.5%, 10%, 15%, and
0 [ 20% has been calculated. In Figure III. B. 3-4, the proportion of datasets
> | with at least x power is plotted against x for effect levels ranging from 1-
I 20% inhibition, and the median power (that is, the power level such that
I half the datasets have greater than that level of power) among those data
I sets to detect each change is indicated on the axis. Only at the level of a
= 10% change is the median power greater than 0.80, which has been a
I conventional goal in designing experiments. Thus, a 10% change in
| mean cholinesterase activity is indeed in the low end of detectability of
| assays for brain cholinesterase activity as they were conducted in the
n [ studies used in this risk assessment.
o I
13 j
0 i
CO 1
"> I
0 I
III.B.3 Page 20
-------
I Figure III.B.3-3. Observed levels of inhibition relative to concurrent control for all
I dose-groups. The solid vertical line indicates 10% inhbition.
CM
O
CD
C
CD
E
CO
CO
CD
V)
GO
CO
ir
CD
D
E
13
O
CL
O
CD
CO
">
0
ACEPHATE
AZINPHOSMETHYL
BENSULIDE
CHLORETHOXYFOS
CHLORPYRIFOS
IHLORPYRIPHOSMETHYL
DIAZINON
DICHLORVOS
DICROTOPHOS
DIMETHOATE
DISULFOTON
ETHOPROP
FENAMIPHOS
FENTHION
FOSTHIAZATE
MALATHION
METHAMIDOPHOS
METHIDATHION
METHYLPARATHION
MEVINPHOS
NALED
OMETHOATE
OXYDEMETONMETHYL
PHORATE
PHOSALONE
PHOSMET
PHOSTEBUPIRIM
PIRIMIPHOSMETHYL
PROFENOFOS
TERBUFOS
TETRACHLORVINPHOS
TRIBUFOS
TRICHLORFON
-0.5 0 0.5
Observed Per Cent. AChE Inhibition
I.B.3 Page 21
-------
CM
O
CD
C
0)
C/)
(/)
0
(/)
C/)
a:
CD
Z5
E
D
O
a.
O
"O
0
GO
">
0
a:
Figure III.B.3-4. Distribution of the power to detect a 1%, 5%, 7.5%, 10%, 15%, and
20% change in mean cholinesterase activity among datasets in the risk
assessment. For each effect of treatment, the curves represent the fraction of
datasets for which the power is at least the value on the x-axis to detect that
effect. For example, half the studies have at least a power of 0.894 to detect a
10% change in mean cholinesterase activity.
1.0 4—
0.8 -
0.6 -
W
T3
13
55
I 0.4
0.2 -
0.0 --
i 1 r
0.0 0.2
0.088
0.4
0.422
Power
0.6
0.8
1.0
0.701
0.894
I.B.3Page22
-------
ii. Standard and formal definition of the full mathematical
exponential model
A formal presentation of the exponential model is included in the
Appendix III.B.1.
CM
O iii. Individual Animal Data: Consequences of Aggregating Data
At the February 5-8, 2002 meeting of the FIFRA SAP, some members
of the panel discussed the fact that the dose-response modeling of
cholinesterase inhibition was based solely on dose group means,
standard deviations, and sample sizes. The discussion centered about
the issue: to what extent would the results of the analysis have differed if
~~ individual animal data had been used? The answer to that question has
££ two parts.
CO 1. The statistical methods used in the analysis depend on the data
CO only through their dose group means, standard deviations, and
3) sample sizes.
< Thus, applying the same analysis to individual animal data would
result in the same numerical estimates as the current analysis. The
following argument shows why this is so. Whether the model fit uses
CO generalized least squares or is a nonlinear mixed effects model (See
' III.B.1), the parameter estimates are the result of optimizing
expressions that depend on the individual data through quadratic
CD forms like:
>
£0
*D
E
Here, y is a column vector of the individual observations
O indexes dose group (in this discussion, "dose group" refers to the
observations on animals of the same sex exposed at the same time
ar|d dose to the same chemical) andy indexes individual within that
Odose group. The vector \\ is the vector of fitted values. Since all
individuals in the same dose group were exposed to the same dose,
~Q the fitted values for each individual in a dose group are all identical.
Q) Finally, the matrix V is symmetric, and has the form D + M, where D is
CO diagonal, and partitioned such that the values corresponding to the
same dose group are identical to each other. M is symmetric and
0 partitioned into blocks that correspond to the dose groups. The values
Q/ within any given block are identical to each other. The partitioning of
the components of V is due to the fact that all the individuals of the
same sex given the same dose in the same study are treated
identically by the model. A direct consequence of the partitioning of u
III.B.3 Page 23
-------
I and V is that the value of the above quadratic form can be expressed
I solely in terms of group means, standard deviations, and sample
| sizes.
i 2. Distribution of the brain cholinesterase data.
CM i
O | The methods used in the dose-response analysis assume the data
I is normally distributed. If the individual cholinesterase activity
I measurements were distinctly non-normal, it would be of interest to
= determine the impact of transformed or trimmed data on the
I benchmark dose estimates used to estimate relative potency.
____ I Individual animal data for female and male rat brain cholinesterase
•ip I activity were available for a small subset of the studies used in the the
92 1 Draft Revised Cumulative Risk Assessment for the OPs. Individual
C 1 animal data were available from 15 studies representing 11 chemicals
CO I (see Table III.B.3-4). Each study included several dose-response data
CO I sets in both males and females; each dose-response data set included
CD I several dose groups. (Note to the reader: Individual animal data form
j£J 1 male and female brain cholinesterase activity used in the following
| analysis have NOT been released to the public).
Jsd 1 i. Test for normality.
"ftp | Each individual dose group (sample sizes ranging from about 5
| to 50) was tested for deviations from normality using the Shapiro-
(D 1 Wilk test for normality (Shapiro and Wilk, 1965). The P-values for
a,> 1 each dose group in a study were then combined using Fisher's
-*•-* I method (Sokal and Rohlf, 1981; section 18.1), giving an overall P-
J5 [ value for deviation from normality for each MRID. Table III.B.3-4
;3 | gives the results of this initial test for normality.
^ | The result of combining all the P-values over all studies was
| highly significant: the P-value is 9*10~8. Thus, there is evidence of
| some deviation from normality, though, given the amount of data
ri 1 available for the test, and the relatively few chemicals for which the
I overall P-value is significant (only 2/15 MRIDs have a significant
| deviation from normality), the overall deviation from normality does
I not seem excessive
CD i
CO I "• Identify the nature of the deviations from normality.
> i
0 | Two possibilities were explored: that the data were such that a
| power transformation (in the form of the Box-Cox transformation;
I Sokal and Rohlf, 1981, section 13.9) would result in a normal
I distribution, and that the data were "contaminated", that is, the bulk
I of the observations are from a normal distribution, with an
[ III.B.3 Page 24
-------
{ occasional too large or too small value (Rosenberger and Gasko,
I 1983). The approach taken in this analysis was to use maximum
I likelihood to estimate the parameters in two models:
| 1). An observation y is sampled from a normal distribution with
C\l I mean u and standard deviation a with probabilty p, and from
O | a normal distribution with mean u and standard deviation
^ | axa, where a > 1 , with probability 1 - p. Here the mean and
| standard deviation are specific to each dose group, but a is
I the same value for all dose groups in a study.
CQ I 2) If the data y were transformed to z by the Box-Cox
1 I y - 1
-*-• | transformation :z = - - (if f *0) or z = log (j^) (if t = 0),
JMM» E '
0 | the transformed data would be normally distributed, with
| separate mean and standard deviation for each dose group
| (but only one power parameter t for each study). When t =
(/) | 1, then z = y- 1, and the original variable y is normally
0 1 distributed.
(/) j
C/5 | The Akaike Information Coefficient (AIC; Burnham and
I Anderson, 1998) was calculated for each of the two hypothetical
| distributions for each study. AIC is useful for comparing different
| probability models fit to the same data sets: smaller AIC values
I indicate better fits. Table III.B.3-5 shows the AIC values that
I resulted from fitting the two models just described to the individual
I animal data from each study. In addition, the power parameter
[ estimated in the Box-Cox model was tested for significant
I difference from one.
_
—j I For eight of the fifteen studies, the AIC for the Box-Cox
E| transformed data was less than that for the contaminated normal.
I Only two of those studies had a Box-Cox parameter significantly
-3 I different from one, indicating that a Box-Cox transformation would
O 1 result in a significantly more normal distribution. In the remaining
^ 1 seven of the fifteen studies, including the two with significant
~~ | Shapiro-Wilk tests, the contaminated normal model provides a
CJ) | better description of the data. The overall AIC for the contaminated
[ normal distribution is less than that for the Box-Cox transformed
| data, showing that the contaminated normal model is superior to
yr$ | the Box-Cox model as a single overall probability model for these
-^ | data.
0 I
cr: i
I.B.3Page25
-------
f iii. Impact of non-normality on the BMD estimates.
| BMD10s were calculated for trimmed and untrimmed data. Table
| III.B.3-6 shows the results of applying the Shapiro-Wilk test to the
I trimmed individual data. The overall P-value for all the data taken
CM I together is 0.056, indicating a substantial improvement
p !
I Aggregated datasets were produced from the original (untrimmed)
I individual data and the trimmed individual data, and both the basic and
I expanded models fit to each set of data for each chemical (See I.B
j and III.B.1). Four chemicals were affected by the trimming:
i I dicrotophos, methamidophos, phorate, and phosalone. Thus,
<•*-» I comparisons between untrimmed and trimmed data is limited to nine
^ I studies from four OPs.
CD |
t | Table III.B.3-7 compares the BMD10 calculated from the original
CO [ data to that calculated using the trimmed data, for both basic and
CO I expanded models. The largest difference is less than 20% of the
CD I untrimmed value, which is reasonably small. The current dose-
J/5 1 response analysis used in the Draft Revised Cumulative Risk
| Assessment of the OPs, based solely on aggregated data, is relatively
| robust to the kinds of deviations from normality identified here.
CO I In summary, since the statistical methods used to fit dose-response
y? 1 models to the data depend on the data only through their means,
•"*" | standard deviations, and sample sizes, the only way an analysis of
0 I individual data might differ from that of aggregated data would be if the
> 1 distribution of the data were substantially non-normal. The distributions of
a subset of the data were examined, resulting in evidence that some
studies did produce data that deviated from normality. When extreme
observations were omitted, the overall distribution of the data became
closer to a normal distribution. However, benchmark doses calculated
using the trimmed data, were quite similar, to those using all the data.
Thus, it is unlikely that using aggregated data has substantially distorted
o
EL
O
ID-
CD
0
the estimates of benchmark doses that would obtain had the analysis
been based on individual animal data.
I.B.3 Page 26
-------
I Table III.B.3-4. Chemicals and studies used in individual animal analysis.
CM
O
CD
i
C
CD
E
CO
CO
0
CO
CO
CO
CD
o
zs
O
CL
O
Chemical
Methamidophos
Methamidophos
Methamidophos
Fenamiphos
Bensulide
ODM
Fosthiazate
Dicrotophos
Phosalone
Phosmet
Terbufos
Phosalone
Phorate
Phorate
Chlorpyrifos-methyl
Study
(MRID no.)
148452
41867201
43197901
44051401
44161101
44189501
44269905
44527802
44801002
44811801
44842302
44852504
44895301
44895302
44906902
Number of
Dose
Groups
20
20
8
8
32
36
14
16
8
16
8
24
8
10
10
Number
Failed
8
2
1
1
4
1
1 .
4
1
1
1
0
1
0
1
Proportion
Failed
0.400
0.100
0.125
0.125
0.125
0.028
0.071
0.250
0.125
0.063
0.125
0.000
0.125
0.000
0.100
Combined
Shapiro-
Wilks
P-value
1 .63e-08
1.08e-01
1.37e-01
1.60e-01
8.14e-02
8.09e-01
3.236-01
1.016-03
6.526-02
4.63e-01
8.986-02
2.606-01
7.556-02
2.716-01
5.476-02
0
CO
0
"Number of Groups" is the total number of dose groups available; "Number Failed" is the number of
individual dose groups for which the Shapiro-Wilks test reported a P-value less than 0.05; "Proportion
Failed" is the proportion of dose groups that failed the test (Number Failed/Number of Groups);
"Combined Shapiro-Wilks P-value" is the overall P-value for each MRID, resulting from using Fisher's
method to combine the P-values for the individual dose-group tests.
I.B.3Page27
-------
1 Table HI.B.3-5. AIC values for the Box-Cox and the contaminated normal models.
CNJ
o
CD
i
c
0
E
CO
CO
0
CO
CO
CO
0
>
JS
"Z5
E
ZJ-
o
DL
o
(D
CO
*t milillHIl
>
0
a:
Chemical
Methamidophos
Methamidophos
Methamidophos
Fenamiphos
Bensulide
ODM
Fosthiazate
Dicrotophos
Phosalone
Phosmet
Terbufos
Phosalone
Phorate
Phorate
Chlorpyrifos-methyl
Study
(MRID no.)
148452
41867201
43197901
44051401
44161101
44189501
44269905
44527802
44801002
44811801
44842302
44852504
44895301
44895302
44906902
Sum:
AIC
Contaminated
Normal
362.46
250.84
64.15
105.23
5586.60
527.11
2242.37
273.37
174.91
382.29
339.77
204.74
708.88
366.05
211.33
11139.44
Box Cox
Transformed
386.90
251.01
62.26
106.66
5578.49
524.59
2239.92
228.71
764.79 *
380.07
340.77
209.10
119.00
359.44 *
208.39
11159.50
| MRID numbers for data that were significantly non-normal by the Shapiro-Wilks test (see Table III.B.3-4)
§ are written in bold. The smaller of the two AIC values for each MRID is written in bold italics. When the
1 Box-Cox power parameter is significantly different from 1, the Box-Cox AIC is followed by an asterisk.
I.B.3Page28
-------
i Table III.B.3-6. P-values for the Shapiro-Wilks test, combined over all dose groups
i in a study for the trimmed individual data.
CM
O
CD
i
C
CD
E
V)
CO
0)
(/)
C/)
o:
CD
Chemical
Methamidophos
Methamidophos
Methamidophos
Fenamiphos
Bensulide
ODM
Fosthiazate
Dicrotophos
Phosalone
Phosmet
Terbufos
Phosalone
Phorate
Phorate
Study
(MRID no.)
148452
41867201
43197901
44051401
44161101
44189501
44269905
44527802
44801002
44811801
44842302
44852504
44895301
44895302
P.value
0.046
0.293
0.137
0.160
0.081
0.809
0.323
0.937
0.065
0.463
0.090
0.863
0.831
0.271
13
E
13
O
CL
O
T5
0
CO
>
0
cc
I.B.3 Page 29
-------
j Table III.B.3-7. Benchmark doses from the basic and expanded models for
j untrimmed (original) and trimmed data.
C\l
O
C
CD
E
to
to
CD
to
to
o>
JE
13
E
:D
O
a.
O
"O
CD
Chemical
Dicrotophos
Methamidophos
Phorate
Phosalone
Data
Treatment
original
trimmed
original
trimmed
original
trimmed
original
trimmed
Female BMD10
Expanded Model
NA
NA
NA
NA
0.215
0.201
6.426
6.313
Basic Model
0.032
0.026
0.080
0.079
0.036
0.037
3.843
3.847
I NA: As shown in I.B, the basic model was used to estimate potency for methamidophos and dicrotophos.
[ References
| Burnham, K. P. and Anderson, D. R. 1998. Model Selection and
I Inference. A Practical information-Theoretic Approach. Springer. New
j York.
| Rosenberger, J. L. and Gasko, M. 1983. Comparing location estimators:
I trimmed means, medians, and trimean. Chapter 10 in Understanding
1 Robust and Exploratory Data Analysis, David C. Hoaglin, Frederick
I Mosteller, and John W. Tukey, eds. Wiley. New York.
I Shapiro, S. S. and Wilk, M. B. (1965). "An analysis of variance test for
f normality (complete samples)", Biometrika, 52: 591-611.
| Sokal, R. R. and Rohlf, F. James. (1981). Biometry, Second Edition.
I Freeman. San Francisco.
0
o:
I.B.3 Page 30
-------
I III. Appendices
1 B. Hazard/RPF
I 4. R Programs for the Revised Analysis
CM I
Of a. Parti: Package RBMDS
o>
RBMDS is a package of utility functions written to facilitate
the analysis of the cholinesterase activity dose-response data.
These functions are made available to the scripts that carry out the
data analysis by including the call "require(RBMDS)" or
"library(RBMDS)" at the beginning of the scripts.
I ### Generalization of exponential decreasing model
| ### A, B, and m are constrained to be strictly positive
w' i cexpB <- function (Dose, A, B, m)
C/> § {
(D i ## exp(A)*(l/(l + exp(-B)) + (exp(-B)/(l + exp(-B)) * exp(-exp(m)*Dose)))
fj\ | .exprl <- exp(A)
7A i .expr3 <- exp(-B)
1 .expr4 <- 1 + .expr3
I .exprG <- exp(m)
I .expr9 <- exp(-.exprG * Dose)
= .exprlO <- .expr3 * .expr9
I .exprll <- .exprlO/.expr4
= .expr!3 <- .exprl * (l/.expr4 + .exprll)
I .exprl4 <- .expr4A2
i .value <- .expr!3
0} | .grad <- array(0, c(length(.value), 3), list(NULL, c("A",
>= "B", "m")))
= .grad[, "A"] <- .expr!3
| .grad[, "B"] <- .exprl * (.expr3/.expr!4 - (.exprll - .exprlO *
1 .expr3/.expr!4))
I .grad[, "m"] <- -.exprl * (.expr3 * (.expr9 * (.exprG *
-J i Dose))/.expr4)
E| attr(.value, "gradient") <- .grad
| .value
I ### cexpBZ: same as above, but gradient includes deriv wrt dose.
