DRAFT - DO NOT CITE OR QUOTE
EPA/635/R-ll/001Bb
United States www.epa.gov/iris
Environmental Protection
Agency
TOXICOLOGICAL REVIEW OF
METHANOL (NONCANCER)
APPENDICES
(CAS No. 67-56-1)
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
May 2013
NOTICE
This document is a Revised External Peer Review Draft. This information is distributed solely
for the purpose of pre-dissemination peer review under applicable information quality guidelines.
It has not been formally disseminated by EPA. It does not represent and should not be construed
to represent any Agency determination or policy. It is being circulated for review of its technical
accuracy and science policy implications.
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This information is distributed solely for the purpose of pre-dissemination peer review
under applicable information quality guidelines. It has not been formally disseminated by EPA.
It does not represent and should not be construed to represent any Agency determination or
policy. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
May 2013 ii Do Not Cite or Quote
-------
CONTENTS TOXICOLOGICAL REVIEW OF METHANOL (NONCANCER)
APPENDICES
(CAS No. 67-56-1)
LIST OF TABLES iv
LIST OF FIGURES v
LIST OF ABBREVIATIONS AND ACRONYMS vi
APPENDIX A. SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC COMMENTS, AND DISPOSITION A-1
A.I. EXTERNAL PEER REVIEW PANEL COMMENTS A-1
A.2. PUBLIC COMMENTS A-40
APPENDIX B. DEVELOPMENT, CALIBRATION, AND APPLICATION OF A METHANOL PBPK MODEL B-l
B.I. SUMMARY B-l
B.2. MODEL DEVELOPMENT B-2
B.3. MONKEY PK DATA AND ANALYSIS B-34
APPENDIX C. HUMAN CASE STUDIES C-l
APPENDIX D. RfC DERIVATION OPTIONS D-l
D.I. BENCHMARK DOSE MODELING SUMMARY D-l
D.2. RFC DERIVATIONS USING THE NEDO (1987) METHANOL REPORT D-2
D.3. RFC DERIVATIONS USING ROGERS ET AL. (1993B) D-25
D.4. RFC-DERIVATIONS USING BURBACHER ET AL. (1999A; 1999B) D-42
APPENDIX E. DOCUMENTATION OF IMPLEMENTATION OF THE 2011 NATIONAL RESEARCH
COUNCIL RECOMMENDATIONS E-l
References R-l
May 2013 iii Do Not Cite or Quote
-------
LIST OF TABLES
Table B-l Parameters used in the rat and human PBPK models B-6
Table B-2 Sensitivity of rat model dose metrics to fitted parameters B-15
Table B-3 Primate Km values reported in the literature B-27
Table B-4 HumanPBPK model sensitivity analysis for steady-state inhalation exposure B-29
Table B-5 PBPK model sensitivity analysis for oral exposure B-31
Table B-6 PBPK model predicted Cmax (Css) and 24-hour AUC for humans exposed to Methanol B-3 3
Table B-7 Monkey group exposure characteristics B-38
Table C-l Mortality rate for subjects exposed to methanol-tainted whisky in relation to their level of
acidosis C-2
Table D-l EPA PBPK model estimates of methanol blood levels (AUC)a in rat dams following inhalation
exposures and reported brain weights of 6 week old male pups D-3
Table D-2 Comparison of BMDi SD results for decreased brain weight in male rats at 6 weeks of age using
modeled AUC above background of methanol as a dose metric D-5
Table D-3 Comparison of BMD05 results for decreased brain weight in male rats at 6 weeks of age using
modeled AUC above background of methanol as a dose metric D-10
Table D-4 EPA PBPK model estimates of methanol blood levels (Cmax) in rat pups at 8 weeks following
inhalation exposures during gestation D-14
Table D-5 Comparison of BMDi SD results for decreased brain weight in male rats at 8 weeks of age using
modeled Cmax above background of methanol as a dose metric D-16
Table D-6 Comparison of BMD05 modeling results for decreased brain weight in male rats at 8 weeks of
age using modeled Cmax above background of methanol as a common dose metric D-21
Table D-7 Methanol blood levels (Cmax above background) in mice following inhalation exposures D-26
Table D-8 Comparison of BMD modeling results for 10% cervical rib incidence in mice using modeled
Cmax above background of methanol as a common dose metric D-27
Table D-9 Comparison of BMD modeling results for 5% cervical rib incidence in mice using modeled
Cmax above background of methanol as a common dose metric D-35
Table D-10 EPA PK model estimates of methanol blood levels (Cmax) above background in monkeys
following inhalation exposures and VDR test results for their offspring D-43
Table D-l 1 Comparison of BMD modeling results for VDR in female monkeys using AUC above
background of blood methanol as the dose metric D-44
Table E-1 National Research Council recommendations that EPA is implementing in the short term E-2
Table E-2 National Research Council recommendations that EPA is implementing in the long-term E-8
May 2013 iv Do Not Cite or Quote
-------
LIST OF FIGURES
Figure B-l Schematic of the PBPK model used to describe the inhalation, oral, and i.v. route
pharmacokinetics of methanol B-3
Figure B-2 Respiratory depression in Sprague-Dawley rats as a function of blood methanol concentration.
The empirical curve fit (solid line) was selected to describe the data with a minimal number of
parameters B-8
Figure B-3 Rat i.v.-route methanol blood kinetics B-10
Figure B-4 Model fits to data sets from inhalation exposures in female Sprague-Dawley rats B-ll
Figure B-5 Model simulations compared to 100 (squares) or 2,500 (diamonds) mg/kg oral methanol data
in female Sprague-Dawley rat (expanded scale in panel B) B-13
Figure B-6 Simulated Sprague-Dawley rat inhalation exposures to 500, 1,000, or 2,000 ppm methanol B-18
Figure B-7 Comparison of model predictions of urine concentration (from bladder compartment), venous
blood, body tissue, and urine concentration data for a 231 ppm, 8-hour exposure. Right axis
provides scale for venous blood and body tissue results B-20
Figure B-8 Urinary methanol elimination concentration (upper panel) and cumulative amount (lower
panel), following inhalation exposures to methanol in human volunteers B-22
Figure B-9 Blood methanol concentrations in subjects exposed for 30 min, 1 hr, or 2 hr at 800 ppm B-23
Figure B-10 Blood methanol concentrations in control (0 ppm) and methanol exposed (200 ppm) subjects B-23
Figure B-ll Oral exposure (10 mg/kg) to methanol in human volunteers (points) B-24
Figure B-12 Inhalation exposures to methanol in human volunteers B-25
Figure B-13 Intravenous exposure (10 mg/kg) to methanol in human volunteers (points) B-26
Figure B-14 Predicted human blood concentrations (increase above background) from total daily exposures
to 10 mg/kg-day methanol, consumed in a series of6 boluses.Time is from the first bolus of the
day. See text for further details B-30
Figure B-15 Blood methanol concentration data from NP and pregnant monkeys B-35
Figure B-16 Chamber concentration profiles for monkey methanol exposures B-37
Figure B-17 PBPK model predictions of changes in blood methanol levels in humans for exposures at the
RfCandRfD B-40
Figure D-l Hill model, BMR of 1 Control Mean S.D. - Decreased Brain weight in male rats at 6 weeks
age versus AUC above background, Fl Generation inhalational study D-8
Figure D-2 Hill model, BMR of 0.05 relative risk - decreased brain weight in male rats at 6 weeks age
versus AUC above background of methanol, Fj generation inhalational study D-13
Figure D-3 Exponential model 4, BMR of 1 control mean S.D. - Decreased brain weight in male rats at 8
weeks of age versus Cmax above background, gestation only inhalational study D-20
Figure D-4 Exponential model 4, BMR of 0.05 relative risk - Decreased brain weight in male rats at 8
weeks age versus Cmax above background, gestation only inhalational study D-25
Figure D-5 Nested logistic model, 0.1 extra risk - Incidence of cervical rib in mice versus Cmax above
background of methanol, GD6-GD15 inhalational study D-34
Figure D-6 Nested logistic model, 0.05 extra risk - Incidence of cervical rib in mice versus Cmax above
background of methanol, GD6-GD15 inhalational study D-42
Figure D-7 Third (3rd) degree Polynomial model, BMR of 1 control mean S.D. - VDR in female
monkeys using Cmax above background of blood methanol as the dose metric D-47
May 2013 v Do Not Cite or Quote
-------
LIST OF ABBREVIATIONS AND ACRONYMS
ACGIH
ADH
ADH1
ADH3
AIC
ALD
ALDH2
ALT
ANOVA
AP
AST
ATP
ATSDR
AUC
P-NAG
Bav
BMD
BMD1SD
BMDL
BMDL1S
1SD
BMDS
BMR
BSD
BUN
BW,bw
Ci pool
C-section
CA
CAR
CASRN
CAT
CERHR
American Conference of Governmental and
Industrial Hygienists
alcohol dehydrogenase
alcohol dehydrogenase-1
formaldehyde dehydrogenase-3
Akaike Information Criterion
aldehyde dehydrogenase
mitochondrial aldehyde dehydrogenase-2
alanine aminotransferase
analysis of variance
alkaline phosphatase
aspartate aminotransferase
adenosine triphosphate
Agency for Toxic Substances and Disease
Registry
area under the curve, representing the
cumulative product of time and
concentration for a substance in the blood
N-acetyl-beta-D-glucosaminidase
oral bioavailability
benchmark dose(s)
BMD for response one standard deviation
from control mean
95% lower bound confidence limit on
BMD (benchmark dose)
BMDL for response one standard deviation
from control mean
benchmark dose software
benchmark response
butathione sulfoximine
blood urea nitrogen
body weight
one carbon pool
peak concentration of a substance in the
blood during the exposure period
Cesarean section
chromosomal aberrations
conditioned avoidance response
Chemical Abstracts Service Registry
Number
catalase
Center for the Evaluation of Risks to
Human Reproduction at the NTP
CH3OH
CHL
CI
Cls
*~inax
CNS
C02
con-A
CR
CSF
Css
CT
CVB
CvBbg
CvBmb
CYP450
d,5,A
D2
DA
DIPE
DMDC
DNA
DNT
DOPAC
DPC
DTH
EFSA
EKG
EO
EPA
ERF
EtOH
F
Fo
Fi
F2
F344
FAD
FAS
FD
methanol
Chinese hamster lung (cells)
confidence interval
clearance rate
peak concentration
central nervous system
carbon dioxide
concanavalin-A
crown-rump length
Cancer slope factor
steady-state concentration
computed tomography
concentration in venous blood
background concentration in venous blood
concentration in venous blood minus
constant background
cytochrome P450
delta, difference, change
dopamine receptor
dopamine
diisopropyl ether
dimethyl dicarbonate
deoxyribonucleic acid
developmental neurotoxicity test(ing)
dihydroxyphenyl acetic acid
days past conception
delayed-type hypersensitivity
European Food Safety Authority
electrocardiogram
Executive Order
U.S. Environmental Protection Agency
European Ramazzini Foundation
ethanol
fractional bioavailability
parental generation
first generation
second generation
Fisher 344 rat strain
folic acid deficient
folic acid sufficient
formate dehydrogenase
May 2013
VI
Do Not Cite or Quote
-------
FP folate paired
FR folate reduced
FRACIN fraction inhaled
FS folate sufficient
FSH follicular stimulating hormone
y-GT gamma glutamyl transferase
g gravity
g, kg, mg, ug gram, kilogram, milligram, microgram
G6PD glucose-6-phosphate dehydrogenase
GAP43 growth-associated protein (neuronal growth
cone)
GD gestation day
GFR glomerular filtration rate
GI gastrointestinal track
GLM generalized linear model
GLP good laboratory practice
GSH glutathione
HAP hazardous air pollutant
HCHO formaldehyde
HCOO formate
Hct hematocrit
HEC human equivalent concentration
HED human equivalent dose
HEI Health Effects Institute
HERO Health and Environmental Research Online
(database system)
HH hereditary hemochromatosis
5-HIAA 5-hydroxyindolacetic acid
HMGSH S-hydroxymethylglutathione
Hp haptoglobin
HPA hypothalamus-pituitary-adrenal (axis)
HPLC high-performance liquid chromatography
HSDB Hazardous Substances Databank
HSP70 biomarker of cellular stress
5-HT serotonin
IL interleukins
i.p. intraperitoneal
IPCS International Programme on Chemical
Safety
IQ intelligence quotient
IRIS Integrated Risk Information System
IUR inhalation unit risk
i.v. intravenous
kj first-order urinary clearance
km
klv
KLH
KLL
Km
Km2
ksl
L, dL, mL
LD50
LDH
LH
LLF
LMI
LOAEL
M, mM, uM
MeOH
MLE
M-M
MN
MOA
4-MP
MRI
mRNA
MTBE
MTX
N2O/O2
NAD+
NADH
NET
NCEA
ND
NEDO
first-order urinary clearance scaling
constant; first order clearance of methanol
from the blood to the bladder for urinary
elimination
first order uptake from the intestine
first order methanol oral absorption rate
from stomach
rate constant for urinary excretion from
bladder
respiratory /cardiac depression constant
keyhole limpet hemocyanin
alternate first order rate constant
substrate concentration at half the enzyme
maximum velocity (Vmax)
Michaelis-Menten rate constant for low
affinity metabolic clearance of methanol
first order transfer between stomach and
intestine
liter, deciliter, milliliter
median lethal dose
lactate dehydrogenase
luteinizing hormone
(maximum) log likelihood function
leukocyte migration inhibition (assay)
lowest-observed-adverse-effect level
molar, millimolar, micromolar
methanol
maximum likelihood estimate
Michaelis-Menten
micronuclei
mode of action
4-methylpyrazole (fomepizole)
magnetic resonance imaging
messenger RNA
methyl tertiary butyl ether
methotrexate
nitrous oxide
nicotinamide adenine dinucleotide
reduced form of nicotinamide adenine
dinucleotide
nitroblue tetrazolium (test)
National Center for Environmental
Assessment
not determined
New Energy Development Organization
(of Japan)
May 2013
VII
Do Not Cite or Quote
-------
NIEHS National Institute of Environmental Health
Sciences
NIOSH National Institute of Occupational Safety
and Health
nmol nanomole
NOAEL no-observed-adverse-effect level
NOEL no-observed-effect level
NP nonpregnant
NR not reported
NRC National Research Council
NS not specified
NTP National Toxicology Program at NIEHS
NZW New Zealand White (rabbit strain)
OR osmotic resistance
ORD Office of Research and Development
OSF oral slope factor
OU oculus uterque (each eye)
OXA oxazolone
P, p probability
PB blood:air partition coefficient
PBPK physiologically based pharmacokinetic
model
PC partition coefficient
PEG polyethylene glycol
PFC plaque-forming cell
PK pharmacokinetic
PMN polymorphonuclear leukocytes
PND postnatal day
POD point of departure
ppb, ppm parts per billion, parts per million
PR body :blood partition coefficent
PWG Pathology Working Group of the NTP of
NIEHS
Q wave the initial deflection of the QRS complex
QCC cardiac output scaling constant
QP pulmonary (alveolar) ventilation
QRS portion of electrocardiogram corresponding
to the depolarization of ventricular cardiac
cells.
R2 square of the correlation coefficient, a
measure of the reliability of a linear
relationship.
RBC red blood cell
RfC reference concentration
RfD reference dose
RNA ribonucleic acid
R0bg zero-order endogenous production rate
ROS reactive oxygen species
S9 microsomal fraction from liver
SAP serum alkaline phosphatase
s.c. subcutaneous
SCE sister chromatid exchange
SD Sprague-Dawley rat strain
S.D. standard deviation
S.E. standard error
SEM standard error of mean
SGPT serum glutamate pyruvate transaminase
SHE Syrian hamster embryo
SOD superoxide dismutase
SOP standard operating procedure(s)
t time
T,/2, t/2 half-life
T wave the next deflection in the electrocardiogram
after the QRS complex; represents
ventricular repolarization
TAME tertiary amyl methyl ether
TAS total antioxidant status
Tau taurine
THF tetrahydrofolate
TLV threshold limit value
TNFa tumor necrosis factor-alpha
TNP-LPS trinitrophenyl-lipopolysaccharide
TRI Toxic Release Inventory
U83836E vitamin E derivative
UF(s) uncertainty factor(s)
UFA UF associated with interspecies (animal to
human) extrapolation
UFD UF associated with deficiencies in the
toxicity database
UFH UF associated with variation in sensitivity
within the human population
UFS UF associated with subchronic to chronic
exposure
Vd volume of distribution
Vmax maximum enzyme velocity
VmaxC maximum velocity of the high-affinity/low-
capacity pathway
VDR visually directed reaching test
VitC vitamin C
VPR ventilation perfusion ratio
v/v volume of solute/volume of solution
VYS visceral yolk sac
May 2013
VIM
Do Not Cite or Quote
-------
WBC white blood cell w/v weight (mass of solute)/volume of solution
WOE weight of evidence %2 chi square
May 2013 ix Do Not Cite or Quote
-------
APPENDIX A. SUMMARY OF EXTERNAL PEER
REVIEW AND PUBLIC COMMENTS, AND
DISPOSITION
A.1. External Peer Review Panel Comments
1 The draft toxicological review of methanol (U.S. EPA, 20lib) has undergone a formal
2 external peer review performed by scientists in accordance with EPA guidance on peer review
3 (U.S. EPA, 2006c), (U.S. EPA, 2000b). An external peer-review meeting was held July 22, 2011.
4 There were seven external peer reviewers. The external peer reviewers were tasked with
5 providing written answers to general questions on the overall assessment and on chemical-
6 specific questions in areas of scientific controversy or uncertainty. At the workshop, they
7 discussed their responses to each of the charge questions and consensus was not sought. A
8 summary of significant comments made by the external reviewers and EPA's responses to these
9 comments arranged by charge question follow. The summary quotes the reviewer comments
10 extensively, but synthesizes and paraphrases in some cases for the sake of clarity and
11 conciseness. The reviewers made a number of editorial suggestions to clarify specific portions of
12 the text. These changes were incorporated in the document as appropriate and are not discussed
13 further. EPA also received comments from the public. These comments and EPA's responses are
14 included in a separate section of this appendix.
A.1.1. "Toxicokinetics and PBPK Modeling"
A.I.1.1. Charge Al. Please comment on the scientific soundness of the PBPK
model used in this assessment.
15 Summary of Comments: In general, four reviewers stated that the PBPK model structure
16 was sound for the purposes of this assessment, a fifth reviewer stated that they noticed no
17 obvious flaws but could not comment on a technical level due to a lack of expertise, and two
18 reviewers did not explicitly state whether the PBPK models were sound but provided comments.
19 Several reviewers commented that the models were comprehensively documented and that stated
20 assumptions are justified for the purposes of this assessment. One reviewer commended the
21 Agency for "developing a consistent model framework and sets of species-specific parameters
22 which have been validated across several somewhat diverse data sets. " Another reviewer
23 commented that "the Sprague-Dawley (SD) rat PBPK model is inappropriately parameterized
May 2013 A-1 Draft - Do Not Cite or Quote
-------
1 (or insufficiently validated) for the inhalation route " and provided specific suggestions to
2 improve or validate the rat, mouse and human PBPK models developed by EPA for the purposes
3 of this assessment. Specific comments or suggestions made by the reviewers with respect to this
4 charge are described below, along with EPA responses.
5 Comment 1: One reviewer asked for "clarification of the process for evaluating the
6 usefulness of each model for the assessment and why the nonhuman primate model was not
7 included."
8 Response: EPA has described a framework (Chiu et al.. 2007: U.S. EPA. 2006a) useful to
9 evaluate models for inclusion in an IRIS assessment. This framework includes review of the
10 model purpose, model structure, mathematical representation, parameter estimation, computer
11 implementation, predictive capacity, and statistical analyses. Currently, there is no specific EPA
12 policy or criteria for PBPK model use in IRIS assessments; consequently, the usefulness of a
13 PBPK model for a given species and assessment is a matter of scientific judgment and a number
14 of EPA PBPK experts are involved in making this judgment. Specific criteria used in evaluating
15 methanol models are presented in Section 3.4.1.2.
16 The ability of the model to fit a wide range of experimental data with a single set of
17 parameters is one of the critical considerations. When a chemical-specific (e.g., methanol) model
18 is able to predict experimental data for a range of doses or exposure conditions, there is
19 confidence that the model can predict chemical-specific (e.g., methanol) pharmacokinetics under
20 exposure conditions for which one does not have data. Confidence that one or more animal
21 species are properly represented by the model increases EPA's confidence that the models can be
22 used to extrapolate test animal exposures to human exposures.
23 Regarding the nonhuman primate model, EPA had incorrectly stated [in Appendix C,
24 Section C.3 of the draft assessment (U.S. EPA, 20 lib)] that external concentrations were used
25 for dose-response modeling for the monkey. However, a nonhuman primate classical PK model
26 (not a PBPK model) was adapted for use in the draft assessment and is used in the final
27 assessment to estimate internal doses (blood methanol Cmax values) for derivation of internal
28 dose BMDLs associated with the Burbacher et al. (2004a; 1999a) monkey study. See Appendix
29 D, Section D.3, and Table D-10 in the final assessment.
30 Comment 2: One reviewer suggested that "the use [of] a bladder compartment is atypical
31 [thus] the EPA should consider receding the model to include a kidney/renal compartment that
32 considers excretion of methanol by the kidney."
33 Response: Urine passes through the bladder, which serves as a storage reservoir between
34 urine voids, so it is biologically realistic to include a compartment that represents this part of the
May 2013 A-2 Draft - Do Not Cite or Quote
-------
1 elimination pathway. Human urinary data are sufficient to identify a bladder residence-time
2 constant, but similar time-course data are not available for rats; therefore, the compartment only
3 impacts the human PBPK model. While only a small fraction of ingested/absorbed methanol is
4 excreted via the urinary elimination pathway, inclusion of a bladder compartment is significant in
5 that it allows a more precise fit to the human urine time-course data, which show a slight
6 nonlinearity in the human dosimetry.
7 Additionally, since the kidney glomeruli filter the blood directly, it is biologically realistic
8 to describe renal excretion as elimination directly from the blood compartment at a rate
9 proportional to the methanol concentration in blood. Inclusion of an explicit kidney compartment
10 of the form typically used for PBPK models would be most beneficial if the kidney was a target
11 tissue for toxicity. Development of a model of glomerular filtration for methanol would require
12 more extensive research, time, and resources. Addition of this type of compartment is not
13 expected to significantly impact the PBPK model predictions that are currently well predicted
14 and validated using the model structure applied in this assessment.
15 Comment 3: One reviewer questioned using a bladder component for only the human
16 model and suggested that the impact of a bladder component on the rodent models should be
17 tested.
18 Response: The bladder compartment is present in the model code used for both rats and
19 humans. The bladder compartment time constant (kM) was identified for humans, and urinary
20 excretion rates were plotted and compared with existing human urine data (Figure B-7). Similar
21 urinary excretion data are not available for rats; hence, the bladder time-constant (kw) cannot be
22 identified for that species. The use of the bladder compartment and kw has no impact on
23 predicted blood concentrations; and hence, no impact on any of the rat model predictions.
24 Comment 4: One reviewer suggested that descriptions (e.g., page 3-45 of the draft
25 assessment) of the two divergent models that were considered (Michaelis-Menten or not) are
26 confusing, and should be clarified.
27 Response: All discussion of this and other considerations in the development and
28 calibration of the human PBPK model have been consolidated into Appendix B, Section B.2.6,
29 and revised for clarification.
30 Comment 5: With respect to the monkey PK model, one reviewer commented that "the
31 description of the chamber volume [page 3-49 of the draft assessment] should be expanded" to
32 clarify the equipment in question and whether there is any evidence that incomplete mixing
33 occurred.
May 2013 A-3 Draft - Do Not Cite or Quote
-------
1 Response: The monkey PK model description and analysis is now found in Appendix B,
2 Section B.3.1 of this final assessment. Information regarding the chamber volume adjustment for
3 the PK model was added to this section. Briefly, the chamber volume was fit to the chamber
4 concentration data to allow for a better fit to the "mixing time" in the chamber (mixing time is
5 the time it takes for the chamber concentration to rise or fall after the inlet concentration is
6 turned on or off) and to account for the volume filled by the monkey and other chamber
7 equipment. The "accessible [chamber] volume" reported by Burbacher et al. (1999a) was
8 1,380 L, and the model fitted volume was 1,220 L. A detailed description of the chamber set-up
9 is included in the original Burbacher et al. (1999a) report and was not included in Appendix B of
10 this assessment.
11 Since methanol chamber time-course data were available for this study, and model
12 predictions did not match the data when the assumption of perfect mixing was used (chamber
13 volume of 1,380 L), EPA considered it reasonable to use the available data to calibrate the
14 residence- or mixing-time for the model of the chamber concentration. The text in Appendix B,
15 Section B.3.1 has been modified to indicate why Vch was varied.
16 Comment 6: With respect to the human PBPK and the monkey PK models, one reviewer
17 stated that EPA "has not clearly articulated why two different fractional absorption values were
18 used based on the same data base (see pages 3-50 (60%) and 3-42 (86.5%) of the draft
19 assessment)."
20 Response: In the external peer review draft document the derivation of the value of
21 FRACIN used for humans was described in detail in Appendix B, p. B-29. The exact mean
22 absorption measured by Sedivec et al. (1981) was 57.7%, but this was based on total ventilation.
23 However the human PBPK model uses alveolar ventilation, assumed to be 2/3 of total
24 ventilation, with the remaining 1/3 of each breath assumed to not enter the gas exchange area.
25 Therefore, to yield the same net uptake as Sedivec et al., 57.7% was divided by 2/3 to yield
26 86.55% for the human model parameter, FRACIN.
27 In contrast the monkey model used total respiration rather than alveolar ventilation. This
28 detail was included at the bottom of p. 3-49, where the monkey model parameter RC was defined
29 as "allometric scaling factor for total monkey respiration" (emphasis added). However the
30 parameter symbol "F" was used for the monkey "fraction of inhaled" to further distinguish it
31 from the human FRACIN, given that they are applied to different portions of total ventilation.
32 Further, in the description of the monkey parameter "F" on p. 3-50 (immediately after the value
33 of 60% is given), the document noted that 60% was the "(rounded)" value from Sedivec et al.
34 (1981) and went on to state, "F and Vmk cannot be uniquely identified, given the model structure,
35 so F was set to the (approximate) human value to obtain a realistic estimate of Vmk. For example,
May 2013 A-4 Draft - Do Not Cite or Quote
-------
1 if both F and Vmk are increased by 50%, then also increasing the fitted Vmax by 50% would yield
2 identical model fits. Any positive value could be assigned to F and it would not affect the
3 resulting model fits. Given that the airways in a 2-4 kg monkey are much smaller than those in a
4 70 kg human, it is unlikely that the transport characteristics for which F and FRACIN account
5 are identical in the two species. But a realistic value was considered desirable for the monkey, so
6 the approximate total ventilation human value was assumed to be sufficient.
7 Since the revised monkey model was ultimately not used in deriving the RfC or RfD, the
8 detailed explanation given here was not considered necessary and was not included in the draft
9 document. The human FRACIN has since been revised to 75% using additional data as noted by
10 another peer reviewer and described in detail in Appendix B, Section B.2. The value used for the
11 monkey model was therefore also revised to 50%, maintaining the 2/3 factor vs. human, and a
12 brief description provided. Reference to the Sedivec et al. (1981) paper was removed from this
13 part of the document, since a different data set is used.
14 Comment 7: Two reviewers noted that the blood methanol levels predicted by the EPA
15 rat PBPK model are much lower than the levels reported for SD rats on recently located
16 (supplied by industry at the peer review meeting) pages from the NEDO (1987) report and
17 Perkins et al. (1995a). One of the commenters further noted that if the fraction inhaled
18 (FRACIN) model parameter is changed from 20% to a value more consistent with the mouse
19 (66.5%) and human (86.6%) estimates, the 1,000 ppm blood prediction is in agreement with
20 Perkins et al. (1995a).
21 Response: In the draft assessment, inhalation data for F344 rats were used. Since the
22 current noncancer assessment does not use any bioassay data from F344 rats, all PK data (and
23 PBPK modeling results) for F344 rats have now been removed. The PBPK model in the final
24 assessment uses the SD rat inhalation data from Perkins et al. (1996a) for calibration, yielding a
25 more appropriate fitted value for FRACIN (81%) and model predictions that are more consistent
26 with the NEDO (1987) reported blood levels (See Sections B.2.2, B.2.4, and B.2.5).
27 Comment 8: One reviewer noted that, with respect to the Burbacher et al. (1999b)
28 monkey data, EPA has not justified why the second trimester group is considered the most
29 representative.
30 Response: When the data and fits shown in Figure 3-14 were evaluated, EPA noted that
31 overall there appears to be no significant or systematic difference among the NP and pregnant
32 groups. The solid lines, in the figure, are model simulations calibrated to only the 2nd trimester
33 data (details below), but they just as adequately represent average concentrations for the NP and
34 3rd trimester data. Likewise, a PK model calibrated to the NP PK data adequately predicted the
May 2013 A-5 Draft - Do Not Cite or Quote
-------
1 maternal methanol concentrations in the pregnant monkeys (results not shown). Since any
2 maternal:fetal methanol differences are expected to be similar in experimental animals and
3 humans (with the maternal:fetal ratio being close to one due to methanol's high aqueous
4 solubility and relatively limited metabolism by the fetus), the predicted levels for the 2nd
5 trimester maternal blood are used in place of measured or predicted fetal concentrations.
6 Thus the primary justification for only showing the results for the 2nd trimester is that it
7 does not matter which stage one selects, since there is not a significant difference in either the
8 data or the model fits among the stages. While there is no clear effect of pregnancy on the PK in
9 monkeys, to the extent that there is some trend (for example, if the AUC decreases slightly with
10 the extent of pregnancy), the value of the metric during the 2nd trimester was expected to be in
11 between the values for the 1st and 3rd trimesters, hence closer to an overall average. In short,
12 because the physiological changes induced by pregnancy are at an intermediate stage in the 2nd
13 trimester relative to the 1st and 3rd, PK parameters were expected to also be intermediate and
14 therefore most representative of the average over all of pregnancy. However, had there been clear
15 time-dependence in the PK data, a quantitative analysis could have been used to incorporate that
16 trend.