Q... I CexpBZ <-
Oi function (Dose, A, B, m)
I {
= .exprl <- exp(A)
I .expr3 <- exp(-B)
i .expr4 <- 1 + .expr3
yf. I .exprG <- .expr3/.expr4
.\H I .expr? <- exp(m)
^> i .exprlO <- exp(-.expr7 * Dose)
0i .expr!3 <- .exprl * (l/.expr4 + .exprG * .exprlO)
| .exprlS <- .expr4A2
fV_ I .value <- .expr!3
i .grad <- array(0, c(length(.value), 4), list(NULL, c("Dose",
.grad[, "Dose"] <- -.exprl * (.exprG * (.exprlO * .expr?))
.grad[, "A"] <- .expr!3
I.B.4 Page 1
-------
.grad[, "B"] <- .exprl * (.expr3/.expr!8 - (.exprG - .exprB *
.expr3/.exprl8) * .exprlO)
.grad[, "m"] <- -.exprl * (.exprS * (.exprlO * (.expr7 *
Dose)))
attr(. value, "gradient") <- .grad
.value
i ### CpkexpB: above model, with the additional assumption that there is
i ### saturable detoxification. This is implemented by composing CexpB
I ### (now assuming that Dose in CexpB refers to internal dose), with
i ### a model that relates administered dose to internal dose:
I ###
= ### idose = 0.5*((oose - S - D) + sqrt((Dose - S - D)AZ + 4*Dose*S))
I ###
| ### This model approaches a line with slope 1 and y-intercept -D as Dose
| ### increases to infinity. The parameter 's' controls the shape of the
»— = ### low-dose part of the curve. For values close to 0, the curve looks
(|) | ### very threshold-like; as S increases, the curve becomes more gradual.
f— I ### S and D must be positive
£Z I ###
CO ^ *** Tne f""na^ function is (Dose refers to administered dose; exponentiation
ft( I ### used to force positive parameters):
l ###
(/) 1 ## exp(A)*(l/d + exp(-B)) + (exp(-B)/(l + exp(-B)) * exp(-exp(m)*0. 5*((Dose
tf\ = - exp(s) - exp(D)) + sqrt((Dose - exp(S) - exp(D))A2 + 4*Dose*exp(S))))))
I CpkexpB <- function (Dose, A, B, m, s, D)
I idose <- CpkB(Dose, S, D)
| .value <- CexpB2(idose, A, B, m)
= .grad <- array(0, c(length(.value), 5),
CD i list(NULL,
.— I .grad[,c("A","B","m")] <-'attr Rvalue/'gradient") [,c("A", "B","m")]
"*± 1 .grad[,c("S","D")] <- attr(.value,"gradient")[,"Dose"] *
CD i attr(idose, "gradient")[,c("s", "D")]
"3T = attr(.value, "gradient") <- .grad
—f | .value
Z3 I ### cexpBS(dose, A, B, m, sex, fixed=NULL) allows fixed, a named list of
O| ### vectors of fixed values for the model CexpB. They can differ by sex.
: JtJtJi iicorl fnr PYamnlp-
used, for example:
n" ' • - • •
f~\ | mrid)
f* = ### nlme(model=chei ~ CexpBS(dose, A, m, sex,
LJL 1 ### fixed=list(B=c(F=-3.01, B=-2.78))), data=mydata, random=A+m~l
i CexpBS <- function(dose, A, B, m, sex, fixed=NULL) {
CD I call List <- vector("list",4)
(/) I names(callList) <- c("Dose","A","B","m")
_ = sex <- switch(length(fixed) +1,
> I sex,
CD f ;5
1 call List[["Dose"]] <- dose
i ### Assume we will always estimate A
I call List[["A"]] <- A
| if ("B" %in% names(fixed)) {
! III.B.4Page2
-------
I callList[["B"]] <- fixed[["B"]][as.character(sex)]
i callList[["m"]] <- if ("m" %in% names(fixed)) {
I fixed[["m"]][as.character(sex)]
I } else {
I B
I } else {
C\( I callList[["B"]] <- B
i callList[["m"]] <- if ("m" %in% names(fixed)) {
I fixed[["m"]][as.character(sex)]
| } else {
= m
CD I .value <- do.call ("CexpB",call List)
I I .grad <- attr(.value, gradient")
| .grad <- .grad[,-match(names(fixed),col names(.grad)),drop=FALSE]
"*^ I attr(.value, "gradient") <- .grad
£-. i .value
0 I >
I ### This is the same model, reparameterized so that, instead of 'm1, we
= ### estimate 'BMD', for a BMR*100% reduction in mean response relative to
, i ### control.
*{ i ###
CD = ### Model is:
(/) I ### ~ exp(A)*(V(l + exp(-B)) + exp(-B)/(l + exp(-B))*exp(log(l - BMR*(1
(/) i exp(B))) * Dose*exp(-BMD)))
[ CexpBwD <- function (Dose, A, B, BMD, BMR=0.10)
V 5 ^
~" | .exprl <- exp(A)
CO = .exprS <- exp(-B)
"f^ | .expr4 <- 1 + .exprS
LL, I .expr6 <- .expr3/.expr4
I .expr? <- exp(B)
Q) = .exprlO <- 1 - BMR * (1 + .exprZ)
"> I .expr!4 <- exp(-BMD)
• — | .exprlS <- log(.exprlO) * Dose * .expr!4
"tl( I .exprlS <- exp(.exprlS)
_VU | .exprl9 <- .exprl *' (l/.expr4 + .expr6 * .expr!6)
— * i .exprZO <- .expr4A2
r£ i .value <- .expr!9
CI 1 .grad <- array(0, c(length(. value) , 3), list(NULL, c("A",
*~ = "B", "BMD
*-* i ## The following returns the hessian, too (for calculating covariances
CD I ## from nlme models).
(/) | CexpBwDH <- function (Dose, A, B, BMD, BMR=0.10)
"•TT" a *•
,> 1 .exprl <- exp(A)
0 | .expr3 <- exp(-B)
I .expr4 <- 1 + .expr3
| .expr6 <- .expr3/.expr4
I .expr? <- exp(B)
i .exprlO <- 1 - BMR * (1 + .expr?)
i .expr!4 <- exp(-BMD)
| .exprlS <- log(. exprlO) * Dose * ,expr!4
[ III.B.4Page3
-------
_-~
0
C
0
if)
(/)
CO
«H»»
m
15
E
13
O
(7)
.exprlG <- exp(.exprlS)
.expr!9 <- .exprl * (l/.expr4 + .exprG * .exprlG)
.expr20 <- .expr4A2
.expr21 <- .expr3/.expr20
.expr22 <- BMR * .expr?
.expr23 <- .expr22/.exprlO
.expr25 <- .expr23 * Dose * .expr!4
.expr26 <- .exprlG * .expr25
.expr28 <- .expr3 * .expr3
.expr30 <- .exprG - .expr28/.expr20
.expr34 <- .exprl * (.expr21 - (.exprG * .expr26 + .expr30 *
.exprlG))
.expr35 <- .exprlG * .exprlS
.expr38 <- -.exprl * (.exprG * .expr35)
.expr40 <- 2 •* (.expr3 * .expr4)
.expr42 <- .expr20A2
.exprSS <- .expr30 * .expr26
.value <- .expr!9
.grad <- array(0, c(length(.value), 3), list(NULL, c("A",
"B", "BMDf')))
.hessian <- array(0, c(length(.value), 3, 3), list(NULL,
C("A", "B", "BMD"), CC'A", "B", "BMD")))
.qrad[, "A"] <- .expr!9
.hessian[, "A", "A"] <- .expr!9
.hessian[, "A", "B"] <- .hessian[, "B", "A"] <- .expr34
.hessian[, "A", "BMD"] <- .hessian[, "BMD", "A"] <- .expr38
.grad[, "B"] <- .expr34
.hessian[, "B", "B"] <- -.exprl * (.expr21 - .expr3 * .expr40/.expr42 +
.exprG * (.exprlG * ((.expr23 + .expr22 * .expr22/.exprlOA2) *
Dose * .expr!4) - .expr26 * .expr25) - .exprSS -
(.exprSS + (.expr30 - ((.expr28 + .expr28)/.expr20 -
.expr28 * .expr40/.expr42)) * .exprlG))
.hessian[, "B", "BMD"] <- .hessian[, "BMD", "B"] <- .exprl *
(.exprSO * .expr35 + .exprG * (.expr26 + .expr35 * .expr25))
.grad[, ''BMD"] <- .expr38
.hessian[, "BMD", "BMD"] <- .exprl * (.exprG * (.expr35 +
.expr35 * .exprlS))
attr(.value, "gradient") <- .grad
attr(.value, "hessian") <- .hessian
.value
### Include the derivative wrt Dose, to combine with the Pk model:
CexpBwD2 <- function (Dose, A, B, BMD, BMR=0.10)
.exprl <- exp(A)
.expr3 <- exp(-B)
.expr4 <- 1 + .expr3
.exprG <- .expr3/.expr4
.expr7 <- exp(B)
.exprlO <- 1 - BMR * (1 + .expr7)
.exprll <- log(.exprlO)
.expr!4 <- exp(-BMD)
.exprlS <- .exprll * Dose * .expr!4
.exprlG <- exp(.exprlS)
.exprl9 <- .exprl * (l/.expr4 + .exprG * .exprlG)
.expr24 <- .expr4A2
.value <- .expr!9
.grad <- array(0, c(length(.value), 4), list(NULL, c("Dose",
"A", "B", "BMD")))
.grad[, "Dose"] <- .exprl * (.exprG * (.exprlG * (.exprll *
.expr!4)))
.grad[, "A"] <- .expr!9
.grad[, "B"] <- .exprl * (.expr3/.expr24 - (.exprG * (.exprlG *
I.B.4Page4
-------
(BMR * .expr?/.expr!0 * Dose * .expr!4)) + (.exprG -
.expr3 * .expr3/.expr24) * .exprlG))
.grad[, "BMD"] <- -.exprl * (.exprG * (.exprlG * .exprlS))
attr(.value, "gradient ) <- .grad
.value
C\| ### Again, with the hessian:.
CexpBwD2H <- function (Dose, A, B, BMD, BMR=0.1)
.exprl <- exp(A)
.expr3 <- exp(-B)
.expr4 <- 1 + .expr3
.exprG <- .expr3/.expr4
.expr? <- exp(B)
.exprlO <- 1 - BMR * (1 + .expr?)
.exprll <- log(.exprlO)
_ .expr!4 <- exp(-BMD)
»— .exprlS <- .exprll * Dose * .expr!4
0 .exprlG <- exp(.exprlS)
E.exprl9 <- .exprl * (l/.expr4 + .exprG * .exprlG)
.expr20 <- .exprll * .expr!4
.expr21 <- .exprlG * .expr20
.expr23 <- .exprl * (.exprG * .expr21)
.expr27 <- BMR * .expr?
,expr28 <- .expr27/.exprlO
(/) .expr32 <- .expr28 * Dose * .expr!4
(/) .expr33 <- .exprlG * .expr32
<.expr37 <- .expr3 * .expr3
.expr38 <- .expr4A2
^, .expr40 <- .exprG - .expr37/.expr38
-^ .expr45 <- .exprlG * .exprlS
.exprSl <- .expr3/.expr38
f^s .exprSG <- .exprl * (.exprSl - (.exprG * .expr33 + .expr40 *
LL .exprlG))
.expr59 <- -.exprl * (.exprG * ,expr45)
0 .exprGl <- 2 * (.expr3 * .expr4)
.expr63 <- .expr38A2
.expr?6 <- .expr40 * .expr33
.value <- .expr!9
.grad <- array(0, c(length(.value), 4), list(NULL, c("Dose",
"A", "B", "BMD")))
.hessian <- array(0, c(length(.value), 4, 4), list(NULL,
c("DOSe", "A , "B", "BMD"), c("DOSe", "A", "B", "BMD")))
.grad[, "Dose"] <- .expr23
.h
essian[, "Dose", "Dose"] <- .exprl * (.exprG * (.expr21 *
.expr20))
^ i .hessian[, "Dose", "A"] <- .hessian[, "A", "Dose"] <- .expr23
n .hessian[, "Dose", "B"] <- .hessian[, "B", "Dose"] <- -.exprl
LL (.exprG * (.exprlG * (.expr28 * .expr!4) + .expr33 *
O.expr20) + .expr40 * .exprZl)
.hessian[, ''Dose", "BMD"] <- .hessian[, "BMD", "Dose"] <- -.exprl *
(.exprG * (.expr21 + .expr45 * .expr20))
.grad[, rA"] <- .exprl9
0 .hessian[, "A", "A"] <- .expr!9
tf\ .hessian[, "A", "B"] <- .hessian[, "B", "A"] <- .exprSG
.— .hessian[, "A", "BMD"] <- .hessian[, "BMD", "A"] <- .expr59
> .grad[, "B"] <- .exprSG
0 .hessian[, "B", "B"] <- -.exprl * (.exprSl - .expr3 * .exprGl/.expr63 +
.exprG * (.exprlG * ((.expr28 + .expr27 * .expr27/.exprlQA2) *
Dose * .expr!4) - .expr33 * .expr32) - .expr76 -
(.expr76 + (.expr40 - ((.expr37 + .expr37)/.expr38 -
.expr37 * .exprGl/.expr63)) * .exprlG))
.hessian[, "B", "BMD"] <- .hessian[, "BMD", "B"] <- .exprl *
(.expr40 * .expr45 + .exprG * (.expr33 + .expr45 * .expr32))
III.B.4Page5
-------
rad[, "BMD"] <- .expr59
essian[, "BMD", "BMD"] <- .exprl * (.exprG * (.expr45 +
.expr45 * .exprlS))
attr(.value, "gradient") <- .grad
attr(.value, "hessian") <- .hessian
.value
i ### CexpB2wD(dose, A, PB, BMD, BMR=0.10) Same as CexpBwD, but PB is on
I ### original scale.
I ### Model is:
i ### ~ exp(A)*(PB + (l-PB)*exp(log((l - BMR - PB)/(1 - PB)) *
= Dose*exp(-BMD)))
i ### Primarily used to be called from CexpBwDS; grad[,"PB"] is not returned.
i CexpB2wD <- function (dose, A,. PB, BMD, BMR=0.1) {
| .exprl <- exp(A)
= .expr2 <- 1 - PB>
{ .expr4 <- 1 - BMR - PB
= > .exprS <- .expr4/.expr2
I .expr9 <- exp(-BMD)
i .exprlO <- log(.exprS) *. dose * .expr9
| .exprll <- exp(.exprlO)
i .exprl4 <- .exprl * (PB + .expr2 * .exprll)
I .value <- .expr!4
= .grad <- array(0, c(length(.value), 2), list(NULL, c("A", "BMD")))
i .grad[, "A"] <- .expr!4
i .grad[, "BMD"] <- -.exprl * (.expr2 * (.exprll * .exprlO))
= attr(.value, "gradient ) <- .grad
i .value
I ### Same as above, but return gradient and hessian, and include PB in both
\ ### (for computing standard errors)
| CexpB2wDH <- function (Dose, A, PB, BMD, BMR)
1 .exprl <- exp(A)
i .expr2 <- 1 - PB
i .expr4 <- 1 - BMR - PB
= .exprS <- .expr4/.expr2
I .exprG <- log(.exprS)
| .expr9 <- exp(-BMD) •
i .exprlO <- .exprG * Dose * .expr9
I .exprll <- exp(.exprlO)
i .expr!4 <- .exprl * (PB + .expr2 * .exprll)
i .exprlS <- .exprG * .expr9
I .exprlG <- .exprll * .exprlS
i .exprlS <- .exprl * (.expr2 * .exprlG)
i .expr23 <- .expr2A2
= .expr25 <- l/.expr2 - .expr4/.expr23
.expr26 <- .expr25/.expr5
.expr30 <- .expr26 * Dose * .expr9
.expr-31 <- .exprll * .expr30
.expr38 <- .exprll * .exprlO
.expr47 <- .exprl * (1 - (.expr2 * .expr31 + .exprll))
.exprSO <- -.exprl * (.expr2 * .expr38)
.exprSl <- l/.expr23
.value <- .expr!4
.grad <- array(0, c(length(.value), 4), list(NULL, c("Dose",
"A", "PB", "BMD")))
.hessian <- array(0, c(length(.value), 4, 4), list(NULL,
c("Dose", "A'', "PB", "BMD"), cC'oose", "A", "PB", "BMD")))
.grad[, "Dose"] <- .exprlS
.hessian[, "Dose", "Dose"] <- .exprl * (.expr2 * (.exprlG *
I.B.4 Page6
-------
I .exprlS))
! .hessian[, "Dose", "A"] <- .hessian[, "A", "Dose"] <- .exprlS
1 .hessian[, "Dose", "PB"] <- .hessian[, "PB", "Dose"] <- -.exprl *
= (.expr2 * (.exprll * (.expr26 * .expr9) + .exprSl * .exprlS) +
= .exprlG)
i .hessian[, Dose", "BMD"] <- .hessian[, "BMD", "Dose"] <- -.exprl *
= (.expr2 * (.exprlG + .exprSS * .exprlS))
CM I .grad[, V] <- .expr!4
o
.hessian
.hessian
.hessian
'A'1] <- .expr!4
"PB"] <- .hessian[, "PB", "A"] <- .expr47
"BMD"] <- .hessian[, "BMD", "A"] <- .exprSO
.grad[, "PB"] <- .expr47
.nessian[, "PB", "PB ] <- -.exprl * (.expr2 * (.exprll *
(((.exprSl + .exprSl - .expr4 * (2 * .expr2)/.expr23A2)/.expr5 +
.expr25 * .expr25/.expr5A2) * Dose * .expr9) - .exprSl *
.exprSO) - .exprSl - .exprSl)
.hessian[, "PB", "BMD"] <- .hessian[, "BMD", "PB"] <- .exprl *
(.exprSS + .expr2 * (.exprSl + .exprSS * .exprSO))
S— i .grad[, ('BMD"] <- .exprSO
0 i .hessian[, "BMD", "BMD"] <- .exprl * (.expr2 * (.exprSS +
E| .exprSS * .exprlO))
i attr(.value, "gradient") <- .grad
tt\ I attr(.value, "hessian") <- .hessian
(/) 1 , 'value
CD i
CO I
(f) = ### Above, but including derivative of dose, for the pk model
^ I ### CexpBwDS(dose, A, B, BMD, sex, fixed=NULL) allows fixed, a named list of
^ i ### vectors of fixed values for the model cexpBwo. They can differ by sex.
-*• i ### used, for example:
CO I ### nlme(model=chei ~ cexpBwos(dose, A, BMD, sex,
IT? I ### fixed=list(PB=c(F=0.05, M=0.06)), data=mydata, random=A+m~l | mrid)
Uu i ### This implementation assumes PB is the only fixed parameter, and is on
I ### its original scale (0 <= PB < 1).
CD i
^> = CexpBwDS <- function(dose, A, BMD, sex, fixed=NULL, BMR=0.10) {
call List <- vector("list",5)
names(callList) <- c("dose","A","PB","BMD","BMR")
call List[["dose"]] <- dose
call List[["A"]] <- A
callList[["PB"]] <- if ("B" %in% names(fixed)) {
B <- fixed[["B"]][as.character(sex)]
B 1/(1 + exp(-B))
IJ i } else {
£^ | fixed[["PB"]][as.character(sex)]
r* I call List[["BMD"]] <- BMD
LL 1 callList[["BMR"]] <- BMR
^*\ | do.call("CexpB2wD",call List)
2 S
"O I ### CpkexpBwD: pk combined with the exp model in terms of BMD:
(f) 1 CpkexpBwD <- function (Dose, A, B, BMD, s, D, BMR=0.10)
> 1 idose <- cpkB(Dose, s, D)
Q) i .value <- CexpBwo2(idose, A, B, BMD, BMR=BMR)
~ .grad <- array(0, c(length(.value), 5),
list(NULL,
.grad[,c("A","B","BMD")] <- attr(.value,"gradient")[,c("A","B","BMD")]
.grad[,c("S","D")] <- attr(.value,"gradient")[,"Dose"] *
attr(idose, "gradient")[,c("S", "D")]
III.B.4Page7
-------
CM
O
CD
i
C-
CD
E
CO
CO
0)
CO
CO
CO
0)
O
CL
O
"O
0)
CO
">
0
a:
I attr(. value, "gradient") <- .grad
i .value
I }
= ###: same as above, but include hessian
CpkexpBwDH <- function (Dose, A, B, BMD, S, D, BMR=0.10)
idose <- CpkBH(Dose, S, D)
.value <- cexpBwD2H (idose, A, B, BMD, BMR=BMR)
.grad <- array(0, c (length (.value) , 5),
list(NULL,
.grad [ , c("A" , "B" , "BMD")] <- attr( . val ue , "gradi ent") [ , c("A" , "B" , "BMD")]
.grad[,c("S","D")] <- attr(. value, "gradient") [, "Dose"] *'
attr(idose, "gradient") [,c("S", "D")]
.hessian <- array(0, c(length(. value) ,5,5) ,
C("A","BV'BMD","S","D"),
.hessian[,c("A","B","BMD"),c("A","B","BMD")] <-
attr(.value,"hessian")[,c("A")"B","BMD"),c("A","B","BMD")]
. hessian [,c("S","D"),c("S","D")] <-
attr(. val ue,"hessi an") [, "Dose", "Dose"] *
(attr(. idose, "gradient") [,c("S","D"),c("S","D")])A2 +
attr(. idose, "gradi ent") [, "Dose"] •*
attr ( . idose , "hessi an") [ , c("s" , "D") , c("S" , "D")]
.hessian[,"A",
attr(.value,
.hessian[,"A",
attr(.value,
.hessian[,"B",
attr(.value,
.hessian[,"B",
attr(.value,
.hessiant,"BMD
attr(.value,
.hessian[,"BMD
attr(.value,
-JG j I IC.3O I O.I I J \_ ) *-\ -J ) U J 1 *~ \
S"] <- .hessian[,"S","A"] <-
'hessi an")[,"Dose","A"] * attr(.i dose,"gradi ent")[,"S"]
'D"] <- .hessian[,"S","D"] <-
'hessian")[,"Dose","A"] * attr(.idose,"gradient")[,"D"]
'S'1] <- .hessian[,"S","B"] <-
'hessian")[,"Dose","B"] * attr(.idose,"gradient")[,"S"]
'D'1] <- .hessian[,"s","D"] <-
'hessian")[,"Dose","B"] * attr(.idose,"gradient")[,"D"]
',"5"] <- .hessian[,"s","BMD"] <-
'hessian")[,"Dose","BMD"] * attr(.idose,"gradient")[,"S"]
',"0"] <- .hessian[,"S","D"] <-
'hessi an")[,"bose","BMD"] * attr(.i dose,"gradi ent")[,"D"]
I attr(. value, "gradient") <- .grad
i attr(. value, "hessian") <- .hessian
1 .value
I }
| CpkexpB2wDH <- function (Dose, A, PB, BMD, s, D, BMR=0.10)
1 idose <- CpkBH(oose, s, D)
1 .value <- CexpBZwDH (idose, A, PB, BMD, BMR=BMR)
I .grad <- array(0, c(length(. value) , 5),
i list(NULL,
.grad[,c("A","PB","BMD")] <- attr(. value, "gradient") [,c("A", "PB","BMD")]
.grad[,c("S","D")] <- attr(. value, "gradient") [, "Dose"] *
attr(idose, "gradient") [,c("S" , "D")]
.hessian <- array(0, c(length(. value) , 5, 5) ,
list(NULL,c("A","PB","BMD","S","D"),
.hessian[,c("A","PB","BMD"),c("A","PB","BMD")] <-
attr ( .val ue , "hessian") [ , C("A" , "PB" , "BMD") , C("A" , "PB" , "BMD")]
.hessi an [,"S","S"] <-
attr(.value,"hessian")[, "Dose", "Dose"] *
attr (idose, "gradi ent") [,"S"] * attr (idose, "gradient") [, "S"] +
attr ( . val ue , rigradi ent") [ , "Dose"] *
attr(idose, *hessian")[,"S","S"]
.hessi an [,"D","D"] <-
attr(. value, "hessian") [."Dose", "Dose"] *
attr(idose, "gradient") [,"D"] * attr(idose, "gradient") [, "D"] +
attr ( .val ue , rigradi ent") [ , "Dose"] *
III.B.4Page8
-------
= attrCidose,"hessi an")[,"D","D"]
= .hessian[,"S","D"] <- .hessian[,"D","s"] <-
i attrC.value,"hessian")[,"Dose","Dose"] *
I attrCidose,"gradient")[,"s"3 * attrCidose,"gradient")[,"DM] +
I attrC.value, gradient")[,"Dose"] *
I attrCidose,shessian")[,"s","D"]
CM I .hessian[,"A","s"] <- .hessian[,"s","A"] <-
— -- attrC.value,"hessian")[,"Dose","A"] * attrCidose, "gradient") [,"s"]
.hessian[,"A","D"] <- .hessian[,"S","D"] <-
at-fi-r waiup "hoccnan'^f "n/-.co" 'A"] * attr0'dose, "gradient") [, "D"]
;PB"] * attrCidose,"gradient")[,"S"]
'PB1'1] * attrCidose,"gradient") [,"DM]
.value,"hessian")[,
.hessian[,"PB","s"] <- .hessian[,
attr(.value,"hessian")[,"Dose",
.hessian[,"PB","D"] <- .hessian[,
attr(.value,"hessian") [,"Dose",
, = .hessian[,"BMD","s"] <- .hessian[,"S'V'BMD"] <-
= attrC.value,"hessian")[,"Dose","BMD"] * attrCidose,"gradient")[,"S"]
•*± I .hessian[,"BMD","D"] <- .hessian[,"s","D"] <-
C = attrC.value,"hessian")[,"Dose","BMD"] * attrCidose,"gradient")[,"DM]
0 i
£= attrC.value, "gradient") <- .grad
I attrC.value, "hessian") <- .hessian
y) | .value
C/) [ }
CD I ### CpkexpB2wD: pk combined with the exp model in terms of BMD. Assume PB,
C/} I ### S, and D
(/) i ### are fixed, and do not return their components of the gradient.
I CpkexpBZwD <- function Cdose, A, PB, BMD, s, D, BMR=0.10)
i idose <- CpkBCdose, S, D)
CO = .value <- cexpB2woCidose, A, PB, BMD, BMR=BMR)
rC? * .grad <- arrayCO, c(lengthC.value), 2),
LL I listCNULL,
i CC"A","BMD")))
Q) i .grad[,cC"A","BMD")] <- attrC.value,"gradient")[,cC"A","BMD")]
*> I attrC.value, "gradient") <- .grad
1 .value
03 1 '
-*t = ### CpkexpBS: pk combined with exp model, with fixed B and/or S values:
| CpkexpBS <- functionCdose, A, B, m, S, D, sex, fixed=NULL) {
I call List <- vector("list",6)
_ i namesCcallList) <- cC'Dose","A","B","m","S","D")
Z5 1 call List[["Dose"]] <- dose
Oi call List[["A"]] <- A
i if ClengthCfixed) == 0) sx <- sex
** i if ClengthCfixed) == 1) sx <- D
LJL | if ClengthCfixed) == 2) sx <- S
Oi if ClengthCfixed) == 3) sx <- m
= if C"B" %in% namesCfixed)) {
_ I callList[["B"]] <- fixed[["B"]][as.characterCsx)]
vJ i callList[["m"]] <- B
01 if C"S" %in% namesCfixed)) {
tr\ I callList[["s"]] <- fixed[["S"]][as.characterCsx)]
.~ I if C"D" %in% namesCfixed)) {
> | callList[["D"]] <- fixed[["D"]][as.characterCsx)]
0 = } else {
%> I callList[["D"]] <- m
UL 1 }
s } else {
. i callList[["s"]] <- m
1 if C"D" %in% namesCfixed)) {
| callList[["D"]] <- fixed[["D"]][as.characterCsx)]
| III.B.4Page9
-------
= } else {
| callList[["D"]] <- S
I } else {
i callList[["B"]] <- B
i callList[["m"]] <- m
C\| i if ("S" %in% names(fixed)) { .
callList[["S"]] <- fixed[["S"]][as.character(sx)]
if ("D" %in% names(fixed)) {
callList[["D"]] <- fixed[["D"]][as.character(sx)]
} else {
callList[["D"]] <- S
— = }
CO I } else {
= T n . • .
callList[["s"]] <- s
if ("D" %in% names (fixed)) {
, callList[["D"]] <- fixed[["D"]][as.character(sx)]
£-. = } else {
(1) I callList[["D"]] <- D
£/) i }
y* 1 do. call("CpkexpB", call List)
CD = ### CpkexpBwDS: pk combined with exp model, with fixed B and/or S values:
(/) | CpkexpBwDS <- function(dose, A, B, BMD, s, D, sex, fixed=NULL) {
tf\ I call List <- vector ("list", 6)
<1 names(callList) <- c("Dose","A","B","BMD","s","D")
i callList[["Dose"]] <- dose
^ i call List [["A"]] <- A
-— • = if (length (fixed) == 0) sx <- sex
CO = if (length (fixed) == 1) sx <- D
'^ i if (length (fixed) == 2) sx <- S
LL = if (length (fixed) == 3) sx <- BMD
i if ("B" %in% names(fixed)) {
CD I callList[["B"]] <- fixed[["B"]][as.character(sx)]
= call List [["BMD"]] <- B
if ("S" %in% names(fixed)) {
+± 1 callList[["S"]] <- fixed[["S"]][as.character(sx)]
TO i if ("D" %in% names(fixed)) {
callList[["D"]] <- fix
else {
callList[["D"]] <- BMD
callList[["D"]] <- fixed[["D"]][as.character(sx)]
— } else {
13 I } else {
Oi callList[["s"]] <- BMD
I if ("D" %in% names (fixed)) {
^ = callList[["D"]] <- fixed[["D"]][as.character(sx)]
LL I } else {
I callList[["D"]] <- S
[ } >
"O I } else {
(D i callList[["B"]] <- B
tr\ = callList[["BMD"]] ,<- BMD
.Sir i if ("s" %in% names(fixed)) {
> = callList[["S"]] <- fixed[["S"]][as.character(sx)]
f\\ I if ("D" %in% names (fixed)) {
^ i callList[["D"]] <- fixed[["D"]][as.character(sx)]
UL I . } else {
| callList[["D"]] <- S
I } else {
1 call List[["s"]] <- S
| III. B.4 Page 10
-------
I if ("D" %in% names(fixed)) {
= callList[["D"]] <- fixed[["D"]][as.character(sx)]
= } else {
| callList[["D"]] <- D
| do.call("CpkexpBwD",call List)
| ### CpkexpB2wDS: pk combined with exp model, with fixed PB, S, and D values:
I CpkexpB2wDS <- function(dose, A, BMD, sex, fixed=NULL) {
i calI List <- vector("list",7)
names(callList) <- c("dose","A","PB", "BMD","s","D","BMR")
= cal1 Li st
= call List
_ I call List
t- i call List
Q) = call List
E= call List
i call List
"dose"]] <- dose
•A"]] <- A
'PB"]] <- fixed[["PB"]][as.character(sex)]
] <- fi
]} <- B
'BMD"]] <- BMD
'S"]] <- fixed[["s"]][as.character(sex)]
'D"]] <- fixed[["D"]][as.character(sex)]
•BMR"]] <- o.l
1 do. cal K"CpkexpB2wD", call List)
0) 1 }
(/) i ### Compute log(BMD) (for a BMR * 100% decrease in the mean) for the
(/) i ### exponential
<1 ### model, based on the parameter transformations used here.
| ### This returns the gradient of log(BMD), to use in computing standard
I ### errors.
(/) I ### log(BMD) <- expression(log(-log(l - BMR*(1 + exp(B)))) - m)
'£"} | ### in terms of the transformation used in the cexpB models.
LL, i ### fixed is a vector of strings, listing the parameters that should not
= ### be in the gradient.
(D i
> | cexplBMD <- function (BMR, m, B, fixed=NULL)
-*± I .exprl <- exp(B)
_VV) 1 .expr4 <- 1 - BMR * (1 + .exprl)
— « I .exprS <- Iog(.expr4)
— J i .value <- log(-.exprS) - m
C- I .grad <- array(0, c(length(. value) , 2), list(NULL, c("m",
^ I "B")))
13 i .grad[, "m"] <- -1
OI .grad[, "B"] <- -BMR * . exprl/. expr4/.expr5
I if (Ms. null (fixed) > 0) {
j-% | .grad <- .grad[,-match(fixed,colnames(.grad)) ,drop=FALSE]
OI attr(. value, "gradient") <- .grad
1 .value
•o 1>
0 = ### Essentially the same function, but takes as input our model object
tf\ i ### (xx) and returns a list with elements IBMD and lBMD.se,
.Zl | ### each with components for "F" and "M".
* explBMD <- function(object, BMR) {
i fit <- object$Fitm
| if (inherits(fit, "nlme")) {
= Sigma <- fit$varFix
1 Coefs <- fit$coefficients$fixed
1 } else {
I III. B.4 Page 11
-------
i Sigma <- fit$varBeta
= coefs <- fit$coefficients
I }
= ### Fixed gives the fixed parameters
1 tmp <- names(fit$can$model [[3]])
= Fixed <- NULL
I if ("fixed" %in% tmp) {
C\i i tmp <- names(fit$can$model[[3]] [["fixed"]])
= Fixed <- c(Fixed,cC"m",11B")[c("m")"B") %in% tmp])
I m.F <- if ("m" %in% Fixed)
1 fit$call$model[[3]][["fixed"]][["m"]][["F"]]
i else
I if ("m.sexF" %in% names(Coefs)) Coefs["m.sexF"] else Coefs["m"]
= m.M <- if ("m" %in% Fixed)
I fit$call$model[[3]][["fixed"]][["m"]][["M"]]
= else
= if ("m.sexM" %in%. names(coefs)) Coefs["m.sexM"] else Coefs["m"]
0 I B.F <- if ("B" %in% Fixed)
r- I fit$call$model[[3]][["fixed"]][["B"]][["F"]]
C. i else
fA i if ("B.sexF" %in% names(Coefs)) Coefs["B.sexF"] else Coefs["B"]
ft( i B.M <- if ("B" %in% Fixed)
%* I fit$call$model[[3]][["fixed"]][["B"]][["M"]]
OP I else
(/) i if ("B.sexM" %in% names(Coefs)) Coefs["B.sexM"] else Coefs["B"]
(/) I ### Females
i lBMD.se = c(F=sqrt(t(grad.F) %*% Sigma.F %*% grad.F),
fl) | M=sqrt(t(grad.M) %*% Sigma.M %*% grad.M)))
I ### Compute BMD (for a BMR * 100% decrease in the mean) for the
i ### exponential
I ### model, based on the parameter transformations used here.
i ### This returns the gradient of BMD, to use in computing standard
[ III.B.4 Page 12
-------
2 ### errors.