17 Comment 9: One reviewer suggested that the Km values estimated by the rat and human
18 models "don't seem to reflect the true Michaelis values of the metabolic enzymes themselves."
19 Response: Since methanol is metabolized by multiple enzymes with differing Km values,
20 at best one would expect the empirical Km values identified here to represent an average of the
21 enzyme-specific values (weighted by the contribution of each enzyme to total metabolism).
22 Further, it is quite typical to find that in vivo PK data are not well-predicted when Km values
23 measured in vitro are used in a model, hence Km values estimated from in vivo data are not
24 expected to be identical to values measured in vitro. The Km values identified for the revised
25 PBPK models described here are 28 mg/L for rats and 40 mg/L for humans. Pollack and Brouwer
26 (1996) analyzed the kinetics of formaldehyde formation in vitro and estimated Km = 39.3 mg/L
27 using nonpregnant adult rat liver homogenates and Km = 35.5 mg/L with GD 20 homogenates:
28 quite similar to the revised value for the rat used in the final assessment. Mani et al. (1970)
29 measured methanol kinetics with human liver ADH and obtained a Km of 48 mg/L. This is
30 likewise quite similar to the Km estimated here with the human PBPK model.
31 Comment 10: With respect to the human PBPK model, one reviewer noted that useful
32 human kinetic studies (Haffner et al., 1992; Schmutte et al., 1988) were overlooked, and that
33 these studies "are potentially quite valuable in model parameterization because they do not
34 involve the inhalation route."
May 2013 A-6 Draft - Do Not Cite or Quote
-------
1 Response: Previously only inhalation data was included for humans. These studies
2 [(Haffner et al., 1992) and (Schmutte et al., 1988)] provide i.v. and oral data, and they have been
3 added to the PBPK analysis (see Appendix B, Section B.2.6). Specifically, the oral data are now
4 used for model calibration, allowing identification of human oral absorption rate constant and
5 bioavailability. The i.v. data from Haffner et al. (1992) are from only four individuals; these data
6 were used to validate the model by comparing model predictions following an i.v. dose with the
7 experimental data (Figure B-13).
8 Comment 11: One reviewer recommended that EPA perform sensitivity analyses of the
9 rat and human PBPK modeling results under conditions approximating the BMDL, stating that
10 "at a minimum, EPA should assess whether or not the model they used in the risk assessment can
11 (adequately) simulate the additional human data identified herein and conduct and provide
12 human model sensitivity analyses at the RfC and RfD."
13 Response: A sensitivity analysis has been conducted and a detailed description of this
14 analysis is included in Appendix B, Section B.2.7. However, such an analysis can only partly
15 inform the question of model adequacy, which is addressed in more detail in the response to
16 Charge Al Comment 1 above.
17 Comment 12: With respect to the mouse PBPK model, one reviewer stated that "it seems
18 odd that, for oral dosing, the mouse blood levels are reported to be insensitive to any parameter
19 related to clearance (e.g., metabolism, blood flow to the liver) (pp B-16 and B-18 of the draft
20 review)," and requested clarification in the text regarding the type of oral dose that is being
21 simulated.
22 Response: Since direct measurements of mouse (CD-I) blood concentrations for
23 bioassay exposures are available (Rogers et al., 1993b) and used for the BMD analysis in this
24 final assessment, the mouse PBPK model is not utilized in the final assessment to estimate an
25 internal dose metric. Therefore the description, analysis, and discussion of the mouse model are
26 not included in the final assessment.
27 Comment 13: One reviewer commented that "the runtime files that should reproduce
28 Figures B-2 and B-5 yield simulations that are slightly off." The reviewer also commented,
29 regarding Figures B-6, B-7, and B-8, that "these files do not accurately reproduce the figures in
30 the document."
31 Response: The figures were produced with the background turned on, while the PBPK
32 runtime files had the background turned off. The current version of the PBPK model inclusive of
33 the runtime files (available electronically on the IRIS website [www.epa.gov/iris]) will exactly
May 2013 A-7 Draft - Do Not Cite or Quote
-------
1 reproduce the figures in the toxicological review, aside from legend placement, which is
2 dependent on acslX window sizes.
3 Comment 14: With respect to the mouse PBPK model sensitivity analysis, one reviewer
4 noted that "EPA does not provide files that fully recreate the sensitivity analyses—only those
5 parameters demonstrated in Figures B-6, B-7, and B-8." This reviewer commented that "the
6 sensitivity analysis does not appear to have been comprehensive," and cited FRACIN as an
7 example of a parameter that was not tested, yet seems to be a parameter to which the mouse
8 PBPK model is sensitive.
9 Response: As stated in the response to Charge Al Comment 12, the mouse model has
10 been removed from the final assessment; thus no sensitivity analysis is included for the mouse
11 PBPK model.
12 Comment 15: One reviewer commented that "it is not clear why two saturable metabolic
13 pathways are needed for the Sprague-Dawley rat and only one for the F344 rat."
14 Response: The liver metabolism in the SD rat is now described using a single saturable
15 rate equation. As discussed in response to Charge Al Comment 7, the analyses of F344 rat PK
16 data has been removed from the toxicological review.
A.I.1.2. Charge A2. Please comment on the scientific justification for the
subtraction of background levels of methanol from the data in relation to the
quantification of non-cancer risks.
17 Summary of Comments: EPA stated two key assumptions for this approach in the peer
18 review draft: "(1) endogenous levels do not contribute significantly to the adverse effects of
19 methanol or its metabolites; and (2) the exclusion of endogenous levels does not significantly
20 alter PBPK model predictions. " Most reviewers were in general agreement with the first
21 assumption, but expressed the need to better characterize background levels of methanol and
22 their relationship to the RfC/D. Three reviewers were concerned that the first assumption, and
23 the subtraction of methanol background levels, gives the impression that endogenous methanol
24 levels are not important. With respect to the second assumption, none of the reviewers disagreed
25 with EPA's determination that the exclusion of endogenous levels does not significantly alter
26 PBPK model predictions; however, two reviewers advocated the use of a PBPK model that
27 incorporates a background term and one reviewer favored the use of the simpler PBPK model
28 (without background levels). Specific comments or suggestions made by the reviewers with
29 respect to this charge are described below, along with EPA responses.
May 2013 A-8 Draft - Do Not Cite or Quote
-------
1 Comment 1: Three reviewers expressed concerns over the first assumption, that
2 endogenous levels do not contribute significantly to adverse effects. One reviewer stated that
3 EPA was giving the impression that "cumulative exposures from different sources are not
4 important." A second reviewer indicated that the first assumption is not met because "the RfC
5 and RfD correspond to blood methanol concentrations in humans squarely in the range of normal
6 'background' levels."." The third reviewer asked "If endogenous levels of methanol do not
7 contribute to adverse effects and an exposure does not produce an increase above background
8 levels, how can that exposure lead to an adverse effect?"
9 Response: The language in the draft assessment may have confused this issue and has
10 been clarified in Section 3.4.3.2. EPA acknowledges that endogenous methanol concentrations
11 can be a contributing factor in health effects that are associated with exogenous methanol
12 exposure. As indicated in response to Comment 2 below, for the sake of obtaining more accurate
13 and reliable toxicokinetic estimates, the PBPK models used in the final assessment incorporate
14 background/endogenous concentrations of methanol. Background estimates on which the models
15 were calibrated are described in Appendix B, Sections B.2.3 (rats), B.2.8 (humans) and B.3.1
16 (monkeys). However, for BMD modeling of laboratory animal dose-response data, the species-
17 specific background estimates were subtracted from the dose metric predicted under the relevant
18 bioassay conditions. This approach takes into account the impact of endogenous levels on the
19 toxicokinetics of methanol, and allows for the derivation of an RfC (or RfD) that is, by
20 definition, a population level estimate (including sensitive populations) of the amount of a
21 substance that a person can inhale (or ingest) every day over the course of a lifetime [above
22 endogenous levels] without an appreciable risk of harm.
23 As pointed out by the 2nd and 3rd reviewers, the relationship between the RfD and RfC
24 and endogenous blood levels is an important consideration. Measured blood concentrations of
25 methanol in humans range between 0.25 mg/L and 5.2 mg/L (see Table 3-1). As described in a
26 new Section 5.3.6, PBPK model estimates of maximum blood level increases of 0.44 mg/L and
27 0.41 mg/L associated with an RfD or RfC, respectively, are within the 0.7 mg/L standard
28 deviation estimated for the average methanol blood levels (1.5 ± 0.7 mg/L) in humans. From this
29 analysis EPA concludes that the estimated increase in blood levels of methanol from exogenous
30 exposures at the level of the RfD or the RfC (or from the RfC + RfD) are distinguishable from
31 natural b ackground vari ati on.
32 Comment 2: Two reviewers advocated the use of a PBPK model that incorporates a
33 background term and one reviewer favored the use of the simpler PBPK model. One of these
34 reviewers indicated that use of a background term would be "more rigorous and appropriate for
35 use in this assessment." The latter reviewer warned that "Including the background levels in the
May 2013 A-9 Draft - Do Not Cite or Quote
-------
1 models necessarily increases the model complexity and like any model enhancement may
2 increase the uncertainty in the final result, especially when as in this case it may be difficult to
3 design a test of its validity."
4 Response: As described in the response to Comment 1 above, EPA re-calibrated the
5 PBPK models to account for species-specific estimates of background/endogenous production of
6 methanol. For humans the model was tuned to have an average background level of 1.5 mg/L
7 determined from the corresponding human data in Table 3-1; for rats the model was tuned to a
8 background level of 3 mg/L from the corresponding (control) rat data in Table 3-5. These revised
9 PBPK models were used in estimation of internal dose metrics for the derivation of the RfD and
10 RfC. This addition did increase the model complexity by including an additional term (Robg, a
11 zero-order endogenous production rate, see Appendix B, Section B.2.1) for endogenous
12 production of methanol; however, this term was estimated using human data for background
13 blood methanol concentrations (Table 3-1). Since the background term was tuned to match
14 average observed background levels in rats and humans for the corresponding models, there
15 should be minimal systematic error or bias due to the incorporation of the background term; i.e.,
16 the average background level is neither under- nor over-predicted by the model. Moreover,
17 adding the term resulted in only minor changes, less than 20%, in model-predicted blood levels
18 at higher exposure levels (i.e., in the range of the bioassays for rats or the HEC and HED values
19 estimated for humans). Hence model predictions are not sensitive to the inclusion of these
20 average background levels vs. no background at all, and so the effect of and uncertainty due to
21 possible small changes (or errors) in the background term will be minimal.
22 Comment 3: One reviewer stated that "the upper bound on background concentrations of
23 methanol in target tissue should be carefully evaluated" and that "the lack of determination of the
24 upper statistical bound on normal physiological concentrations of methanol in relevant species,
25 including humans, can be considered to be a major deficiency of the reviewed document."
26 Response: Statistical bounds on normal physiological concentrations of methanol cannot
27 be determined for all tissues and species. The most complete dataset exists for blood levels of
28 methanol in humans. A discussion of endogenous background levels of methanol and their
29 relationship to the RfC and RfD has been added to Chapter 5 (Section 5.3.6) and elsewhere in the
30 toxicological review. There is a scarcity of data for endogenous methanol levels in the general
31 population. Also, the existing data (Table 3-1) is from populations with various (e.g., age, gender,
32 cultural) characteristics that were asked to adhere to a variety of diets, generally restricted of
33 food and drink that contain or convert to methanol. Measured values have been documented as
34 low as 0.25 mg/L and as high as 5.2 mg/L. From the data gathered for this document (Table 3-1),
May 2013 A-10 Draft - Do Not Cite or Quote
-------
1 EPA has estimated a mean background methanol level of 1.5 mg/L with an approximated
2 standard deviation of 0.7 mg/L (see Section 5.3.6).
3 Comment 4: One reviewer noted that, "in the simulations whose results are listed in the
4 Table B-5, a background level of 2 mg/L has been set to model human internal concentration
5 from inhalation (page B-92; line 29) but not from the oral exposure (page B-92; line 55)."
6 Response: This inconsistency was corrected and the values in Table B-5 have been
7 updated for both inhalation and oral exposures to reflect concentrations above endogenous
8 background.
9 Comment 5: One reviewer noted that EPA did not adequately explain the modest
10 differences in HED and HEC predictions from the PBPK models when background levels of
11 methanol were included or excluded.
12 Response: This comment was made in relation to a discussion of why including
13 background in the PBPK models might not be necessary. That discussion was removed from the
14 assessment because, as described above in response to Charge A2 Comment 2, the final versions
15 of the PBPK models do include background.
A.I.1.3. Charge A3. The PBPK modeling effort assumed similar methanol
pharmacokinetics between pregnant and non-pregnant animals. Please
comment on the adequacy of the dose-metric extrapolation based on a PBPK
model for non-pregnant adults (i.e., no fetal compartment) for predicting
risks associated with fetal/neonatal brain concentrations of methanol.
16 Summary of Comments: All reviewers agreed that the existing literature supports the
17 assumption of similar pharmacokinetics between pregnant and nonpregnant animals. Specific
18 comments or suggestions made by the reviewers with respect to this charge are described below,
19 along with EPA responses.
20 Comment 1: One reviewer stated they understood the rationale for omitting a fetal
21 compartment in the PBPK model, but felt that "for PBPK modeling to be effective, a fetal
22 compartment will ultimately be needed." This reviewer noted that "PBPK modeling is most
23 useful when the proximate form of the toxicant and mode of action are known, which is
24 unfortunately not the case with developmental effects of methanol."
25 Response: EPA agrees that a PBPK model with a fetal compartment would be ideal, and
26 that more mode of action information, including the identification of the proximate toxicant,
27 would be helpful. However, studies have shown, and reviewers have agreed, that methanol
28 pharmacokinetics between pregnant and nonpregnant animals are similar and, absent additional
May 2013 A-11 Draft - Do Not Cite or Quote
-------
1 information, provide a reasonable justification for extrapolation based on a PBPK model for non-
2 pregnant adults. If there are studies to the contrary, or studies that provide insight into fetal
3 metabolism or the embryotoxic moiety of methanol, a fetal compartment may be considered in
4 the future.
5 Comment 2: One reviewer expressed concern over the model's ability to predict neonatal
6 blood levels, stating that "this issue is important since the critical study used by EPA to derive an
7 RfC involved combined gestational and lactational (inhalational) exposure of neonates" and that
8 "the use of an adult-based PBPK model could under predict potentially 'toxic' blood methanol
9 concentrations."
10 Response: It is recognized that neonatal blood levels will likely be higher than maternal
11 blood levels of methanol. Therefore, the ratio of blood concentrations between a human infant
12 and its mother is not expected to be significantly greater than the approximate 2-fold difference
13 that has been observed between rat pups and dams. Further, as stated in the final version of
14 Section 5.1.3.2.2, "the health-effects data indicate that most of the effects of concern are due to
15 fetal exposure, with a relatively small influence due to postnatal exposures." For these reasons
16 and because EPA has confidence in the ability of the PBPK model to accurately predict adult
17 blood levels of methanol, the maternal blood methanol levels for the estimation of HECs from
18 the NEDO (1987) study were used as the dose metric.
A.I.1.4. Charge A4. EPA assumes limited methanol metabolism in the fetus
because of limited alcohol dehydrogenase (ADH) activity in the human fetus,
limited catalase and ADH activity in fetal rodents, and existing
pharmacokinetic data that show nearly equal concentrations in maternal
blood vs. the fetal compartment. Please comment on the validity of this
assumption given the lack of data regarding potential alternate metabolic
pathways in the fetus.
19 Summary of Comments: All reviewers agreed that this is a reasonable assumption given
20 the limited data available. Specific comments or suggestions made by the reviewers with respect
21 to this charge are described below, along with EPA responses.
22 Comment 1: Two reviewers thought that the assumption of limited methanol metabolism
23 in the fetus was valid based on the methanol pharmacokinetic data, but one of the reviewers
24 noted that embryotoxicity from methanol may be influenced by fetal catalase in mice as
25 demonstrated by a recent study (Miller and Wells, 2011). This reviewer further stated that this
26 and another study (Sweeting et al., 2011) suggest that fetal methanol concentrations in rodents
27 may not be a "good predictor of teratogenic responses in different species.".
May 2013 A-12 Draft - Do Not Cite or Quote
-------
1 Response: While these studies provide insights into the fetal metabolism of methanol, it
2 is unknown if fetal catalase is a controlling factor for methanol teratogenicity in mice.
3 Furthermore, a recent in vivo study (Siu et al., 2013) suggests that high catalase activity does not
4 protect against methanol teratogenicity in the strains of mice tested. EPA evaluated these studies
5 and, as described in Section 5.3.5 "Choice of Species/Gender," concluded that the available
6 evidence related to fetal catalase and methanol's teratogenicity in mice is contradictory and
7 inadequate to suggest that rodent effects should not be used in an assessment of methanol's
8 potential to cause developmental effects in humans. Also, because the critical gestational window
9 for developmental effects could be different for rabbits versus mice, the claim that rabbits are
10 resistant to teratogenic effects of methanol needs to be verified over several gestational days, as
11 has been done for mice.
12 Comment 2: One reviewer commented that, "the assumption of limited methanol
13 metabolism in the fetus is probably justified based on the existing studies showing low levels of
14 ADH and catalase in fetal tissues" but added that "these studies have technically measured these
15 proteins using indirect measures such as immunoblotting showing protein amounts or activity
16 measures with ethanol as the substrate."
17 Response: EPA agrees with this comment. An activity measurement using methanol as
18 the substrate would be ideal. However, lacking such studies, it is reasonable to assume low
19 activity of methanol metabolism in fetal tissues from relevant, indirect studies.
A.I.1.5. Charge A5. Please comment on the scientific justification of the
extrapolation approach from rats to humans for in-utero and neonatal
lactational and inhalation exposures.
20 Summary of Comments: Four reviewers agreed that a reasonable approach was taken
21 given the data available, though one of these reviewers reiterated that issues identified in Charge
22 Al with respect to the rat and human models need to be addressed. A fifth reviewer reiterated
23 comments made in response to Charge Al regarding the need for "clarification of the process for
24 evaluating the usefulness of each model for the assessment and why the nonhuman primate
25 model was not included" and noted that the use of the NEDO rat studies which included
26 neonatal exposures is "problematic, given the lack of data on lactational and early postnatal
27 inhalation exposure to methanol. " A sixth reviewer suggested that the model should be modified
28 to include gestational and lactational components and expressed concern over the use of rodent
29 data for estimating human risk from developmental effects. A seventh reviewer suggested that
30 EPA's assumption that rats and humans would have similar maternal/offspring methanol
31 concentration ratios is a significant source of uncertainty.
May 2013 A-13 Draft - Do Not Cite or Quote
-------
1 Response: EPA agrees with the majority of the reviewers that the extrapolation approach
2 employed is justified given the available data. The reviewer concerns regarding the model
3 evaluation process and the perceived lack of a nonhuman primate model are addressed in
4 response to Charge Al Comment 1. The lack of data on lactational and early postnatal inhalation
5 exposure to methanol is a recognized data gap that led to the current approach. As discussed in
6 response to comments under Charge A3 above, gestational and lactational compartment may be
7 considered in a future assessment, but they are not necessary at this time for the purposes of this
8 toxicological review. Concerns over use of rodent studies stem from the Sweeting et al. (2011)
9 study. The relevance of the Sweeting et al. (2011) to these concerns is discussed in Section 5.3.5
10 and elsewhere in the toxicological review and in response to Comment 1 of Charge A4 and
11 Comment 1 of Charge D2. As discussed in response to Charge A3 Comment 2 and Section
12 5.1.3.2.2 of the toxicological review, the uncertainty surrounding the assumption of similar
13 maternal/offspring methanol concentration ratios between rats and humans is recognized, but the
14 ratio of blood concentrations between a human infant and its mother is not expected to be
15 significantly greater than the approximate 2-fold difference that has been observed between rat
16 pups and dams. Clarifications have been added in this regard to Sections 5.1.3.2.2 and 5.3.5.
A.1.2. Charge B: "Inhalation Reference Concentration (RfC) for Methanol"
A.l.2.1. Charge Bl. A chronic RfC for methanol has been derived from a
perinatal inhalation study of the effects from exposing rat dams and pups to
methanol during gestation and lactation (NEDO, 1987). Reference values
from mouse (Rogers et al., 1993b) and monkey (Burbacher et al., 2004a;
Burbacher et al., 1999a) developmental studies, were also derived and
discussed, but were not chosen for the RfC. Please comment on whether the
selection of the principal study has been scientifically justified.
17 Summary of Comments: Two reviewers indicated that selection of NEDO (1987) as the
18 principal study was scientifically justified. Two reviewers stated that choice of the NEDO rat
19 study was based on "practical/technical grounds" or "policy" (i.e., use of the lowest RfC),
20 rather than scientific considerations. One reviewer did not explicitly state whether the use of the
21 NEDO rat study was scientifically justified, but stated that selection of the principal study is
22 contingent on the determination of the HEC/HED after the implementation of suggested model
23 revisions (e.g., Charge Al Comment 7). Two reviewers suggested that the NEDO rat
24 developmental study was not the most appropriate study for RfC derivation. Specific comments
25 or suggestions made by the reviewers with respect to the advantages and limitations of each of
26 the three studies addressed in the charge are described below, along with EPA responses.
May 2013 A-14 Draft - Do Not Cite or Quote
-------
1 Comment 1: Two reviewers indicated that the selection of NEDO (1987) as the principal
2 study was scientifically justified and noted the following scientific advantages:
3 • "The nearly continual exposure (20-22 hours per day depending on the study) represents the
4 types of exposures relative to the RfC/RfD (i.e. the daily exposure over the lifetime)."
5 • Selection of the NEDO study "is in accordance with the usual guidelines recommending use
6 of the study with the best data quality (including, in this case, availability of a validated
7 PBPK model) and greatest sensitivity."
8 Two reviewers did not think that the NEDO study was the most appropriate choice and
9 noted the following concerns:
10 • The prior, EPA-sponsored peer review of the NEDO study questioned "procedures used in
11 the NEDO study (in utero and postnatal exposures, litter effects, etc) that make it difficult to
12 evaluate the study for RfC derivation."
13 • "The discussion on page 5-10 regarding the complications that arise from using the NEDO
14 study where exposure was both gestational and postnatal postulates a number of assumptions
15 that are supported by little or no data."
16 • "Data on lactational transfer and early postnatal inhalation exposures are limited."
17 • The neonatal brain weight response has not been replicated in other studies.
18 • "The analysis provided by the NEDO authors showed a gender difference (effects seen in
19 males but not female rats)."
20 • "The NEDO study relied on multiple t-tests as opposed to a more appropriate use of an
21 ANOVA to evaluate gender and treatment responses."
22 • There were no "corroborating clinical or pathological observations of depressed CNS activity
23 noted in the rats in the NEDO study."
24 The remaining three reviewers did not address the scientific merits of the NEDO study.
25 However, two of these reviewers suggested that its selection was based on it resulting in the
26 lowest RfC, and one stated that selection of the principal study is contingent on the determination
27 of the HEC/HED and suggested that (due to possible error in the PBPK model) the HEC/HED
28 value for the NEDO rat study "could be on the order of 6-fold too low."
29 Response: In addition to the advantages of the NEDO (1987) developmental rat study
30 noted by several reviewers (e.g., relevant exposure route and duration, validated PBPK model
31 estimates of internal dose, and a sensitive response endpoint), the NEDO study offers other
32 advantages (described further below) such as the identification of an endpoint that (1) is
May 2013 A-15 Draft - Do Not Cite or Quote
-------
1 biologically significant, (2) is observed at a sensitive developmental stage, (3) has been
2 replicated in adult rats and (4) is in an organ system for which suggestive pathology has been
3 observed in adult primates that received acute and chronic exposure to methanol via the same
4 exposure route. While reviewer comments on EPA's choice of the NEDO (1987) developmental
5 rat study were mixed, only two of seven reviewers indicated that its selection was not justified.
6 Though EPA recognizes that the NEDO (1987) study has limitations, these limitations do not
7 preclude its use as the principal study for RfC derivation (see Section 5.3.1 and below).
8 EPA agrees that data on lactational transfer and early postnatal inhalation exposures are
9 limited, and this is largely the reason that maternal blood levels were used as a dose metric in the
10 analysis of the neonatal brain weight endpoint. The related discussion that was on page 5-10 of
11 the draft assessment has been revised to clarify the Agency's justification and address reviewer
12 concerns regarding the use of maternal (versus neonatal) blood levels of methanol as a basis for
13 the benchmark dose analysis of these data. In essence, the ratio of the difference in blood
14 concentrations between a human infant and mother is assumed to be similar to the approximate
15 two-fold difference that has been observed in rats. Further, while rat studies indicate that
16 postnatal exposure to methanol can impact brain weight, fetal exposure has been shown to have
17 the greatest influence on this endpoint. For these reasons and because EPA has confidence in the
18 ability of the PBPK model to accurately predict adult blood levels of methanol, the maternal
19 blood methanol levels for the estimation of FtECs from the NEDO (1987) study were used as the
20 dose metric. EPA has added text to Sections 4 and 5 to further clarify and discuss the limitations
21 of NEDO (1987).
22 NEDO (1987) observed brain weight reductions in the Fl and F2 generations of their two
23 generation study, in the Fl generation of the supplementary developmental study to the two
24 generation study and in a separate teratogenicity study. They also observed potentially adverse
25 histopathology (astrocytes) in the brains of monkeys receiving acute, subchronic and chronic
26 exposure to methanol (see further discussion in Section 4.4.2). While brain weight reduction has
27 not been observed in developmental bioassays of other laboratories, it has been observed in adult
28 rats exposed to methanol (TRL, 1986). Also, brain weight reduction is not an endpoint that has
29 been extensively measured or focused on in other developmental studies of methanol (e.g., the
30 Rogers et al. mouse studies).
31 EPA agrees that the multiple t-tests applied in the NEDO study are not optimal for the
32 evaluation of the dose-response data from this study. For this reason, EPA did not rely on this
33 information and, instead, relied on the results of the more definitive benchmark dose analysis of
34 this data (Appendix D), as described in Section 5 of the final methanol toxicological review.
May 2013 A-16 Draft - Do Not Cite or Quote
-------
1 With respect to the use of absolute brain weight change without clinical or pathological
2 corroboration, the Agency's neurotoxicity guidelines (U.S. EPA, 1998a) states that a "change in
3 brain weight is considered to be a biologically significant effect," and further states that "it is
4 inappropriate to express brain weight changes as a ratio of body weight and thereby dismiss
5 changes in absolute brain weight" and that "changes in [absolute] brain weight are a more
6 reliable indicator of alteration in brain structure than are measurements of length or width in
7 fresh brain, because there is little historical data in the toxicology literature."
8 With respect to the basis for EPA's study choice, while it is true that EPA guidelines
9 generally promote use of the more sensitive endpoint, the relative strengths of candidate studies
10 are not to be ignored (U.S. EPA, 2002, 1994b). As discussed above and in Chapter 5 of the
11 methanol toxicological review, the NEDO (1987) study limitations were considered, but do not
12 preclude its use for the derivation of a candidate RfC/D. On the other hand, questions concerning
13 the Burbacher et al. (2004a; 1999a) monkey study dose-response are considered serious enough
14 to not use this study for RfC/D derivation, despite the possibility that a lower BMDL POD would
15 have been derived from this study (see Section 5.3.1 and Appendix D).
16 Comment 2: Regarding the Burbacher et al. monkey study, four reviewers had no
17 comment on its potential for use as the principal study and two reviewers stated the following
18 reasons why it should not be used as the principal study:
19 • "The lack of a dose-response function for the major effects."
20 • No "convincing evidence of an effect, given the inconsistencies in dose-response, multiple
21 comparisons, and the potential for unreliable identification of 'effects' in small studies."
22 However, one reviewer suggested that it would be a better choice than the NEDO rat
23 study because it "uses the most appropriate species (monkey) and examined a wide range of
24 reproductive and neurotoxicological endpoints and significant pharmacokinetic data," and two
25 reviewers suggested that the following limitations noted in the toxicological review were
26 overstated:
27 • Inclusion of wild-caught monkeys
28 • Influence of C-sections on results
29 • Not being relevant to persons who are folate deficient
30 • Lack of a dose-response for VDR in the male monkeys
31 Response: EPA agrees with reviewer comments regarding the significant difficulties of
32 assessing the dose-response data from the Burbacher et al. (2004a; 1999a) monkey study. These
33 concerns are addressed in Section 5.3.1 of the final review. In response to reviewer concerns,
May 2013 A-17 Draft - Do Not Cite or Quote
-------
1 EPA's attempt at performing a benchmark dose analysis of the Visually Directed Reaching
2 (VDR) endpoint from this study (described in Appendix D) are no longer presented in Chapter 5
3 (i.e., in Table 5-4 or 5-5) alongside the benchmark dose analyses of critical effects from the
4 candidate principal mouse and rat studies. With respect to the concerns that limitations in this
5 study were overstated, EPA has taken the following action:
6 • Inclusion of wild-caught or feral-born monkeys - One section of the draft toxicological
7 review inadvertently referred to monkeys from this study as being "wild" and this statement
8 has been removed. In two sections, they were referred to as "a mixture of feral-born and
9 colony-bred animals." Since the Burbacher et al. (2004a; 1999a) study investigated for and
10 found no effects that were dependent on origin, EPA agrees that this statement is unnecessary
11 and it has been removed from the review.
12 • Influence ofC-sections on results - EPA agrees with the commenters and the toxicological
13 review has been edited to reflect that Ceasarean section (C-section) deliveries performed in
14 the methanol exposure groups did not impact the "decreased length of pregnancy" finding
15 (decreased length of pregnancy was observed in vaginally delivered animals).
16 • Not being relevant to persons who arefolate deficient - EPA agrees that this statement could
17 be made about most of the methanol studies reviewed. Hence the statement has been
18 removed.
19 • Lack of a dose-response for VDR in the male monkeys - While the ANOVA test in the male
20 monkeys suggests a statistical significant VDR change at 600 ppm (p = 0.007), there was no
21 significant difference between responses and/or variances (indicating lack of a dose-response
22 trend) among the dose levels for males only (p = 0.321), even when the high dose group is
23 excluded (p = 0.182). However, there was a significant dose-response trend for females only
24 (p = 0.0265). This is largely because the females had a larger overall sample size across dose
25 groups than males (21 females versus 13 males). Hence, only the VDR response for females
26 only exhibited a dose-response that could be adequately modeled (see Appendix D).