= ###
I ### BMD <- expression(-loq(l - BMR*(1 + exp(B)))*exp( - m ))
| ### in terms of the transformation used in the CexpB models.
= ### fixed is a vector of strings, listing the parameters that should not
\ ### be in the gradient.
| CexpBMD <- function (BMR, m, B, fixed)
i .exprl <- exp(B)
| .expr4 <- 1 - BMR * (1 + .exprl)
I .exprS <- Iog(.expr4)
| .exprS <- exp(-m)
I .value <- -.exprS * .exprS
= .grad <- array(0, c(length(.value), Z), list(NULL, c("m",
I "B")))
= .grad[, "m"] <- .exprS * .exprS
I .grad[, "B"] <- BMR * .exprl/.expr4 * .exprS
v- | if (Ms.null (fixed) > 0) {
ft) i .grad <- .grad[,-match(fixed,col names(.grad)),drop=FALSE]
El }
i attr(.value, "gradient") <- .grad
(A | .value
0 | ### Compute IBMD and its standard error for pkexpBS
C/5 | ### This calculates the value for both sexes at the same time
<| pkexpSlBMD.se <- function(object.BMR) {
i fitpk <- object$Fitpk
^ I if (inherits(fitpk, "nltne")) {
-*• \ Sigma <- fitpkSvarFix
(/) = Coefs <- fitpk$coefficients$fixed
" I } else {
i Sigma <- fitpk$varBeta
I Coefs <- fitpkScoefficients
i ### Fixedl gives the fixed parameters for the CexpBMD part (i.e., 'B')
_. | ### FixedZ gives the fixed parameters for the CpkBMD part (i.e., 'S')
IHJ I tmp <- names(fitpk$call$model[[3]])
i Fixedl <- FixedZ <- NULL
I if ("fixed" %in% tmp) {
i tmp <- names(fitpk$call$model[[3]][["fixed"]])
I Fixedl <- c(Fixedl,c("m","B")[c("m","B") %in% tmp])
| FixedZ <- c(FixedZ,c("S","D")[c("S","D") %in% tmp])
CL I S.F <- if ("S" %in% FixedZ)
Ol fitpk$call$model[[3]][["fixed"]][["s"]] [["F"]]
= else
I if ("S.sexF" %in% names(Coefs)) Coefs["S.sexF"] else Coefs["S"]
I S.M <- if ("S" %in% FixedZ)
0 I fitpk$call$model[[3]][["fixed"]][["S"]][["M"]]
(/) I else
~ I if ("S.sexM" %in% names(coefs)) Coefs["S.sexM"] else Coefs["S"]
m I D.F <- if ("D" %in% FixedZ)
I fitpk$call$model[[3]][["fixed"]][["D"]][["F"]]
i else
I if ("D.sexF" %in% names(Coefs)) Coefs["D.sexF"] else Coefs["D"]
i D.M <- if ("D" %in% FixedZ)
I fi tpkScal1$model[[3]][["fi xed"]][["D"]][["M"]]
i else
I III.B.4 Page 13
-------
I if ("D.sexM" %in% names(Coefs)) Coefs["D.sexM"] else coefs["D"]
I m.F <- if ("m" %in% Fixedl)
| fitpk$can$model[[3]][["fixed"]][["m"]][["F"]]
i else
i if ("m.sexF" %in% names(Coefs)) coefs["m.sexF"] else Coefs["m"]
i m.M <- if ("m" %in% Fixedl)
CM I fitpk$call$model[[3]]C["fixed"]][["m"]][["M"]]
i else
= if ("m.sexM" %in% names(Coefs)) Coefs["m.sexM"] else Coefs["m"]
I B.F <- if ("B" %in% Fixedl)
i fitpk$call$model[[3]][["fixed"]][["B"]][["F"]]
* else
i if ("B.sexF" %in% names(Coefs)) Coefs["B.sexF"] else Coefs["B"]
I B.M <- if ("B" %in% Fixedl)
| fitpk$call$model[[3]][["fixed"]][["B"]][["M"]]
I else
L. = if ("B.sexM" %in% names(Coefs)) Coefs["B.sexM"] else Coefs["B"]
CD i
l ### Females
i idose.F <- cexpBMD(BMR,m.F,B.F,fixed=Fixedl)
f/j i IBMD.F' <- CpkBi(as.vector(idose.F),s.F,D.F,fixed=Fixed2) +
fft i log(object$Dosescale)
0 I ### Males
C/) i idose.M <- cexpBMD(BMR,m.M,B.M,fixed=Fixedl)
.(/) | IBMD.M <- CpkBi(as.vector(idose.M),S.M,D.M,fixed=Fixed2) +
| log(object$Dosescale)
i ## Get the standard errors
•*: i ##1) Get the 'Female' and 'Male' Sigmas
:*? = "indx <~ match (c("m.sexF-","B. sexF","s ,"D") , rownames (Sigma) ,nomatch=0)
i Sigma.F <- sigma[indx,indx,drop=FALSE]
| ## Strip off the '.F' from the rownames and col names
i rownames(Sigma.F) <- col names(Sigma.F) <-
{ sub("\\.sexF","",rownames(Sigma.F))
I indx <- match(c("m.sexM","B.sexM","S","D"),rownames(Sigma),nomatch=0)
| Sigma.M <- Sigma[indx,indx,drop=FALSE]
I ## Strip off the '.M' from the rownames and col names
= rownames(Sigma.M) <- col names(Sigma.M) <-
| sub("\\.sexM","",rownames(Sigma.M))
f\ I ## 2) Get the 'Female' and 'Male' gradients
n I grad.F <- c(attr(idose.F,"gradient")*attr(lBMD.F,"gradient")[,"idose"],
UL = attr(lBMD.F,"gradient")[,-match("idose ,
Oi col names(attr(lBMD.F,
I "gradient"))),
— I drop=FALSE])
v-' = nm <- c(colnames(attr(idose. F, "gradient")) ,
0 I colnames(attr(lBMD.F,"gradient")))
(/) | names(grad.F) <- nm[-match("idose",nm)]
2> i ## Make sure the elements are in the right order for sigma.F
| grad.F <- grad.F[rownames(Sigma.F)]
1 grad.M <- c(attr(idose.M,"gradient")*attr(lBMp.M,"gradient")[,"idose"],
i attr(1BMD.M,"gradient")[,-match("idose ,
i colnames(attr(lBMD.M,
| "gradient"))),
[ . III.B.4 Page 14
-------
I - drop=FALSE])
i nm <- c(colnames(attr(idose.M,"gradient )),
I colnames(attr(lBMD.M,"gradient")))
i names(grad.M) <- nm[-match("idose",nm)]
I ##Make sure the elements are in the right order for Sigma.M
I grad.M <- grad.M[rownames(Sigma.M)]
CM I
i list(lBMD=c(F=as.vector(lBMD.F),M=as.vector(lBMD.M)),
i lBMD.se=c(F = sqrt(t(grad.F) %*% Sigma.F %*% grad.F),
i M = sqrt(t(grad.M) %*% Sigma.M %*% grad.M)))
CD
I ### Compute IBMD and its standard error for pkexpBwDS
L. i ### This calculates the value for both sexes at the same time
CD i
E= pkexpwDSlBMD.se <- function(object,BMR) {
I fitpk <- object$Fitpk
//\ | if (inherits (fitpk, "nlme")) {
ff( = Sigma <- fitpk$varFix
~£ i coefs <- fitpk$coefficients$fixed
CD i } else {
(/) = Sigma <- fitpk$varBeta
(/) i Coefs <- fitpkScoefficients
1 }
i ### Fixedl gives the fixed parameters for the cexpBMD part (i.e., 'B')
| ### FixedZ gives the fixed parameters for the CpkBMD part (i.e., 'S')
= Fixedl <- c(Fixedl,c("BMD","B")[c("BMD","B") %in% tmp])
! FixedZ <- c(Fixed2,c("S","D")[c("S","D") %in% tmp])
Extract the parameter estimates
.F <- if ("Sn %in% Fixed2)
Cu = ##
S.
fitpk$call$model[[3]][["fixed"]][["S"]][["F"]]
if ("S.sexF" %in% names(Coefs)) Coefs ["S.sexF"] else Coefs["S"]
ZJ I S.M <- if ("S" %in% FixedZ)
Ol fitpk$call$model [[3]] [["fixed"]] [["S"]] [["M"]]
i else
~ I if ("S.sexM" %in% names(Coefs)) Coefs ["S.sexM"] else Coefs ["S"]
Oi D.F <- if ("D" %in% FixedZ)
I fitpk$call$model [[3]] [["fixed"]] [["D"]] [["F"]]
i else
TJ i if ("D.sexF" %in% names(Coefs)) Coefs ["D.sexF"] else Coefs ["D"]
0 I D.M <- if ("D" %in% FixedZ)
(A i fitpk$call$model[[3]][["fixed"]][["D"]][["M"]]
— i else
> | if ("D.sexM" %in% names (Coefs)) Coefs ["D.sexM"] else Coefs ["D"]
r?s ^ BMD.F <- if ("BMD" %i n% Fixedl)
Q_ | fitpk$call$model[[3]][["fixed"]][["BMD"]][["F"]]
\ else
i if ("BMD.sexF" %in% names(coefs)) coefs ["BMD. sexF"] else coefs["BMD"]
2 BMD.M <- if ("BMD" %i n% Fixedl)
| f i tpkScal 1 Smodel [ [3] ] [ ["f i xed"] ] [ ["BMD"] ] [ ["M"] ]
I III. B.4 Page 15
-------
i else
I if ("BMD.sexM" %in% names(Coefs)) Coefs["BMD.sexM"] else Coefs["BMD"]
| ### Calculate the log BMDs
| ## Females
i idose.F <- exp(BMD.F)
CM I IBMD.F <- CpkBi(as.vector(idose.F),s.F,D.F,fixed=Fixed2) +
| log(object$Dosescale)
I ### Males
= idose.M <- exp(BMD.M)
| IBMD.M <- CpkBi(as.vector(idose.M),S.M,D.M,fixed=Fixed2) +
| log(object$Dosescale)
i I ### calculate the standard errors
I ## 1) Get the 'Female' and 'Male' Sigmas
~J± i indx <- match(c("BMD.sexF","S","D").rownames(sigma),nomatch=0)
S— I Sigma.F <- sigma[indx,indx,drop=FALSE]
CD i
El ## Strip off the '.F' from the rownames and colnames
i rownames(Sigma.F) <- col names(Sigma.F) <-
(/) I sub("\\.sexF","",rownames(Sigma.F))
^ = indx <- match(c("BMD.sexM","S","D"),rownames(sigma),nomatch=0)
w i Sigma.M <- sigma[indx,indx,drop=FALSE]
CO I
(/)=## Strip off the '.M' from the rownames and colnames
= drop=FALSE])
! nm <- c("BMD",
I col names(attr(lBMD.F,"gradient")))
i names(grad.F) <- nm[-match("idose",nm)]
| ## Make sure the elements are in the right order for Sigma.F
i grad.F <- grad.F[rownames(Sigma.F)]
-J I
Oi grad.M <- c(idose.M*attr(lBMD.M,"gradient")[,"idose"],
I attr(1BMD.M,"gradient )[,-match("idose",
r* 1 colnames(attr(lBMD.M,
LL, I "gradient"))),
OI drop=FALSE])
i nm <- c("BMD",
__ = colnames(attr(lBMD.M,"gradient")))
**3 | names(grad.M) <- nm[-match("idose",nm)]
CP I ##Make sure the elements are in the right order for Sigma.M
.«.— | grad.M <- grad.M[rownames(Sigma.M)]
J^ i
fl) i list(lBMD=c(F=as.vector(lBMD.F),M=as.vector(lBMD.M)),
fZZ i lBMD.se=c(F = sqrt(t(grad.F) %*% Sigma.F %*% grad.F),
LJL | M = sqrt(t(grad.M) %*% Sigma.M %*% grad.M)))
I ### CpkB: a low-dose modification to simulate the effect of first-pass
i ### metabolism on the relationship between administered dose and
j III.B.4 Page 16
-------
i ### internal dose.
I ### Model expression is:
I ### ~0.5*(dose - exp(S) - exp(D) + sqrt((dose - exp(S) - exp(D))A2 +
| 4*dose*exp(S)))
i ### This maps administered dose (dose) to internal dose (CpkB)
i "CpkB" <-
= function (dose, S, D)
CNI I {
i .exprl <- exp(S)
i .expr3 <- exp(D)
i .expr4 <- dose - .exprl - .exprS
i .expr7 <- 4 * dose * .exprl
i .exprS <- .expr4A2 + .expr?
I .exprlS <- .expr8A-0.5
= .value <- 0.5 * (.expr4 + sqrt(.exprS))
| .grad <- array(0, c(length(.value), 2), list(NULL, c("S",
! .grad[, "S"] <- 0.5 * (0.5 * ((.expr? - 2 * (.exprl * .expr4)) *
v—. i .exprlS) - .exprl)
0 I .grad[, "D"] <- -0.5 * (0.5 * (2 * (.exprS * .expr4) * .exprlS) +
El .exprS)
i attr(.value, "gradient") <- .grad
y) I .value
CO I }
CD = ### CpkBH: just like CpkB, but also returns the hessian
CO 1 CpkBH <- function (dose, S, D)
<| .exprl <- exp(S)
1 .expr3 <- exp(D)
^ | .expr4 <- dose - .exprl - .exprS
•~- | .expr7 <- 4 * dose * .exprl
CO . 1 .exprS <- .expr4A2 + .exprZ
"JTZ 1 .expr!2 <- .exprl * .expr4
UL, i .expr!4 <- .expr? - 2 * .expr!2
I .exprlS <- .expr8A-0.5
0 | .expr25 <- .expr8A-l.5
"^ I .expr36 <- .expr3 * .expr4
= .expr37 <- 2 * .expr36
\ .expr39 <- -0.5 * (.expr37 * .expr25)
= .value <- 0.5 * (.expr4 + sqrt(.exprS))
I .grad <- array(0, c(length(.value), 2), list(NULL, c("S",
= "D")))
I .hessian <- array(0, c(length(.value), 2, 2), list(NULL,
| C(«S" "D"), C("S--F »[,»)))
13 i .grad[, "S"] <- 0.5 * (0.5 * (.expr!4 * .exprlS) - .exprl)
Ol .hessian[, "S", "S"] <- 0.5 * (0.5 * ((-expr7 - 2 * (.expr!2 -
= .exprl * .exprl)) * .exprlS + .expr!4 * (-0.5 * (.expr14 *
n | .expr25))) - .exprl)
LL | .hessian[, "S", "D"] <- .hessian[, "D", "S"] <- 0.5 * (0.5 *
O= (2 * (.exprl * .expr3) * .exprlS - ,expr!4 * .expr39))
I .grad[, "D"] <- -0.5 * (0.5 * (.expr37 * .exprlS) + .expr3)
__ I .hessian[, "D", "D"] <- -0.5 * (0.5 * (2 * (.expr36 - .expr3 *
U | .expr3) * .exprlS - .expr37 * .expr39) + .expr3)
0 = attr(.value, "gradient") <- .grad
(/) i attr(.value, "hessian") <- .hessian
= .value
> | }
I ### and this is the inverse function: maps internal dose to the
| corresponding
1 ### administered dose.
= ### CpkBi <- expression((idoseA2 + idose*(exp(S) + exp(D)))/(idose +
I exp(S)))
| ###
\ - III.B.4 Page 17
-------
: I ### since I want to give log(BMD), use:
I ### CpkBi <- expression(log(idoseA2 + idose*(exp(S) + exp(D))) - log(idose +
I exp(s)))
= ### fixed: a vector of strings of parameters that will not be included in
I ### the gradient
i CpkBi <- function (idose, S, D, fixed=NULL)
C\| I .expr2 <- exp(S)
| .expr3 <- exp(D)
I .expr4 <- .expr2 + .exprS
i .exprG <- idoseA2 + idose * .expr4
^ 1 .exprS <- idose + .expr2
3 1 .value <- log(.exprG) - log(.exprS)
,,"^; \ .grad <- array(0, c(length(.value), 3), list(NULL, c("idose",
CO = "S", "D")))
I i .grad[, "idose"] <- (2 * idose + .expr4)/.exprG - l/.expr8
| .grad[, "S"] <- idose * .expr2/.exprG - .expr2/.expr8
•*±I I .grad[, "D"] <- idose * .expr3/.exprG
L» | if (lis.null(fixed))
(p 1 .grad <- .grad[,-match(fixed,colnames(.grad)),drop=FALSE]
f- i attr(.value, "gradient") <- .grad
C I .value
tA I "mypkPredict" <- function (object, newdata, level) {
•ff | fparms <- object$coefficients$fixed
rparms <- object$coefficients$random
1/3 = B.est <- !("fixed" %in% names(object$call$model
•7) = !("B" %in% names(object$call$model[[3"
„*• i D.est <- !("fixed" %in% names(object$call$model
"i, s ! ("p" %in% names(object$call$model [[3"
,. i S.est <- !("fixed" %in% names(object$call$model
!("S" %in% names(object$call$model[[3'
])) II
"fixed"]]))
])) II
"fixed"]]))
])) II
•fixed"]]))
CD = B.sex <- if (B.est) "B.sexF" %in% names(fparms) else NA
i»*3 I indx <- paste("A.s.U", as.character(newdataSs.u), sep = "")
LJL i A <- fparms [indx]
I B <- if (B.est) {
0 | indx <- if (B.sex) {
*> i paste("B.", "sex", as.character(newdata$sex), sep = "")
j= i > else t
"l_- i rep("B",nrow(newdata))
"^ I fparms [indx]
"Z. ^ } else {
— i eval(object$call$model[[3]][["fixed"]])$B[as.character(newdata$sex)]
s>— | } •
13 = indx'<- paste("BMD.", "sex", as.character(newdata$sex),
O i ?eP = '"">
x«/ -
<- fparms[indx]
*-» 5 ### sis always fixed in my application
SJL I s <- if (S.est) {
O= rep(fparms["S"],nrow(newdata))
1 } else {
eval(object$cal1$model[[3]][["fixed"]][["S"]])[as.character(newdata$sex)]
~~:- I ### for simplicity, I'm assuming what we actually did, that is, model D ~
> i ### but fixed=list(D=c(F=,M=))
fl) i D <- if (D.est) {
= rep(fparms["D"],nrow(newdata))
| } else {
I eval(object$call$model[[3]][["fixed"]][["D"]])[as.character(newdata$sex)]
| if, (level >= 1) {
| ' III.B.4 Page 18
-------
I grpname <- names(rparms) [1]
i if ("A. (intercept) %in% co~lnames(rparms[[l]]))
I A <- A + rparms[[l]] [as. character (newdata[, grpname] ),
i "A. (intercept)"]
I if (B.est) {
i B <- B + if (B.sex) {
i rparms[[l]] [as. character(newdata[, grpname]) , "B. (Intercept)"]
C\j I } else {
| rparms[[l]] [as. character(newdata[, grpname]) , "B"]
| *
i BMD <- BMD + rparms[[l]] [as. character(newdata[, grpname]) ,
I "BMD. (intercept)"]
1 if (level == 2) {
I indx <- paste(as.character(newdata$mrid) , as.character(newdata$set) ,
I sep = "/")
L- = if ("A. (intercept)" %in% colnames(rparms[[2]]))
fl) = A <- A + rparms[[2]] [indx, "A. (intercept)"]
E = if (B.est)
i B <- B + if (B.sex) {
tt\ = rparms[[2]] [indx, "B. (intercept)"]
fA 'l > else *
W I rparms[[2]][indx, "B"]
(/) I BMD <- BMD + rparms[[2]] [indx, "BMD. (intercept)"]
(/) i >
| CpkexpBwD(newdata$Dose. scaled, A, B, BMD, S, D)
I ### The following function is not general, but applies only to the model
I ### as used in the analysis of the OP data
I mypkPredict2 <- function (object, newdata, level) {
i fparms <- object$coefficients$fixed
i rparms <- obnect$coefficients$random
CD I
•> I ## Fixed Effects
*Z~ i indx <- paste("A.s.M.t", as.character(newdata$s.M.t), sep = "")
"*± I A <- fparms [indx]
"mj I PB <-
-£ i eval (object$call$model [[3]] [["fixed"]])$PB[as.character(newdata$sex)]
~ i indx <- paste("BMD.", "sex", as. character (newdata$sex) ,
13 i sep = "")
O| BMD <- fparms [indx]
I ### S is always fixed in my application
n I s <-
UL 1 eval (object$call$model[ [3] ] [["fixed"]] [["S"]]) [as. character (newdataSsex)]
i ### for simplicity, I'm assuming what we actually did, that is, model D ~ 1
__ I ### but fixed=list(D=c(F=,M=))
w = D <-
0 I eval (object$cal 1 $model [ [3] ] [ ["f i xed"] ] [ ["D"] ] ) [as . character (newdataSsex)]
(/) 1 if (level >= 1) {
-- i grpname <- names(rparms) [1]
> | BMD <- BMD + rparms[[l]] [as. character(newdata[, grpname]),
0 | "BMD. (intercept)"]
I if (level == 2) {
= indx <- paste (as. character (newdataSmri d) , as. character (newdataSset) ,
i sep = V")
= BMD <- BMD + rparms [[2]] [indx, "BMD. (intercept)"]
I III. B.4 Page 19
-------
I CpkexpB2wD(newdata$Dose.scaled, A, PB, BMD, S, D)
= myPredict <- function(object, newdata,level) {
I fparms <- object$coefficients$fixed
I rparms <- object$coefficients$random
i ## Did we estimate Bs?
I B.est <- !("fixed" %in% names(object$call$model[[3]]))
C\! I ## Did tne Bs differ between sexes"
Q i B.sex <- if (B.est) "B.sexM" %in% names(fparms) else NA
irr i B.REname <- if (B.est) {
-_ I if (B.sex) "B.(intercept)" else "B"
• i } else NA
y~ i ## is this 'm' or 'BMD1?
-J5: i is.m <- "m.sexF" %in% names(fparms)
• i ## rake care of the fixed effects
| indx <- paste("A.","s.U",as.character(newdata$s.u),sep="")
**Zl i A <- fparms [indx]
L. i if (B.est) {
Q) i B <- if (B.sex) {
E= indx <- paste("B.","sex",as.character(newdata$sex),sep="")
i fparms[indx]
GO i } else {
tii \ rep(fparms["B"] ,nrow(newdata))
0 I } else {
CO | B <- object$call$model[[3]]["fixed"][[l]]$B[as.character(newdata$sex)]
I if (is.m) {
i indx <- paste("m.","sex",as.character(newdata$sex),sep="")
i thirdcol <- "m. (Intercept)"
I > else {
v/3 = indx <- paste("BMD.sex",as.character(newdata$sex),sep="")
| thirdcol <- "BMD. (intercept)"
i m <- fparms[indx]
i if (level >= 1) {
i qrpname <- names(rparms)[1]
| if ("A. (intercept)" %in% col names (rparms [[!]]))
= A <- A + rparms[[l]][as.character(newdata[,grpname]),"A.(intercept)"]
1 if (B.est)
i B <- B + rparms[[l]][as.character(newdata[,grpname]).B.REname]
| m <- m + rparms[[l]][as.character(newdata[,grpname]).thirdcol]
I if (level == 2) {
Z3 i indx <- paste(as.character(newdataSmrid),
Of as.character(newdata$set),sep="/")
i if ("A.(intercept)" %in% colnames(rparms[[2]]))
r* I A <- A + rparms[[2]][indx,"A.(Intercept)"]
LJL I if (B.est)
Oi B <- B + rparms[[2]][indx,B.REname]
| m <- m + rparms[[2]][indx,thirdcol]
"O 1 if (is.m) {
CD i cexpB(newdata$Dose.scaled,A,B,m)
(/) | } else {
"— i cexpBwD(newdata$Dose.scaled,A,B,m)
> I }
0 I ,
= •»
i myPredict2 <- function(object, newdata,level) {
I fparms <- object$coefficients$fixed
| rparms <- object$coefficients$random
| ## Take care of the fixed effects
[ III.B.4 Page 20
-------
CM
O
CD
i
C
0)
E
c/)
indx <- paste("A.s.M.t",as.character(newdata$s.M.t),sep="")
A <- fparms[indx]
B <- objectScal1Smodel[[3]]["fixed"][[!]][[1]][as.character(newdata$sex)]
indx <- paste("BMD.sex",as.character(newdata$sex),sep="")
thirdcol <- "BMD.(Intercept)"
m <- fparms[indx]
if (level >= 1) {
grpname <- names(rparms)[1]
if ("A.(intercept)r %in% col names(rparms[[!]]))
A <- A + rparms[[l]][as.character(newdata[,grpname]),"A.(intercept)"]
m <- m + rparms[[!]][as.character(newdata[,grpname]).thirdcol]
if (level == 2) {
indx <-
paste(as.character(newdata$mrid),as.character(newdata$set),sep="/")
if ("A:(Intercept)" %in% col names(rparms[[2]]))
A <- A + rparms[[2]][indx,"A.(intercept)"]
m <- m + rparms[[2]][indx,thirdcol]
if ("B" %in% names(object$caH$model [[3]] ["fixed"] [[!]]))
CexpBwD(newdata$Dose.sealed,A,B,m)
else
~' = cexpB2wD(newdata$Dose.scaled, A, B, m)
C/) I Expr <- expression(exp(A)*(l/(l + exp(-B)) + (exp(-B)/(l +
(/) i exp(-B)))*exp(log((l - BMR - B)/(l-B))*Dose/BMD)))
### Function to produce a phony dataset for input into gls and gnls
### Dose, N, M, and SD can be vectors of the same lengtn
(/) I ### DoseName and RespName are strings that give the names for the
1TJ I ### corresponding
LL. I ### variables
(1) i PhonyDF <- function(Dose,N,M,SD,DoseName,RespName,chem=NULL,chemName=NULL){
^> I tmp <- data.frame(rep(Dose,N),
.— = qlnorm01(N,M,SD))
**lf I names(tmp) <- c(DoseName,RespName)
U5 | if (Ms.null (ChemName)) tmp[,ChemName] <- factor(rep(chem,N))
—* | tmp
C I ### Possible values are "Best","Biggest","Both"."None"
*r i options(BMDSplot="Best")
Oi assign("%inint%",function(x,interval) (min(interval) <= x && max(interval)
I >= x5)
** = plot.BMDS <- function(x, which=getOption("BMDSplot"),
LL = Logx=c("auto","log","linear"),
OI ...) {
i dots <- list(...)
__ I reduceddots <- dots
*-> 1 if ((indx <- match("ylim",names(reduceddots),nomatch=0)) > 0)
Q) i reduceddots <- reduceddots[-indx]
(/) I if ((indx <- match("xlim".names(reduceddots),nomatch=0)) > 0)
"— i reduceddots <- reduceddots [-indx]
> = if ((indx <- match("Title",names(reduceddots),nomatch=0)) > 0)
fl^ i reduceddots <- reduceddots[-indx]
i switch(which,
i Best={
= Agg <- !is.null(X$Data[,x$varNames["SS"]])
| if (!Agg) {
i ## if there is replication, make an aggregated dataset
| ## and set MyAgg to be true
1 III.B.4 Page 21
-------
I tmp <- rle(sort(x$Data[,x$varNames["Dose"]]))$lengths
1 MyAgg <- sum(tmp > 3) >= length(tmp) - 2
1 if-(MyAgg) {
| indx <- orcier(x$Data[,x$varNames["Dose"]])
i dose <- unique(x$Data[indx,X$varNames["Dose"]])
i resp <- tapp1y(x$Data[indx,x$varNames["Resp"]],
I factor(x$Data[indx,x$varNames["Dose"]]),
CXI * mean)
I sd <- tapply(X$Data[indx,x$varNames["Resp"]],
I factor(x$Data[indx,x$varNames["Dose"]]),
i function(x) sqrt(var(x)))
I MyData <- data.frame(dose,resp,sd,tmp)
i names(MyData) <- x$varNames[c("Dose","Resp","SD","SS")]
= } else MyData <- x$Data
= } else MyData <- x$Data
I LogX <- match.arg(LogX)
i if (LogX == "auto") {
1 if (Agg I I MyAgg) {
L- i dose <- sort(MyData[,x$varNames["Dose"]])
<1) I nd <- length(dose)
r~ i Logx <- if ((dose[nd] - dose[l])/(dose[2] - dose[l]) > 20)
C i Tog" else "linear"
f/j I } else {
ff* \ ## This is a copout for epi data, but works for now
[fr I Logx <- "linearr'
co i >}
{/) | drange <- range(c(MyData[,X$varNames["Dose"]],X$BMD))
0]
*s I if (!is.null(x$Fit)) {
—^ i newdata <- data.frame(Dose=doses/x$DoseScale)
C/3 i names(newdata) <- x$varNames["Dose"]
IT"? = predcrv <-predict(X$Fit,newdata=newdata)*X$RespScale
LL i if (!is.null(X$RR)) {
\ newdata <- data.frame(c(0,x$BMD/x$DoseScale))
I]} i names(newdata) <- x$varNames["Dose"]
"> = critresp <- predict(X$Fit,newdata=newdata)*x$RespScale
I BMR <- critresp[l]*(l - x$RR)
i else {
| critresp <- BMR <- NULL
= }
I } else predcrv <- critresp <- BMR <- NULL
«- = if (Agg I I MyAgg) {
.«J i mn <- MyDatal,x$varNames[ Resp ]]
Oi ss <- MyData[,x$varNames["SS"]]
i sd <- MyData[,x$varNames["SD"]]
-i i mnlcl <- mn - qt(0.975,ss - l)*sd/sqrt(ss)
LL | mnucl <- mn + qt(0.975,ss - l)*sd/sqrt(ss)
i mnlcl <- mnucl <- MyData[,x$varNames["Resp"]]
| ylim <- if (is.null(dots$ylim))
Q} i range(c(predcrv,mnlcl,mnucl,BMR))
(/) | else dots$ylim
-— i xlim <- if (is.null (dots$xlim)) drange else dots$xlim
*> i Title <- if (is.null(dots$Titie)) X$RunName else dots$Title
0) i if (Agg 11 MyAgg)
oJJ i do.call("plotCl",c(list(MyData[,x$varNames["Dose"]],
LL i Mypata[,x$varNames["Resp"]],
| aui=mnucl,ali=mnlcl,err="y",
i yli m=yli m,xli m=xli m,
i xlab=x$varNames["Dose"],
| ylab=X$varNames["Resp"]),
j III.B.4 Page 22
-------
CM
O
CD
i
reduceddots))
else
do.cal1("plot",c(list(MyData[,X$varNames["Dose"]],
MyData[,X$varNames["Resp"]],
yIi m=yli m,xli m=xli m,
xlab=x$varNames["Dose"],
ylab=x$varNames["Resp"]),
reduceddots))
if (MyAgg) points(x$Data[,x$varNames["Dose"]],
x$Data[,x$varNames["Resp"]])
if (!is.null(x$Fit)) lines(doses,predcrv,col="green")
### } else {
### }
if (!is.null(x$RR) && Ms.null (X$Fit)) {
segments(par("usr")[1],BMR,x$BMD,BMR,col="red")
segments(x$BMD,par("usr")[3],x$BMD,BMR,col="red")
segments(x$BMDL,par("usr")[3],x$BMDL,BMR,col="red")
C 1 ti.tle(main=Title,sub=X$ModelName)
0 i },
E| Biggest={
i if (is.null(x$FitAllDoses)) plot(x,which="Best",...)
ffl | else plot(x$FitAllDoses,which="Best",...)
tr\ \ }>
;:; i Both={
0 = plot(x,which="Best",...)
C/) i if (!is.null(X$FitAllDoses)) plot(X$FitAllDoses,which="Best",...)
CO I >• r
<= None={
i invisible(NULL)
^ h "
CO i ### Function to generate N approximately lognormal quantiles with exactly
JT"J! | ### mean M and sd SD. Does reasonable things if N,M,and SD are vectors
LL. i ### that have the same length, and exits otherwise, if N[i] == 1,
§ ### then just returns the value of M[i]. If SD[i] == 0, returns
Q) i ### rep(M[i] ,i
I qlnormOl <- function(N,M,SD) {
I if((lenqth(N) != length(M) || (length(N) != length(SD))))
i stop( N, M, and SD must have the same length")
i out <- NULL
1 for (i in l:length(N)) {
I if (N[i] > 1) {
I if CSD[I] > 0) {
13 i sdlog <- sqrt(l9g(SD[i]A2/M[i]A2 + 1))
Oi mnlog <- log(M[i]) - sdlogA2/2
i y <- qlnorm(ppoints(N[i]),mnlog,sdlog)
o I sd <- sqrt(var(y))
LL. = mn <- mean(y)
OI . out <- c(out,SD[i]*(y - mn)/sd + M[i])
i } else {
^-, | out <- c(out,rep(M[i],N[i]))
01 } else out <- c(out, M)
CO I >
-— = out
> I }
0 I
CC i
I.B.4 Page 23
-------
b. Part 2: Modifications to the package nlme.
CM
CD
i
C
0
E
CO
CO
0
CO
CO
CO
0
CO
"b
E
ZJ
O
CL
O
"O
0
CO
HI m ium! in
>
0
Two small modifications to the package nlme (version 3.1-24
was used here) were required to facilitate convergence of
the models.
i R/reStruct.R
1 300c300
i < PACKAGE = "n~lme")$log"lik
I > PACKAGE = "nlme",NAOK=TRUE)$loglik
i src/nlme.c
| 292C292
= <
i >
-1. /*d~lt*/, pow(epsm, 1.0/3.0) /*gradtl*/, 0. /*stepmx*/,
-1. /*dlt*/, pow(epsm, 1.0/3.0) /*gradtl*/, 1-0 /*stepmx*/,
c. Part 3: Estimating parameters.