27 Comment 3: Regarding the Rogers et al. mouse study, two reviewers supported the use
28 of this study over the NEDO rat study and noted the following advantages:
29 • "The study is scientifically sound and robust."
30 • "Exposures are limited to the prenatal period and the outcomes are clear."
31 • "The Rodgers (sic) study has undergone independent peer review, documents responses
32 reported by other laboratories, and has quite robust group sizes."
May 2013 A-18 Draft - Do Not Cite or Quote
-------
1 Response: EPA agrees with reviewer comments regarding the advantages of the Rogers
2 et al. mouse developmental study. EPA also agrees with reviewer comments (see Charge Bl
3 Comment 1 above) regarding the advantages of the NEDO rat study, including the use of a
4 continuous, nearly full day exposure regimen and the adequacy of the reported response data for
5 dose-response analysis. As a result, EPA decided to treat both the Rogers and NEDO studies as
6 candidate principal studies and derived candidate RfCs and RfDs for the most sensitive endpoint
7 from each study (see Section 5.1.1.2).
A.l.2.2. Charge B2. Reduction of brain weight at 6 weeks postnatally as
reported in the NEDO (1987) developmental rat study was selected as the
critical effect. Please comment on whether the rationale for the selection of
this critical effect has been scientifically justified. Please identify and provide
the rationale for any other endpoints (e.g., other reproductive and
developmental effects reported in mouse and monkey studies) that should be
considered in the selection of the critical effect.
8 Summary of Comments: Four reviewers indicated that the use of brain weight change
9 was justified, but one of these and one other reviewer questioned the use of the 6-week time
10 point. Two reviewers suggest using the cervical rib endpoint from the Rogers et al. mouse study.
11 One reviewer expressed a preference for endpoints from the Burbacher et al. monkey study or
12 Rogers et al. mouse study over the NEDO rat study. Specific comments or suggestions made by
13 the reviewers with respect to this charge are described below, along with EPA responses.
14 Comment 1: Two reviewers suggested that the increased incidence of cervical ribs
15 should serve as the critical effect for RfC derivation, with one stating that "the increases in
16 cervical ribs and supernumerary ribs observed in this [Rogers et al. (1993b)] study could be
17 considered a more scientifically justified critical effect."
18 Response: EPA agrees that the Rogers et al. (1993b) study is of high quality. In the final
19 assessment, it is considered a candidate principal study. The preference of these two reviewers
20 for the cervical rib endpoint seems to be based in part on perceived problems with the brain
21 weight change endpoint in rats (NEDO, 1987). The reviewer who stated that the cervical rib
22 endpoint "could be considered a more scientifically justified critical effect" pointed out that the
23 NEDO (1987) developmental rat study did not note "abnormal brain histopathology or functional
24 deficits" and a statistical analysis of the brain weight changes was performed that was questioned
25 in a separate peer review of this study that was conducted for EPA (ERG, 2009). As discussed in
26 the response to Charge Bl Comment 1, EPA neurotoxicity guidelines allow for the treatment of
27 absolute brain weight change as an adverse neurological effect regardless of the existence of
May 2013 A-19 Draft - Do Not Cite or Quote
-------
1 corroborating histopathological or functional observations, and EPA used benchmark dose
2 analyses in lieu of the statistical test results reported by the authors for this endpoint.
3 Comment 2: One reviewer commented that there is a "general lack of transparency"
4 regarding the basis for the selection of the critical effect and stated that EPA chose "the one that
5 led to the lowest RfC," without regard to the limitations of the NEDO study, including
6 inappropriate use of statistical methods as described in a 2009 EPA-sponsored external peer
7 reviewer of the NEDO study. Another reviewer also commented that EPA had not acknowledged
8 errors in the NEDO statistical analysis and recommended that EPA conduct its own analysis of
9 variance (ANOVA) to determine "if there were an overall effect on brain weight" and "which
10 time frame and which methanol level are used in the BMD analysis."
11 Response: The basis for the selection of the candidate principal studies and effects are
12 primarily described in Section 5.1.1, Choice of Principal Study and Critical Effect(s). The critical
13 effects considered for the derivation of an RfC and RfD were chosen because they were reported
14 in studies of adequate quality, are considered relevant to humans, evidence a clear dose-response
15 and are sensitive indicators of alterations in important organ systems. The dose-response data for
16 the effects that meet these criteria (in this case the mouse cervical ribs and rat brain weight
17 effects) were considered for the derivation of the RfC and RfD. If EPA had based its selection on
18 the effect that "led to the lowest RfC" an endpoint in the Burbacher et al. or NEDO monkey
19 studies might have been chosen as some of the endpoints in these studies suggested a lower
20 NOAEL or BMDL. However, as described in Sections 4.2.2.3, 4.4.2, and 5.1.1, these monkey
21 studies did not meet all of the criteria necessary for an effect to be considered a critical effect. As
22 discussed in response to Charge Bl Comment 1, the limitations of the NEDO rat developmental
23 study, including the inappropriate use of statistical methods, are not serious enough to preclude
24 its consideration as a candidate principal study. There is no need for the Agency to perform an
25 ANOVA analysis because a benchmark dose analysis was performed in accordance with EPA
26 guidelines (U.S. EPA, 2012a) for all postnatal time frames (3, 6 and 8 weeks).
27 Comment 3: Two reviewers were concerned that EPA did not consider other postnatal
28 time points besides 6 weeks, with one stating that this approach "weakens the potential statistical
29 power for a response that appears stable over a wide range of time points (3 to 8 weeks)."
30 Response: A benchmark dose analysis of brain weight reductions in male and female rats
31 was performed for all postnatal time frames (3, 6 and 8 weeks). In accordance with EPA
32 guidelines (U.S. EPA, 2002, 1998a), the most sensitive developmental time point in the most
33 sensitive gender was used as the basis for the RfC/D. In order to achieve an increase in statistical
34 power by combining data together from separate ages, the animals must be exchangeable
35 (required for Bayesian statistics) or represent the same population (i.e., the brain weights from 3-
May 2013 A-20 Draft - Do Not Cite or Quote
-------
1 8 week old SD rats would have to represent the same population; required of frequentist
2 statistics). The data for the more sensitive gender (males) suggests that each age represents a
3 separate subpopulation (3 wks (mean ± standard deviation): 1.45g ± 0.06; 6 wks: 1.78g ± 0.07; 8
4 wks: 1.99g ± 0.06). Randomly permuting the individuals across the groups would not yield the
5 same conclusions, proving a lack of exchangeability. Thus, combining these samples together
6 would in all likelihood violate the exchangeability and independent and identically distributed
7 (i.i.d) assumptions required of Bayesian and frequentist methods, respectively.
8 Comment 4: Two reviewers were concerned over the "lack of histological or functional
9 follow-up for this [brain weight] response."
10 Response: See response to Charge Bl Comment 1 regarding brain weight.
11 Comment 5: Two commenters noted that EPA may need to reevaluate the endpoint
12 selection if modification of the PBPK analysis for SD rats significantly alters the relative
13 sensitivity (based on HECs) of the rat, mouse and monkey studies.
14 Response: EPA agrees with this comment. Consideration has been given to whether the
15 modified PBPK model results warrant a change in the critical effect. While the final candidate
16 RfDs and RfCs from SD rat brain weight response (NEDO, 1987) and the CD-I mouse cervical
17 rib response (Rogers et al., 1993b) are similar, the PBPK model modifications do change the
18 relative sensitivities such that the mouse study now serves as the basis for the methanol RfD. The
19 RfC is still based on the brain weight changes observed in the rat study.
A.l.2.3. Charge B3. Benchmark dose modeling of decreased pup brain
weight relative to maternal internal methanol doses predicted by the PBPK
model was used to derive the point of departure (POD) for the RfC. Has the
BMD/PBPK approach been appropriately conducted? Has adequate
justification been provided for the selected internal dose metric, i.e., area
under the curve (AUC) for methanol, in the blood of dams? Please identify
and provide the rationale for any alternative approaches for the
determination of the POD, including choice of another dose metric (e.g.,
methanol metabolized), and discuss whether such approaches are preferred
to EPA's approach.
20 Summary of Comments: Four reviewers indicated that the BMD analysis was
21 appropriate and appropriately applied. Three reviewers said that this was not their area of
22 expertise. Five reviewers accepted the choice of A UC as the dose metric, but noted limitations in
23 the data available (MOA and empirical information) for making that choice. One reviewer
24 preferred Cmax over A UC as the dose metric and one reviewer did not comment on the selected
May 2013 A-21 Draft - Do Not Cite or Quote
-------
1 dose metric. Specific comments or suggestions made by the reviewers with respect to this charge
2 are described below, along with EPA responses.
3 Comment 1: With respect to the selection of AUC as the dose metric for the BMD
4 analyses of the brain weight endpoint from the NEDO developmental study in which rats were
5 exposed gestationally and postnatally, one reviewer stated that "Without understanding of the
6 exact mechanism of action of the chemical, selection of any surrogate dose metric is somehow
7 speculative." A second reviewer commented that "The Agency has not adequately explained its
8 rationale for the use of AUC rather than Cmax (e.g., see literature related to methanol and 2-
9 methoxyethanol)." A third reviewer noted that justification for the AUC is "tenuous" because
10 "brain weight does not differ between the 3, 6 and 8 week periods."
11 Response: When performing BMD analyses, it is important to choose a reliably
12 measured or estimated dose metric that has a close relationship to the health effects under
13 consideration. For the BMD analyses of the mouse cervical rib endpoint, which has been shown
14 to result from just one day of gestational exposure, it is assumed that the level of exposure is
15 more important than duration. Internal methanol blood concentrations reported by Rogers et al.
16 (1993b) for the dams of each dose group at day 6 of gestation were assumed to be approximately
17 equivalent to Cmax levels and were used as the modeled dose metric. For the BMD analyses of
18 the rat brain weight endpoint following gestational and lactational exposure, PBPK model
19 estimates of AUC methanol in blood for the dams of each dose group were used as the modeled
20 dose metric. As described in Section 5.1.2.1, the decision to use AUC as the dose metric for the
21 gestationally and postnatally exposed rats was made because under this exposure regimen, brain
22 weight is susceptible to both the level and duration of exposure. It is true that the results of
23 NEDO (1987), described in Section 4.4.2 and shown in Table 4-13, indicate that there is not an
24 obvious cumulative effect of ongoing exposure on brain-weight decrements in rats exposed
25 postnatally for 3, 6 and 8 weeks. However, there is a greater brain-weight effect in rats exposed
26 postnatally versus only during organogenesis (GD7-17). Further, brain weight reductions have
27 been observed in adult rats that were exposed for 90 days beginning no earlier than 30 days of
28 age (TRL, 1986). That brain weight is susceptible to continued exposure beyond gestation
29 suggests that a dose metric that incorporates a time component would be more appropriate. For
30 this reason, and because it is more typically used in internal-dose-based assessments and better
31 reflects total exposure within a given day, daily AUC (measured for 22 hours exposure/day) was
32 chosen as the most appropriate dose metric for modeling the effects of methanol exposure on
33 brain weights in rats exposed throughout gestation and continuing into the Fl generation.
May 2013 A-22 Draft - Do Not Cite or Quote
-------
1 Comment 2: One reviewer asked why, for the purposes of Table 5-2 and the PBPK
2 estimation of AUC methanol in rat dam blood, the AUC was calculated with a 5 day 22 hr/day
3 simulation.
4 Response: The full text of the subject footnote is "AUC values were obtained by
5 simulating 22 hr/day exposures for 5 days and calculated for the last 24 hours of that period."
6 Simulations were run for 5 days as this was sufficient to reach "periodicity" when the daily time-
7 course is the same from one day of exposure to the next. From Figure B-13 it can be seen that
8 model predictions for the second day and beyond are essentially identical, but because the blood
9 level does not drop to zero during the 2 hour "off period, the AUC is higher on the 2nd day and
10 beyond than the first day. More importantly, the AUC was calculated for a single day of
11 simulated exposure, which happened to be the fifth day. With the PK parameters used for those
12 simulations, the same results would have been obtained if the simulation had only been run for
13 3 days, or for 30 days.
14 Comment 3: With respect to the alternative hypothesis that formaldehyde is the
15 teratogenic moiety and that increased effects of methanol in GSH-depleted animals are due to
16 decreased formaldehyde elimination, one reviewer noted that GSH depletion does not necessarily
17 imply formaldehyde involvement because "depletion of GSH, as the major cellular antioxidant,
18 will also increase the accumulation of reactive oxygen species (ROS)."
19 Response: The toxicological review has been revised (Section 4.7) to reflect that the
20 impact of GSH depletion can support both formaldehyde and ROS involvement in the
21 teratogenic effects of methanol. However, this reviewer and another reviewer agreed with the
22 Agency's position that methanol would play a key transport role in either case, with the latter
23 reviewer stating that "even if the metabolism-related formation of ROS or formaldehyde are
24 important contributors to the observed toxic effects, a methanol-based dose metric is applicable
25 when the downstream metabolic processes such as removal of ROS or formaldehyde are much
26 faster than the rate-limiting oxidation of methanol."
27 Comment 4: One reviewer commented that "Neither the 5% nor the 10% BMR have any
28 particular a priori justification for continuous data: the default assumption in this case is the
29 BMR of 1 standard deviation of the control dataset (as preferred here). In any case the data need
30 to be examined to determine an appropriate BMR representing a minimal detection level or
31 threshold of biologically significant response: this especially applies for continuous data."
32 Response: The reviewer's comments with respect to the selection of a BMR are correct
33 and consistent with EPA BMD Technical Guidance (U.S. EPA. 2012a). In the case of the
34 methanol toxicological review, all BMR levels considered for RfC or RfD derivation lie well
May 2013 A-23 Draft - Do Not Cite or Quote
-------
1 within the range of the dose-response observations. As indicated in the EPABMD guidance (U.S.
2 EPA, 2012a), a series of papers (Allen etal., 1994a, b; Faustman et al., 1994) suggest that a 5%
3 BMR is appropriate for dichotomous response data from well designed nested developmental
4 studies such as the Rogers et al. (1993b). For continuous response data, EPA guidance (U.S.
5 EPA, 2012a) suggests that "if there is an accepted level of change in the endpoint that is
6 considered to be biologically significant then that amount of change is the BMR." For
7 continuous response data from developmental studies, comparisons with the NOAEL showed
8 that several cutoff values, including a 5% change in mean fetal weight, could be used to give
9 values similar to the NOAEL (Kavlock et al., 1995). If a 5% change in fetal weight is considered
10 biologically significant, it is reasonable to assume that a 5% brain weight change should also be
11 considered biologically significant. However, in a recent report on the statistical power in the
12 analyses of brain weight measures in pesticide neurotoxicity testing, Weichenthal et al. (2010)
13 state that "if toxicological experts ultimately decide that brain weight changes in the range of 5%
14 are physiologically meaningful, a larger [than 10 per dose group] sample size will be needed to
15 consistently achieve reasonable power to detect this magnitude of effect." EPA BMD guidance
16 (U.S. EPA, 2012a) states that "in the absence of any other idea of what level of response to
17 consider adverse, a change in the mean equal to one control SD from the control mean can be
18 used." Because there is no clear biological basis for choosing one over the other, both are
19 considered and deference is given to the BMR that results in the lower RfC or RfD (see Tables 5-
20 4 and 5-5).
A.l.2.4. Charge Question B4. Please comment on the rationale for the
selection of the uncertainty factors (UFs) applied to the POD for the
derivation of the RfC. It is assumed that these UFs account for variability in
methanol dosimetry among human newborns following gestational and
lactational exposure, and for uncertainty regarding the ratio of newborn-
dose to maternal-dose in humans. Please comment on these assumptions and
on the scientific justification for the selected UFs.
21 Summary of Comments: In general, four reviewers indicated that the selected UFs are
22 adequate and consistent with EPA policy and three reviewers did not agree with certain UFs. Of
23 the four reviewers that generally agreed with EPA's proposed UFs, one suggested that some
24 further examination and discussion of the UFn would be helpful and another noted that a strong
25 argument could be made for eliminating the UFr>. Of the three reviewers that expressed
26 disagreement, all three stated that the 3-fold UFo was not necessary, and one suggested that the
27 UFH of 10 is not warranted. With respect to where the UFs are applied, one reviewer supported
28 the Agency's practice of applying UFs to the HEC and one advocated application of UFs to
May 2013 A-24 Draft - Do Not Cite or Quote
-------
1 BMDLs (before HEC derivation). Specific comments or suggestions made by the reviewers with
2 respect to this charge are described below, along with EPA responses.
3 Comment 1: Regarding the UFH, one reviewer stated that a full UFH of 10 is not
4 warranted because "at the level of the proposed RfC and RfD, intraspecies differences in
5 disposition of exogenous methanol in humans will likely have no meaningful impact on the body
6 burden of 'total' methanol." This reviewer recommended that EPA perform a sensitivity analysis
7 of the human PBPK model to identify variability and/or uncertainty in parameters which have an
8 impact on methanol levels predicted. Also, this and another reviewer stated that the UFH does not
9 need to account for uncertainty regarding the sensitivity of children because the critical study is
10 in neonates, two-generation study exists and "no particular developmental susceptibility of
11 humans versus test species is expected." Another reviewer questioned whether a UFH of 10 was
12 "sufficient in the general case" and recommended further examination and discussion to establish
13 "the limits of the data available to inform the decision on the value for UFH."
14 Response: A sensitivity analysis of the human PBPK model has been performed (see
15 Appendix B), and the results suggest that parameter variability is not likely to result in methanol
16 blood level estimates that vary more than 3-fold, the toxicokinetic portion of the 10-fold UFn.
17 However, one needs to also consider the variation in endogenous levels of methanol (Table 3-1),
18 and variation in toxicodynamics, because both may affect the impact of an exogenous methanol
19 exposure. Overall, the extent of human interindividual variation in (endogenous and exogenous)
20 methanol toxicokinetics and toxicodynamics would be very difficult to quantify given the
21 significant uncertainties that exist regarding endogenous levels and methanol's mode of action.
22 Toxicodynamic variability can only be discussed qualitatively. As discussed in Section
23 4.9, there are a number of issues that may lead to sensitive human subpopulations. Potentially
24 sensitive subpopulations would include individuals with polymorphisms in the enzymes involved
25 in the metabolism of methanol and individuals with significant folate deficiencies. The effects
26 used to derive the candidate RfCs are observed in a potentially susceptible and sensitive
27 fetal/neonatal subpopulation. However, there is also variability across fetuses and neonates that
28 need to be taken into account. Children vary in their ability to metabolize and eliminate methanol
29 and in their sensitivity to methanol's toxic teratogenic effects. Consequently, there exists
30 considerable uncertainty pertaining to human population variability in methanol metabolism,
31 which provides justification for the 10-fold intraspecies UF used to derive the RfC and RfD.
32 Comment 2: Regarding the UFA, one reviewer stated that "it is surprising that the EPA
33 used the same interspecies UFA for rodent and nonhuman primate studies - given the fact that
34 significant species difference exist between rodents and humans and less so between monkeys
35 and people (use UF = 1)."
May 2013 A-25 Draft - Do Not Cite or Quote
-------
1 Response: As discussed in response to Charge Bl Comment 2, due to uncertainties in the
2 dose-response data for the monkey studies, EPA has removed the alternative RfC derivation for
3 monkeys from the toxicological review.
4 Comment 3: Regarding the UFo, three reviewers provided the following reasons why a
5 3-fold UFD was not needed:
6 • Methanol has a "very rich toxicology database."
7 • "There is never enough data to be certain regarding a risk 'assessment' (that is why it is
8 called risk assessment not a risk determination)."
9 • Conservative assumptions are "always used," including:
10 o use of a single SD for BMDL rather than a 4 or 10% changes as commonly used
11 in some noncancer risk assessments (e.g., see page 5-23),"
12 o "PBPK assumes the most conservative scenarios,"
13 o "BMD analysis itself favors the conservative numbers" and
14 o "when given the choice of alternative BMD numbers such as those obtained from
15 the 3 versus 6 versus 8 week data, the lowest (i.e., most conservative) number is
16 chosen."
17 • "The key endpoint is developmental toxicity, which has been evaluated in multiple species,
18 including primate, and special endpoints such as neurotoxicity and immunotoxicity have
19 been evaluated."
20 • "There is no need to have a UF because 'there is uncertainty regarding which test species is
21 most relevant to humans'—the lowest, high-quality point of departure was used."
22 • "There is also no need to have a UFD for "dose spacing" because the BMD analysis counters
23 this potential design deficiency.
24 Response: The database uncertainty factor accounts for the potential to underestimate
25 noncancer hazard as a result of data gaps. EPA agrees that the database for methanol toxicity is
26 quite extensive: there are chronic and developmental toxicity studies in rats, mice, and monkeys,
27 a two-generation reproductive toxicity study in rats, and neurotoxicity and immunotoxicity
28 studies. However, as discussed in Section 5.1.1.1, chronic and developmental studies in
29 monkeys, the species most likely to best represent the potential for developmental effects in
30 humans, were considered inadequate or inferior to the candidate principal rodent studies for the
31 purposes of RfC/D derivation. As discussed in Sections 5.1.3.2.3 and 5.3.6, the lack of a
32 quantifiable monkey study is an important data gap given the potential relevance to humans and
May 2013 A-26 Draft - Do Not Cite or Quote
-------
1 the uncertainties raised by existing monkey studies regarding this species sensitivity to
2 reproductive effects (e.g., shortened pregnancies discussed in Section 4.3.2), CNS degeneration
3 (e.g., stellate cell fibrosis described in Section 4.4.2) and delayed neurobehavioral development
4 (e.g., VDR response described in Section 4.4.2) from methanol exposure. In addition, a full
5 developmental neurotoxicity test (DNT) in rodents has not been performed and is warranted
6 given the critical effect of decreased brain weight in rats and the suggestive (but quantitatively
7 inconclusive) DNT results in monkeys. For these reasons, an UF of 3 was applied to account for
8 deficiencies in the database.
9 Comment 4: Regarding the application of all UFs, one commenter stated that EPA
10 should "apply the uncertainty factors to the internal dose point of departure, prior to interspecies
11 extrapolation with the pharmacokinetic model to account for non-linearities in external versus
12 internal dose relationships." This reviewer suggested that EPA should discuss their choice of
13 applying UFs to the HEC/D rather than the BMDL. The commenter estimated that if UFs are
14 applied first to the mouse cervical rib BMDLos, then converted to the candidate RfC using the
15 PBPK model, the candidate RfC would increase by more than 2-fold. Another reviewer noted
16 that the application of UFs to the HEC/D values is the standard procedure and is "preferred to
17 alternative suggestions that the UFs be applied to intermediate measures such as blood
18 concentrations or AUCs."
19 Response: The first commenter is correct in that, after modifications were made to the rat
20 PBPK model (see Response to Charge Al Comment 7), BMDL estimations from both the rat and
21 mouse candidate principal studies are not within the linear range of EPA's PBPK model
22 predictions. EPA has reevaluated the analysis and applied the UFs prior to HEC/D derivation as
23 suggested. This approach results in more scientifically reliable model predictions by lowering the
24 BMDLs to within the more linear, calibrated range of the human PBPK model. Clarifying text
25 has been added to the Sections 5.1.3.2 and 5.2.2.3.
26 The concern expressed by the second reviewer regarding departure from EPA practice is
27 recognized, given the uncertainty associated with dividing internal dose BMDLs by UFs that are
28 at least partially based on empirical analyses of ratios of NOAELs obtained from external oral
29 exposures (U.S. EPA, 1994b: Dourson and Stara, 1983). In the methanol (noncancer) assessment,
30 the general EPA practice of applying the human PBPK model to derive HEC/D values prior to
31 applying UFs (U.S. EPA. 2002. 1994b) would result in RfC/Ds lower than if the PBPK model
32 was used to derive HEC/D estimates after dividing the BMDL internal doses by UFs. However,
33 this general practice if applied to methanol would result in greater model uncertainty because the
34 HECs (1,042 to 1,604 mg/m3) and HEDs (133 to 220 mg/kg-day) estimated from the BMDLs by
35 the revised PBPK model are well above the inhalation concentrations (655 mg/m3) and oral
May 2013 A-27 Draft - Do Not Cite or Quote
-------
1 exposures (50 mg/kg-day) for which there are human data to calibrate the PBPK model (see
2 Appendix B, Section B.2.7, Table B-6).
A.1.3. Charge C: "Oral Reference Dose (RfD) for Methanol"
A.l.3.1. Charge Cl. EPA concluded that the oral RfD should be derived
using a route-to-route extrapolation from the more extensive inhalation
database given the paucity of oral toxicity data. Please comment on whether
the rationale for this approach has been scientifically justified and clearly
explained. Please identify and provide the rationale for any alternative
approaches for the determining the RfD and discuss whether such
approaches are preferred to EPA's approach.
3 Summary of Comments: Six reviewers indicated that the approach taken by EPA was
4 appropriate and one reviewer did not comment due to a lack of expertise. Specific comments or
5 suggestions made by the reviewers with respect to this charge are described below, along with
6 EPA responses.
7 Comment 1: One reviewer recommended that EPA "provide alternative RfC estimates
8 that would be derived using traditional approaches."
9 Response: Since this comment was made in response to the oral RfD charge, it is
10 assumed that the reviewer is requesting that EPA provide traditional RfD estimates and not
11 "RfC" estimates. None of the oral studies provided sufficient dose-response data for a dose-
12 response analysis and none of the developmental toxicity studies identified a NOAEL for use in a
13 traditional RfD estimate. The only NOAEL identified was 500 mg/kg-day from the subchronic
14 oral study in adult rats (TRL, 1986). As discussed in Section 5.2.4, the previous IRIS assessment
15 of methanol divided this NOAEL by a 1,000-fold uncertainty factor to obtain an RfD of
16 0.5 mg/kg-day. This value is lower than the current proposed RfD of 2 mg/kg-day, largely
17 because a 10-fold higher uncertainty factor was employed in the previous assessment.
18 Comment 2: One reviewer stated that "Human model validation using the oral data of
19 Schmutte et al. (1988) (see Charge D2) could further strengthen confidence in the route-to-route
20 extrapolation."
21 Response: EPA agrees and as discussed in response to Charge Al Comment 11, this oral
22 study has been added to EPA's PBPK analysis and used in the validation of the oral human PBPK
23 model.
May 2013 A-28 Draft - Do Not Cite or Quote
-------
A.l.3.2. Charge C2. A PBPK model was used to derive the RfD via a route-
to-route extrapolation, in which the internal-dose POD used for the
derivation of the RfC based on data from the NEDO (1987) study was
extrapolated to human oral exposure levels using the human PBPK model.
Please comment on whether the rationale for this approach has been
scientifically justified. Has adequate justification been provided for the
selected internal dose metric, i.e., AUC for methanol, in the blood of dams? Is
the PBPK model suitable for extrapolation of fetal and neonatal endpoints to
human oral exposures? Please provide a detailed explanation.
1 Summary of Comments: Five reviewers stated that the approach taken by EPA was
2 appropriate and a sixth reviewer did not comment due to a lack of expertise. A seventh reviewer
3 cited the lack of gestational and lactational components as a weakness in the EPA approach. All
4 reviewers either referred to or repeated previous comments on the PBPK model and the RfC
5 derivation approach.
6 Summary Response: The reviewers did not offer any new comments in response to this
7 charge question that were not covered in response to previous charge questions.
A.l.3.3. Charge C3. EPA applied the same UFs to the POD for the
derivation of the RfD as for the RfC. Please comment on the rationale for the
selection of the UFs.
8 Summary of Comments: All but one reviewer agreed with the use of the same UFs for
9 the RfD as for the RfC. One reviewer stated that this was "unexpected" because the database for
10 oral and inhalation are very different.
11 Summary Response: The critical effects were systemic, developmental effects that are
12 assumed to be dependent on blood concentrations of methanol. EPA was able to use methanol
13 blood concentrations in its benchmark dose analyses of the critical effects in the candidate
14 principal studies because blood levels were either reported in the study or could be estimated
15 using a validated PBPK model. After application of UFs, a validated human PBPK model was
16 then used to convert the adjusted benchmark dose estimates to an RfD and RfC. For these
17 reasons, EPA was able to derive the oral RfD and inhalation RfC with a similar degree of
18 confidence using the same data set, endpoint, BMD methods and PBPK model.
May 2013 A-29 Draft - Do Not Cite or Quote
-------
A.1.4. Charge D: "General Charge Questions"
A.l.4.1. Charge Dl. Is the Toxicological Review logical, clear and concise?
Has EPA clearly synthesized the scientific evidence for noncancer and cancer
hazards?
1 Summary of Comments: In general, reviewers commented that the Toxicological review
1 was logical, comprehensive and clear, but not concise. One reviewer stated that the review "is
3 thorough and well written, and takes care to provide descriptions of the available evidence in a
4 clear, complete and unbiased form " and "presents a careful and well justified synthesis of these
5 data. " However, five of the seven reviewers criticized the repetitive or redundant nature of the
6 review and four reviewers were critical of the review format, with one reviewer stating that "a
7 different format could be much more effective in conveying critical information, interpretations,
8 and decisions regarding available, relevant toxicological literature. " Specific major (non-
9 editorial) suggestions made by the reviewers with respect to this charge are described below,
10 along with EPA responses.
11 Comment 1: One reviewer suggested that EPA add a "decision tree" to make choices for
12 major decisions more transparent.
13 Response: In light of this concern, an Executive Summary has been added to the
14 beginning of the toxicological review which makes the choices for major decisions more readily
15 apparent and transparent. Exposure-Response arrays for oral and inhalation toxicity studies were
16 also added as Figures 4-1 and 4-2 to better depict the relationship of NOAELs and LOAELs in
17 the overall database of studies.
18 Comment 2: Two reviewers suggested that EPA make edits consistent with recent NRC
19 (2011) recommendations for the draft EPA formaldehyde assessment to:
20 a. Reduce text volume, narrative approach, redundancies and inconsistencies,
21 b. rely more heavily on tables and not repeat individual study descriptions,
22 c. include "inclusion and exclusion criteria" for cited references, and
23 d. reduce "extraneous information" contained in Appendices.