Parameters for both the basic and expanded models were
estimated by maximizing a profile likelihood. 'Since this
required repeated optimizations, each of which could take a
significant amount of time, the optimizations were carried out
on a cluster of computers. A single "master" script passed
out tasks to be completed to "slave" scripts running on the
different nodes of the cluster. The R package rpvm, which is
an interface to the software package pvm, was used to
facilitate communication between the master program and
the slaves. After the virtual machine was initiated using the
pvm console program, the simulations were started using
the command line:
I R --slave < master.R > master.Rout 2>&1 &
d. Part 3a: Basic model.
Before running 'master.R', another script, 'getStartVals.R'
was used to set up directory structures and some basic data
structures that were then used by the simulation programs.
After running this sequence of scripts once, the grid is
refined, 'master.R1 and 'slave.R' are updated, and the whole
thing run again.
I.B.4 Page 24
-------
I ### GetStartvals.R
i ### Construct skeletons of the Fit files that includes everything we can
i ### determine in adyance: data set, pseudodata, etc, including skeletons
i ### of the LL and Fits vectors for holding the results. Reorder the Bgrid
i ### data frame so that we start at the lower left corner (pf = pm = 0.001)
| ### and end up in the upper right corner, when we do the fits, the start
\ ### value for the fit will be the previous set of estimates
f\J i #'## in particular, set up a structure that holds the initial estimates
| ###
\ ### From the pre-SAP-review data, extract the best parameter estimates from
i ### the basic model, and build a list.
I ###
i ### This is a list instead of a dataframe because the number of parameters
I ### will vary, because of the number of background parameters, and whether
= ### we have one or two estimates of B. The entry for each chemical is a
I ### list of two elements, named "A", and "BMD". Each element is
1 ### a vector of parameter estimates, named appropriately.
C 1 ### Make sure parameters are expressed on the original data
Q) i ### scale, not the values based on the rescaled data. New datasets may
| ### change the scale! New datasets may also add new background values.
tr\ | ### The following function takes as its argument one of the Basic Model
7/( I ### objects and returns a list with the proper format.
%{ I ###
CU = ### Finally, we execute this code only once for any chemical. Another
(/) I script
(/) 1 ### will be used in case we need to updata data for some chemicals and rerun
require(nlme)
y> | require(RBMDS)
(/) I getparms <- function(x.dta) {
IT? | ## Extract the parameters
fi = parms <- if (Ms.null (x)) {
i if (inherits(x$FitBMD,"nlme")) {
Q) i x$FitBMD$coefficients$fixed
5> I } else {
.— = x$FitBMD$coefficients
} } else {
NULL
## Break them into A, and BMD
_ out <- vector("list",2)
Z3 I names(out) <- c("A","BMD")
02 ## Get the values of s.M.t
jj out[["A"l] <- log(unlist(unclass(by(dta,dta$s.M.t,function(x)
mean(x$chei[x$dose == min(x$dosej],na.rm=TRUE)
O= nm <- names(out[["A"]])
i out[["A"]] <- as.vector(out[["A"]])
__ | names(out[["A"]]) <- nm
CD I
(/)=## Get BMD from x if possible, else figure it is about 1/4 way between
— i ## the maximum and minimum dose
> | if (Ms.null (parms)) {
Q) i tmp <- parms[grep("ABMD",names(parms))]
i ## Rescale using x$Dosescale
= out[["BMD"]] <- tmp + log(x$Dosescale)
1 } else {
i out[["BMD"]] <- log(rep(max(dta$dose)/4,2))
| names(out[["BMD"]]) <- c("BMD.sexF","BMD.sexM")
! III.B.4 Page 25
-------
## There is no point is getting "B", since that will be fixed.
out
| attach(". ./Data/opdata. rda")
C\J I Nchems <- length (chemicals)
i BasicStarts <- vector ("1 i st" , Nchems)
I names (BasicStarts) <- chemicals
I Nsteps <- 16
i Bgrid <- expand. grid(B.F=seq(0. 001,0. 8, length=Nsteps) ,
I B.M=seq (0.001, 0.8, length=Nsteps))
= Bgrid$LL <- numeric(nrow(Bgrid))
i Bgrid$LL[] <- NA
i for (chem in Chemicals) {
i if (file. exists(file.path("Skeletons",paste(chem, "rda", sep=".")))) next
i— i f <- file.path(". ./. ./pre-Feb-02/BasicModel Fits/Basic-Model",
CD '1 paste(chem,"rda",sep="."))
Ei if (file.exists(f)) load(f) else xx <- NULL
i xxsave <- xx
tft | seldata <- BRAlNdata[sel <- (BRAlNdataSchemical == chem),]
ffl 1 seldata$s.M.t <- factor(seldata$s.M.t)
Jfr I seldata$sex <- factor (seldata$sex)
CP = seldata$block <- factor(seldata$block)
(/) I
(/) | ### loop through seldata (by block), using PhonyDF to expand it to
| ### synthetic individual data.
-,*.. i tmp <_ by(seldata,list(seldata$block) ,function(x) {
yl i PhonyDF(x$dose,x$n,x$chei,x$sd,"Dose","AChE",
1T> I chem=rep(as.character(x$chunit[l]),nrow(x)),
LL i ChemName="L)nit")
= for (i in seq(along=tmp)) {
fids <- unlist(strsplit(names(tmp[i]),"-"))
N <- nrow(tmp[[i]])
JS
"5
tmp
tmp
tmp
Smrid <- rep(fids[2],N)
$sex <- rep(flds[4],
$set <- rep(paste(flds[2],flds[5],flds[6],sep=":"),N)
Ss.M.t <- paste(flds[4],flds[2],flds[5],sep=f;-")
LI tmp
5 "L
13 I
f ") = Pseldata <- do.call ("rbind" ,unclass(tmp))
^*-* \ row.names(Pseldata) <- seq(along=Pseldata[[l]])
f\ I Pseldata$set <- factor(PseldataSset)
LL, | Pseldata$s.M.t <- factor(Pseldata$s.M.t)
Oi Respscale <- max(seldata$chei)
i Dosescale <- max(seldata.$dose)
"O | Pseldata$AChE.scaled <- Pseldata$AChE/Respscale
® I Pseldata$Dose.scaled <- PseldataSDose/oosescale
(/) i ### Random Effects
.—. = RnDoMl <- if (length(levels(Pseldata$mrid)) == 1) {
> i if (length(levels(Pseldata$set)) == 1) {
f11 = NULL
I } else {
| list(set=pdDiag(form=BMD ~ 1)) ## BMD-1 | set
I } else {
i if (length(levels(Pseldata$set)) > length(levels(Pseldata$mrid))) {
| list(mrid=pdDiag(form=BMD ~ 1),
[ 111.B.4 Page 26
-------
set=pdDiag(form=BMD~l)) ## BMD~l|mrid/set
} else {
list(mrid=pdDiag(form=BMD ~ 1)) ## BMD~l|mrid
CM
C~ I xx <- 1 ist(Chemical=chem,start=getparms(xxsave,sel data) ,Data=sel data,
__ I Pdata=Pseldata,Respscale=Respscale,
I Dosescal e=Dosescal e , Random=RnDoMl)
^ | save(xx,file=fi1e.pathC"Skeletons1I>paste(chem,"rda",sep=".")))
CD I ### master. R
, = ### Assume that pvm has already been started, and assign work to each node
| ### in the cluster. , This is a general purpose master; be sure to change
•*lJ I ### the value of slavename. Also, the value of chemicals can be reassigned
C- | ### if you want to do just a subset. All the real work is done in the slave
Q) | ### program.
yy | require(rpvm)
*^ I attach ("/home/setzer/tasks/CumRisk/post-Feb-SAP/Data/opdata.rda")
CD i griddb <- read. csv("griddb.csv", row. names=l)
(/) I Chemicals <- row.names(griddb[griddb$Rerun == 1,])
(/) i cat("Running for Chemical s:\n")
I print(Chemicals)
I MOREWORK <- 22
I TASKFAIL <- 99
(/) i HERESWORK <- 33
JT"? | workingdir <- outputdi r <- getwdQ
LL. i Hosts <- "/home/setzer/.xpvm_hosts"
i pri nt ( . PVM . start . pvmd (hosts=Hosts , bl ock=TRUE))
Q) § ### Here is where changes are most likely
.— | slavename <- "slave" # name of program to do the work for a particular
"*li = # chemical
f ### TO here
[
I cfg <- .PVM.configQ
I nodenames <- as. character (cfgSname)
LL, = #ntasks <- max (mi n(n row (cfg) ,length(Chemicals)-l) ,1)
/~\ I ntasks <- min(nrow(cfg) ,length(Chemicals))
_ | mytid <- .PVM.mytidO
CD 1 children <- NULL
(/) | # for (i in seq(along=ntasks)) {
• — i # tmp <- .PVM.spawn(task="slaveR.sh" ,ntask=ntasks[i] ,
> i # flag="Host", where=nodenames[i] ,
Q\ i # arglist=c(s1avename,workingdi r,
f*/ i # outputdi r))
LL = # cat (tmp, fill=TRUE)
i # if (length(tmp) > 0) children <- c(children,tmp)
1 # }
= children <- .PVM.spawn(task="slaveR.sh",ntask=ntasks,
| flag="Default",
i III.B.4Page27
-------
argli st=c(siavename,workingdi r,
butputdi r))
I Sys.sleep(10)
1 if (length (Chemicals) > 1) {
i tmp <- .PVM.spawn(task="slaveR.sh",ntask=l,flag="Host",
C\| i where="gandalf .local domain",
i argl ist=c(sl avename, worki ngdi r,outputdi r))
i children <- c (children, tmp)
\ ### check for and delete any -1's (system errors)
1 if (any(children == -1)) {
= warning(paste(sum(children == -1), "failed starts out of",length(children) ,
, = "potential"))
i children <- children[children != -1]
c !}
CD I if ((Nchild <- length(children)) == 0) stopC'No children started\n")
I ### Get the nodenames of the tasks
ff\ | NodeNames <- character(length(children))
jit I for (i in seq(along=NodeNames))
CD i NodeNames [i] <-
(/) I as . character (cfg$name [match ( . PVM . tasks (where=chi 1 dren [i ] ) $host ,
C/) 1 cfg$host.id)])
I ### Request notification of task exiting
i . PVM. notify(msgtag=TASKFAlL,what="ExitTask", children)
IT^ | ### start looping
= i <- 0
0 I j <- 1
2> i Nrunning <- Nchild
i repeat {
**± I i <- (i %% Nchild) + 1
_vU | while (buf <- .PVM.nrecv(-l, TASKFAIL) > 0) {
— » i tmp <- .PVM.upkintQ
s kk <- match (tmp, children)
I if (lis.na(kk)) {
| cat(paste("\n>» Task", tmp, "on" .NodeNames [kk] ,"has exited\n"))
Z3 i sys.sleep(S)
Oi children[kk] <- .PVM.spawn(task="slaveR.sh",ntask=l,
= flag="Host", where=NodeNames[kk] ,
**. | arglist=c(slavename,workingdi r,
LL = outputdir))
OI cat(paste("Replaced by task", chil dren [kk] , "\n\n"))
i sys.sleep(S)
__ 1 . PVM . noti f y (msgtag=TASKFAIL , what="Exi tTask" , chi 1 dren [kk] )
U i next
CD I } }
— 1 buf <- . PVM. nrecv(chil dren [i ], MOREWORK)
> I if ( buf > 0) {
m = # tmp <- .PVM.upkstrvecQ
rZ i # cat(paste("\n++", tmp, "completed at",date() , "\n"))
LJ_ = if (j <= length(Chemicals)) {
H cat(paste( \n++ Sending" , Chemical s[j] , "to task" .children [i] ,
| "on", NodeNames [i] , "at",date() , "\n"))
| .PVM.initsendQ
| .PVM.pkstrvec(Chemicals[j])
[ III. B.4 Page 28
-------
CM
O
CD
i
E
co
CO
0)
CO
CO
0
^
13
.PVM.send(chi1dren[i],HERESWORK)
j <- j + 1
} else {
Nrunning <- Nrunning - 1
if (Nrunning <= 0) {
break
}
cat(paste("\n!!!!!! All Done!",date(),"\n"))
for (i in seq(along=children)) .PVM.kill(children[i])
rm(Chemicals) ## so we can use the one in opdata.rda
source("piotProfiles.R")
.PVM.exitQ
C,
CD
### slave.R
### Profile likelihoods for exponential model and male,female values of B
###
### Get the command line arguments (to specify which chemicals to run):
invisible(options(echo=FALSE))
require(rpvm)
cat(11\n===—======~====\n")
mytid <- .PVM.mytidQ
cat(paste("l am task",mytid,"on",system("uname -n",intern=TRUE),"\n"))
cat (r\n========~========\n\n ")
attach("/home/setzer/tasks/CumRisk/post-Feb-SAP/Data/opdata.rda")
require(RBMDS)
requi re(nlme)
setwd("/home/setzer/tasks/CumRisk/post-Feb-SAP/ProfilesForB")
savepath <- "Fits"
griddb <- read.csv("griddb.csv",row.names=l)
MOREWORK <- 22
TASKFAIL <- 99
HERESWORK <- 33
myparent <- .PVM.parentQ
chem <- "START"
i ### BMR is set here
« [ BMR <- 0.1
| if (myparent <= 0) stopC'PVM error!")
1 repeat {
= .PVM.initsendQ
0 = # .PVM.pkstrvec(chem)
(f\ \ .PVM. send (myparent, MOREWORK)
— E buf <- . PVM. recv(myparent, HERESWORK)
> s chem <- .PVM.upkstrvec()
m 1 cat(paste(" ------- \n" ,chem,"\n\n"))
= OldModel <- file.path("Skeletons", paste(chem, "rda" ,sep="."))
i load (OldModel)
i Firststart<- c(xx$Start[["A"]] - loq(xx$Respscale) ,
| xx$start[["BMD"]] - Tog(xx$Dosescale))
i start <- Firststart
| ### Compute the grid of values.
1 III. B.4 Page 29
-------
i for (i in l:nrow(xx$Bgrid)) {
i xx$Models[[i]] <- eval(substitute(AChE.scaled ~
= CexpBwDS(Dose.scaled,A,BMD,sex,
i fixed=x),
| list(x=list(B=c(F=as.vector(xx$Bgrid$B.F[i]),
C\| i M=as.vector(xx$Bgrid$B.M[i]))))))
p f }
i Mrids <- unique(xx$Data$mrid)
| if (length(Mrids) > 1) {
i Power <- rep(0.5, length(Mrids))
1 names(Power) <- as.character(Mrids)
i weights <- varComb(varldent(form=~l[factor(mrid)),
I varPower(form=~fitted(.)[factor(mrid),
I fixed=Power))
I } else {
L. i weights <- varPower(form=~fitted(.),fixed=l)
CD [ }
i for (i in l:nrow(xx$Bgrid)) {
ff( I cat(paste("\n \n",i,":",xx$Bgrid$B.F[i],xx$Bgrid$B.M[i],date(),"\n\n"))
;** I Model <- xx$Models[[i]]
CD i RnDoMl <- xx$Random
(0 .1
<| xx$Fits[[i]] <- if (!is.null(RnDoMl)) {
i try(eval(substitute(nlme(Model,data=xx$Pdata,
i fixed=list(A ~ s.M.t - 1, BMD ~ sex-1),
| random=xxxx,
W | start=start,
"f^s i weiqhts=weights,
LJL | method="ML"),
CD ^ 1i st(Model=Model,xxxx=RnDoMl,Wei ghts=wei ghts))))
> I } else {
.— i try(eval (substitute(gnls(Model, data=xx$Pdata,
"tlf I params=list(A ~ s.M.t - 1, BMD ~ sex-1),
— ^ start=start,
—* i weights=weights),
—' | list(Model=Model,Weights=weights))))
I xx$Bgrid$LL[i] <-
13 i if (!inherits(xx$Fits[[i]], "try-error")) logLik(xx$Fits[[i]]) else NA
Oi j <- if (i %% xx$Nsteps == 0) {
= i - xx$Nsteps + 1
CL 1 Mlse{
i start <- if (!inherits(xx$Fits[[j]],"try-error")).{
_,_ I if (inherits(xx$Fits[[j]],"nlme")) xx$Fits[[j]]$coefficients$fixed
w | else xx$Fits[[j]]$coefficients
CD I } else Firststart
-;— | save(xx,file=file.path(savepath,paste(chem,"rda",sep=".""
m I. cat(paste(" ",chem,"complete",date(),"\n\n"))
### makefinegriddb.R
loadpath <- "FineFits"
griddb <- read.csv("griddb.csv",row.names=l)
III.B.4 Page 30
-------
CN
| attach("•./Data/opdata.rda")
griddb$Npoints[] <- 11
griddb$Rerun[] <- 1
for (chem in Chemicals) {
"1 oad (file, path (loadpath, paste (chem," rda", sep=".")))
GO
0
griddb
griddb
griddb
griddb
;chem,"B.Fmin
;chem,"B.Fmax";
'chem,"B.Mmin""
:chem,"B.Mmax""
- min(xx$Bgrid$B.F)
- max(xx$Bgrid$B.F)
- min(xx$Bgrid$B.M)
- max(xx$Bgrid$B.M)
f~. write.table(griddb,file="finegriddb.csv",sep=",",col .names=NA)
viJ
e. Part 3b: Expanded model.