24 Response: In response to this comment EPA has made the following format and text edits
25 to the toxicological review:
26 a. In response to the suggestion to "reduce text volume, narrative approach,
27 redundancies and inconsistencies," EPA has extensively condensed Section 3.4
28 (e-g-, abbreviating the discussion of model structure and deleting detailed
May 2013 A-30 Draft - Do Not Cite or Quote
-------
1 discussion of model parameter and model calibration in deference to Appendix
2 B), combined Section 4.6 "MECHANISTIC DATA AND OTHER STUDIES IN
3 SUPPORT OF THE MO A" with Section 4.8 "NONCANCER MOA
4 INFORMATION" into a new Section 4.7, deleted or merged redundant portions
5 of Section 5.3 "UNCERTANTIES IN THE INHALATION RFC AND ORAL
6 RFD" with Section 5.1.3.2 "Application of UFs," removed portions of Section 5
7 that were unnecessarily redundant with Appendix D, revised and consolidated
8 portions Section 5 related to the justification of the dose metric (AUC versu Cmax
9 versus total metabolites) employed for the BMD analyses of candidate principal
10 studies and removed Section 6 (in lieu of a new Executive Summary).
11 b. In response to the suggestion to "rely more heavily on tables and not repeat
12 individual study descriptions," EPA has created new tables for whole embryo
13 studies described in Section "4.3.3 Other Reproductive and Developmental
14 Studies" and the i.p. studies described in Section "4.4.3 Neurotoxicity Studies
15 Employing In Vitro and I.P. Methanol Exposures" and has edited all Sections to
16 reduce unnecessary repetition of individual study descriptions. EPA has also
17 added an exposure-response array (Figure 4-1) to the assessment.
18 c. In response to the suggestion to include 'inclusion and exclusion criteria' for cited
19 references," EPA has added text to the Preface of the methanol toxicological
20 review that describes how EPA evaluates the quality of studies.
21 d. In response to the suggestion to "reduce 'extraneous information' contained in
22 Appendices," EPA has removed Appendix D, having determined it to be
23 extraneous due to the stronger biological basis for the choice of a Cmax dose
24 metric described in Section 5.1.2.1, and removed the source code text from
25 Appendix B. The source code text will be posted on the EPA IRIS website, along
26 with the final methanol (noncancer) assessment, as part of a Windows zip file
27 containing a complete package of acslX code necessary to run all of the
28 developed models.
29 Comment 3: One reviewer suggested that "Table 3-3 should include the Dorman
30 cynomolgus monkey study with a clear indication that it involved lung only exposure of
31 anesthetized monkeys."
32 Response: Table 3-3 has been revised to include the Dorman cynomolgus monkey study
33 in response to this comment.
May 2013 A-31 Draft - Do Not Cite or Quote
-------
1 Comment 4: One reviewer asked whether EPA considers the Fagan test results from the
2 Burbacher et al. monkey study to be "biologically significant despite the lack of a statistically
3 significant response?"
4 Response: There is uncertainty regarding both the biological and statistical significance
5 of the Fagan test results from the Burbacher et al. monkey study. As explained in Section 4.4.2,
6 "Unlike the VDR results discussed previously, results of this test did not appear to be
7 gender specific and were neither statistically significant (ANOVA p = 0.38) nor related to
8 exposure concentration. The findings indicated a cohort effect which appeared to reduce
9 the statistical power of this analysis. The authors' exploratory analysis of differences in
10 outcomes between the 2 cohorts indicated an effect of exposure in the second cohort and
11 not the first cohort due to higher mean performance in controls of cohort 2 (70% + 5%
12 versus 55% ± 4% for cohort 1). In addition, this latter finding could reflect the inherent
13 constraints of this endpoint. If the control group performs at the 60% level and the most
14 impaired subjects perform at approximately the 50% chance level (worse than chance
15 performance would not be expected), the range over which a concentration-response
16 relationship can be expressed is limited."
17 However, the Fagan test results cannot be ignored and, as described in Section 5.1.1.2.2:
18 "Although not statistically significant and not quantifiable, the results of this test
19 need to be considered, in conjunction with VDR test results and brain weight
20 changes noted in the NEDO (1987) rat study, as a possible indication of CNS
21 effects."
22 Comment 5: One reviewer found Section 3.4.2.4 confusing, and suggested that other
23 models that have been developed with inhaled manganese (Schroeter et al., 2011; Yoon et al.,
24 2011; Yoon et al., 2009a, b) "could form the basis for a gestational and lactational model."
25 Response: To reduce text volume in response to Charge Dl Comment 2a, Section 3.2.4
26 has been removed from the toxicological review. It contained an unnecessary discussion of the
27 rat and human isopropanol models described by Gentry et al. (2003; 2002) and Clewell et al.
28 (2001). It was originally included because it was thought to be a possible guide had EPA decided
29 to develop a more complex gestational and lactational model. The reviewer is right in that, had
30 EPA decided to take this approach, other gestational and lactational models, such as the one
31 developed for manganese, could have been considered. However, EPA has determined, and the
32 peer reviewers generally agreed (see "Summary of Comments" under Charges A3 and A5), that
33 such a model was unnecessary for the purposes of the methanol toxicological review.
May 2013 A-32 Draft - Do Not Cite or Quote
-------
1 Comment 6: One reviewer stated that the "the discussion of a two compartment stomach
2 (page 3-28 and elsewhere) for rodents need additional justification (squamous and epithelial
3 portions?)" and questioned whether this structure is "appropriate for people (as indicated on page
4 3-51)."
5 Response: EPA agrees with the reviewer and, in response, has simplified the GI
6 absorption model and revised the associated text in the toxicological review. In particular, the GI
7 model for humans has been reduced to a single, first order compartment and rate (see Appendix
8 B, Section B.2.6).
9 Comment 7: One reviewer commented that the use of "terms that describe model fits as
10 'quite poor' (e/g/. see page 3-40 and elsewhere)" need to be "better clarified (visual inspection,
11 goodness of fit, other?)."
12 Response: Except where numerical measures of fit are given, all such references to
13 model fit reflect visual inspection. This has been clarified in the toxicological review.
14 Comment 8: One commenter requested that EPA "pick one set of units (ppm would be
15 preferred until calculation of the actual RfC value)."
16 Response: In general, both units are given, with mg/m3 values provided parenthetically
17 after the ppm values, except for RfC/D and point of departure (e.g., BMDL) values discussed in
18 Section 5.
19 Comment 9: One reviewer requested a discussion of the use of alcohol dehydrogenase
20 inhibitors as a clinical 'antidote' on "page 4-7 (and possibly elsewhere)."
21 Response: Explanatory text has been added on page 4-1 and 4-4 to explain that infusion
22 of ADH1 inhibitors such as ethanol or fomepizole (4-methylpyrazole) can serve as treatment for
23 methanol poisoning.
24 Comment 10: One reviewer asked whether the folate deficiency described on page 4-40
25 affects methanol concentrations significantly, and "which data support this conclusion?"
26 Response: Folate is the coenzyme of tetrahydrofolate synthetase, an enzyme that is rate
27 limiting in the removal of formate. However, there is limited evidence regarding how folate
28 deficiency would impact methanol and formaldehyde levels. Hence, the statement on page 4-40
29 of the draft assessment, that "Folate deficiency would be expected to cause potentially toxic
30 levels of methanol, formaldehyde, and formate to be retained" has been revised in the final
31 assessment (Section 4.3.2), to read" "Folate deficiency would be expected to cause potentially
32 toxic levels of formate to be retained."
May 2013 A-33 Draft - Do Not Cite or Quote
-------
1 Comment 11: One reviewer recommended that EPA remove the Section 4.1 discussion
2 of the CNS effects produced by acute methanol overdosing because it could be perceived as an
3 inappropriate and "biased way to validate the subsequent choice of the NEDO study (decrease in
4 brain weights suggesting a methanol-induced CNS effect)."
5 Response: Because of the limited usefulness of human case study information to this
6 assessment, this portion of Section 4.1 was moved to a new Appendix C. However, the remainder
7 of Section 4.1 is retained because it contains important information relevant to the acute toxicity
8 of methanol and is one of the only sections in the toxicological review for which human data are
9 available. It is recognized that the CNS effects from acute exposure to methanol are likely the
10 result of a different mode of action than methanol's developmental effects. This is discussed in
11 several places in the toxicological review, particularly Section 4.7 on the MOAfor noncancer
12 effects.
13 Comment 12: One reviewer suggested that EPA needs to improve the synthesis of SD rat
14 toxicokinetic data for purposes of PBPK model development.
15 Response: This has been done and Appendix B has been revised accordingly.
16 Comment 13: One reviewer suggested that EPA correct inconsistencies between the
17 toxicokinetics section of Section 3 and Appendix B.
18 Response: To avoid redundancy and address inconsistencies, the PBPK discussions in
19 Section 3 have been removed and the reader is referred to Appendix B for technical details.
20 Comment 14: One reviewer noted that the "clarity of the document is hampered by the
21 lack of a clear synthesis of evidence regarding plausible modes of action for developmental
22 toxicity."
23 Response: The mode of action discussions previously divided between Section 4.6 and
24 4.8 have been revised for clarity and consolidated into Section 4.7.
A.l.4.2. Charge D2. Please identify any additional studies that would make a
significant impact on the conclusions of the Toxicological Review and should
be considered in the assessment of the non-cancer health effects of methanol.
25 Summary of Comments: Three reviewers identified seven additional studies for EPA to
26 consider (Miller and Wells, 2011; Leavens et al, 2006; Dorman et al, 1995; Bolon et al, 1994;
27 Bolon et al., 1993; Haffner et al., 1992; Schmutte et al., 1988). Specific comments or suggestions
28 made by the reviewers with respect to this charge are described below, along with EPA
29 responses.
May 2013 A-34 Draft - Do Not Cite or Quote
-------
1 Summary Response: The identified papers were evaluated and are now discussed and
2 referenced in the final assessment. As discussed in response to Charge Al Comment 10, oral data
3 from two of these studies (Haffner et al., 1992) and (Schmutte et al., 1988)1 are now used for
4 model calibration, allowing identification of human oral absorption rate constant and
5 bioavailability. The most informative of the remaining studies may be the in-vitro study of Miller
6 and Wells (2011) which demonstrated that methanol-induced developmental effects are enhanced
7 in mouse embryos with low catalase activity and reduced in mouse embryos with high catalase
8 activity. The authors propose that this observation is related to methanol's impact on the ability
9 of catalase to control the damaging effects of reactive oxygen species (ROS) activity, which
10 would be greater in mouse embryos with low catalase activity. As discussed in Section 5.3.5,
11 there are several problems with this interpretation, including that in vivo results from the same
12 laboratory (Siu et al., 2013) do not support the Miller and Wells (2011) in vitro findings. Further,
13 these observations do not preclude alternative explanations that involve a more direct interaction
14 between methanol and the embryo.
15 Comment 1: One reviewer suggested that the University of Toronto rabbit studies
16 published by Sweeting and coworkers "were not considered in the EPA's consideration of inter-
17 species differences (i.e., are rat or mice studies appropriate)." Another reviewer commented that
18 the discussion of the University of Toronto studies, especially the "publication regarding the role
19 of ROS in mediating the effects of methanol," needs to be improved and included in the
20 "sections related to choice of POD, critical effect, etc."
21 Response: EPA has added additional discussion of the University of Toronto (Miller and
22 Wells, 2011; Sweeting et al., 2011) research to the toxicological review. A detailed discussion of
23 the University of Toronto findings and hypotheses regarding species differences and the role of
24 ROS following methanol exposure has been added to Section 5.3 "UNCERTANTIES IN THE
25 INHALATION RFC AND ORAL RFD" of the toxicological review (see Section 5.3.5 "Choice
26 of Species/Gender"). Miller and Wells (2011) have suggested that developmental studies in
27 rodents may not be suitable for assessing human risk, and Sweeting et al. (2011) have suggested
28 that rabbits would be a more appropriate test species than mice and that rabbits are resistant to
29 methanol teratogenicity. A developmental study in rabbits via an appropriate route of exposure
30 would be of interest, particularly if it involved an investigation of effects over a broad set of
31 gestational days. However, more research is needed before it can be definitively stated that rabbit
32 developmental study would be more relevant to humans than rodent studies and that rabbits are
33 resistant to methanol teratogenicity.
May 2013 A-35 Draft - Do Not Cite or Quote
-------
1 Comment 2: One reviewer stated that "there are also other studies, including work in
2 monkeys, with aspartame that may be supportive (e.g., Reynolds). Since Table 3-2 includes
3 results from aspartame exposure this does not seem to be a clear exclusion criterion."
4 Response: A review of the aspartame literature is beyond the scope of this toxicological
5 review. The aspartame exposure studies have been removed from Table 3-2.
6 Comment 3: One reviewer noted that "the ethanol teratology literature has been largely
7 ignored despite some similarities in teratogenic response" and that "this larger literature may
8 help inform the MOA discussions in the draft document and help guide whether formaldehyde
9 should be considered as the proximate teratogen."
10 Response: A review of the ethanol literature is beyond the scope of this toxicological
11 review.
12 Comment 4: One reviewer stated that "search terms and databases examined have been
13 poorly defined" and that "there is a lack of inclusion and exclusion criteria" for references.
14 Response: EPA has added text to the Preface of the methanol toxicological review that
15 describes how literature searches are performed and how studies are evaluated and selected.
A.l.4.3. Charge D3. Please discuss research likely to substantially increase
confidence in the database for future assessments of methanol.
16 Summary of Comments: Reviewers suggested the following research to increase
17 confidence in the database for a future assessment:
18 • A proper study should be performed "to confirm the low activity of methanol metabolism in
19 fetal tissues."
20 • "Future studies using different animal models from rodents to primates should focus on
21 outcomes related to reproductive function, early sensorimotor development and object
22 memory as well as changes in brain architecture and size.
23 • "Development of a PBPK model that considers gestation and lactational exposure. "
24 • Studies that "replicate the findings of the critical study used by NEDO including the
25 inclusion of additional neuropathological and neurobehavioral assessments " and using
26 "NEDO-type " exposures.
27 • "Although additional monkey studies could be considered the Burbacher study is extremely
28 robust and should receive more attention by EPA. "
May 2013 A-36 Draft - Do Not Cite or Quote
-------
1 • Studies using "dual labeled material to confirm fetal exposure " and "designed to resolve
2 whether formaldehyde is involved in the developmental effects following perinatal methanol
3 exposure."
4 • "Completion of surveys to examine blood methanol concentrations in the US population. "
5 • "A study that fully characterizes methanol metabolism [including estimates the Km and Vmax]
6 in the intact fetus and the dam using the rat as model... (as opposed to the existing studies
7 that only assess protein levels or activities using ethanol as substrate). "
8 • "Studies of the role ofADH and catalase in the metabolism of methanol by F-344 and
9 Sprague-Dawley rats " to "clarify why there might be two saturable pathways in one strain
10 but only one in the other (as implied by the PBPK model) "
11 • An oral developmental study of methanol sufficient for use in the derivation of an RfD.
12 • "Research to explain the basis for differences in species/strain developmental effects " and
13 determine "the proximate toxicant and mode of action for developmental toxicity. "
14 • "Further studies to illuminate the relative sensitivity of rodents and primates to chronic
15 methanol toxicity, especially with regard to developmental and neurotoxicity endpoints. "
16 • Studies to elicit better "inhalation kinetic data for Sprague-Dawley rats. "
17 • "Monkey studies with longer exposure durations and similar endpoints. "
18 • "Additional mode-of-action motivated [including in-vitro] studies "
19 Summary Response: EPA agrees that these studies could enhance the methanol
20 toxicological review. However, EPA is not planning to, and none of the reviewers suggested that
21 EPA should, delay the completion of this assessment pending the completion of any of these
22 future study suggestions.
A.l.4.4. Bonus Charge Question: Please comment on the proposed RfD and
RfC values for their intended use in risk assessment. Are these numbers
more conservative than they need to be to protect public health? Note:
During the external review panel meeting an additional charge question was
developed by the chair of the panel with input from some panel members. This
charge question relates to the RfC/D and their relationship to endogenous
background blood levels. While not apart of the EPA Charge to the external
review panel, most of the panel members responded to this "Bonus Charge
Question" as discussed here.
23 Summary of Comments: The six reviewers that provided comments seem to be in
24 agreement that there needs to be more discussion of the relation of the RfD and RfC to existing
May 2013 A-37 Draft - Do Not Cite or Quote
-------
1 endogenous blood levels. Five of six reviewers suggested that the RfD andRfC values were more
2 conservative (lower) than necessary. One reviewer pointed out "the RfC and RfD are specifically
3 defined as levels at which the risk assessor can be reasonably confident that adverse effects will
4 not appear " and are "not threshold levels at which effects might start to appear. "
5 Two reviewers suggested that estimates of the increased blood levels associated with the
6 RfD/C values should be compared with either an upper bound or a standard deviation for
7 existing or normal physiological background levels ofmethanol. However, another reviewer
8 warned that "in view of the uncertainties as to fetal metabolism, mode of action and contribution
9 of diet and individual metabolic or toxicodynamic differences which are identified in the report it
10 seems very unwise to conclude that high-end [of the distribution of background] exposures which
11 are apparently safe for some individuals are necessarily safe for all. "
12 One reviewer supported the NTP CERHR (2003) opinion that a blood methanol
13 concentration of<10 mg/L would not be associated with adverse developmental effects. Another
14 reviewer cited the NTP CERHR (2003) report as indicating that "common exposures " are not a
15 concern for developmental toxicity, and suggested that this presents a credibility problem for the
16 proposed RfD and RfC values, which have been likened to common exposures such as a glass of
17 orange juice. Another reviewer expressed concern that the assumption that common exposures or
18 "current background levels " are safe has not been analytically investigated, and suggested that
19 the uncertainty factors applied are needed to reflect these concerns, "which therefore indicates
20 that the proposed values for RfC and RfD are not necessarily unreasonable. "
21 Summary Response: The RfC and RfD have increased by several-fold due to PBPK
22 model revisions made in response to the comments received during external peer review. The
23 final RfD of 2 mg/kg-day and RfC of 20 mg/m3 are not overly conservative because they (1) are
24 well above the levels associated with common exposures to methanol such as from a glass of
25 orange juice and (2) need to account for uncertainty regarding the sensitivity of primates to the
26 reproductive and developmental neurotoxic effects ofmethanol.
27 EPA addressed the recommendation of a reviewer that estimates of the increased blood
28 levels associated with the RfD and RfC be compared with a standard deviation for existing or
29 normal physiological (endogenous blood) background levels ofmethanol. As described in
30 Section 5.3.6, the methanol blood levels of individuals receiving both an RfC and RfD exposure
31 would increase by a daily maximum of 0.86 mg/L and a daily average of 0.59 mg/L. As shown in
32 Figures 5-3, 5-4 and B-17, these increases are comparable to the 0.7 mg/L standard deviation
33 estimated for the average methanol blood levels (1.5 ± 0.7 mg/L) in humans. Thus, the estimated
34 increase in blood levels of methanol from exogenous exposures at the level of the RfD or the
35 RfC (or from the RfC + RfD) are distinguishable from natural background variation. These RfC
May 2013 A-38 Draft - Do Not Cite or Quote
-------
1 and RfD methanol blood level increases are also more than 100-fold higher than the increase that
2 would be associated with a "common exposure" such as from a glass of pasteurized orange juice
3 and about 10-fold higher than the increase that would be associated with exposure to a glass of
4 unpasteurized orange juice (note that this is a relatively rare exposure and FDA requires warning
5 labels on unpasteurized juice that state "This product has not been pasteurized and therefore may
6 contain harmful bacteria that can cause serious illness in children, the elderly, and persons with
7 weakened immune systems"). Hence there is consistency with NTP CERHR (2003) in this
8 regard.
9 However, there is uncertainty with respect to the NTP CERHR (2003) statement that
10 methanol blood levels below 10 mg/L would not be associated with adverse developmental
11 effects. As discussed in Sections 5.1.3.2.3 and 5.3.6, there is uncertainty as to whether rodents
12 are as sensitive as monkeys and humans to the reproductive and developmental neurotoxic
13 effects of methanol. The lack of a reliably quantifiable monkey study is an important data gap
14 given the potential relevance to humans and the uncertainties raised by existing monkey studies
15 regarding monkey sensitivity to reproductive effects (e.g., shortened pregnancies discussed in
16 Section 4.3.2), CNS degeneration (e.g., stellate cell fibrosis described in Section 4.4.2) and
17 delayed neurobehavioral development (e.g., VDR response described in Section 4.4.2) from
18 methanol exposure. In the Burbacher et al. (2004a; 1999a) study, statistically significant
19 shortened pregnancy duration was observed in monkeys exposed to 200 ppm and statistically
20 significant VDR delay was observed in male monkey infants exposed to 600 ppm methanol for
21 just 2 hours per day. EPA estimates that these exposures raised the methanol blood levels over
22 endogenous methanol blood levels in these monkeys to peak values of just 3 and 10 mg/L,
23 respectively (see Appendix D, Table D-10), corresponding to total blood levels of 5 and 12 mg/L,
24 respectively. Also, NEDO (1987) observed potential signs of CNS degeneration in
25 histopathology reported for monkeys exposed chronically to 100 ppm for 21 hours per day,
26 which is estimated to be associated with an increase in methanol blood levels over endogenous
27 levels of approximately 1 mg/L (based on EPA monkey model), corresponding to total methanol
28 blood levels of roughly 3 mg/L (assuming an endogenous background in these monkeys of
29 2 mg/L).
30 Regarding the comment warning that it should not be assumed that "high-end [of the
31 distribution of background] exposures [endogenous methanol blood levels] which are apparently
32 safe for some individuals are necessarily safe for all", EPA agrees some individuals may have a
33 high background level of methanol and/or high susceptibility. However, for the purposes of this
34 assessment, EPA assumes that endogenous blood levels of methanol in a human population with
35 normal background variation do not elicit adverse health effects. This greatly simplifies the
May 2013 A-39 Draft - Do Not Cite or Quote
-------
1 derivation of an RfD and an RfC which are, by definition, population level estimates (including
2 sensitive populations) of the amount of a substance that a person can inhale (or ingest) every day
3 over the course of a lifetime [above endogenous levels] without an appreciable risk of harm.
4 As discussed in response to Charge A2 Comment 1, the discussion of the RfC and RfD
5 and their relation to endogenous background blood levels has been clarified in the revised draft
6 assessment (see Section 5.3.6). In summary, EPA does not feel that the RfD of 2 mg/kg-day and
7 RfC of 20 mg/m3 are overly conservative. They are well above the levels associated with
8 common exposures to methanol and they appropriately account for uncertainty regarding the
9 sensitivity of primates to the reproductive and developmental neurotoxic effects of methanol.
A.2. Public Comments
10 Comment 1: "EPA has arbitrarily decided to establish these reference levels to identify
11 risks ONLY for exposure to methanol that increases the body burden of methanol or its
12 metabolites."
13 Response: As discussed above in response to external peer review Comment 1 of Charge
14 A2 and comments associated with the Bonus Charge Question, the decision to base the RfC/D on
15 exposures that increase the body burden of an individual above their naturally occurring
16 endogenous blood levels was not arbitrary. The PBPK models used in the final assessment
17 incorporate background/endogenous concentrations of methanol; however, for BMD modeling,
18 the PBPK model estimate of background concentration is subtracted from the predicted dose
19 metric under bioassay conditions. This approach for dealing with endogenous background
20 concentrations of methanol and it metabolites avoids the issue of whether or not individuals
21 experience health effects from endogenous concentrations of methanol or its metabolites because
22 only the risk due to exposures above background is thereby evaluated. This greatly simplifies the
23 dose-response assessment for methanol and the derivation of an RfC (or RfD), which is by
24 definition a population level estimate (including sensitive populations) of the amount of a
25 substance that a person can inhale (or ingest) every day over the course of a lifetime [above
26 endogenous levels] without an appreciable risk of harm.
27 Comment 2: "The recommended reference levels represent a very small addition to the
28 average person's body burden of methanol and by implication suggest that half the population is
29 at risk from their own background level of methanol."
30 Response: The approach taken by EPA in deriving the RfC and the RfD assumes that
31 endogenous blood levels of methanol in a human population with normal background variation
32 do not elicit adverse health effects. There is currently little evidence, epidemiological or
May 2013 A-40 Draft - Do Not Cite or Quote
-------
1 otherwise, to challenge this assumption. Given this assumption and lack of evidence to the
2 contrary, if the 2 mg/kg-day RfD or 2x 101 mg/m3 RfC were so low that the resulting (predicted)
3 change in methanol blood levels was only a small fraction of the normal variation in background
4 levels (e.g., 1% of one standard deviation), one could argue that this would be indistinguishable
5 from natural variation and lexicologically irrelevant. Therefore, a comparison of the expected
6 increase in methanol levels in blood resulting from exposure to methanol at the level of the RfC
7 or RfD to the variation in endogenous (i.e., background) levels of methanol observed in humans
8 is provided in Section 5.3.6 to determine if this might be the case. As shown in Figures 5-3, 5-4
9 and B 17, the estimated increase in blood levels of methanol resulting from exposure to methanol
10 at the RfC alone, at the RfD alone, or at the RfC + RfD combined is comparable to the variability
11 (represented as one standard deviation) observed in the average estimated methanol blood levels
12 (1.5 ± 0.7 mg/L) in humans (see Table 3-1 and Section 5.3.6). This then demonstrates that the
13 estimated increase in levels of methanol from the RfD or the RfC (or from the RfC + RfD) are
14 distinguishable from natural background variation, but the overall derivation of the RfD and RfC
15 ensures that these increases will not significantly increase adverse health outcomes.
16 Comment 3: "EPA incorrectly supports its decision to ignore the naturally-occurring
17 background levels of methanol in human blood by citing the results of its PBPK modeling."
18 Response: As explained above in response to external peer review Comment 2 of Charge
19 A2, EPA has re-calibrated the PBPK models to account for background levels and has used them
20 to derive the revised RfD and RfC. Hence, the justification for not including a background term
21 in the PBPK models has been removed from the toxicological review.
22 Comment 4: "While the increment of a reference level dose of methanol is a small
23 percentage of the average background blood level in humans, the intake of certain common
24 foods can easily exceed EPAs recommended reference level."
25 Response: Due to changes in the rat PBPK model, the RfC and RfD are several-fold
26 higher than the previously proposed values and, as explained above in response to comments
27 related to the external peer review Bonus Charge Question, the increase in an individual's
28 methanol blood levels after an exposure equivalent to the final RfC or RfD is expected to be well
29 in excess of the increase that would be associated with a "common exposure" such as from a
30 gl as s of orange j ui ce.
31 Comment 5: "Application of a physiologically based pharmacokinetic (PBPK) model to
32 these study data that is inappropriate for modeling exposures to pregnant animals, neonates, and
33 weanling rats, and that is based on a data set that severely underestimates the likely exposures in
34 both studies."
May 2013 A-41 Draft - Do Not Cite or Quote
-------
1 Response: As explained above in response to external peer review Comment 2 of Charge
2 A2, EPA recognizes that neonatal blood levels will likely be higher, approximately 2-fold higher
3 for rats, than maternal blood levels of methanol. However, the ratio of blood concentrations
4 between a human infant and its mother is not expected to be significantly greater than the
5 approximate 2-fold difference that has been observed between rat pups and dams. Further, the
6 health-effects data indicate that most of the effects of concern are due to fetal exposure, with
7 only a small influence due to postnatal exposures. As stated in Section 5.1.3.2.2, for these
8 reasons and because EPA has confidence in the ability of the PBPK model to accurately predict
9 adult blood levels of methanol, the maternal blood methanol levels for the estimation of HECs
10 from the NEDO (1987) study were used as the dose metric.
11 Comment 6: "Failure to confirm the results of EPAs PBPK modeling against blood
12 methanol concentration data collected in both [Rogers et al. (1993b) and NEDO (1987)1 studies."
13 Response: The mouse PBPK model has now been removed from the assessment and the
14 blood concentration data from Rogers et al. (1993b) used directly for the benchmark dose
15 analysis. For example, the BMDLio for the mouse cervical rib effect based on the blood Cmax
16 metric has thereby changed from 94.3 mg/L to 90.9 mg/L (both representing concentration
17 increases above background). Data were not available for validating the rat PBPK model
18 predictions prior to the methanol external peer review. Subsequent to the methanol external peer
19 review, EPA received blood measurements from the NEDO (1987) rat study and has validated
20 model predictions against them (Appendix B, Sections B.2.3, B.2.4 and B.2.5).
21 Comment 7: "Recent research by Dr. Peter Wells of the University of Toronto, which we
22 detail in these comments, raises serious questions about the use of rodent models for hazard
23 assessment of methanol in humans because rodents and humans metabolize methanol very
24 differently."
25 Response: As explained above in response to comments made under external peer review
26 Charge D2, a detailed discussion of the University of Toronto findings have been added to
27 Section 5.3 "UNCERTANTIES IN THE INHALATION RFC AND ORAL RFD" of the
28 toxicological review (see Section "5.3.5 Choice of Species/Gender"). The in-vitro study of Miller
29 and Wells (2011) demonstrated that methanol-induced developmental effects are enhanced in
30 mouse embryos with low catalase activity and reduced in mouse embryos with high catalase
31 activity. The authors propose that this observation is related to methanol's impact on catalase's
32 ability to control the damaging effects of reactive oxygen species (ROS) activity, which would be
33 greater in mouse embryos with low catalase activity. However, as discussed in Section 5.3.5,
34 there are several problems with this interpretation, including that in vivo results from the same
35 laboratory (Siu et al., 2013) do not support the Miller and Wells (2011) in vitro findings. The
May 2013 A-42 Draft - Do Not Cite or Quote
-------
1 University of Toronto studies are informative, but do not demonstrate conclusively that rodent
2 developmental studies are irrelevant to humans.
3 Comment 8: "Severe reporting deficiencies in the two-generation reproductive study,
4 including a lack of mean and individual animal data in the main study and the absence of details
5 regarding methods and data related to maternal or gestational outcomes in the supplementary
6 study."