C
0
(/) i For profile likelihood estimation of S and D in the expanded
(/) | model, first 'master.R' and 'slave.R' are run, which use the
0 I best values found in the estimation of PB for the basic model
GO I as initial values. Then successive versions of 'master' and
| 'slave' are produced (see below, 'finemaster.R',
| 'fineslave.R') that result from successive refinements of the
I grid.
1 ### master.R
i ### Assume that pvm has already been started, and assign work to each node
I ### in the cluster. This is a general purpose master; be sure to change
I ### the value of slavename. Also, the value of Chemicals can be reassigned
I ### if you want to do just a subset. All the real work is done in the slave
I ### program.
cs ! . , ,
— 5 require(rpvm)
attach("/home/setzer/tasks/cumRisk/post-Feb-SAP/Data/opdata.rda")
griddb <- read.csv("griddb.csv",row.names=l)
Chemicals <- row.names(griddb[griddb$Doit == 1,])
cat("Running for Chemicals:\n")
print(Chemicals)
f^ 1 MOREWORK <- 22
= TASKFAIL <- 99
HERESWORK <- 33
GETSTARTED <- 1
workingdir <- outputdir <- getwdQ
Hosts <- "/home/setzer/.xpvm_hosts"
#print(.PVM.start.pvmd(hosts=Hosts,block=TRUE))
#Sys.sleep(10)
### Here is where changes are most likely
slavename <- "slave" # name of program to do the work for a particular
# chemical
### TO here
III.B.4 Page 31
-------
I cfg <- .PVM.configQ
| nodenames <- as. character (cfg$name)
1 #ntasks <- max (mi n(n row (cfg) .length (Chemical s)-l) ,1)
I ntasks <- min(nrow(cfg) , length (Chemicals))
CM I
= myti d <- . PVM . myti d ()
1 #children <- NULL
| # for (i in seq(along=ntasks)) {
I # tmp <- .PVM.spawn(task="slaveR.sh",ntask=ntasks[i] ,
I # flaig="Host", where=nodenames[i] ,
I # arglist=c(slavename,workingdi r ,
= # • outputdir))
= # cat(tmp, fill=TRUE)
\ # if (length (tmp) > 0) children <- c (children, tmp)
•
.
L— = children <- .PVM. spawn (task="slaveR.sh",ntask=ntasks,
d) i flag="Default",
p- 1 arc|list=c(slavename,workingdi r ,
C i outputdir))
fft | Sys. sleep (10)
,-A i print(children)
Q) I if (length (Chemicals) > 1) {
C/) i tmp <- .PVM. spawn (task='" si aveR.sh" ,ntask=l,flag="Host" ,
(/} 1 where="gandalf . local domain",
<• I arglist=c(slavename,workingdi r.outputdi r))
| children <- c(children,t:mp)
= sys. sleep (10)
I print(children)
I ### check for and delete any -1's (system errors)
I if (any(children == -1)) {
d) i warning(paste(sum(children == -1),"failed starts out of".length(children),
> I ' "potential")) .
= children <- children[children != -1]
[ if ((Nchild <- length(children)) == 0) stop("No children started\n")
I ### Get the nodenames of the tasks
i NodeNames <- character(length(children))
I for (i in seq(along=NodeNa.mes)) {
I NodeNames[i] <-
*% I as.character(cfg$name[match(.PVM.tasks(where=children[i])$host,
LL = cfg$host.id)])
_ § }
i ### Request notification of task exiting
| .PVM.notify(msgtag=TASKFAIL,what="ExitTask",children)
{£} I ### Start them
> i .PVM.initsendQ
m i .PVM.pkintvec(l:3)
| .PVM.mcast(children,GETSTARTED)
1 ### start looping
I i <- 0
I J <- 1
I.B.4 Page 32
-------
I Nrunninq <- Nchild
i repeat {
I i <- (i %% Nchild) + 1
| while (buf <- . PVM.nrecv(-l, TASKFAIL) > 0) {
I tmp <- . PVM.upkintQ
i kk <- match(tmp, children)
I if (Ms.na(kk)) {
C\j i cat(paste("\n>» Task" ,tmp, "on" ,NodeNames[kk] ,"has exited\n"))
Sys.sleep(S)
chi
ildren[kk] <- .PVM.spawn(task="slaveR.sh",ntask=l,
| flaq="Host", where=NodeNames[kk],
I arglist=c(slavename,workingdir,
i outputdir))
I cat(paste("Replaced by task",children[kk],"\n\n"))
I Sys.sleep(3)
I .PVM.noti fy(msgtag=TASKFA!L,what="Exi tTask",chi1dren[kk])
| next
c j }}
CD i buf <- .PVM.nrecv(children[i], MOREWORK)
El if ( buf > 0) {
i tmp <- .PVM.upkstrvecQ
(A | cat(paste("\n++",tmp,"completed at",date(),"\n"))
ffi I if (j <= length (Chemicals)) {
V' 1 cat(paste( \n++ Sending",Chemicals[j],"to task",children[i],
CD 1 "on",NodeNames[i],"at",date(),"\n"))
(/) = .PVM.initsendQ
(/) § .PVM.pkstrvec(Chemicals[j])
i } else {
-*• = Nrunmng <- Nrunmng - 1
if (Nrunning <= 0) {
= break
0 | } >
.— I cat(paste("\n!!!!!! All Done!",date(),"\n"))
"*^ | for (i in seq(along=children)) .PVM.kill (children[i])
"mj I rm(Chemicals) ## so we can use the one in opdata.rda
—' | source("plotProfiles.R")
_ I .PVM.exitO
ZJ | ### slave.R
^-^ | invisible(options(echo=FALSE))
** i require(rpvm)
LL = cat("\n=================================\n")
O= mytid <- .PVM.mytidQ
i cat(paste("l am task",mytid,"on"Isystem("uname -n".intern=TRUE)."\n"))
^ § cat(K\n=================================\n\n")
0 i
C/) i
— i attach("/home/setzer/tasks/CumRisk/post-Feb-SAP/Data/opdata.rda")
> I require(RBMDS)
Q\ i requi re(nlme)
I setwd("/home/setzer/tasks/CumRisk/post-Feb-SAP/ProfilesForSD")
i savepath <- "Fits"
| griddb <- read.csv("griddb.csv",row.names=l)
I ### walks diagonals of a grid, starting at the upper right hand corner.
| walkgrid <- function(Nsteps) {
[ III.B.4 Page 33
-------
i mx <- matrix(l: (Nsteps*Nsteps) ,nrow=Nsteps,byrow=TRUE)
1 Gridlndex <- numeric(Nsteps*Nsteps)
1 indx <- 1
I for (i in 1: Nsteps) {
= for (k in l:i) {
1 if (i %% 2 == 1) {
i • ii <- i - k + 1
CM I Jj <- Nsteps - k + 1
O i } ^se {,
••sr = 11 <- k
I jj <- Nsteps - i + k
| Gridlndex[indx] <- mx[ii,jj]
rj? i indx <- indx + 1
CD i >
i for (i in 2: Nsteps) {
"*Zl 1 for (k in i:Nsteps) {
C | if ((Nsteps - i + 1) %% 2 == 1) {
d) I ii <- Nsteps + i - k
r~ I jj <- Nsteps - k + 1
o) ! > ?!se -t,
fX i n <- k
«g i }jj<-k-i + i
(/) I Gridlndex [indx] <- mx[ii,jj]
tf\ i indx <- indx + 1
v^ s -,
•* i Gridlndex
}
! MOREWORK <- 22
I TASKFAIL <- 99
i HERESWORK <- 33
I GETSTARTED <- 1
| myparent <- .PVM.parentQ
JTO I chem <- "START"
—. I .PVM.recv(myparent, msgtag=GETSTARTED)
—J i tmp <- .PVM.upkintvecQ
~ i if (myparent <= 0) stop("PVM error")
_3 i repeat {
Ol .PVM.initsendO
i .PVM.pkstrvec(chem)
n | .PVM.send(myparent,MOREWORK)
LL | buf <- .PVM.recv(myparent,HERESWORK)
O\ chem <- . PVM.upkstrvecQ
! fname <- paste(chem,"rda",sep=".")
__ I cat(paste("\n ",chem," \n\n"))
CD I load(file.path("../ProfilesForB/FinerFits",fname))
,,_ i ### Get the best fitting model
D> i indx <- which.max(xx$Bgrid$LL)
0 | StartLL <- xx$Bgrid$LL[indx]
Q_ i ### Get initial start value, the coefficients from xx$Fit[[indx]]
I start <- if (inherits(xx$Fit[[indx]], "nlme")) {
'i xx$Fi t[[i ndx]]Scoeffi ci ents$fi xed
I } else {
| xx$Fit[[indx]]$coefficients
-------
I Bests <- as.vector(c(xx$Bgrid[indx,"B.F"],xx$Bgrid[indx,"B.M"]))
I ### Get the model for the random parameters used in the basic model
1 RandomParms <- xx$Random
CM I
Pseldata <- xx$Pdata
### compute the grid of values.
SDgrid <- expand.grid(S = seq(griddb[chem,"Smin"],
?riddb[chem,"Smax"],
ength=gri ddb[chem,"Npoi nts"]),
D = seq(griddb[chem,"Dmin"],
?riddb[chem, "Dmax"],
ength=gri ddb[chem,"Npoi nts"]))
SDgrid$LL <- numeric(nrow(SDgrid))
Q) 1 Spgrid$LL[] <- as.numeric(NA)
El Fits <- vector("list",nrow(SDgrid))
| Gridlndex <- walkgrid(griddb[chem,"Npoints"])
(f( 1 Mrids <- unique(xx$Data$mrid)
!:: I if (length(Mrids) > 1) {
Qj | Power <- rep(0.5, length(Mrids))
(/) 1 names(Power) <- as.character(Mrids)
(/) I weights <- varComb(varldent(form=~l|factor(mrid)),
<1 varPower(form=~fitted(.)|factor(mrid),
I fixed=Power))
^ I > else {
•*• = weights <- varPower(form=~fitted(.),fixed=l)
I xx$Fit <- Fits
I xx$Bgrid <- NULL
Q5 i . xx$SDgrid <- SDgrid
I ## If our Dmin exceeds 0.001, walk up from 0.001 to Dmin in steps of 0.05
CO [ if (griddb[chem,"Dmin"] > 0.001) {
—» = cat("\n-----— Walking up to the parameter grid \n")
I Dseq <- seq(0.001, gnddb[chem, "Dmin"], by=0.05)
I Smax <- griddb[chem,"Smax"]
= for (i in l:length(Dseq)) {
I cat("\n—-. \n")
! cat(paste("i :",i ,"D[i] :",Dseq[i],", S:",Smax,"\n"))
n 1 cat(K\n \n")
LL. = ## set up the model
O= Model <- eval(substitute(AChE.scaled ~
| CpkexpB2wDS(Dose.scaled,A,BMD,sex,
~O = fixed=list(PB=c(F=xxxx,
01
(/) = M=yyyy),
• — = S=c(F=zzzz,
> i M=zzzz),
fl 1 D=c(F=wwww,
1 M=WWWW))),
= list(xxxx=Bests[l],yyyy=Bests[2],
= zzzz=log(Smax),
| wwww=log(Dseq[i]))))
1 ## estimate it
| if (!is.null(RandomParms)) {
| III.B.4 Page 35
-------
try(fitpk <-
eval (
eval (substi tute(nlme(Model , data=Pseldata,
fixed=list(A ~ s.M.t - 1, BMD ~ sex - 1) ,
I random=xxxx,
I start=zzzz,
i weights=weights,
i method="ML") ,
C\| = 1 1st (Model =Model ,xxxx=RandomParms,
>
C/) I for (iii in l:nrovy(SDgrid)) {
"JT^ i i <- Gridlndex[iii]
LL, i cat("\n ------------- \ri")
I cat(paste("i:",i,"D[i]:",SDgrid$D[i],", S[i] :" ,SDgrid$S[i] ,date() , "\n")
0 i cat(r'\n ------------- \n")
| ## set up the model
1 Model <- eval (substitute (AChE. scaled ~
| cpkexpB2wDS(Dose. scaled, A, BMD, sex,
| fixed=list(PB=c(F=xxxx,
i M=yyyy),
| S=C(F=ZZZZ, M=ZZZZ)
-J i . D=c(F=wwww,
1i st(xxxx=Bests[1],yyyy=Bests[2],
zzzz=log(SDgrid$S[i]),
f \ | M=wwww))),
LJL I wwww=log(SDgrid$o[i]))))
Oi ## estimate it
= if (lis.null(RandomParms)) {
-r- I try(fitpk <-
^J = eval(substi tute(nlme(Model,data=Pseldata,
0 i fixed=list(A ~ s.M.t - 1, BMD ~ sex - 1),
(/) | random=xxxx,
••— i start=zzzz,
^H> i weights=Weights,
0 | method="ML"),
I list(Model=Model,xxxx=RandomParms,
i zzzz=start,
I weights=weights))))
i } else {
I fitpk <-
| try(eval(substitute(gnls(Model, data=Pseldata,
| III.B.4 Page 36
-------
i params=list(A ~ s.M.t - 1, BMD ~ sex - 1) ,
i start=zzzz,
I weights=weights) ,
| 1 ist (Model =Model ,
= zzzz=start,
| weights=weights))))
CM I }
/— \ = if (!inherits(fitpk, "try-error")) {
irr i xx$SDgrid$LL[i] <- logLik(fitpk)
T— i cat(paste("\nLL:"Ixx$SDgrid$LL[i]I"\n"))
_ i ## Use the current fit to give the start value for the next
3~~ = start <- if (inherits(fitpk, "n~lme")) {
/"£; | fitpk$coefficients$fixed
CD = } else {
, | fitpkScoefficients
•*-• I }
C 1 xx$Fit[[i]] <- fitpk
0 I save(xx)file=file.path("Fits",paste(chem)"rda",sep=".")))
El }
= cat(paste(chem, "finished", dateQ , "\n\n"))
10 i >
& 1 ### plotPr9files.R
d? I require(akima)
(/) I pdf(file="Prof iles-4-SD2.pdf")
(/) | par(xpd=NA)
i "if (! fil e. exists (file, path (di rname, fname))) next
.— i tmp <- try(load(file.path(di rname, fname)))
•^f I if (inherits (tmp, "try-error")) {
JO i cat(paste("Problem reading", chem, "\n"))
— « I next
i LLgrid <- xx$SDgrid
H LLgrid <- na.omit(LLgrid)
Z3 I if (length(LLgrid$LL) == 0) {
Oi plot(c(0,l),c(0,l),type="n",axes=FALSE, xlab="", ylab="")
I text(0.5,0.5,paste(chem,"no fits") ,adj=0. 5)
n I > else {
LL. i indx <- which. max(LLgrid$LL)
O= BestLL <- max(LLgrid$LL[indx] .B.BestLL)
= scaledLL <- 2*(BestLL - LLgrid$LL)
._. = out <- try(interp(LLgrid$S,LLgrid$D,ScaledLL,
v~> i xo=seq(min(LLgr-id$S), max(LLgrid$S) ,length=100) ,
d) i yo=seq(min(LLgrid$D) , max(LLgrid$D) ,length=100)))
(/) i if (!inherits(out, try-error")) {
-- i res <- try(image(out,xlab="S",ylab="D",main=chem,
> i col = rev (heat. colors (9) ),
= breaks =
1 c(0,qchisg(c(0.05,.10,.25,.50,.75,.90,.95,.99),2)1le26)))
i if (inherits(res, "try-error")) {
I plot(c(0,l),c(0,l),type="n",axes=FALSE, xlab="", ylab="")
i text(0.5,0.5,paste(chem,"not enough fits") ,adj=0. 5)
I } else {
| critx2 <- qchisq(0.95,2)
I III. B.4 Page 37
-------
I points(LLgrid$S,Ll.grid$D,
I pch=ifelse(2*(BestLL - LLgrid$LL) < critx2, 19, 3))
| points(LLgrid$S[indx],LLgrid$D[indx],pch=4,cex=l.5)
I } else {
i plot(c(0,l),c(0,l),type="n",axes=FALSE, xlab="", ylab="")
i text(0.5,0.5,paste(chem,"not enough fits"),adj=0.5)
C\l [ }
I dev.offO
| ### makeFineGrid.R —
i ### Uses the information in the files in ./Fits to create a new set of
i templates
I ### in FineFits. Each new template already contains SDgrid and Fits
| ###
= ### The best expanded fit for the following chemicals is essentially the
s— 1 basic model,
0 I ### so they are excluded from further action. The criteria were:
Ei ### 1) Pvalue for the difference in log likelihoods was greater than 0.05
1 AND
(A | ### 2) BOTH BMDs were no more than 10% different between the expanded and
*f. I basic models
:ff I ### with the expanded model being the point of comparison (in the
CD I denominator).