7 Response: As described in Section 5.1.2.2, the supplementary study to the NEDO two
8 generation study provides sufficient dose and response information for a benchmark dose
9 analysis. Uncertainties associated with this study as they relate to the benchmark dose analysis,
10 including the absence of a detailed reporting of methods and maternal or gestational outcomes,
11 are discussed in Section "5.3.1 Choice of Study/Endpoint." Though the methods for this
12 supplementary study are not described, the methods for the parent two-generation study are
13 adequately described and it is reasonable to assume that the supplementary study was performed
14 under the same protocol starting with a number of FO females appropriate for a one-generation
15 developmental study (see response to public Comment 11 below). While data related to maternal
16 or gestational outcomes in the supplementary study are not given, signs of overt maternal
17 toxicity were not reported in the two-generation study at similar exposure levels and it is
18 reasonable to assume that they did not occur, and would have been reported had they been
19 observed, in the supplementary study. While this supplementary study no longer forms the basis
20 of the RfD, it does form the basis for the RfC because its limitations are not considered serious
21 enough to preclude its consideration as a candidate principal study and because it documents a
22 clear dose-response for a relevant endpoint for a critical organ system, brain weight reduction,
23 which is consistent with its parent two-generation study and with other teratogenicity (NEDO,
24 1987) and subchronic (TRL, 1986) study findings with respect to the effect of methanol exposure
25 on brain weight.
26 Comment 9: "The lack of utility of the NEDO (1987) reproductive study for the purpose
27 of human health risk assessment, as judged by other authoritative bodies."
28 Response: By "authoritative bodies" the commenter refers specifically to a 2002 report
29 from the National Toxicology Program's Center for the Evaluation of Risks to Human
30 Reproduction (NTP CERHR) on the reproductive and developmental toxicity of methanol. The
31 methanol toxicological review cites the subsequently peer reviewed and published version of this
32 report (CERHR, 2004). In the more recent, published version of the report, the CERHR panel
33 states that "a summary of a two-generation rat reproductive toxicity study done by the Japanese
34 NEDO was received, but data were not available in sufficient detail for Expert Panel review."
35 The NEDO summary reviewed by the CERHR panel did not contain the more detailed
May 2013 A-43 Draft - Do Not Cite or Quote
-------
1 supplementary study data, with pup brain weight means and standard deviations that EPA
2 evaluated in its benchmark dose analysis (Appendix D). This information, as well as
3 supplemental methanol blood measurements, was obtained by EPA after the CERHR panel
4 completed its report. In addition, EPA sponsored an external peer review (ERG, 2009) of the
5 NEDO (1987) report that contained this information, along with two other NEDO chronic rat and
6 mouse studies (NEDO, 1985a, b). This expert peer review panel of five scientists was asked
7 specifically to "Describe the reliability of the subject NEDO studies for consideration in the
8 derivation of EPA IRIS quantitative health benchmarks." With respect to this charge and the two-
9 generation study, including the supplementary study, the main concerns expressed by the peer
10 reviewers were that they "may be useful for RfD derivation if brain weight changes persist when
11 normalized by body weight" and that the authors "should have used ANOVA plus multiple
12 comparison tests to analyze these data." With respect to the former concern, NEDO only reported
13 means and standard deviations for absolute brain weight change and did not report body weight
14 data for the supplementary study. However, body weight data reported for the parent, two-
15 generation study did not indicate a body weight effect in the exposed Fl or F2 generation pups.
16 Further, the absolute brain weights reported by NEDO are an appropriate basis for a dose-
17 response assessment. With respect to the concern over the lack of appropriate statistical testing,
18 EPA did not rely on the NEDO statistical determinations, but performed its own more definitive
19 benchmark dose analysis of the data (see response to external peer review Comment 1 of Charge
20 Bl). Hence, while the NEDO report of the two-generation supplementary study results has
21 limitations, particularly with respect to reporting of methods (see discussion in Section 5.3.1), it
22 was not evaluated by the NTP CERHR (2004) expert panel, and it was not deemed inadequate,
23 for the purposes of RfC derivation, by a panel of expert peer reviewers (ERG, 2009).
24 Comment 10: "The use of exposure regimens in both the reproductive study and the
25 24-month rat study that confound the estimates of exposure."
26 Response: The stated basis for this comment is the concern over (1) "consumption of
27 [methanol] contaminated feed," (2) "ingestion of methanol during the act of preening," (3)
28 dermal absorption in adult rats, and (4) "increased dermal absorption of neonatal animals [over
29 adults]" because "the epidermal layers of neonatal rats are thinner than those of adult animals,
30 they lack fur for the first week after birth." With respect to the first concern, EPA estimates that
31 data on a rodent breeder diet (http://www.labdiet.com/pdf/5013.pdf) indicates 10% moisture
32 content. If one assumes that chamber methanol concentrations equilibrate with this moisture
33 content, using the blood:air PC for methanol, and uses a typical pregnancy food consumption of
34 30 g/day in rats, then the amount of methanol ingested in the chow would be about 3% of that
35 inhaled during a 22 hr/day exposure. This is likely an upper bound since it would take some time
May 2013 A-44 Draft - Do Not Cite or Quote
-------
1 for methanol to diffuse into and through a container of chow (equilibration with the chow would
2 take time), and fresh chow is provided each day. Thus the amount ingested by this route is not
3 considered significant and dosimetry calculations have not been adjusted to reflect that
4 possibility.
5 With respect to the second, third and fourth concerns, methanol is not known to adhere to
6 or be absorbed by rat dermal surfaces in amounts that would significantly impact model
7 predictions. According to Perkins et al. (1996b page 160):
8 "The method [flow-through chamber exposure], like all whole-body methods, exposes
9 the animal to the vapor at all dermal surfaces. For the very water-soluble vapors, such as
10 methanol, dermal exposure is not significant; indeed, when taring the chamber by
11 inserting a dead rat versus just opening an empty chamber for the same length of time, no
12 difference in methanol loss was noted except at 20,000 ppm, in which case the steady-
13 state loss was 27% higher with a dead rat than an empty chamber. This higher steady-
14 state loss at 20,000 ppm methanol may be related to physical properties of the compound;
15 at the high vapor level, somewhat more methanol may condense and become adsorbed to
16 the fur. Further experimentation is required to clarify-the significance of this
17 observation."
18 Since the concentrations used in the NEDO rat studies were well below the 20,000 ppm
19 level at which Perkins et al. (1996b) observed a difference in methanol loss in the chamber,
20 methanol dermal absorption is not expected to significantly impact model predictions of
21 methanol blood levels in the NEDO rat studies [also, EPA presumes that the empty-chamber loss
22 rate, of which the dead rat caused a 27% increase, was fairly small. The actual loss rate was not
23 reported by Perkins et al.(1996b)]. Though the Perkins et al. conclusion was based on an adult
24 rat, there is no scientific basis for the belief that neonatal rats would absorb a significantly
25 greater amount of methanol.
26 Comment 11: "Use of an insufficient number of parental animals in the supplementary
27 reproductive study (from which EPA derives its RfC) to support proper statistical evaluation."
28 Response: The basis provided by the commenter for this statement is that "EPA and
29 Organisation for Economic Cooperation and Development (OECD) guidelines recommend
30 evaluation of at least 20 litters per group in a two-generation reproduction toxicity test, in order
31 to ensure sufficient statistical power in the study (OECD. 2001: U.S. EPA. 1998b)." The number
32 of FO parental animals used in the NEDO two-generation study (30 males and 30 females per
33 dose group) was appropriate and in accordance with both EPA and OECD two-generation
34 reproduction toxicity test guidelines. The supplementary study performed by NEDO does not fall
35 under these guidelines because it was not a two-generation reproduction study. According to
May 2013 A-45 Draft - Do Not Cite or Quote
-------
1 NEDO (1987 page 201 ), the purpose of the supplementary study was "to confirm its [decreased
2 brain weight] relationship with the treatment and to know from what period after birth such
3 changes would appear" and, therefore, the test rats were only exposed "from Day 0 of gestation
4 throughout the Fl generation." This type of study and purpose would more appropriately fall
5 under the Agency's developmental neurotoxicity guidelines (U.S. EPA, 1998b), which state that
6 "on postnatal day 11, either 1 male or 1 female pup from each litter (total of 10 males and 10
7 females per dose group) should be sacrificed" and that "brain weights should be measured in all
8 of these pups." The number of FO parental animals included per group in the supplemental
9 experiment was not reported. However, the number of pups per dose group was reported and it is
10 reasonable to assume that, consistent with the standard culling protocol used for both the Fl and
11 F2 generations of the two-generation study (NEDO, 1987 pages 185 and 189 ), each dose group
12 pup came from a different litter (to avoid problems associated with litter correlation). Hence, by
13 examining more than 10 male and 10 female litter-specific pups per dose group at three time
14 points (3, 6 and 8 weeks), the NEDO supplementary study actually went well beyond EPA
15 recommendations for this type of study.
16 Comment 12: "Use of statistical methods in both the reproductive study and the 24-
17 month rat study that, by today's standards, are considered inadequate."
18 Response: As mentioned above in response to public Comment 9, EPA did not rely on
19 the NEDO statistical determinations, but performed its own more definitive benchmark dose
20 analysis of the NEDO rat two-generation an teratogenicity data (see response to external peer
21 review Comment 1 of Charge Bl).
22 Comment 13: "Derivation of an RfC based on absolute brain weight data without
23 considering the significance of other gestational outcome data (including body-weight data) that
24 would put these data in proper context for risk assessment purposes."
25 Response: As mentioned above in response to public Comment 9, the absolute brain
26 weights reported by NEDO in a supplementary developmental study are an appropriate basis for
27 a dose-response assessment (also see response to external peer review Comment 1 of Charge
28 Bl). Other gestational outcome data, including body weight data, were not provided for the
29 supplementary developmental study. However, body weight data reported for the parent, two-
30 generation study did not indicate a body weight effect in the exposed Fl or F2 generation pups.
31 The commenter argued that relative brain weights are important for neonates. While it would
32 have been helpful to have the body weight information for the neonates from the supplementary
33 study, the two-generation data indicate that methanol does not significantly impact pup body
34 weight at the exposure levels of concern. Further, because brain weights are conserved in both
May 2013 A-46 Draft - Do Not Cite or Quote
-------
1 neonates and adults, a dose-related reduction in absolute brain weight is an important
2 consideration for both neonates and adults.
3 Comment 14: "Lack of proper consideration of species differences in sensitivity to
4 developmental toxicity due to methanol exposures in the RfC derivation."
5 Response: The commenter cites differences in breathing rates, minute volumes and
6 metabolism (i.e., the preference for metabolism via catalase over ADH that is unique to rodents)
7 as factors that are not properly considered. The first two factors are accounted for by the
8 Agency's rat and human PBPK models. The latter factor is considered extensively in the
9 toxicological review (e.g., Section 5.3.5) and is discussed above in response to external peer
10 review Comment 1 of Charge A4, Comment 1 of Charge D2, and public Comment 7. There is
11 currently not enough known about methanol's teratological mode of action to conclude that
12 rodent developmental studies are not relevant to humans.
13 Comment 15: "Failure of EPA to consider more robust developmental toxicity data in
14 derivation of an RfC value." Specifically, the commenter suggests that the Rogers et al. (1993b)
15 study would be the more appropriate study on which to base an assessment of the developmental
16 toxicity of methanol.
17 Response: EPA agrees that the Rogers et al. (1993b) study is an appropriate study, and
18 the final RfC and RfD are derived from quantitative analyses of the Rogers et al. (1993b) and
19 NEDO (1987) studies.
20 Comment 16: "Sweeting et al. demonstrated a large difference in developmental toxicity
21 between mice and rabbits; minor differences in number of stillbirths or postpartum mortality do
22 not equal developmental effects and the EPA's reliance on them is not appropriate."
23 Response: EPA is not relying on the Sweeting et al. (2011) study results as evidence of
24 teratogenic effects in rabbits, but simply points out that their claim that rabbits are resistant
25 relative to mice to the teratogenic effects of methanol needs to be verified over several
26 gestational days, as has been done for mice, because the critical gestational window for
27 developmental effects could be different for rabbits versus mice. Under different study
28 conditions, the observed increase in postpartum lethality (11% versus 5% in controls) and
29 stillbirths (4% versus 0% in controls) may prove significant given that postpartum lethality
30 ("wasting syndrome") and a shortened gestational period were possible adverse outcomes
31 observed in methanol exposed monkeys (see discussion of Burbacher, et al., (2004a; 1999a) in
32 Section 4.3.2).
33 Comment 17: "The draft assessment states that Sweeting et al. (2011) suggests that low
34 ADH activity in mouse embryos could lead to a 'greater depletion of catalase.' The assessment
May 2013 A-47 Draft - Do Not Cite or Quote
-------
1 further states (line 7,8) 'If ROS accumulation due to this catalase consumption...' Sweeting et al.
2 do not postulate a depletion or consumption of catalase."
3 Response: The text in the final assessment has been clarified.
4 Comment 18: "The [draft] IRIS Assessment should note here [page 9, paragraph 1] that
5 Sweeting et al. postulated that methanol and/or its metabolites may enhance the embryonic
6 production of ROS (by mechanisms that do not involve catalase)."
7 Response: The text in the final assessment has been clarified.
8 Comment 19: "The use of the citation from the Tran et al. study to suggest that embryos
9 are in danger of development effects of methanol draws overly broad conclusions from a very
10 limited study."
11 Response: EPA is not making a broad conclusion regarding the danger of methanol to
12 human fetuses based on the Tran et al. (2007) study. The Tran et al. (2007) study lends
13 uncertainty to the hypothesis presented by others, including Sweeting et al. (2011), that
14 developmental studies in mice are not relevant to humans because human infants do not rely on
15 catalase to metabolize methanol as do mice. The Tran et al. (2007) study provides limited
16 evidence that catalase may play a role in the metabolism of alcohols in neonates.
17 Comment 20: "The embryo culture model [used in the Miller and Wells (2011) study]
18 removes the confounding effects of maternal catalase activity, and specifically the maternal
19 peroxidative activity of catalase responsible for metabolizing methanol."
20 Response: This is offered by the commenter as an explanation for why the in-vivo
21 studies of Siu et al. (2013) did not observe the enhanced embryopathies in aCat (catalase-
22 deficient) mice that were reported in the in-vitro studies of Miller and Wells (2011). As discussed
23 in Section 5.3.5, Miller and Wells (2011) acknowledge that aCat mice in the in-vivo study of Siu
24 et al. (2013) "appeared resistant to methanol teratogenicity." However, they suggest that the in-
25 vivo results were confounded by "maternal factors, including the metabolism of methanol and its
26 formic acid metabolite by maternal catalase (Dorman et al., 1995)," which would presumably
27 reduce the methanol body burden to levels that do not competitively inhibit embryonic catalase
28 antioxidant activity. Alternatively, maternal factors could be protecting the embryo from a more
29 direct interaction with methanol, the compound which this assessment assumes to be the toxic
30 agent.
May 2013 A-48 Draft - Do Not Cite or Quote
-------
APPENDIX B. DEVELOPMENT, CALIBRATION,
AND APPLICATION OF A METHANOL PBPK
MODEL
B.1. Summary
1 This appendix describes the development, calibration, and approach for application of
2 PBPK models for adult (non-pregnant) Sprague-Dawley (SD) rats and humans to extrapolate rat
3 methanol inhalation-route internal dose metrics to human equivalent inhalation exposure
4 concentrations (HECs) or oral exposure doses (HEDs) that result in the same internal doses. This
5 model is a revision of the model reported by Ward et al. (1997), reflecting significant
6 simplifications (removal of compartments for placenta, embryo/fetus, and extraembryonic fluid)
7 and several elaborations (details follow), which allow the model to describe methanol blood
8 kinetics. The reasoning for removal of the pregnancy description is given in Section 3.4.1.2, so is
9 not reiterated here.
10 The model includes compartments for lung/blood methanol exchange, liver, fat, and the
11 rest of the body. A single set of parameters was identified for each species modeled, whereas
12 Ward et al. (1997) employed a number of data-set specific parameters. Fitting parameters to each
13 data set make it difficult at best to apply the model to bioassay conditions (i.e., to extrapolate the
14 model to exposure scenarios not used for model calibration). Other biokinetic methanol models
15 that were considered as starting points for the current model also used varied parameters by data
16 set to achieve model fits to the data. For example, the model of Bouchard et al. (2001) used
17 different respiratory rates and fractional inhalation absorbed for different human exposures.
18 Thus, model re-calibration using a single set of parameters was considered necessary for use in a
19 health assessment.
20 The model structure common to rats and humans is described in further detail in Section
21 B.2.1. Three model features are species-specific:
22 • (1) A term to account for observed decreases in respiration rate (and assumed
23 corresponding decrease in cardiac output) was used to match rat data for rats reported by
24 Pollack and Brouwer (1996). Human exposures used for model calibration, and for which
25 model application is expected, are assumed to be low enough that the term is inactive for the
26 human model.
27 • (2) A urinary bladder compartment is used to simulate urine excretion time-course data in
28 humans. Human urinary data are sufficient to identify a bladder residence-time constant, but
29 no such data are available for rats. Urinary elimination is included in the rat model, but the
30 kinetics of methanol appearance in rat urine were not analyzed.
May 2013 B-1 Draft - Do Not Cite or Quote
-------
1 • (3) For rats, the body:blood PC had to be adjusted to match the short-time i.v. data (i.e., the
2 data indicated that the volume-of-distribution predicted by assuming that body:blood
3 partitioning was identical to muscle:blood was incorrect). But once this was done, the oral
4 data were well predicted with 100% bioavailability. However for humans, since there was a
5 single limited i.v. data set which could not be used to calibrate the body:blood PC, the value
6 for muscle was used, but it was then found that less than 100% oral bioavailability must be
7 used to match the oral PK data.
8 Further details of and justification for these features are given in corresponding Sections below.
9 Algebraic functions which approximate the full human PBPK model to within -1% are
10 also presented. These functions allow one to calculate human oral methanol doses (HEDs) and
11 inhalation concentrations (HECs) yielding internal dose(s) equal to specified maximum
12 concentrations (Cmax values) or area-under-the-curve (AUC) values, specifically to match
13 internal doses (internal PODs) determined from rat dose-response data.
B.2. Model Development
B.2.1. Model Structure
14 The model structure is shown in Figure B-l. A gas-exchange model for inhalation
15 exposure was added, with an adjustment factor (FRACIN) for methanol absorption/desorption in
16 the conducting airways, as was done by Fisher et al. (2000), to describe delivery of methanol to
17 blood as a function of ventilation, partitioning, and blood flow rather than the less standard
18 approach used by Ward et al. (1997). A second (non-physiological) GI compartment was added
19 to better describe oral uptake in rats. For humans the limited oral PK data were not sufficient to
20 identify the two additional parameters associated with the second GI compartment, but the data
21 were consistent with less than 100% bioavailability. (Rat data were consistent with 100%
22 bioavailability, with the second GI compartment included.) The kidney was lumped with the
23 body compartment because the blood:tissue partition coefficients for these tissues were similar
24 and recent practice in PBPK modeling is to treat urinary clearance as occurring from the blood
25 compartment rather than the kidney tissue [for example, (Loccisano et al., 2011)]. In particular
26 this reflects the biological reality that renal excretion is initiated by filtration of blood flowing
27 through the glomeruli, rather than a partitioning form kidney tissue into the nephrons. A fat
28 compartment was included because it is the only tissue with a tissue:blood partitioning
29 coefficient appreciably different than unity, and the liver is included because it is the primary site
30 of metabolism. Background levels of methanol are included through use of a zero-order rate of
May 2013 B-2 Draft - Do Not Cite or Quote
-------
1 infusion, R0bg. Equations in the model code allow R0bg to be calculated as a function of other
2 model parameters to match a user-specified background blood or urine concentration.
Inhalation Fracin
exposure
Exhaled
air
Bladder
(human only)
Urine
Metabolism
Oral
exposure
Endogenous
production
Note: Parameters: Fracin (FRACIN), fraction of exposure concentration reaching gas exchange region in lungs; Bav,
oral bioavailability; kas, first-order oral absorption rate from stomach; kai, first-order uptake from 2nd GI
compartment; ksl, first-order transfer between stomach and 2nd GI; Vmax and Km Michaelis-Menten rate constants for
metabolism in liver; ku first-order rate constant for urinary elimination; kM, rate constant for urinary excretion from
bladder. For the rat only, high levels of methanol in the body compartment lead to respiratory and cardiac
depression, indicated by the dashed line. Rat data were consistent with Bav = 100% but humans with Bav = 83%.
Figure B-l Schematic of the PBPK model used to describe the inhalation, oral, and i.v.
route pharmacokinetics of methanol.
3 Methanol is well absorbed by the inhalation and oral routes, and is readily metabolized to
4 formaldehyde, which is rapidly converted to formate in both rodents and humans. Although the
5 primary enzymes responsible for metabolizing formaldehyde are different in rodents (CAT) and
6 adult humans (ADH); the metabolite, formate, is the same, and the metabolic rates are similar
7 (Clary, 2003). The published rodent kinetic models for methanol differ in how they describe the
8 metabolism of methanol (Bouchard et al.. 2001: Fisher et al.. 2000: Wardetal.. 1997: Horton et
9 al., 1992). Ward et al. (1997) used one saturable and one first-order pathway for mice, and
10 Horton et al. (1992) applied two saturable pathways of metabolism to describe methanol
11 elimination in rats. Bouchard et al. (2001) employed one metabolic pathway and a second
12 pathway described as urinary elimination in rats and humans, both being first-order.
May 2013
B-3
Draft - Do Not Cite or Quote
-------
1 Since metabolic reactions are known to be saturable - the rate is ultimately limited by the
2 amount of enzyme present - and metabolism is known to be the primary route of elimination in
3 rats and humans, the starting point for both rats and humans was to assume the simplest model
4 form consistent with this biochemistry: a single saturable pathway, described by Michaelis-
5 Menten kinetics. This model structure provided a reasonable fit to a range of data, with the first-
6 order urinary pathway included. If the human PK data, in particular, were completely linear, then
7 attempts to fit this structure would result in a lack of parameter convergence, with the saturation
8 constant (Km) approaching infinity, which did not occur. Parameter estimation converged to a
9 reasonable value for Km when the model was fit to the human data, and the resulting fit to the
10 data was slightly but clearly improved versus a forced first-order function (evaluated by both a
11 quantitative measure of fit and visual inspection). Further, when human model optimization was
12 begun with larger values of Km, which would make the model predictions more linear, the
13 optimization still converged to the value reported here, clearly indicating that the human data are
14 more consistent with a saturable metabolic description than a first-order description. The impact
15 of uncertainty in this model choice is discussed in the corresponding section of the review
16 (Section B.2.6.1). However, that uncertainty would not be reduced by assuming strictly first-
17 order metabolism, given the data available and results described here.
18 Inclusion of a second metabolic pathway was tested for the rat model, but was found to
19 create problems with parameter convergence and not found to significantly improve model fits.
20 For the rat, suppression of respiration rate at higher exposure levels was reported by
21 (Perkins et al., 1996a). Therefore, an empirical function was fit to the respiration rate vs. blood
22 data from Perkins et al. (Perkins et al., 1996a) and, assuming this indicates a parallel depression
23 in both cardiac output and ventilation, the function was applied to the rat cardiac output with
24 ventilation-perfusion-ratio fixed. Further details are given in Section B.2.3 on rat model
25 calibration below.
26 While the PBPK model explicitly describes the concentration of methanol, it only
27 describes the rate of metabolism or conversion of methanol to its metabolites. Distribution and
28 metabolism of formaldehyde is not considered by the model, and this model does not track
29 formate or formaldehyde. The data needed to parameterize or validate a specific description of
30 either of these metabolites is not available. Since the metabolic conversion of formaldehyde to
31 formate is rapid (< 1 minute) in all species (Kavet and Nauss, 1990), the methanol metabolism
32 rate should approximate a formate production rate, though this has not been verified. Thus the
33 rate of methanol metabolism predicted by the model can be used as a dose metric for either or
34 both of these metabolites, but scaling of that metabolic rate metric to humans requires that the
35 rate be normalized to BW°'75, (i.e., scaled rate = mg/kg0'75 - time), to account for the general
36 expectation metabolic elimination of the metabolites scales as BW°'75, hence is slower in
May 2013 B-4 Draft - Do Not Cite or Quote
-------
1 humans. First-order rate constants were scaled as 1/BW0'25, since the resulting rate is also
2 multiplied by tissue volume which scales as BW1.
3 The model was initially coded in acslXtreme vl .4 and was subsequently updated in acslX
4 v 3.0.2.1 (The AEgis Technologies Group, Inc., Huntsville, AL). Most procedures used to
5 generate this report, except those for the optimization, may be run by executing the
6 corresponding .m files. The model code (acslX .csl file) and supporting .m files are available
7 electronically on the IRIS website (www.epa.gov/iris). A key identifying .m files associated with
8 figures and tables in this report is also provided in the supporting materials.
B.2.2. Model Parameters
9 Physiological parameters such as tissue volumes, blood flows, and ventilation rates were
10 obtained from the open literature (Table B-l). Parameters for blood flow, ventilation, and
11 metabolic capacity were scaled as BW°'75, according to the methods of Ramsey and Andersen
12 (1984). Pulmonary air-flow (Qp) was coded as the product of cardiac output (Qc) and a
13 ventilation-perfusion ratio (VPR) in order to facilitate coding of changes in these quantities due
14 to exercise or respiratory depression in rats. In particular it was generally assumed that VPR
15 remained constant, so Qp and Qc varied in proportion to one another during such changes, unless
16 data specifically indicated otherwise.
17 As briefly described in the summary, when published partition coefficients (PCs) were
18 used for all body compartments for the rat, the predicted blood levels immediately following i.v.
19 doses were not well estimated. Since those blood levels only depend on the tissue partitioning
20 and the rest-of-body compartment is compromised of multiple tissues which have differing
21 partition coefficients, it was therefore decided to initially fit the body:blood PC to the i.v. data
22 and then to the total PK data set in global parameter estimation. This approach is validated by the
23 observation that the resulting fitted PC was in the range of those measured for other tissues and
24 the rat model was then consistent with 100% oral bioavailability. Rat PCs were taken as
25 measured for that species by Horton et al. (1992) for liverblood and blood:air. The "slow-to-
26 blood" PC (1.1) in rats reported by Horton et al. (1992), is inconsistent with the value for
27 fatblood (0.083) in mice from Ward et al. (1997), and that determined for rat fatblood (0.11)
28 partitioning of ethanol by Pastino and Conolly (2000): these other results indicate much lower
29 partitioning of alcohols into fat. Therefore the Ward et al. (1997) PC for mouse fatblood, was
30 used.
May 2013 B-5 Draft - Do Not Cite or Quote
-------
Table B-l Parameters used in the rat and human PBPK models
SDRat
Human
Data Source
Body weight (kg)
0.2753
70
Measured/estimated
Tissue volume (% body weight)
Liver
Arterial blood
Venous blood
Fat
Lung
Rest of body
3.7
1.85
4.43
7.0
0.50
73.9
2.6
1.98
5.93
21.4
0.8
58.3
Brown et al. (1997)
Calculated15
Flows: Total
Cardiac output (QCC; L/hr/kg° 75)c
Ventilation-perfusion ratio (VPR)°
16.4
1
16.5
1.45
Brown et al. (1997); Perkins et al.
(1995a); U.S. EPA QOQOa)
Blood Flows: (% Cardiac Output)
Liver
Fat
Rest of body
25.0
7.0
68
22.7
5.2
72.1
Calculated
Biochemical constants'1
VmaxC (mg/hr/kg075)
Km (mg/L)
kjCCkg025/^)
21.4
29
0.153
41
36
0.034
Fitter! except rat kjC which is
calculated from Pollack and Brouwer
(1996).
Oral absorption
kas (hr-1)
k.i (hr-1)
kai (hr-1)
Bav (fraction)
12.8
3.1
0.38
1
0.21
3.17
3.28
0.79
Rat: fitted, except Bav assumed =
Human. kas, ksl, and kai aie foi eth;
[from (Sultatos et al., 2004)1; Bav
1
uiol
fitted.
May 2013
B-6
Draft - Do Not Cite or Quote
-------
SDRat
Human
Data Source
Partition coefficients
LiverBlood
FatBlood
Blood:Air
Body:Blood
Lung:Blood
Bladder time-constant (kbi, hr *)f
Inhalation fractional availability
(FRACIN, %)
1.6
0.083
1350
0.89
1
NA
0.81
0.583e
0.142
1626
0.805
1.07
0.76
0.75
Human: Fiserova-Bergerova & Diaz
(1986) (human "body" assumed =
muscle);
Rat' Hnrtnn et al C\ 999Vpxrent rat
fat:blood assumed equal to mouse
[Ward et al. (1997)1. body:blood was fit
to data (estimated), and lung:blood
assumed (approximately equal to
human)
Fitted (human)
Fitted
aThe midpoints of rat weights reported for each study was used and ranged from 0.22 to 0.33 kg
bfhe volume of the other tissues was subtracted from 91% (whole body minus a bone volume of approximately 9%) to derive the
volume of the remaining tissues.
°In the model cardiac output (QC; L/hr) was set as the primary constant, via the scaling constant QCC (QC/BW0 75), and
pulmonary ventilation (QP) was defined as the product of QC and the ventilation-perfusion ratio, VPR. QCC and VPR for humans
were obtained (VPR calculated) from U.S. EPA (2000a).
dVmax and Km represent asaturable metabolic process assumed to occur solely in the liver. Vmax used in the model =
VmaxC (mg/kg° 75-hr)XBW075. l^C is the first-order urinary elimination constant (from the blood compartment), ki used in the
model = kjC/BW0 25.
eHuman liverblood partition coefficient estimated by Fiserova-Bergerova and Diaz (1986) from correlation to measured
fatblood partition coefficient, based on data from 27 other solvents.
fkbi - a first-order rate constant for elimination from the bladder compartment, used to account for the difference between blood
kinetics and urinary excretion data as observed in humans.
NA - Not applicable for that species.
B.2.3. Rat Model Calibration
1 The SD rat model was calibrated to fit blood concentration data from intravenous,
2 inhalation, and oral exposures. However, the urinary clearance constant and the respiratory
3 depression function were specified outside of the PBPK model, using separate data. Pollack and
4 Brouwer (1996) used linear regression of urine excretion rates versus blood concentration in
5 non-pregnant rats to obtain a clearance constant, ki = 0.00916 L/hr/kg. This was converted to the
6 equivalent kiC in the PBPK model by normalizing to the venous blood fraction of BW,
7 0.0443 L/kg, and multiplying by an average SD BW of 0.3 kg (raised to 0.25) to obtain the
8 allometric constant:
0.25 /
0.25 /
9 kiC = (0.00916 L/hr/kg) x (0.3 kg)a/7 (0.0443 L/kg) = 0.153 kga/7hr.
10 As mentioned above, suppression of respiration rate in the rat at higher exposure levels
11 was reported by (Perkins et al., 1996a). An empirical function was therefore fit to the respiration
12 rate versus blood data from that source, shown in Figure B-2. It was assumed that cardiac output
May 2013
B-7
Draft - Do Not Cite or Quote
-------
\5.5->-l
1 decreased proportionately with ventilation, so the inhibition term (1+ ([MeOH])/3,940) }" , was
2 applied to the rat cardiac output with ventilation-perfusion-ratio fixed. However, when the
3 response was assumed to occur instantaneously due to changes in mixed venous blood
4 concentration (i.e., the mixed venous blood concentration was used for [MeOH]), the model
5 predicted an unreasonable level of suppression immediately after i.v. dosing because of the short-
6 term spike in blood levels predicted to occur. If instead the concentration in venous blood exiting
7 the "body" compartment was used for [MeOH], reasonable model simulations resulted. Since
8 some (short) time is likely needed for methanol to interact with the neurons involved in
9 respiratory and cardiac control, and for neural processing of the resulting signal to the heart and
10 lungs, the use of this body-tissue-blood concentration, for which the methanol concentration
11 changes are slightly delayed and "smoothed" relative to the mixed venous blood, seems a
12 reasonable option.
25
.0
JS 20
1
oj 15
— 10
o
I 5
V0/[l + ([MeOH]/3940)A5.5]
data
0
1000 2000 3000
Blood methanol (mg/L)
4000
Source: Perkins et al. (1996b).
Figure B-2 Respiratory depression in Sprague-Dawley rats as a function of blood methanol
concentration. The empirical curve fit (solid line) was selected to describe the
data with a minimal number of parameters.
13 All of the data available for SD rats are reported with background levels subtracted.
14 These data are from the laboratory of Gary M. Pollack (then at the University of North Carolina,
15 Chapel Hill) and most are also presented in the thesis of Keith W. Ward (1995). The original
16 source reported only values with background subtracted, and neither Dr. Ward nor Dr. Pollack
17 have retained any other records of these experiments (personnel communications). Therefore, the
May 2013
B-8
Draft - Do Not Cite or Quote
-------
1 methanol blood levels reported by NEDO (NEDO, 1987) in control animals, 3 mg/L, was
2 assumed for all PK experiments analyzed. Rather than adding this number to the reported data,
3 however, this background was subtracted from model simulation result obtained with this
4 background level set to match the reported data. Specifically, model simulations were run with a
5 zero-order endogenous production rate, Robg, set to produce a concentration in venous blood
6 (CVB) of 3 mg/L in the absence of exposure. This background level is denoted CvBbg and is a
7 constant in the model code. A secondary variable was defined in the code: CvBmb = CVB - CvBbg,
8 i.e., the concentration predicted including background, CVB, minus that constant background.
9 Since the rates of metabolism, including saturation, calculated in the model all used the total
10 concentration, which includes that produced from the zero-order term, this approach accounts for
11 background methanol in the animals to the extent possible, given the data, without adjusting the
12 data using an otherwise assumed background level. All of the plots which follow, demonstrating
13 model fits to various data, then show model predictions of CvBmb versus the data as reported in
14 the various publications. Total blood concentrations, CVB, are listed in tables of internal metrics
15 and show in plots depicting internal dosimetry under bioassay conditions.
16 Initial values VmaxC, Km, and the body:blood partition coefficient (PR) were then
17 obtained by fitting the model to the 100 and 2,500 mg/kg i.v. data provided in the command file
18 of Ward et al. (1997) (holding other parameters constant). As mentioned previously, if PR was
19 not also adjusted, the predicted concentration immediately following the distribution phase,
20 which are only dependent on the partition coefficients, were discrepant from the data. Without
21 adjusting PR, this then created a bias in the metabolic parameters to correct for the error in the
22 distribution phase. Model predictions were also compared to 500 mg/kg i.v. data in the command
23 file of Ward et al. (1997), with additional early time-points reported by Pollack and Brouwer
24 (1996). With PR adjusted this way to fit the 100 and 2,500 mg/kg data, the model matched the
25 initial time points of the 500 mg/kg data quite well (see Figure B-3). However, the subsequent
26 clearance rate fit to these other two dose levels was inconsistent with the 500 mg/kg data. All
27 three data sets with globally-fit model parameters are shown in Figure B-3. If one compares the
28 clearance rate of the 500 mg/kg data at 20 hours and beyond, when the concentration range is the
29 same as the 100 mg/kg data, it is clear that the two data sets are discrepant. Thus no model with a
30 single set of parameters could simultaneously match both data sets. That the model does fit the
31 2,500 mg/kg data quite well indicates that the discrepancy is not due to a simple dose-
32 dependency. Since it is most important that the model describe the low-dose data well, in the
33 range of the point-of-departure for toxicity extrapolation, while capturing as much of the high-
34 dose dependency as possible, the 500 mg/kg i.v. data were not used in subsequent model
35 calibration.
May 2013 B-9 Draft - Do Not Cite or Quote
-------
I
o
0)
T3
o
jO
CD
10,000
1,000
100
10
— 2500 mg/kg sim
— 500 mg/kg sim
— 100 mg/kg sim
• 2500 mg/kg data
* 500 mg/kg data
100 mg/kg data
10 15 20 25 30
Time (hr)
35
40
45
50
Source: Ward et al. (1997: squares'): with additional 500 mg/kg points from Pollack and Brouwer (1996).
Note: MeOH (methanol) was infused into female Sprague-Dawley (SD) rats at target doses of 100, 500, or
2,500 mg/kg. Data points are measured blood concentrations and lines are PBPK model simulations with metabolic
parameters fit to a large set of SD rat data (see text for further details).
Figure B-3 Rat i.v.-route methanol blood kinetics.
1 The initial value for the inhalation fractional absorption constant, FRACIN, was then set
2 by fitting it to inhalation PK time-course data from Perkins et al. (1996a), holding other
3 parameters constant. Model fits to these data with the final set of parameters are shown in panel
4 A of Figure B-4. Model predictions are also shown versus end-of-exposure concentrations
5 reported by Pollack and Brouwer (1996) in panel B of Figure B-4. This second data set was not
6 used in model fitting. It is worth noting that while the model significantly under-predicts the
7 1,000 and 5,000 ppm data shown in panel A of Figure B-4, the model almost exactly matches the
8 end-of-exposure concentration at 1,000 ppm in panel B, and only slightly under-predicts then
9 5,000 ppm measurement from that data set. Also the downward curvature which is noticeable in
10 the 20,000 ppm simulation and to a lesser extent at 15,000 ppm (panel A), and above ~
11 12,000 ppm in panel B, is due to the respiratory depression term.
May 2013
B-10
Draft - Do Not Cite or Quote
-------
CD
• 1000 ppm data
* 5000 ppm data
A 10000 ppm data
V 15000 ppm data
O 20000 ppm data
1000 ppm sim
5000 ppm sim
10000 ppm sim
15000 ppm sim
20000 ppm sim
A
o
v
O
<Ł,
V
..--A"
45
Time (hr)
00
en
4,000
3,500
3,000-
J=, 2,500-
c
.9 2,000-
Ł 1,500-
u
g 1,000-
u
O
J±!
CD
500-
0-
B
0
5,000 10,000 15,000
Exposure concentration (ppm)
20,000
Source: (Panel A): Perkins et al. (19%a); (Panel B): Pollack and Brouwer (1996).
Note: (A) Model fits to time-course data for 1,000-20,000 ppm exposures reported by Perkins et al. (1996a).
(B) Model predictions versus end-of-exposure data, for 8-hr exposures; data from Pollack and Brouwer (1996), not
used for parameter estimation. Model results are with globally fit parameters. The noticeable downward curvatures
seen in the 20,000 ppm model prediction (panel A) and above ~ 12,000 ppm in panel B are the due to the inclusion
of the respiratory depression term in the PBPK model.
Figure B-4 Model fits to data sets from inhalation exposures in female Sprague-Dawley
rats.
May 2013
B-11
Draft - Do Not Cite or Quote
-------
1 Oral absorption parameters were first fit to the lower dose (100 mg/kg) oral absorption
2 data reported by Ward et al. (1997) (with other parameters held constant). The initial fit, with a
3 single GI compartment was not very good, even with the oral bioavailability adjusted at the same
4 time (dashed line in panel B of Figure B-4). Therefore, an empirical (non-physiological) second
5 GI compartment was considered, like that used by Sultatos et al. (2004) for ethanol. With
6 bioavailability fixed at 100%, use of this second compartment gave an excellent fit to the data
7 (solid line in panel B of Figure B-4). Therefore the two-compartment GI structure was used.
8 While the fit to the 100 mg/kg oral data was quite good, the fit to the 2,500 mg/kg data
9 exhibited a much faster and higher peak than shown by the data and under-predicted the data
10 between 24 and 31 hr, during the clearance phase (Figure B-5, panel A). Notably, the two-
11 compartment model reproduces these high-dose data much better than the single-compartment
12 model. Even when the model was fit to both the high- and low-concentration data
13 simultaneously, the fit to the high-concentration data could not be significantly improved without
14 completely degrading the low-concentration fit (not shown). Several variations of the GI
15 compartment rate equations were tested, in part reflecting data available from the ethanol
16 literature, but none could significantly improve the fit to the 2,500 mg/kg data without
17 introducing otherwise untested parameters and hypotheses. Since the primary concern is with
18 fitting low-dose data, which produce blood concentrations near the point of departure, it was
19 therefore decided to not use the 2,500 mg/kg data for parameter estimation, though comparisons
20 of model predictions to those data are still presented, and to use first-order kinetics with the
21 empirical, two-compartment GI model shown in Figure B-l.
22 A final set of fitted model parameters for the rat was obtained by allowing all of the
23 adjusted parameters (VmaxC, Km, PR, kas, ks;, ka;, and FRACIN) to the data sets as described
24 above: 100 and 2,500 mg/kg i.v. doses [Figure B-3; (WardetaL 1997). squares]; 1,000 to
25 20,000 ppm inhalation time-course data [Figure B-4, panel A, (Perkins et al., 1996a)], and
26 100 mg/kg oral dose data [Figure B-5; (Ward et al., 1997)1. The resulting parameter values are
27 listed in Table B-l and the simulations with solid lines in Figures B-3 to B-5 all use this global
28 set. Although the model does not fit all of the data as well as one might like, particularly the
29 1,000 and 5,000 ppm data in Figure B-4, panel A, the overall quality of the fits is considered
30 good. The number of parameters adjusted is considered modest, since could reduce the number
31 of parameters by keeping PR at a value measured for muscle or using a one-compartment GI
32 model. But either of these choices significantly degrades the model fits (shown for GI model),
33 which indicates that the number and variety of data available are sufficient to inform these seven
34 fitted parameters. One can consider the urinary excretion constant, kiC, and the two parameters
35 used to define the level of respiratory depression, as additional adjusted parameters. These two
36 parameters were fit using additional data on urinary excretion and respiration rate, respectively.
37 So the total number of fitted parameters is considered to be well supported by the corresponding
May 2013 B-12 Draft - Do Not Cite or Quote
-------
1 data used to determine their values, and hence a fairly good level of confidence should be held
2 for model predictions of bioassay dosimetry. To further elucidate the level of confidence one can
3 place in model predictions, evaluation of model sensitivity to these parameters was conducted as
4 described in Section B.2.4.
01
- — '
I
O
O
0
01
E
o
(D
2,700-
2,400 -
2,100-
1,800-
1,500-
1,200-
900-
600-
300-
0
100-
80-
60-
40 -
A
— 2500 mg/kg two-compartment GI model
2500 mg/kg single GI compartment model
^ 2500 mg/kg data
— 100 mg/kg two-compartment GI model
D 100 mg/kg data
o
O
O
^ 20-J
0
10
B
20 30
Time (hr)
40
50
— 100 mg/kg two-compartment GI model
•••• 100 rng/kg single GI compartment model
D 100 mg/kg data
345
Time (hr)
Source: Ward et al. (1997).
Note: Thick solid lines are PBPK model results using a two-compartment GI model (oral bioavailability = 100%);
thin dotted lines use a single GI compartment (bioavailability allowed to vary below 100%).
Figure B-5 Model simulations compared to 100 (squares) or 2,500 (diamonds) mg/kg oral
methanol data in female Sprague-Dawley rat (expanded scale in panel B).
May 2013
B-13
Draft - Do Not Cite or Quote
-------
B.2.4. Rat Model Sensitivity Analysis
1 An evaluation of the importance of selected parameters on rat model estimates of blood
2 methanol concentration was performed. Since the rat model was only used to evaluate internal
3 doses during inhalation exposures, the sensitivity to the oral uptake parameters was not
4 evaluated. The parameters which can affect inhalation dosimetry that were identified by
5 matching to PK (and respiratory response) data were VmaxC, Km, kiC [estimated by Pollack and
6 Brouwer (1996)1, PR (body:blood partition coefficient), FRACIN, and kiv (respiratory/cardiac
7 depression constant). For the purpose of comparison, the blood:air partition coefficient (PB) was
8 also included. Sensitivity of the dose metrics, Cmax and AUC (both above background) was
9 estimated under conditions of the NEDO bioassay (NEDO, 1987), 22 hr/day inhalation exposure,
10 at the bounding levels of 200 and 5,000 ppm. The analysis was conducted by measuring the
11 change in each metric resulting from a ±1% change in a given model parameter when all other
12 parameters were held fixed. The normalized sensitivity coefficient is then:
13 SC = (A metric/metrico)/ (A p/p0),
14 where metrico and p0 are the values of the metric and parameter, respectively with the
15 unchanged (as fitted) values and A metric and A p are the differences between the values obtained
16 with p increased by 1% and decreased by 1%.
17 A normalized sensitivity coefficient of 1 indicates that there is a one-to-one relationship
18 between the fractional change in the parameter and model output; values close to zero indicate a
19 small effect on model output. A positive value for the normalized sensitivity coefficient indicates
20 that the output and the corresponding model parameter are directly related while a negative value
21 indicates they are inversely related. Results are listed in Table B-2.
May 2013 B-14 Draft - Do Not Cite or Quote
-------
Table B-2
Sensitivity of rat model
dose metrics to fitted
parameters
Exposure level, metric
200 ppm
Parameter"
vmaxc
Km
PR
PB
kjC
FRACIN
Kiv
^max
-l.l
0.7
0.0
0.0
0.0
1.2
0
AUC
-1.0
0.7
0.0
0.0
0.0
1.2
0
5,000 ppm
^max
-0.2
0.0
0.0
0.3
-0.2
0.8
0.4
AUC
-0.2
0.0
0.0
0.4
-0.2
0.8
0.4
aValues are normalized sensitivity coefficients (SCs), as explained in text, for a 22 hr/day inhalation exposure to the
concentrations indicated. Parameters with SC absolute values greater than 0.2 are generally considered to be sensitive.
1 The sensitivity analysis results are mostly not surprising. At the lower concentration of
2 200 ppm, metabolic elimination has a significant influence, with both VmaxC and Km having high
3 SCs. The SC for VmaxC is negative since an increase in its value decrease blood concentration,
4 while Km is positive for the opposite reason. At 5,000 ppm VmaxC is only marginally significant
5 and Km not at all, but urinary elimination (kiC) becomes significant, though only slightly. The
6 one somewhat surprising result is that the body:blood partition coefficient, PR, has very little
7 influence on the inhalation dose predictions. However, the analysis was conducted on conditions
8 near steady-state with 22 hr/day exposure. As shown by Chiu and White (2006), the steady-state
9 level predicted in blood by a PBPK model depends on only a small number of parameters: those
10 affecting absorption, elimination (metabolic), and the blood:air partition coefficient (PB). For
11 this model, at 200 ppm the rate of absorption by inhalation is likely limited by respiration rate,
12 hence PB has little influence at that concentration, but it does significantly impact uptake at
13 5,000 ppm. More importantly, since PR has so little effect on these predictions means that any
14 uncertainty in its value is inconsequential to the outcome of this assessment. (PR is expected to
15 more strongly influence non-steady-state conditions, such as when oral ingestion occurs in
16 boluses.)
17 The fraction inhaled (FRACIN) is highly sensitive at both dose levels. The respiration
18 inhibition constant, k;v, has no influence at 200 ppm but is sensitive at 5,000 ppm. Since
19 increasing kiv decreases the level of inhibition - increases respiration - its coefficient is positive.
20 Differences in the sensitivities of the two metrics existed in the second decimal place, but
21 otherwise the two are closely correlated for this exposure scenario, hence the SCs are effectively
22 identical.
May 2013 B-15 Draft - Do Not Cite or Quote
-------
1 Thus, all of the adjusted parameters except PR have a significant influence on model
2 predictions over part of the relevent range of concentrations. Of these fitted parameters, kiC and
3 kiv were fit to independent data sets, not used to fit any other parameters. Hence a good degree of
4 confidence can be given to their values. Because of the wide range of doses, particularly by the
5 i.v. route, used for the PK data, VmaxC and Km can also be considered fairly well identified.
6 However the model's inability to fit the 500 mg/kg i.v. data (Figure B-3) and 1,000 and
7 5,000 ppm inhalation data (Figure B-4, panel A) create some level of uncertainty in their values
8 and that of FRACIN. That the model fits rather well both the 100 and 2,500 mg/kg i.v. data,
9 makes it difficult to come up with a simple explanation for the lack-of-fit to the intermediate
10 dose. Since the clearance observations at 500 ppm go beyond 24 hr, it is possible that there is a
11 time-dependent process that reduces clearance in that time range. The 2,500 mg/kg i.v. dose
12 clearance was only measured to 43 hours, when it had just dropped to ~ 100 mg/L, so one cannot
13 say if the clearance from then on would have been more like the 100 mg/kg data or the
14 500 mg/kg data.
15 For FRACIN, the poor fit to the lower two inhalation exposures (Figure B-4, panel A)
16 suggests a concentration-dependence; (i.e., FRACIN is higher at low concentrations. However,
17 even if FRACIN is set to 100%, the later time points for the 1,000 and 5,000 ppm concentration
18 curves are still under-predicted (results not shown). One hypothesis is that at low concentrations,
19 deposition in the conducting airways leads to a significant amount of absorption, not accounted
20 for in the standard gas-exchange model used here. Including such a mechanism would increase
21 model complexity significantly, and such a hypothesis should be tested by also comparing model
22 predictions to methanol gas uptake experiments, which would clearly show if methanol is being
23 taken up more efficiently at low concentrations versus an error in the model's description of
24 metabolic elimination or some other systemic process.
25 This consideration of possible model errors and potential future improvements (with
26 necessary data) should be balanced against the observation that the model-predicted blood level
27 at 8 hours from a 1,000 ppm exposure, 81 mg/L, almost exactly matches the measured
28 concentration reported in Pollack and Brouwer's (1996) (Table 16): 83 ± 15 mg/L. The
29 discrepancy between that result and the value obtained from a plot (digitized) in the same report
30 which also appears in Perkins et al. (1996a), ~ 290 mg/L, can only be attributed to experimental
31 variability, which no model can fully describe. Since the model does fit the lower-concentration
32 8-hr data (Figure B-4, panel B) fairly well, it is considered adequate for use in the assessment as
33 is, without further complication and additional parameters, and FRACIN is assumed to provide a
34 reasonable adjustment to the internal doses with the value obtained here.
May 2013 B-16 Draft - Do Not Cite or Quote
-------
B.2.5. Rat Model Simulations
1 A range of adverse developmental effects was noted in rat pups exposed to methanol
2 throughout embryogenesis (NEDO, 1987). In particular, model simulations were conducted for
3 SD rats in utero over different periods of pregnancy and as neonates via inhalation. Inhalation
4 exposures to methanol were carried out for 18-22 hours, depending on the exposure group.
5 Simulations of predicted Cmax and 22-hour exposures to 500, 1,000, and 2,000 ppm methanol are
6 shown in Figure B-6. Although the exposures in these studies are to rats over long periods and in
7 some cases exposures of the newborn pups, the model simulations are for NP adult rats only, and
8 do not take into account changes is body weight or composition. These simulated values are
9 presumed to be a better surrogate for and predictor of target-tissue concentrations in developing
10 rats, and the corresponding estimated human concentrations a better predictor of developmental
11 risk in humans than would be obtained using the applied concentration or dose and default
12 extrapolations. The logic here is simply that the ratio of actual target tissue concentration (in the
13 developing rat pup or human) to the simulated concentration in the NP adult is expected to be the
14 same in both species and hence, that proportionality drops out in calculating a HEC.
15 Figure B-6 depicts rat model simulations to determine internal doses for 22 hours/day
16 inhalation exposures at 500, 1,000, or 2,000 ppm. Atypical BW of 0.3 kg was used, since
17 predicted inhalation dosimetry is usually insensitive to the exact BW. Simulation results for
18 continuous inhalation exposures are shown for contrast. The simulations show that for all but the
19 highest dose (2,000 ppm) steady-state is reached within 22 hours, and that "periodicity," where
20 the concentration time course is the same for each subsequent day, is reached by the 3rd day of
21 exposure. At 2,000 ppm, however, steady state is not reached until after 8 days for the continuous
22 exposure. Therefore, the Cmax and 24-hour AUC were calculated by simulating 22 hours/day
23 exposures for 12 days, with the AUC calculated over the last day (24 hours) of that period. The
24 AUC values shown in Figure B-6 are calculated from the concentration increase above the
25 background or endogenous level; (i.e.,
AUC = I (C - Cbg)dt
Jo
26 where the integration is over 24 hours, C is the instantaneous blood concentration, and Cbg is the
27 endogenous/background level, set to 3 mg/L for the rat).
May 2013 B-17 Draft - Do Not Cite or Quote
-------
6 8
Time (days)
12
Exposure concentration
(ppm)
500
1,000
2,000
Cmax; (mg/L)
28.7
118
783
Cmax-Cbg;(mg/L)
25.7
115
780
AUC (C - Cbg); (mg-hr/L)
547
2,310
17,500
Note: Rat B W was set to 0.3 kg. Simulations are shown for both continuous (thin, dashed/dotted lines in plot) and
22 hours/day exposures (thick, solid lines in plot). Simulations shown are total blood concentration (including
endogenous^ackground methanol, Cbg). Cmax and AUC are determined from the 22 hour/day simulations, run for a
total of 12 days (288 hours), with the AUC calculated from the total concentration minus background for the last
24 hours of the simulation.
Figure B-6 Simulated Sprague-Dawley rat inhalation exposures to 500,1,000, or 2,000
ppm methanol.
B.2.6. Human Model Calibration
1 The rat model was scaled to human body weight (70 kg or study-specific average), using
2 human tissue compartment volumes and blood flows, and calibrated to fit the human inhalation-
3 exposure data available from the open literature, which comprised data from four publications
4 (Ernstgard et al.. 2005: Batterman et al.. 1998: Osterloh et al.. 1996: Sedivec et al.. 19811 and a
5 single data set for oral exposure (Schmutte et al., 1988). Model predictions are also compared to
6 an i.v. data set that was not used for parameter estimation (Haffner et al., 1992). Since the bulk of
7 the human data were from inhalation exposures, the approach to identifying parameters was to
May 2013
B-18
Draft - Do Not Cite or Quote
-------
1 first fit the metabolic (and endogenous level) parameters to those data sets. Initial estimates for
2 the oral uptake parameters were then obtained by fitting the oral PK data with other parameters
3 held constant. Finally a global fit over the inhalation and oral data sets combined was performed
4 to obtain final values. The two key differences in model structure and parameters adjusted are
5 discussed below, followed by a more detailed description of the calibration against specific data.
6 More specifically, the human model calibration differed from the rat calibration in two
7 ways. First, a bladder compartment was included (calibrated) to better describe the kinetics of
8 human urinary data, where both the rise and the drop in excretion rate is slower than the
9 predicted decline in blood and tissue methanol and hence rate of metabolite production. This
10 difference is shown in Figure B-7 for the 231 ppm exposure data of Sedivec et al. (1981). The
11 model-predicted venous blood and body tissue concentration curves show the pattern typical for
12 PBPK models which use the common venous-equilibration equations for tissue distribution (used
13 in this model) for fixed-duration inhalation exposures: an asymptotic rise in concentration during
14 the exposure period and then a sharp decline starting the moment that exposure ends. If urinary
15 excretion was assumed to be proportional to the body tissue concentration (which includes the
16 kidney tissue) or a separate kidney compartment was used wih the same venous-equilibration
17 equations, then the shape of the predicted time-course would simply mirror that of the tissue
18 level shown in Figure B-7, which is clearly a poor representation of the data. However, fitting the
19 one additional parameter introduced for the bladder compartment, the bladder clearance constant,
20 kbi, allows the model to reproduce the distinct kinetics of urinary excretion quite well. Thus this
21 addition is considered both biologically realistic and well justified.
22 The second difference from the rat calibration is that the body:blood partition coefficient
23 (PR) was not adjusted but the oral bioavailability (Bav) was adjusted. In particular, PR was not
24 adjusted because only limited i.v. dosing data were available (a single dose level with actual data
25 only available for one subject). Instead the value measured for muscle by Fiserova-Bergerova &
26 Diaz (1986) was used for PR without adjustment. However, when attempting to match the model
27 to the oral PK data, model predictions then significantly over-predicted those data (with
28 parameters otherwise consistent with the inhalation data). Therefore the oral bioavailability was
29 allowed to vary to less than 100% to fit the oral PK data.
30 In summary, the set of key parameters fit for the human model were the metabolic (VmaxC
31 and Km) and urinary elimination (kiC and kbi) constants, the inhalation fraction (FRACIN), and
32 the oral bioavailability (Bav). In addition, the endogenous background concentration and an
33 increment in background over time were fit to control data from Osterloh et al. (1996). A detailed
34 description of each data set and the parameter(s) that it primarily informs follow. However, as
35 with the rat, the final set of parameters was obtained by global optimization: varying all
36 parameters while fitting all of data sets simultaneously. Other human parameters were set as
37 reported in Table B-1.
May 2013 B-19 Draft - Do Not Cite or Quote
-------
10 H
Model urine from bladder
Venous blood
Body tissue
• Urine data
CD
12
Time (hr)
16
20
24
Source: Sedivec et al. (1981).
Figure B-7 Comparison of model predictions of urine concentration (from bladder
compartment), venous blood, body tissue, and urine concentration data for a
231 ppm, 8-hour exposure. Right axis provides scale for venous blood and body
tissue results.
1 The first-order rate of clearance of methanol from the blood to urine, kiC, and first-order
2 bladder compartment time constant, kbi, were used to describe urinary methanol elimination. (See
3 Section B.2.1 on the reasoning for treating urinary elimination as occurring from the blood
4 compartment versus a kidney tissue compartment.) The inhalation-route urinary methanol kinetic
5 data described by Sedivec et al. (1981) (Figure B-8) were used to inform these parameters. The
6 urine methanol concentration data reported by the authors were converted to amount in urine by
7 assuming 0.5 mL/hr/kg total urinary output (Horton et al., 1992). Since the resulting values of
8 kiC and kbi (Table B-l) are only calibrated using a small data set, they should be considered an
9 estimate. Urine is a minor route of methanol clearance in humans, with little impact on total
10 blood methanol concentration, but changes in urine levels are expected to closely reflect
11 corresponding changes in blood levels, hence the slight nonlinearity in the urine data also inform
12 the metabolic saturation constant, Km. The potential for this information is lost, however, if the
13 kinetics of urinary elimination are not well matched; i.e., if the bladder compartment is not used.
14 To estimate both the Michaelis-Menten (hepatic) and first-order (urinary) clearance rates,
15 all human inhalation data under nonworking conditions were used (Batterman et al., 1998;
16 Osterloh et al.. 1996: Sedivec et al.. 1981).
May 2013
B-20
Draft - Do Not Cite or Quote
-------
1 • The initial urine concentrations from Sedivec et al. (1981) (reported at time = 0; see
2 Figure B-8) were assumed to represent endogenous background levels, and therefore were
3 used to set a (constant) endogenous level for each exposure level to match that urinary level
4 (i.e., the endogenous blood level that must exist to match the observed urine concentration,
5 given the urinary clearance constant, kiC). The endogenous blood concentrations so
6 estimated were 0.6-0.74 mg/L.
7 • Batterman et al. (1998) subtracted background levels before reporting their results, but also
8 included the exposure-specific background (pre-exposure) concentrations in a separate table.
9 Therefore those background levels were added back to the reported exposure-group values
10 and treated as actual blood concentrations. Results of model fits to the Batterman et al.
11 (1998) data are shown in Figure B-9.
12 • Osterloh et al. (1996) measured and reported (plotted) blood methanol in nonexposed
13 controls (data shown in Figure B-10). The data for Osterloh et al. (1996) clearly show a time-
14 dependent trend which is close to linear. Therefore, the endogenous methanol production rate
15 was assumed to increase at a constant rate over time when simulating the Osterloh et al.
16 (1996) data (both controls and methanol-exposed), with the rate of increase fit to the control
17 data set. The results shown in Figure B-10 (solid lines) include this increase. For comparison,
18 the thin dashed line shows results for the 200 ppm exposure if the endogenous production is
19 assumed to be constant.
May 2013 B-21 Draft - Do Not Cite or Quote
-------
i
o
Urine concentration
• 231 ppm
O 157 ppm
A 78 ppm
12
Time (hr)
I
O
fj
Total urinary excretion
231ppm
v 157 ppm
A 78 ppm
12
Time (hr)
16
20
24
Source: Sedivec et al. (1981).
Note: Data points in lower panel represent estimated total urinary methanol elimination from humans exposed to
78 (diamonds), 157 (triangles), and 231 (circles) ppm methanol for 8 hours, and lines represent PBPK model
simulations. Solid lines are model results with the saturable equation for hepatic metabolism.
Figure B-8 Urinary methanol elimination concentration (upper panel) and cumulative
amount (lower panel), following inhalation exposures to methanol in human
volunteers.
May 2013
B-22
Draft - Do Not Cite or Quote
-------
345
Time (hr)
Source: Batterman et al. (1998).