CO I
ff) | K2IJ <- function(k, Npoints) {
= indxmin <- function(i, N) {
"> 1 if (i == 1) i else i - 1
•— I }
"tlf i indxmax <- function(i, N) {
w | if (i == N) i else i +1
—' = Neighbors <- function(k, Npoints) {
i] <- K2lJ(k, Npoints)
— - i <~ ij[l]
13 I j <- ij[2]
Oi iseq <- indxmin(i.Npoints):indxmax(i,Npoints)
| jseq <- indxmin(j, Npoints):indxmax(j, Npoints)
UL I as.vector(apply(data.matrix(expand.grid(i=iseq,j=jseq)),l,function(x)!J2K(x[
l],x[2],Npoints)))
0
0 I require(RBMDS)
(/) | requi re(nlme)
DropChems <-
C("ACEPHATE","CHLORPYRIPHOSMETHYL","DICROTOPHOS","DIMETHOATE","ETHOPROP",
"FENTHION","METHAMIDOPHOS","METHIDATHION","NALED","OXYDEMETONMETHYL",
"PIRIMIPHOSMETHYL","PROFENOFOS")
attach("•./Data/opdata.rda")
savedir <- "Skel2r'
I.B.4 Page 38
-------
I finegriddb <- read.csv("griddb.csv",row.names=l)
I finegriddb[DropChems,"Dolt"] <- 0
finegriddb$Npoints[] <- 5
Dochem <- which(finegriddb$Doit == 1)
CM
O
CD
i
•4— •
c
CD
E
V)
C/3
CD
CO
o:
CD
for (chem in Chemicals) {
fname <- paste(chem,"rda",sep=".")
load(file.path("Fits",fname))
## 'xx1 is the new version
if (chem %in% DropChems) {
save(xx, fi1e=fi1e.path("FineFits",fname))
next
} else {
oldxx <- xx
xx$Nsteps <- 5
xx$Fit <- listQ
Slist <- sort(unique(oldxx$SDgrid$S))
Dlist <- sort(unique(oldxx$SDgrid$D))
## Build the new SDgrid centered around the maximum LL on the old one
indx <- which.max(oldxx$SDgrid$LL)
SDindx <- K2lJ(indx,oldxx$Nsteps)
Smin <- Slist[indxmin(SDindx[l],oldxx$Nsteps)]
Smax <- Slist[indxmax(SDindx[l],oldxx$Nsteps)]
Dmin <- Dlist[indxmin(SDindx[2],oldxx$Nsteps)]
if (SDindx[2] < oldxx$Nsteps) {
Dmax <- Dlist[indxmax(SDindx[2], oldxx$Nsteps)]
} else {
## If the old D was on the upper border of the grid, expand it by one
step
Delta <- (max(Dlist) - min(Dlist))/oldxx$Nsteps
Dmax <- max(Dlist) + Delta
finegriddb
finegriddb
finegriddb
finegriddb
ichem, "Smin";
;chem,"Smax";
;chem,"Dmin";
;chem,"Dmax";
<- Srtnn
<- Smax
<- Dmin
<- Dmax
xx$SDgrid <- expand.grid(S=seq(Smin,Smax,length=xx$Nsteps),
D=seq(Dmi n,Dmax,1ength=xx$Nsteps))
xx$SDgrid$LL <- numeric(nrow(xx$SDgrid))
xx$SDgrid$LL[] <- NA
xx$Start <- if (inherits(9ldxx$Fit[[indx]],"
oldxx$Fi t[[i ndx]]Scoeffi ci ents$fi xed
} else {
oldxx$Fi t[[i ndx]]Scoeffici ents
}
xx$Fit <- vector("list",nrow(xx$SDgrid))
E
13
O
CL
O
## Finally, fill in the LLs and Fits we already know.
*
-------
I j <- which(al & a2)
= if (length(i) == 1) {
i xx$SDgrid$LL[j] <- oldxx$SDgrid$LL[K]
i xx$Fit[[j]] <- oldxx$Fit[[KJ]
\ }
C\J i save(xx,file=file.path(savedir.fname))
I write.tab!e(finegriddb,file="finegriddb.csv",sep=",",col.names=NA)
1 ### master.R
I ### Assume that pvm has already been started, and assign work to each node
= ### in the cluster. This is a general purpose master; be sure to change
i ### the value of slavename. Also, the value of Chemicals can be reassigned
_ I ### if you want to do just a subset. All the real work is done in the slave
v— i ### program.
CD I
I require(rpvm)
(A I attach("/home/setzer/tasks/CumRisk/post-Feb-SAP/Data/opdata.rda")
Jfr I griddb <- read.csv("finegriddb.csv",row.names=l)
vU I chemicals <- row.names(griddb[griddb$Dolt == 1,])
(/) I cat("Running for chemicals:\n")
(/) = print(Chemicals)
i MOREWORK <- 22
= TASKFAIL <- 99
I HERESWORK <- 33
_CQ | GETSTARTED <- 1
Q_ i workingdir <- outputdir <- getwdQ
I Hosts <- "/home/setzer/.xpvm_hosts"
(D i #print(.PVM.start.pvmd(hosts=Hosts,block=TRUE))
"> I #Sys.sleep(10)
i ### Here is where changes are most likely
Cw i slavename <- "fineslave" # name of program to do the work for a particular
""HZ. \ # chemical
### To here
*"— ' | cfg <- .PVM.configQ
Q- I nodenames <- as.character(cfg$name)
OI
| #ntasks <- max(min(nrow(cfg) ,length(Chemicals)-l) ,1)
~_ | ntasks <- min(nrow(cfg) ,length(Cnemical s))
0 I mytid <- .PVM.mytidQ
(n | #children <- NULL
._ = # for (i in seq(along=ntasks)) {
> | # tmp <- . PVM. spawn (task="sl aveR. sh", ntask=ntasks [i ],
m = # flag="Host", where=nodenames[i] ,
i # argtist=c(slavename,workingdi r ,
I # outputdir))
I # cat (tmp, fill=TRUE)
= # if (length(tmp) > 0) children <- c (children, tmp)
- # }
I children <- .PVM.spawn(task="slaveR.sh",ntask=ntasks,
j III. B.4 Page 40
-------
Osi
o
flag="Default",
arglist=c(siavename,workingdi r,
outputdir))
Sys.sleep(lO)
if (length(Chemicals) > 1) {
tmp <- .PVM.spawn(task="slaveR.sh",ntask=l,flag="Host",
where="gandalf.1 oca!domai n",
arglist=c(slavename,workingdi r.outputdi r))
| children <- c(children,tmp)
| Sys.sleep(lO)
| ### Check for and delete any -1's (system errors)
! if (any(children == -1)) {
= warning(paste(sum(children == -1),"failed starts out of",length(children),
I "potential"))
t~ i children <- children[children != -1]
0. | }
! if ((Nchild <- length(children)) == 0) stopC'No children started\n")
¥2 | ### Get the nodenames of the tasks
CD | NodeNames <- character(length(children))
(/) I for (i in seq(a1ong=NodeNames)) {
(/) | NodeNames [i] <-
<= as.character(cfg$name[match(.PVM.tasks(where=chi1dren[i])$host,
| cfg$host.id)])
-*k | ### Request notification of task exiting
C/J i
.PVM.noti fy(msgtag=TASKFA!L,what="Exi tTask",chi1dren)
### Start them
CD I
*> = .PVM.initsend()
.2-. i .pVM.pkintvec(l:3)
~*± | .PVM.mcast(children.GETSTARTED)
"« I ### Start looping
El i <- 0
-, M <- i
~j = Nrunning <- Nchild
Oi repeat {
i i <- (i %% Nchild) + 1
n = while (buf <- .PVM.nrecv(-l, TASKFAIL) > 0) {
LL | tmp <- .PVM.upkintQ
Ol kk <- match(tmp, children)
i if (Ms.na(kk)) {
= cat(paste("\n>» Task",tmp,"on",NodeNames[kk],"has exited\n"))
i Sys.sleep(S)
i children[kk] <- .PVM.spawn(task="slaveR.sh",ntask=l,
| flag="Host", where=NodeNames[kk],
= arglist=c(s1avename,worki ngdi r,
i outputdir))
i cat(paste("Replaced by task",children[kk],"\n\n"))
= Sys.sleep(B)
i .PVM.noti fy(msgtag=TASKFA!L,what="Exi tTask",chi1dren[kk])
= next
| buf <- .PVM.nrecv(children[i], MOREWORK)
f III.B.4 Page 41
-------
I if ( buf > 0) {
I tmp <- .PVM.upkstrvecQ
I cat (paste ("\n++", tmp, "completed at",date() , "\n"))
| if (j <= length (chemicals)) {
i cat (paste (\n++ Sending" ,Chemicals[j] , "to task",children[i] ,
= "on",NodeNaines[i],"at",dateO,"\n11))
i .PVM.initsendO
C\i i . PVM.pkstrvec(Chemicals[j])
i . PVM . send (chi 1 d ren [i ] , HERESWORK)
I j <- j + 1
§ } else {
| Nrunm'ng <- Nrunning - 1
i >
i if (Nrunm'ng <= 0) {
I break
i}}
C i cat(paste("\n!!!M! All Done! " ,date() ,"\n"))
0 | for (i in seq(a~long=children)) .PVM. kill (children [i])
rm(Chemicals) ## so we can use the one in opdata.rda
yy source("p1otProfiles.R")
!;;: I .PVM.exitO
CD i ### slave. R
C/) I
(/) = invisible(options(echo=FALSE))
l require(rpvm)
i cat("\n=================================\n")
mytid <- .PVM.mytidQ
i cat(paste("l am task",mytid,"on",system("uname -n",intern=TRUE),"\n"))
C/5 = catr\n===================:==============\n\n")
ir I
0 I attach("/home/setzer/tasks/CumRisk/post-Feb-SAP/Data/opdata.rda")
2> E require(RBMDS)
"— i require(nlme)
"Jzf I setwd("/home/setzer/tasks/cumRisk/post-Feb-SAP/ProfilesForSD")
J-v i savepath <- "FineFits" •
^ 1 ### Walks diagonals of a grid, starting at the upper right hand corner.
*"- I MOREWORK <- 22
Z3 i TASKFAIL <- 99
Ol HERESWORK <- 33
I GETSTARTED <- 1
** | myparent <- .PVM.parentO
Ol chem <- "START"
| .PVM.recv(myparent, msgtag=GETSTARTED)
___ | tmp <- . PVM.upkintvecO
0 I if (myparent <= 0) stop("PVM error")
(A | repeat {
—i. = .PVM.initsendO
"?> i . PVM. pkstrvec (chem)
0 i .PVM.send(myparent,MOREWORK)
I buf <- .PVM.recv(myparent,HERESWORK)
i chem <- .PVM.upkstrvecO
i fname <- paste(chem,"rda",sep=".")
| cat(paste("\n ",chem," \n\n"))
| Ioad(file.path("skel2",fname))
I III.B.4 Page 42
-------
### Get initial start value, the coefficients from xx$Fit[[indx]]
start <- xx$Start
### and the estimates of PB to use
indx <- which.max(xx$SDgrid$LL)
Bests <- as.vector(eval(xx$Fit[[indx]]Seal1$model[[3]]$fixed$PB))
(Nj ### Get the model for the random parameters used in the basic model
RandomParms <- xx$Random
Pseldata <- xx$Pdata
Mrids <- unique(xx$Data$mrid)
if (length(Mrids) > 1) {
Power <- rep(0.5, length(Mrids))
names(Power) <- as.character(Mrids)
weights <- varcomb(varldent(form=~l[factor(mrid)) ,
varPower(form=~fitted(.)|factor(mrid),
fixed=Power))
} else {
weights <- varpower(form=~fitted(.),fixed=l)
CO
0 for (i in l:nrow(xx$SDgrid)) {
CO if (!is.na(xx$SDgrid$LL[i])) next
cat("\n \n")
cat(paste("i:",i,"D[i]:">xx$SDgrid$D[i],",
S[i]:",xx$SDgrid$S[i]IdateO,"\n")5
^ cat("\n \n")
CO I ## set up the model
JT^ | Model <- eval(substitute(AChE.scaled ~
LL. | CpkexpB2wDS(Dose.scaled,A,BMD,sex,
fixed=list(PB=c(F=xxxx,
w=yyyy),
S=C(F=ZZZZ, M=ZZZZ),
D=c(F=wwww,
M=wwww))),
list(xxxx=Bests[l],yyyy=Bests[2],
zzzz=log(xx$SDgrid$S[i]),
wwww=log(xx$SDgrid$D[i]))))
## estimate it
fitpk <- if (lis.null(RandomParms)) {
Otry(eval(substitute(nlme(Model,data=Pseldata,
fixed=list(A ~ s.M.t - 1, BMD ~ sex - 1),
f* random=xxxx,
LL start=zzzz,
Oweights=weights,
method="ML"),
list(Model=Model,xxxx=RandomParms,
zzzz=start,
0 weights=weights))))
(/) } else {
• •—' try(eval(substitute(gnls(wodel, data=Pseldata,
>• params=list(A ~ s.M.t - 1, BMD ~ sex - 1),
start=zzzz,
weights=Weights),
1 ist (Model =Model,
zzzz=start,
Wei ghts=wei ghts))))
III.B.4 Page 43
-------
I if (!inherits(fitpk, "try-error")) {
i xx$SDgrid$LL[i] <- logLik(fitpk)
\ cat (paste ("\nl_L :",xx$SDgrid$LL[i] ,"\n"))
I xx$Fit[[i]] <- fitpk
| save(xx,fi1e=fi1e.path("Fi neFi ts",paste(chem,"rda",sep=".")))
CSJ i cat(paste(chem,"finished",date(),"\n\n"))
i ### makeFine2Grid.R -- .
I ### Uses the information in the files in ./Fits to create a new set of
= templates
i ### in Fine2Fits. Each new template already contains SDgrid and Fits
1 ###
I ### The best expanded fit for the following chemicals is essentially the
| basic model,
_ I ### so they are excluded from further action. The criteria were:
£- i ### 1) Pvalue for the difference in log likelihoods was greater than 0.05
-------
I finegriddb <- read.csv("griddb.csv",row.names=l)
I finegriddb[DropChems,"Dort:"] <- 0
i finegriddb$Npoints[] <- 5
i Dochem <- which(finegriddb$Dolt == 1)
CM !
for (chem in Chemicals) {
fname <- paste(chem,"rda",sep=".")
load(file.path("FineFits",fname))
## 'xx' is the new version
if (chem %in% DropChems) {
save(xx, file=file.path("Fine2Fits",fname))
next
} else {
oldxx <- xx
xx$Nsteps <- 5
L- = xx$Fit <- listQ
Q) = Slist <- sort(unique(oldxx$SDgrid$S))
£- 1 Dlist <- sort(unique(oldxx$SDgrid$D))
(f\ | ## Build the new SDgrid centered around the maximum LL on the old one
CO
0
0
DC
indx <- which.max(oldxx$SDgrid$LL)
SDindx <- K2lJ(indx,oldxx$Nste
0— *JIS I IIV4AX -X IX^JU
| Smin <- Slist
CO I smax <- Slist
tf\ I Dmin <- Dlist
Vx ; ..- /-__• i..r-»n
;indxmin(SDindx[l
;i ndxmax(SDi ndx [1
^indxmin(SDindx[2
if (SDindx[2] < 9ldxx$Nsteps)
Dmax <- Dlist[indxmax(SDindx
s)
,oldxx$Nsteps)]
,oldxx$Nsteps)]
,oldxx$Nsteps)]
], oldxx$Nsteps)]
i } else {
i ## if the old D was on the upper border of the grid, expand it by one
CO 1 step
i Delta <- (max(Dlist) - min(Dlist))/oldxx$Nsteps
= Dmax <- max(Dlist) + Delta
finegriddb
finegriddb
finegriddb
finegriddb
;chem, "Smin";
;chem,"Smax";
;chem,"Dmin";
;chem,"Dmax";
<- Smin
<- Smax
<- Dmin
<- Dmax
Ol } else {
i l xx
0. 1 >
xx$SDgrid <- expand. grid(S=seq(Smin, Smax , 1 ength=xx$Nsteps) ,
D=seq(Dmin,pmax,length=xx$Nsteps))
xx$SDgrid$LL <- numeric(nrow(xx$SDgrid))
xx$SDgrid$LL[] <- NA
xx$Start <- if (inherits(9ldxx$Fit[[indx]] ,"nlme")) {
oldxx$Fit[[indx]]$coefficients$fixed
ol dxx$Fi t [ [i ndx] ] Scoef f i ci ents
xx$Fit <- vector("list",nrow(xx$SDgrid))
## Finally, fill in the LLs and Fits we already know.
## what are all the indexes?
nearby <- Neighbors(indx,oldxx$Nsteps)
CO
## There must be a better way to do this, but I'm tired ...
for (i in seq(along=nearby)) {
K <- nearby[i]
al <- sapply(xx$SDgrid$S,
f unction(x)i denti cal(al1.equal(x,
oldxx$SDgrid$S[K]),TRUE))
a2 <- sapply(xx$SDgrid$D,
function(x)i denti cal(al1.equal(x,
oldxx$SDgrid$D[K]),TRUE))
III.B.4 Page 45
-------
sz
CD
CO
CO
0
(/>
CO
<
J^
CO
C£
0)
o
a.
O
"O
0
CO
">
0)
cr
I j <- which(al & a2)
1 if Clength(j) == 1) {
= xx$SDgrid$LL[j] <- oldxx$SDgn'd$LL[K]
| xx$Fit[[j]] <- oldxx$Fit[[K]]
[ }
C\| I save(xx,fi1e=fi1e.path(savedir,fname))
I wri te.tab!e(fi negri ddb,fi1e="fi ne2gri ddb.csv",sep=",",col.names=NA)
\ =.
CO I
I
I.B.4Page46
------- |