Note: Pre-exposure blood background levels as measured for each exposure group were used: 2.0 mg/L for 30 min
group; 1.3 mg/L for 1 hr group; and 1.8 mg/L for 2 hr group.
Figure B-9 Blood methanol concentrations in subjects exposed for 30 min, 1 hr, or 2 hr at
800 ppm.
6
^ 5
.§ 4
I
2 3
s
o ^
m
• 200 pprn
^ 0 ppm
—- Constant bgd
345
Time(h)
Source: Osterloh et al. (1996).
Note: Symbols are data and lines are model simulations. An initial endogenous background level was set using a
constant rate of appearance of methanol in the liver, but this rate was increased linearly over time to match the non-
constant level in controls (diamonds); assumed to also apply to exposed subjects (squares). Thin dashed line is a
model simulation with this time-dependence turned off.
Figure B-10 Blood methanol concentrations in control (0 ppm) and methanol exposed
(200 ppm) subjects.
May 2013
B-23
Draft - Do Not Cite or Quote
-------
1 Oral PK data from Schmutte et al. (1988) from a 10 mg/kg dose was used to set an oral
2 bioavailability for humans and to test the assumption that human oral absorption of methanol
3 could otherwise be described using the simple two-compartment GI model of Sultatos et al.
4 (2004), with parameters fit by Sultatos et al. to ethanol PK data. Sultatos et al. (2004) included a
5 rate of metabolism for ethanol in the stomach, which would reduce the systemic bioavailability
6 of that compound from 100%. Lacking the data to fit a specific rate constant for methanol
7 metabolism in the GI, the simulated dose was simply reduced using a bioavailability constant
8 (Bav), but the mechanism for less than 100% availability could also be metabolism in the GI.
9 A value of Bav = 0.79 was obtained and the simulation curve matches the data of Schmutte et al.
10 (1988) fairly well (Figure B-l 1). The initial condition was set to the reported pre-exposure
11 background by Schmutte et al. (1988) (1.1 mg/L). The model reproduces the data well,
12 considering that only one parameter is adjusted for the oral dose route. Data were only collected
13 for 1.5 hours: a longer sampling time would have provided a better evaluation of the model's
14 ability to predict longer-term kinetics from oral exposures.
„ 10-
o
E
4->
O
O
•u
o
_0
CO
6-
4-
2-
0-1
0.25 0.5 0.75 1 1.25 1.5
Time (hr)
Source: Schmutte et al. (1988)
Note: The endogenous background was set to match the reported pre-exposure blood concentration of 1.1 mg/L and
the bioavailability (Bav) was calibrated to fit the data (Bav = 0.79). Otherwise the ethanol absorption parameters for
ethanol from Sultatos et al. [(2004),see Table B-l] were used.
Figure B-ll Oral exposure (10 mg/kg) to methanol in human volunteers (points).
15 The data from Ernstgard et al. (2005) was used to assess the use of the model parameters
16 with a data set collected under conditions of light work. Historical measures of VPR (2.023) and
17 QcC (26 L/hr/kg0'75) for individuals exposed under conditions of 50 W of work from that
18 laboratory (Ernstgard, 2005; Corley et al., 1994; Johanson et al., 1986) were used for the 2-hour
19 exposure period (Figure B-12). Also, a linear rate of increase in the endogenous production rate
20 was fit to the control data set, as this set showed an increasing trend over time, like Osterloh
May 2013
B-24
Draft - Do Not Cite or Quote
-------
1 et al. (1996), and the initial background level was set to match the observed value at time = 0 for
2 each data set. Otherwise, there were no changes in the model parameters (no fitting to these
3 data). The results are remarkably good, given the lack of parameter adjustment to data collected
4 in a different laboratory, using different human subjects than those to which the model was
5 calibrated.
9
10
11
12
13
14
15
16
17
58
en 7
O
c
rc
4J
T!
O
^
CD
Ernstgard et al. (2005)
a Control data
* 100 ppm data
200 ppm data
Control simulation
100 ppm simulation
200 ppm simulation
6 8
Time (h)
10
12
Source: Ernstgard et al. (2005)
Note: Data are average measured blood methanol concentrations from 4 men and 4 women exposed to 100
(98.4) ppm or 200 (192.4) ppm target (actual) methanol for 2 hours during light physical activity. Smooth lines are
PBPK model simulations using actual concentrations and an estimated BW of 75 kg (see text). The initial
concentration for each exposure group was matched to the measured level. A small constant rate of increase in
endogenous production was calibrated to fit the control data, but otherwise model parameters were not fitted. For the
first 2 hours, a VPR of 2.023 (unitless) and a QCC of 26 L/hr/kg°75 were used to match the subjects' light exercise,
after which QCC is reduced to 15 L/hr/kg075 and VPR to 1.0 (Corlev et al.. 1994: Johanson et al.. 1986).
Figure B-12 Inhalation exposures to methanol in human volunteers.
A final set of data used for model validation is provided by Haffner et al. (1992) who
observed blood kinetics in 4 volunteers after i.v. injections of 10 mg/kg methanol in a 10-minute
infusion. Model simulations based on this dosing regimen, with an assumed average BW of 70
kg, are shown in Figure B-13 versus reported data for Subject A and simulated data using
reported regression results for Subject B-D. Haffner et al. (1992) only showed data for the first
subject, but gave exponential regression equations that they fit to the data for the other subjects.
The "regression" results in Figure B-13 are calculated from regression equations provided in the
paper for each subject, at the same time points as Subject A data. The model simulation during
the first hour of exposure poorly matches the data, with the maximum blood level predicted to
only be 27.6 mg/L versus 116 mg/L observed. After 1 hour the simulation matches the data for
Subject A well, but over-predicts the regression curves for the other subjects. It is possible that
the perfusion-limited PBPK model over-predicts the rate of distribution from the blood to
May 2013
B-25
Draft - Do Not Cite or Quote
-------
1 various body tissues and hence under-predicts blood concentrations in this time period; i.e., that
2 distribution to body tissues is diffusion-limited, with the effect being significant at shorter times.
3 The slope of the simulation line closely matches that of Subject B, indicating similar clearance
4 kinetics. Subjects C and D exhibited faster elimination kinetics than predicted by the model. The
5 authors report elimination rate constants of 0.259, 0.325, 0.406, and 0.475 hr"1 for Subjects A-D,
6 respectively, so Subject D has 60% higher clearance than A.
~ 100-
o
o
g
o
o
I
10-
1-1
* Subject A (data)
• Subject B (regression)
• Subject C (regression)
A Subject D (regression)
PBPK simulation
23
Time (hr)
Source: Haffner et al. (1992)
Note: Points for Subject A are actual data. Only regression parameters were reported for Subjects C-D, so simulated
data were estimated from the regression results (points shown) at the same times as Subject A's data. See text for
further details.
Figure B-13 Intravenous exposure (10 mg/kg) to methanol in human volunteers (points).
B.2.7. Discussion and Sensitivity Analysis of Human Model.
7 Horton et al. (1992) employed two sets of metabolic rate constants to describe human
8 methanol disposition, but in vitro studies using monkey tissues with non-methanol substrates
9 were used as justification for this approach. Although Bouchard et al. (2001) described their
10 metabolism using Michaelis-Menten metabolism, Starr and Festa (2003) reduced that to an
11 effective first-order equation and showed adequate fits. Perkins et al. (1995a) estimated a Km of
12 320 + 1,273 mg/L (mean + S.E.) by fitting a one-compartment model to data from a single oral
13 poisoning to an estimated dose. In addition to the extremely high standard error, the large
14 standard error for the associated Vmax (93 + 87 mg/kg/hr) indicates that the set of Michaelis-
May 2013
B-26
Draft - Do Not Cite or Quote
-------
1 Menten constants was not uniquely identifiable using this data. Other Michaelis-Menten
2 constants that have been used to describe methanol metabolism in various models for primates
3 are given in Table B-3. Because the Km calculated by Perkins et al. (1995a) from the high-dose
4 oral exposure is 320 mg/L, while the highest observed concentration in the data sets considered
5 here is 14 mg/L (Batterman et al., 1998), forcing the model to use this higher Km would simply
6 result in fits that are effectively indistinguishable from the linear model. The value obtained in
7 this analysis, 40 mg/L, allows the model to describe the slight nonlinearity that exists in the data.
8 For example, the peak urine concentration observed by Sedivec et al. (1981) after the 231 ppm
9 exposure was increased 8.66 mg/L above the time zero value, while that observed after the
10 78 ppm exposure was 2.63 mg/L above the time zero value, so a 3-fold increase in exposure lead
11 to a 3.3-fold increase (8.66/2.63) increase in the urinary excretion above background. It is
12 possible that a much higher Km pathway is also operant in humans, but is only significant at
13 much higher concentrations than evaluated here.
Table B-3 Primate
Km (mg/L)
320 + 1,273
716 + 489
278
252 + 116
33.9
0.66
36a
Km values reported in
Reference
Perkins et al. (1995a)
Perkins et al. (1995a)
Perkins et al. (1995a)
Perkins et al. (1995a)
Horton et al. (1992)
Fisher et al. (2000)
(This analysis.)
the literature
Note
Human: oral poisoning, estimated dose
Cynomolgus monkey: 2 g/kg dose
Rhesus monkey: 0.05-1 mg/kg dose
Cynomolgus monkey: 1 g/kg dose
PBPK model: adapted from rat Km
PBPK model, Cynomolgus monkey: 10-900 ppm
PBPK model, human: 100-800 ppm
Note: The values from Perkins et al. (j_995a) are + S.E.
afhis Km was optimized while also varying Vmax, k{C, kbl, Bav, FRACIN, and parameters to fit the time-varying control data
(endogenous) of Osterloh et al. (1996) (used only for simulating that study), to the full data set.
14 Sedivec et al. (1981) estimated a fractional uptake of 57.7%, based on total amount
15 inhaled. Since the PBPK model uses alveolar rather than total ventilation and this is typically
16 assumed to be 2/3 of total ventilation, one might correct this value by dividing by 2/3 to obtain a
17 value for FRACIN of 0.8655. Ernstgard et al. (2005) also estimated a fractional uptake, 51% at
18 100 ppm and 49.3% at 200 ppm under light exercise. It is reasonable to expect uptake efficiency
19 to decrease with more rapid breathing due to exercise, since an inhaled volume element of air
20 spends less time in the respiratory tract, allowing less time for uptake, as respiration increases.
21 Also, while Ernstgard et al. (2005) based their calculation on estimated pulmonary ventilation,
22 they used the difference between inhaled air concentration and exhaled air concentration.
23 Exhaled air will be a mixture of air that was taken into the pulmonary airways and air that only
24 entered the conducting airways. Very little methanol would be absorbed from the later air and
May 2013
B-27
Draft - Do Not Cite or Quote
-------
1 hence the mixed exhaled concentration will be higher than that which exits the pulmonary region
2 and the resulting calculation will then under-estimate the fraction of methanol absorbed from
3 pulmonary air. Thus the "fraction inhaled" estimated from a given data set will depend on which
4 flow rates and concentrations are being used in the calculation, or to which it might be applied;
5 i.e., the value depends on the model "context" in which it is used. Therefore, EPA decided to fit
6 FRACIN with the other parameters estimated in the context of the PBPK model used here, as
7 was done for the rat, and obtained a value of 0.75. This indicates that the concentration entering
8 the pulmonary space is reduced by 25% due to deposition in the conducting airways (with that
9 material assumed to desorb on exhalation), and is not the fraction removed in the pulmonary
10 space. At 200 ppm, for example, the model predicts that 99.9976% of the methanol entering the
11 pulmonary region is absorbed. The value is slightly less than estimated for the rat (rat FRACIN =
12 81%) which seems reasonable since the larger human airways would reduce uptake efficiency
13 somewhat. Assuming that 2/3 of inhaled air goes to the pulmonary region, the total rate of
14 inhalation would be 1.5*Qp*CONC (rate of inhalation through nose and mouth at air
15 concentration CONC), and the amount removed in the pulmonary region roughly
16 0.75*QP*CONC (using FRACIN = 0.75), so the fraction of each breath absorbed is predicted by
17 the model to be:
18 (0.75*QP*CONC)/ (1.5*QP*CONC) = 50%,
19 which closely matches the estimates of Ernstgard et al. (2005). Considering that EPA did not fit
20 FRACIN to the Ernstgard et al. data, this appears to be a good validation of the value obtained
21 for this parameter.
22 Considering the model simulations versus the data of Haffner et al. (1992) (Figure B-13),
23 it is first evident that the model is not capturing the short-term kinetics shown for Subject A.
24 Since Haffner et al. (1992) did not indicate that the data for this subject were discrepant from the
25 other subjects in the first hour it is assumed that these data represent human distribution, hence
26 that the model does not describe well what happens immediately after such an exposure. Given
27 that i.v. exposures are not a route for which risk is estimated, model failure is not considered
28 critical here; however, it does suggest an area for future research and model improvement. The
29 model does track the longer-term clearance data for Subject A quite well. Since the model is
30 intended to represent an average adult human, it is also not alarming that it does not match so
31 well the individually-fitted clearance curves for Subjects B-D, which indicates a range of human
32 variability. In particular the results for those other subjects indicate that some people will clear
33 methanol more quickly than predicted by the model, which means that the model will somewhat
34 over-estimate internal doses and health effects for those individuals. Since other data to which
35 the model is fit are averages among individuals, and the model does not show a strong bias with
36 regard to those data (Figures B-8 to B-l 1), neither does it appear that the model is systematically
May 2013 B-28 Draft - Do Not Cite or Quote
-------
1 under-predicting clearance for most of the population. Therefore, the model predictions are
2 expected to provide reasonably good estimates of average adult human methanol PK under long-
3 term exposure scenarios. Caution is suggested, though, in potential use of the model to estimate
4 internal doses shortly after accidental exposures.
5 A sensitivity analysis for human model predictions to the primary fitted parameters was
6 conducted for continuous inhalation exposures, and results are shown in Table B-4. Normalized
7 sensitivity coefficients are calculated using the method described for the rat (see B.2.4). To
8 bracket the range of likely concern for human exposures, inhalation sensitivities were evaluated
9 at 10 and 200 ppm concentration. The bladder time constant, kbi, was not included in the analysis
10 since it has no influence on blood concentrations. The resulting coefficients (Table B-4) are not
11 surprising. VmaxC and Km both strongly influence model predictions. At these exposure levels the
12 urinary pathway (kiC) has little effect on blood level. There is essentially a 1:1 correspondence
13 with FRACIN, which follows from the fact that close to 100% of what enters the gas-exchange
14 compartment is absorbed. That all of the sensitivities are slightly higher at 200 ppm than at
15 10 ppm is due to the slight metabolic saturation.
Table B-4 Human PBPK model sensitivity analysis for steady-state inhalation exposure
Parameter
vmaxc
Km
kjC
FRACIN
Exposure level"
10 ppm
-0.81
0.74
-0.0024
1.00
200 ppm
-0.93
0.75
-0.0029
1.11
aNormalized sensitivity coefficients for steady-state blood levels (increase above background) at the indicated concentrations.
16 For oral exposures ingestion is assumed to occur in a series of six boluses over the course
17 of the day, with the fraction of the total daily dose and respective times ingested being: 25% at 7
18 a.m., 10% at 10 a.m., 25% at 12 p.m., 10% at 3p.m., 25% at 6 p.m., and 5% at 9 p.m. The pattern
19 is meant to be representative of human ingestion patterns, recognizing that this will vary among
20 the population. The impact of changing the pattern on estimated AUC values is fairly small,
21 since the total ingestion remains the same. However the pattern will clearly influence the peak
22 concentration, since an assumption of ingestion in a single bolus would lead to the highest
23 predicted daily peak, while assumption of continuous ingestion would lead to the minimal peak
24 possible, for a given total daily exposure. With the pattern used here, the blood concentration
25 profile predicted at 10 mg/kg-day is shown in Figure B-14 (time is in hours from first bolus). In
26 particular, the model predicts a following peak to occur ~ 45-55 minutes after each bolus
27 (depending on dose size), with the overall daily peak occurring just before 1 p.m. (~ 6 and 30 hr
May 2013 B-29 Draft - Do Not Cite or Quote
-------
1 time-points in Figure B-14). While the boluses assumed to be ingested at 7 a.m. and 12 p.m. are
2 both 25%, because methanol is predicted to accumulate somewhat over the morning, the later
3 bolus leads to a peak that is roughly 30% higher than the first of the day. At this exposure level
4 there is a very small residual blood level at 24 hr, about 1% of the mid-day peak. At higher
5 exposure levels more significant day-to-day accumulation would be predicted until a state of
6 "periodicity" is reached, when the day-to-day pattern no longer changes. For example, at
7 200 mg/kg-day the blood level just prior to the next day's ingestion is predicted to approach 3%
8 of the daily peak (~ 80 mg/L). At 500 mg/kg-day, the model predicts that it will take about 2
9 weeks to reach periodicity, where the peak during the first day is ~ 350 mg/L, but this increases
10 to 740 mg/L after two weeks, and the end-of-day minimum is 460 mg/L.
11 The model sensitivities to the key fitted parameters at 0.2 and 10 mg/kg-day under this
12 oral exposure scenario are listed in Table B-5. As with the inhalation sensitivity analysis, these
13 exposure levels are selected to bracket the range of primary concern for this assessment. The
14 results are qualitatively the same as for inhalation exposure (Table B-4), with oral bioavailability
15 (Bav) having an effect essentially identical to that of FRACIN for inhalation. The metabolic
16 parameters have slightly less impact for these exposure levels; probably due to blood-flow and
17 oral-absorption limitation, and the increase in sensitivity from 0.2 to 10 mg/L is not as large as in
18 going from 10 to 200 ppm inhalation concentrations.
I
O
01
5!
TJ
O
jO
CD
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48
Time (hr)
Figure B-14 Predicted human blood concentrations (increase above background) from
total daily exposures to 10 mg/kg-day methanol, consumed in a series
of6 boluses.Time is from the first bolus of the day. See text for further details.
May 2013
B-30
Draft - Do Not Cite or Quote
-------
Table B-5 PBPK model sensitivity analysis for oral exposure
Exposure level, metric"
0.2 mg/kg-day 10 mg/kg-day
Parameter C^ AUC C^ AUC
VmaxC -0.67 -0.98 -0.68 -1.01
Km 0.62 0.91 0.59 0.89
kjC -0.0014 -0.0024 -0.0015 -0.0025
Bav 1.00 1.00 1.03 1.04
"Normalized sensitivity coefficients for methanol blood levels at the indicated oral exposure rates. Human oral
exposures are assumed to occur in a series of boluses, with a blood concentration profile as shown in Figure B-14.
See text for further details.
1 Considering the multiple data sets used for human model calibration and validation, there
2 is fairly high confidence in the fitted metabolic/clearance parameters, VmaxC, Km, and klC.
3 Since the pharmacokinetics are mostly linear in the range of interest, it is really VmaxC/Km that
4 is the critical determinant of predicted internal doses, but as that is equally true of the model fits
5 to the data, this does not decrease confidence in model predictions. Where more uncertainty and
6 concern exists is with the oral bioavailability, since it is only estimated from a small data set
7 [4 individuals; (Schmutte et al., 1988)1, with measurements only extending to 90 minutes after
8 ingestion. However, bioavailability can be no more than 100%, a 25% increase over the fitted
9 value (79%). Hence any under-prediction of human dosimetry after oral exposure should be no
10 greater than that factor, well within the general variability and uncertainty expected for human
11 dosimetry (for which the UFH of 10 is used).
B.2.8. Inhalation Route HECs and Oral Route HEDs
12 The atmospheric methanol concentration resulting in a human daily blood methanol AUC
13 (hr x mg/L) or Cmax (mg/L) equal to that occurring in experimental animals following exposure
14 at the POD concentration is termed the HEC. Similarly, the oral dose (rate) resulting in human
15 daily blood methanol AUC (hr x mg/L) equivalent to that occurring in an experimental animal at
16 the POD concentration is termed the HED. For humans these estimates are made using long-term
17 exposure patterns, after a steady-state is reached from continuous inhalation exposures, or
18 otherwise there is no longer a variation from day-to-day in the blood concentration profile, given
19 an assumed consistent exposure pattern, as indicated in Figure B-14. Internal concentration
20 PODs in mice were estimated by BMD analysis applied to measured (peak) blood concentrations
21 (Cmax values), as described in Section 5. For the rat, internal Cmax and AUC values were
May 2013 B-31 Draft - Do Not Cite or Quote
-------
1 estimated using the rat PBPK model as described in B.2.5 for bioassay exposures prior to BMD
2 analysis.
3 To estimate the HEC for specific blood methanol Cmax and 24-hour AUC values,
4 continuous 1,000-hour exposures were simulated, to assure steady state was achieved, for which
5 the human Cmax was the steady state blood methanol concentration (Css) so predicted and the
6 AUC calculated from the last 24 hours of that period. (AUC = 24*CSS.). For oral exposure, the
7 daily ingestion pattern described in B.2.6.1 was used, simulations were again run for 1,000
8 hours, the Cmax selected as the maximum achieved over the resulting time-course, and the AUC
9 calculated over the last 24 hours. Results for selected exposure levels are given in Table B-6.
10 While the PBPK computational code was used to derive the HECs and HEDs used in this
11 assessment, using a computational script that will be described below, an alternative approach
12 was developed that provides an initial approximation, which also allows non-PBPK model users
13 to estimate methanol HECs and HEDs from BMDs in the form of Cmax (or Css) and AUC values.
14 This approach uses algebraic equations describing the relationship between predicted methanol
15 Cmax or 24-hour AUC and the inhalation exposure level (i.e., an HEC in ppm) (Equations 1 or 2
16 below) or oral exposure rate (i.e., an HED in mg/kg-day) (Equations 3 or 4 below). The
17 equations were derived by generating tables of exposure-dose values like Table B-6, but with
18 more entries to define the relationship, then selecting and fitting equations to interpolate among
19 the simulated points from that table. The resulting approximations match the exact PBPK model
20 results to within a few percent. To use the equations to derive an HEC or HED, the target human
21 Cmax or AUC is simply plugged into to the appropriate equation.
May 2013 B-32 Draft - Do Not Cite or Quote
-------
Table B-6 PBPK model predicted Cr
Methanol
(Css) and 24-hour AUC for humans exposed to
Inhalation exposure"
Concentration
(ppm)
1
5
10
50
100
200
500
1,000
2,000
5,000
AUC
(mg-hr/L)
0.65
3.27
6.56
33.5
68.7
145
437
1,380
15,400
115,000
c = c
^max *^ss
(mg/L)
0.03
0.14
0.27
1.39
2.86
6.04
18.2
57.3
639
4,810
Dose
(mg/kg-day)
0.1
1
10
50
100
200
500
1,000
2,000
5,000
Oral exposure"
AUC
(mg-hr/L)
0.21
2.14
22.2
130
320
984
15,000
80,600
216,000
625,000
^max
(mg/L)
0.02
0.22
2.27
12.9
29.9
81.1
751
3,610
9,520
27,300
"Values are increases above background, with an assumed endogenous background of 1.5 mg/L. For example, at
10 ppm inhalation, the total blood steady-state concentration is predicted to be 1.5 + 0.27 = 1.77 mg/L. Human
simulation results are considered uncertain above 500 ppm (inhalation) or 50 mg/kg-day (oral), since the blood
levels predicted rise above those for which there are calibration data at higher exposures.
HEC( ppm) = 0.5 54 x Css +
1734 x Css
45.73 +Css
HEC(ppm) = 0.02308 x AUC +
//ED(mg/kg-day) = 0.1 904 x Cmax
\734xAUC
1098 +AUC
4-4-0
09.9 + C max
Equation 1
Equation 2
Equation 3
HED(mg/kg-day) = 0.007257x AUC + 4\9-°xAUC Equation 4
55 y 1098 + AUC
1 In Equations 1-4 above, AUC, Css, and Cmax are above endogenous background. The endogenous
2 background blood concentration (Cmax or Css) was set to 1.5 mg/L, so the endogenous
3 background AUC =1.5 (mg/L)x24 (hr) = 36 mg-hr/L. So to identify an HEC or HED that lead to
4 a total daily AUC of 50 mg-hr/L, for example, one would then plug 50 - 36 = 14 mg-hr/L into
5 Equation 2 or 4.
6 While the preceding equations approximate the PBPK model fairly well, an exact
May 2013
B-33
Draft - Do Not Cite or Quote
-------
1 solution is preferred if the full PBPK model can be run. For example, for Cmax = 20 mg/L (above
2 background), Equation (1) estimates HEC = 538.6 ppm, but running the PBPK model at this
3 exposure level predicts a (peak) blood level of 20.2 mg/L. An exact HEC (to 4 significant
4 figures) is 535.6 ppm. Two .m file scripts were created as part of the acslX PBPK model
5 workspace for methanol, which calculate HEC and HED values through a simple search
6 algorithm (U.S. EPA, 2012b). These were used to generate all of the HEC and HED values
7 reported in Section 5 of this assessment.
B.3. Monkey PK Data and Analysis
8 In order to estimate internal doses (blood Cmax values) for the monkey health-effects
9 study of Burbacher, et al. (1999a) and further elucidate the potential differences in methanol
10 pharmacokinetics between NP and pregnant individuals (2nd and 3rd trimester), a focused
11 reanalysis of the data of Burbacher, et al. (1999b) was performed. The monkeys in this study
12 were exposed for 2.5 hours/day, with the methanol concentration raised to approximately the
13 target concentration for the first 2 hours of each exposure and the last 30 minutes providing a
14 chamber "wash-out" period, when the exposure chamber concentration was allowed to drop to 0.
15 Blood samples were taken and analyzed for methanol concentration at 30 minutes, 1, 2, 3, 4, and
16 6 hours after removal from the chamber (or 1, 1.5, 2.5, 3.5, 4.5, and 6.5 hours after the end of
17 active exposure). These data were analyzed to compare the PK in NP versus pregnant animals,
18 and fitted with a simple PK model to estimate blood Cmax values above background for each
19 exposure level. Dr. Burbacher graciously provided the original data, which were used in this
20 analysis.
21 Two cohorts of monkeys were examined, but the data (plots) did not indicate a systematic
22 difference between the two, so the data from the two cohorts were combined. The data from the
23 scatter plots of Burbacher, et al. (1999b) for the NP (pre-pregnancy), first pregnancy (2nd
24 trimester), and second pregnancy (3rd trimester) studies are compared in Figure B-15, along with
25 model simulations (explained below). Since the pregnancy time points were from animals that
26 had been previously exposed for 87 days plus the duration of pregnancy to that time point, the
27 pre-exposed NP animals were used for comparison, rather than naive animals, with the
28 expectation that effects due to changes in enzyme expression (i.e., induction) from the
29 subchronic exposure would not be a distinguishing factor. Note that each exposure group
30 included a pre-exposure baseline or background measurement, also shown. To aid in
31 distinguishing the data visually, the NP data are plotted at times 5 minutes prior to the actual
32 blood draws and the 3rd trimester at 5 minutes after each blood draw.
May 2013 B-34 Draft - Do Not Cite or Quote
-------
200 ppm
10.0 -i
o>
E
.0 -
6.0 -
o
D
I 2.0 H
-0.0-
x
x
x -
x
s
x
x
i
-
-200
-100
100
200
300
400
600 ppm
24.0 -i
x
X Pre-preg nancy
-i- 2ndTrimester
- 3rdTrim ester
Simulation
x
xf
-200 -100 0
1800 ppm 64.0 n
=5,48.0 4
O 30 -
100
200
300
400
16.0 -
-0.0
-200
-100
100 200
Time fmin)
300
400
Source: Burbacher, et al. Q999b).
Note: NP and 3rd trimester data are plotted, respectively, at 5 minutes before and after actual collection times to
facilitate comparison. Solid line is from simple PK model, fit to 2nd trimester data only.
Figure B-15 Blood methanol concentration data from NP and pregnant monkeys.
May 2013
B-35
Draft - Do Not Cite or Quote
-------
B.3.1. PK Model Analysis for Monkeys
1 To analyze and integrate the PK data of Burbacher, et al. (1999b), the one-compartment
2 model used by Burbacher, et al. (1999a) and Burbacher, et al. (1999b) was extended by the
3 addition of a chamber compartment to capture the kinetics of concentration change in the
4 exposure chamber, as shown in Figure B-16. The data in Figure B-16 [digitized from Figure 5 of
5 Burbacher, et al. (1999b)] show an exponential rise to and fall from the approximate target
6 concentration during the exposure period. The use of a single-compartment model for the
7 chamber allows this dynamic exposure period to be captured, so that the full concentration-time
8 course is used in simulating the monkey internal concentration rather than an approximate step
9 function (i.e., rather than assuming an instantaneous rise and fall). The pair of equations
10 representing the time-course in the chamber and monkey are as follows (bolded parameters are
11 fit to data):
12 Chamber: dCch/dt = [(CCM»S - Cch) • Fch - Rinh]/Vch
13 Monkey: dCWdt = [Rmh - Vmax.Cmk/ (Km + C^)]/ (V^.BW)
14 with Rmh = Cch.Rc (l,000-BW)a74.F and Cnet = C^ + Cbg.
15 d: delta, change
16 Cch: instantaneous chamber concentration (mg/L)
17 t: time (hour)
18 CCM: chamber in-flow methanol concentration (mg/L), which was set to the concentrations corresponding to those
19 reported in Table 2 of Burbacher, et al. (1999b). using the "Breeding" column for the NP (87 days pre-
20 exposed; values in Table B-7)
21 S: exposure switch, set to 1 when exposure is on (first 2 hours) and 0 when off
22 Fch: chamber air-flow, 25,200 L/hr, as specified by Burbacher, et al. (2004b) and Buibacher, et al. (2004a)
23 Rj^: net rate of methanol inhalation by the monkeys (mg/hr)
24 Vch (1,220 L): chamber volume, initially set to 1,380 L ("accessible volume" stated by Burbacher, et al. (2004b) and
25 Burbacher, et al. (2004a), but allowed to vary below that value to account for volume taken by equipment,
26 monkey, and to empirically fit the mixing time to the observed data (Figure B-16).
27 Cmk: instantaneous inhalation-induced monkey blood methanol concentration (mg/L); this is added to the measured
28 background/endogenous concentration before comparison to data
29 Vmax (32.5 mg/hr): fitted (nonscaled) Michaelis-Menten maximum elimination rate
30 Km (14.4 mg/L): fitted (nonscaled) Michaelis-Menten saturation constant
31 Vmk (0.623 L/kg): fitted volume of distribution for monkey
32 BW: monkey body weight (kg); for NP monkeys set to group average values in data of Burbacher, et al. (1999b) and
33 Burbacher, et al. (1999a)
34 Rc: allometric scaling factor for total monkey respiration (0.12 L/hours/g°74 = 2 mL/minute/g0 74), as used by
35 Burbacher, et al. (1999b; 1999a) (note that scaling is to BW in g, not kg)
36 F: fractional absorption of inhaled methanol; set to 0.5 (50%), 2/3 the value fitted for humans using the human
37 PBPK model (see AppendixB, Section B.2); for the monkey F and V^ cannot be uniquely identified, given
May 2013 B-36 Draft - Do Not Cite or Quote
-------
1 the model structure; since the monkey model uses total ventilation (defined by Rc) as the driver, while the
2 human model uses alveolar ventilation which is assumed to be 2/3 of total ventilation, F was set to 2/3 the
3 human value to obtain a realistic estimate of Vmk
4 Cnet: net blood concentration, equal to sum of the inhalation-induced concentration (Cmk) and the background blood
5 level (Cbg) (mg/L)
6 Cbg: background (endogenous) methanol concentration, set to the pre-exposure group-specific mean from the data of
7 Burbacher, et al. (1999b) and Burbacher, et al. (1999a)
2000 -
1800 -
.=.1600 -
0-1400 -
&
1800 ppm
Chamber concentration profile
600 ppm
200 ppm
-+—-+-—-+-—+-—+--—+—-+—-+—^ V, '
0.5
1 1.5
Time(hr)
Source: Burbacher, et al. Q999b).
Note: Lines are model simulations. Indicated concentrations are target concentrations; measured concentrations
differed slightly (see Table 3-9).
Figure B-16 Chamber concentration profiles for monkey methanol exposures.
8 The model was specifically fit to the 2nd trimester monkey data, assuming that the
9 parameters were the same for all the exposure groups and concentrations. While the data show
10 little difference between the NP and two pregnancy groups, the 2nd trimester group was
11 presumed to be most representative of the average internal dosimetry over the entire pregnancy.
12 Further, the results of Mooney and Miller (2001) show that developmental effects on the monkey
13 brain stem following ethanol exposure are essentially identical for monkeys exposed only during
14 early pregnancy versus full-term, indicating that early pregnancy is a primary window of
15 vulnerability.
16 Model simulation results are the lines shown in Figures B-15 and B-16. The model
17 provides a good fit to the monkey blood and chamber air concentration data. The chamber
18 volume was treated as a fitted parameter, decreasing the "accessible volume" of 1,380 L,
19 provided by Burbacher, et al. (1999b) to 1,220 L, which calibrated the mixing time in the
20 chamber to match the chamber concentration data (Figure B-16). An adjustment of the
May 2013
B-37
Draft - Do Not Cite or Quote
-------
1 "accessible volume" also accounts for any volume filled by the monkey and other chamber
2 equipment. A detailed description of the chamber set-up is found in Burbacher, et al. (1999b).
3 The model does an adequate job of fitting the data for all exposure groups without group-specific
4 parameters. In particular, the data for all exposure levels can be adequately fit using a single
5 value for the volume of distribution (Vmk) as well as each of the metabolic parameters. While one
6 may be able to show statistically distinct parameters for different groups or exposure levels (by
7 fitting the model separately to each), as was done by Burbacher, et al. (1999b), it is unlikely that
8 such differences are biologically significant, given the fairly large number of data points and the
9 large variability evident in the blood concentration data. Thus, the single set of parameters listed
10 with the parameter descriptions above will be used to estimate internal blood concentrations
11 (Cmax above background) for the dose-response analysis described in Appendix D. The chamber
12 concentrations for "pregnancy" exposures recorded by Burbacher, et al. (1999b: Table 2) and
13 average body weights for each exposure group at the 2nd trimester time point were used along
14 with the model to calculate Cmax values above background (Table B-7).
Table B-7 Monkey group exposure characteristics
Exposure concentration (ppm)a Group average BW (kg)b Cmax above background (mg/L)c
0 3.93 0
206 3.46 2.87
610 4.08 10.38
1,822 3_83 38.51
Trom Burbacher, et al. (l_999b) and Burbacher, et al. [(l_999a), Table 2, "pregnancy" exposure.]
bFrom Burbacher, original data (personal communication).
°The two-compartment PK model described above was used to estimate Cmax above background [i.e., max (C^)].
B.3.2. Conclusions and Discussion
15 Rat and human methanol PBPK models have been developed and calibrated to data in the
16 open literature. EPA developed its own model because none of the existing models satisfactorily
17 fulfilled all of the criteria specified in Section 3.4.1.2. Further, none of the existing models had
18 been calibrated or tested against the larger collection of data considered for each species here. As
19 a result, while each model may fit the subset of the data to which it had been calibrated better
20 than the final model described here, without adjustment of parameters from those published,
21 each model either had features which made it incompatible with risk extrapolation (e.g.,
22 parameters which vary with dose in an unpredictable way) or had an inadequate fit to other data
23 considered critical for establishing overall model soundness. The EPA model simplifies the
May 2013 B-38 Draft - Do Not Cite or Quote
-------
1 structure used by Ward et al. (1997) in some aspects while adding specific refinements (e.g., a
2 standard lung compartment and a two-compartment GI tract).
3 Although the developmental endpoints of concern are effects, which result from in utero
4 and (to a lesser extent) lactational exposure, it is not necessary for a methanol PBPK model to
5 specifically describe pregnancy (i.e., specify a fetal/gestational/conceptus compartment) and
6 lactation in order for it to provide better cross-species extrapolation of risk than default methods.
7 Representation of the unique physiology of pregnancy and the fetus/conceptus would be
8 necessary if methanol pharmacokinetics differed significantly during pregnancy or if the
9 observed partitioning of methanol into the fetus/conceptus versus the mother showed a
10 concentration ratio significantly greater than or less than 1. Further details on the reasoning for
11 not including a pregnancy description are given in Section 3.4.1.2.
12 While lactational exposure is less direct than fetal exposure and blood or target-tissue
13 levels in the breast-feeding infant or rat pup are likely to differ more from maternal levels, the
14 health-effects data indicate that most of the effects of concern are due to fetal exposure, with
15 only a small influence due to postnatal exposures. Separating out the contribution of postnatal
16 exposure from prenatal exposure to a given endpoint in a way that would allow the risk to be
17 estimated from estimates of both exposure levels would be extremely difficult, even if one had a
18 lactation/child PBPK model that allowed for prediction of blood (or target-tissue) levels in the
19 offspring. Target tissue concentrations in the offspring would still be expected to be closely
20 related to maternal blood levels (which depend on ambient exposure and determine the amount
21 delivered through breast milk), with the relationship between maternal levels and those in the
22 offspring being similar across species.
23 Therefore, the development of a lactation/child PBPK model appears not to be supported,
24 given the minimal change that is likely to result in risk extrapolations and use of (NP) maternal
25 blood levels as a measure of risk in the offspring is still considered preferable over use of default
26 extrapolation methods. In particular, the existing human data allow for accurate predictions of
27 maternal blood levels, which depend strongly on the rate of maternal methanol clearance. Failing
28 to use the existing data (via PBPK modeling) for human methanol clearance (versus that in other
29 species) would be to ignore this very important determinant of exposure to breast-fed infants.
30 And since bottle-fed infants do not receive methanol from their mothers, they are expected to
31 have lower or, at most, similar overall exposures for a given ambient concentration than the
32 breast-fed infant, so that use of maternal blood levels for risk estimation should also be
33 adequately protective for that group.
34 During model development, several inconsistencies between experimental blood
35 methanol kinetic data embedded in the Ward et al. (1995) model and the published figures first
36 reporting these data were discovered. Therefore, data were digitized from the published literature
37 when a figure was available, and the digitized data was compared to the provided data. When the
May 2013 B-39 Draft - Do Not Cite or Quote
-------
1 digitized data and the data embedded in the computational files (i.e., provided to Battelle under
2 contract from EPA) were within 3% of each other, the provided data was used; when the
3 difference was greater than 3%, the digitized data was used. Often, using the published figures as
4 a data source resulted in substantial improvements of the fit to the data in the cases where the
5 published figures were different from the embedded data.
6 The final methanol PBPK model fits inhalation-route blood kinetic data from separate
7 laboratories in rats and humans fairly well. The low-dose exposures of all routes were considered
8 the most important for model optimization, since these doses are most relevant to a health
9 assessment.
10 Figure B-17 illustrates the changes in blood methanol concentrations predicted by the
11 human PBPK model for exposures to either the RfC or RfD alone, or a combined exposure to
12 both the RfC and RfD, with oral exposure assumed divided into six daily boluses as described
13 previously (Section B.2.7) and inhalation exposure assumed to be continuous. The predictions
14 are shown for an individual starting with an average background level of 1.5 mg/L blood
15 methanol, relative to one standard deviation of the background (grey area).
Ut
Ł
C
o
ro
4-1
C
Of
o
C
o
o
TJ
O
o
2.4-1
2.1
l.S
1.5
1.2
0.9
0.6
0.3
0
RfC only
RfD only
RfC & RfD
mean background
background +/- 1 SD
12 24 36
Time (hr)
48
60
72
Note: The horizontal grey lines and band show the mean background blood concentration (1.5 mg/L) ± one standard
deviation (1 SD; 0.7 mg/L). The thin, solid, red curve shows the predicted change in blood concentration given a
continuous exposure to the RfC alone, simply rising over ~ 10 hr to a new steady state at 1.91 mg/L. The thin,
dashed, blue curve shows the predicted change given ongoing exposure to the RfD, with ingestion divided among
six daily boluses (see Section B.2.7 for details), with a resulting daily pattern which has a peak concentration of 1.94
mg/L (differs slightly from the RfC due to round-off) and average level of 1.68 mg/L. The thick, solid, green curve
(upper most) shows the predicted change due to simultaneous exposure to both the RfD (six daily boluses) and RfC
(continuous), with a peak predicted concentration of 2.36 mg/L and average concentration of 2.09 mg/L.
Figure B-17 PBPK model predictions of changes in blood methanol levels in humans for
exposures at the RfC and RfD.
May 2013
B-40
Draft - Do Not Cite or Quote
-------
APPENDIX C. HUMAN CASE STUDIES
1 An extensive library of case reports has documented the consequences of acute
2 accidental/intentional methanol poisoning. Nearly all have involved ingestion, but a few have
3 involved percutaneous and/or inhalation exposure. The associated effects were discussed in
4 Section 4.1.1. As many of the case reports demonstrate, the association of Parkinson-like
5 symptoms with methanol poisoning is related to the observation that lesions in the putamen are a
6 common feature both in Parkinson's disease and methanol overexposure. A brief discussion of
7 the terms cited in case report literature follows.
8 Basal ganglia, a group of interconnected subcortical nuclei in each cerebral hemisphere,
9 refers to various structures in the grey matter of the brain that are intimately involved, for
10 example, in coordinating motor function, maintaining ocular and respiratory function, and
11 consciousness. The connectivity within the basal ganglia involves both excitatory and inhibitory
12 neurotransmitters such as dopamine (associated with Parkinson's disease when production is
13 deficient).
14 The structures comprising the basal ganglia include but are not limited to: the putamen
15 and the globus pallidus (together termed the lentiform nuclei), the pontine tegmentum, and the
16 caudate nuclei. Dystonia or involuntary muscle contraction can result from lesions in the
17 putamina; if there are concomitant lesions in the globus pallidus, Parkinsonism can result (Bhatia
18 and Marsden, 1994). Bhatia and Marsden (1994) have discussed the various behavioral and
19 motor consequences of focal lesions of the basal ganglia from 240 case-study reports. Lesions in
20 the subcortical white matter adjacent to the basal ganglia often occur as well (Airas et al., 2008;
21 Rubinstein et al., 1995; Bhatia and Marsden, 1994). In the case reports of Patankar et al. (1999),
22 it was noted that the severity and extent of necrosis in the lentiform nuclei do not necessarily
23 correlate with clinical outcome.
24 In one of the earliest reviews of methanol overexposure, Bennett et al. (1953) described a
25 mass accidental poisoning when 323 persons, ranging in age from 10 to 78 years, in Atlanta,
26 Georgia, consumed "whisky" adulterated with as much as 35-40% methanol. In all, 41 people
27 died. Of the 323 individuals, 115 were determined to be acidotic with symptoms (visual
28 impairment, headache [affecting ~62%], dizziness [affecting ~30%], nausea, abdominal pain and
29 others) beginning around 24 hours post exposure. Visual impairment was mostly characterized
30 by blurred or indistinct vision; some who were not acidotic experienced transient visual
31 disturbances. The cardiovascular parameters were unremarkable. The importance of acidosis to
32 outcome is shown in Table C-l. Among the key pathological features were cerebral edema, lung
33 congestion, gastritis, pancreatic necrosis, fatty liver, epicardial hemorrhages, and congestion of
34 abdominal viscera.
May 2013 C-1 Draft - Do Not Cite or Quote
-------
Table C-l Mortality rate for subjects exposed to methanol-tainted whisky in relation to
their level of acidosis
Subjects" Number Percent deaths
All patients 323 6.2
Acidotic (CO2 <20 mEq) 115 19
Acidotic (CO2 <10 mEq) 30 50
aThese data do not include those who died outside the hospital or who were moribund on arrival.
Source: Bennett et al. (1953).
1 Riegel and Wolf (1966), in a case report involving a 60-year-old woman who ingested
2 methanol, noted that nausea and dizziness occurred within 30 minutes of ingestion. She
3 subsequently passed out and remained unconscious for 3 days. Upon awakening she had
4 paralysis of the vocal cords and was clinically blind in one eye after 4 months. Some aspects of
5 Parkinson-like symptoms were evident. There was a pronounced hypokinesia with a mask-like
6 face resembling a severe state of Parkinson's disease. The patient had difficulty walking and
7 could only make right turns with difficulty. There was no memory loss.
8 Treatment of a 13-year-old girl who ingested an unspecified amount of a windshield-
9 washer solution containing 60% methanol was described by Guggenheim et al. (1971). She
10 displayed profound acidosis; her vital signs, once she was treated for acidosis, were normal by
11 36 hours after hospital admission. During the ensuing 6 months after discharge from the hospital,
12 visual acuity (20/400, both eyes) worsened, and she experienced muscle tremors, arm pain, and
13 difficulty in walking. A regimen of levodopa treatment greatly improved her ability to function
14 normally.
15 Ley and Gali (1983) also noted symptoms that are Parkinson like following methanol
16 intoxication. In this case report respiratory support was needed; the woman was in a coma. Once
17 stabilized, she exhibited symptoms similar to those noted in other case study reports, such as
18 blurred vision, movement difficulty, and tremors. Computerized Axial Tomography scan findings
19 highlighted the central nervous system (CNS) as an important site for methanol poisoning.
20 Rubinstein et al. (1995) presented evidence that a methanol blood level of 360 mg/L is
21 associated with a suite of CNS and ocular deficits that led to a 36-year-old man (who
22 subsequently died) becoming comatose. CT scans at 1-2 days following ingestion were normal.
23 However, MRI scans at day 4 revealed lesions in the putamen and peripheral white matter of the
24 cerebral and cerebellar hemispheres. Bilateral cerebellar cortical lesions had been reported in an
25 earlier case of methanol poisoning by Chen et al. (1991).
26 Finkelstein and Vardi (2002) reported that long-term inhalation exposure of a woman
27 scientist to methanol without acute intoxication resulted in a suite of delayed neurotoxic
28 symptoms (e.g., hand tremor, dystonia, bradykinesia, and other decrements in body movement).
May 2013 C-2 Draft - Do Not Cite or Quote
-------
1 Despite treatment with levodopa, an increase in the frequency and severity of effects occurred.
2 Exposure to bromine fumes was concomitant with exposure to methanol.
3 Hantson et al. (1997a) found, in four cases, that MRI and brain CT scans were important
4 tools in revealing specific brain lesions (e.g., in the putamina and white matter). The first subject
5 was a 57-year-old woman who complained of blurred vision, diplopia, and weakness 24 hours
6 after ingesting 250 mL of a methanolic antifreeze solution. Upon hospital admission she was
7 comatose and in severe metabolic acidosis. An MRI scan at 9 days indicated abnormal
8 hyperintense foci in the putamina (decreased in size by day 23) and subtle lesions (no change by
9 day 23) in the white matter. Upon her discharge, bilateral deficits in visual acuity and color
10 discrimination persisted.
11 Similar deficits (metabolic acidosis, visual acuity, and color discrimination) were seen in
12 a man who ingested 300 mL of 75% methanol solution. His blood methanol level
13 wasl,630 mg/L. An MRI administered 24 hours after hospital admission revealed abnormal
14 hyperintense foci in the putamina, with less intense lesions in the white matter. Like the first
15 subject, a subsequent MRI indicated the foci decreased in size over time, but visual impairments
16 persisted.
17 The third individual, a male, ingested an unspecified amount of a methanolic solution.
18 His blood methanol level was 12,900 mg/L, and he was in a coma upon hospital admission. An
19 MRI revealed lesions in the putamina and occipital subcortical white matter. A follow-up CT
20 scan was performed after 1 year and showed regression of the putaminal lesions but no change in
21 the occipital lesions. Upon his discharge, severe visual impairment remained but no
22 extrapyramidal signs were observed.
23 The last case was a man who became comatose 12 hours after ingesting 100 mL
24 methanol. His blood methanol level at that time was 600 mg/L. An MRI revealed lesions in the
25 putamina; at 3 weeks these lesions were observed to have decreased in size. Upon his discharge,
26 the neurological signs had improved but optic neuropathy (in visual evoked potential) was
27 observed.
28 In a separate publication, Hantson et al. (1997b) reported a case of a 26-year-old woman
29 who had ingested 250-500 mL methanol during the 38th week of pregnancy. Her initial blood
30 methanol level was 2,300 mg/L (formate was 336 mg/L), yet only a mild metabolic acidosis was
31 indicated. No distress to the fetus was observed upon gynecologic examination. Six days after
32 therapy was initiated (methanol was not present in blood), she gave birth. No further
33 complications with either the mother or newborn were noted.
34 There have been several case reports involving infant or toddler exposures to methanol
35 (De Brabander et al.. 2005: WuetaL 1995: Brent etal.. 1991: Kahn and Blum. 1979). The
36 report by Wu et al. (1995) involved a 5-week-old infant with moderate metabolic acidosis and a
37 serum methanol level of 11,480 mg/L, a level that is ordinarily fatal. However, this infant
May 2013 C-3 Draft - Do Not Cite or Quote
-------
1 exhibited no toxic signs and survived without any apparent permanent problems. De Brabander
2 et al. (2005) reported the case of a 3-year-old boy who ingested an unknown amount of pure
3 methanol; at 3 hours after ingestion, the blood methanol level was almost 300 mg/L. Ethanol
4 infusion as a therapeutic measure was not well tolerated; at 8 hours after ingestion, fomepizole
5 (4-methylpyrazole) was administered to inhibit the metabolism of methanol by ADH1, and blood
6 methanol levels stabilized below 200 mg/L, a level above which is considered to be toxic by the
7 American Academy of Clinical Toxicology (Barceloux et al., 2002). Neither metabolic acidosis
8 nor visual impairment was observed in this individual. Hantson et al. (1997a), in their review,
9 touted the efficacy of fomepizole over ethanol in the treatment of methanol poisoning
10 Bilateral putaminal lesions, suggestive of nonhemorrhagic necrosis in the brain of a man
11 who accidentally ingested methanol, were reported by Arora et al. (2005). Approximately
12 10 hours after MRI examination, he developed blurred vision and motor dysfunction. After
13 5 months, visual deficits persisted along with extrapyramidal symptoms. Persistent visual
14 dysfunction was also reported in another methanol poisoning case (Arora et al., 2007): the vision
15 problems developed -46 hours subsequent to the incident.
16 Vara-Castrodeza et al. (2007) applied diffusion-weighted MRI on a methanol-induced
17 comatose woman. Diffusion-weighted MRI provides an image contrast distinct from standard
18 imaging in that contrast is dependent on the molecular motion of water (Schaefer et al., 2000).
19 The neuroradiological findings were suggestive of bilateral putaminal hemorrhagic necrosis,
20 cerebral and intraventricular hemorrhage, diffuse cerebral edema, and cerebellar necrosis.
21 Diffusion-weighted MRI allows for differentiation of restricted diffusion which is indicative of
22 nonviable tissue. In this case, treatment for acidosis (blood methanol levels had risen to
23 1,000 mg/L) was unsuccessful and the patient died.
24 Emergency treatment was unable to save the life of a 38-year-old man who presented
25 with abdominal pain and convulsions after methanol intoxication (Henderson and Brubacher,
26 2002). A review of a head CT scan performed before the individual went into respiratory arrest
27 revealed bilateral globus pallidus ischemia.
28 Discrete lesions of the putamen, cerebral white matter, and corpus callosum were
29 observed upon MRI (8 days post ingestion) in a man exposed to methanol (blood level
30 370 mg/L) complaining of vision loss (Keles et al., 2007). Standard treatments corrected the
31 acidosis (pH 6.8), and at 1-month follow-up, his cognitive function improved but blindness and
32 bilateral optic atrophy were described as permanent. The follow-up MRI showed persistent
33 putaminal lesions with cortical involvement.
34 Fontenot and Pelak (2002) described a case of a woman who presented with persistent
35 blurred vision and a worsening mental status 36 hours after ingestion of an unspecified amount
36 of methanol. The initial CT scan revealed mild cerebral edema. The blood methanol level at this
37 time was 860 mg/L. A repeat CT scan 48 hours after presentation showed hypodensities in the
May 2013 C-4 Draft - Do Not Cite or Quote
-------
1 putamen and peripheral white matter. One month after discharge, cognitive function improved,
2 and the patient experienced only a mild lower-extremity tremor.
3 Putaminal necrosis and edema of the deep white matter (the corpus callosum was not
4 affected) was found upon MRI examination of a 50-year-old woman who apparently ingested an
5 unknown amount of what was believed to be pure laboratory methanol (Kuteifan et al., 1998).
6 Her blood methanol level was 1,272 mg/L upon hospital admission and dropped to 1,020 mg/L at
7 10 hours and to 710 mg/L at 34 hours. The woman, a chronic alcoholic, was in a vegetative state
8 when found and did not improve over the course of a year.
9 MRI and CT scans performed on a 51-year-old man with generalized seizures who had a
10 blood methanol level of 3,044 mg/L revealed bilateral hemorrhagic necrosis of the putamen and
11 caudate nuclei (Gaul etal., 1995). In addition, there was extensive subcortical necrosis and
12 bilateral necrosis of the pontine tegmentum and optic nerve. The patient died several hours after
13 the scans were performed.
14 The relation of methanol overexposure to brain hemorrhage was a focus of the report by
15 Phang et al. (1988), which followed the treatment of 7 individuals, 5 of whom died within
16 72 hours after hospital admission. In two of the deceased individuals, CT scans and autopsy
17 revealed putaminal hemorrhagic necrosis. The investigators postulated that the association of
18 methanol with hemorrhagic necrosis may be complicated by the use of heparin during
19 hemodialysis treatment for acidosis
20 Treatment of two men who had drunk a solution containing 58% methanol and presented
21 with impaired vision, coma, and seizures was discussed in a case report by Bessell-Browne and
22 Bynevelt (2007). A CT scan, on one individual, revealed bilateral putaminal and cerebral lesions.
23 Blood methanol levels were 21 mg/L. This individual, despite standard treatments, never
24 regained consciousness. The second individual, upon MRI, showed scattered hemorrhage at the
25 grey-white interface of the cerebral hemispheres.
26 There have been case reports that involved percutaneous and inhalation exposure (Adanir
27 et al., 2005; Downieetal., 1992). Use of a methanol-containing emollient by a woman with
28 chronic pain led to vision loss, hyperventilation and finally, coma (Adanir et al., 2005).
29 Subsequent to standard treatment followed by hospital discharge, some visual impairment and
30 CNS decrements remained. The methanol blood threshold for ocular damage and acidosis
31 appeared to be -20 mg/L. Dutkiewicz et al. (1980) have determined the skin absorption rate to be
32 0.192 mg/cm2/minute. In the case report of Aufderheide et al. (1993), two firefighters were
33 transiently exposed to methanol by inhalation and the percutaneous route. Both only complained
34 of a mild headache and had blood methanol levels of 230 and 160 mg/L, respectively.
35 Bebarta et al. (2006) conducted a prospective observational study of seven men who had
36 purposefully inhaled a methanol-containing product. Four had a blood methanol level upon
37 hospital presentation of >240 mg/L; the mean formic acid level was .71 mg/L. One individual
May 2013 C-5 Draft - Do Not Cite or Quote
-------
1 had a blood methanol level of 860 mg/L and a blood formic acid level of 250 mg/L upon hospital
2 admission. This latter individual was treated with fomepizole. No patient had an abnormal
3 ophthalmologic examination. All seven stabilized quickly and acidosis was normalized in 4
4 hours.
5 Numerous other case reports documenting putaminal necrosis/hemorrhage and/or
6 blindness have been reported (Blanco et al., 2006; Feany et al., 2001; Hsu et al., 1997; Pelletier
7 etal.. 1992: Chenetal.. 1991).
8 Hovda et al. (2005) presented a combined prospective and retrospective case series study
9 of 51 individuals in Norway (39 males and 12 females, many of whom were alcoholics) who
10 were hospitalized after consuming tainted spirits containing 20% methanol and 80% ethanol. In
11 general, serum methanol concentrations were highest among those most severely affected. The
12 poor outcome was closely correlated with the degree of metabolic acidosis. It was noted by the
13 investigators that the concomitant consumption of ethanol prevented more serious sequelae in
14 2/5 individuals who presented with detectable ethanol levels and were not acidotic despite 2
15 having the highest blood methanol levels. However, others with detectable levels of ethanol
16 along with severe metabolic acidosis (two of whom died) presumably had subtherapeutic levels
17 of ethanol in their system.
18 In a later report, Hovda et al. (2007) focused on formate kinetics in a 63-year-old male
19 who died 6 days after being admitted to the hospital with headache, vomiting, reduced vision,
20 and dizziness. The investigators speculated that the prolonged metabolic acidosis observed (T1/2
21 for formic acid was 77 hours before dialysis, compared to a typical normal range of
22 2.5-12 hours) may have been related to retarded formate elimination.
23 Hovda and colleagues (Hunderi et al., 2006) found a strong correlation between blood
24 methanol concentration and the osmolal gap (R2 = 0.92) among 17 patients undergoing dialysis
25 after consuming methanol-contaminated spirits. They concluded that the osmolal gap could be
26 taken as a priori indication of methanol poisoning and be used to guide initiation and duration of
27 dialysis. As they indicated, many hours of dialysis could be safely dispensed with. The osmolal
28 gap pertains to the effect that methanol (and other alcohols) has on the depression of the freezing
29 point of blood in the presence of normal solutes. Braden et al. (1993) demonstrated in case
30 studies that the disappearance of the osmolal gap correlates with the correction of acidosis; they
31 cautioned that methanol and ethanol should not be assumed to be the main factors in causing
32 osmolal gap as glycerol and acetone and its metabolites can as well. A more detailed discussion
33 of the anion and osmolal gap has been provided by Henderson and Brubacher (2002).
34 Hassanian-Moghaddam et al. (2007) compiled data on the prognostic factor relating to
35 outcome in methanol-poisoning cases in Iran. They examined 25 patients, 12 of whom died; 3 of
36 the survivors were rendered blind. There was a significant difference in mean pH of the first
37 arterial blood gas measurements of those who subsequently died compared with survivors. It was
May 2013 C-6 Draft - Do Not Cite or Quote
-------
1 concluded that poor prognosis was associated with pH <7, coma upon admission, and >24-hours
2 delay from intake to admission.
3 The use of blood methanol levels as predictors of outcome is generally not recommended
4 (Barceloux et al., 2002). These investigators cited differences in sampling time, ingestion of
5 ethanol, and levels of toxic (e.g., formic acid) metabolites among the complicating factors. As an
6 illustration, the case report by Prabhakaran et al. (1993) cites two women who ingested a
7 methanol solution (photocopying diluent) at about the same time, were admitted to the hospital
8 about the same time (25-26 hours after ingestion) and had identical plasma methanol
9 concentrations (830 mg/L) upon admission, but different outcomes. Patient #1 was in metabolic
10 acidosis and had an unstable conscious state even after treatment. Upon discharge at day 6, there
11 were no apparent sequelae. Patient #2 had severe metabolic acidosis, fixed and dilated pupils,
12 and no brain stem reflexes. This patient died at day 3 even though therapeutic measures had been
13 administered.
14 In a discussion of 3 fatal methanol-overexposure cases, Andresen et al. (2008) found
15 antemortem blood methanol levels of 5,400 and 7,400 mg/L in two individuals. At autopsy brain
16 stem blood levels were 7,380 and 10,080 mg/L, respectively. These brain levels were much
17 higher than blood levels postmortem. Autopsy revealed brain and pulmonary edema in all three
18 individuals; in the two who had the longer survival times, there was hemorrhagic necrosis of the
19 putamen and hemorrhages of the tissue surrounding the optic nerve. In their study of 26 chronic
20 users of methylated spirits, Meyer et al. (2000) found that the best predictor of death or a poor
21 outcome in chronic abusers was a pH <7.0; there was no correlation between blood methanol
22 levels and outcome. Mahieu et al. (1989) considered a latency period before treatment exceeding
23 10 hours and a blood formate level >500 mg/L as predictive of possible permanent sequelae. Liu
24 et al. (1998) in their examination of medical records of 50 patients treated for methanol
25 poisoning over a 10-year period found that: (1) deceased patients had a higher mean blood
26 methanol level than survivors; and (2) initial arterial pH levels <7.0 (i.e., severe metabolic
27 acidosis). Coma or seizure was also associated with higher mortality upon hospital admission.
28 Numerous cases of methanol poisoning have been documented in a variety of countries.
29 In Tunisia, 16 cases of methanol poisoning were discussed by Brahmi et al. (2007). Irreversible
30 blindness occurred in two individuals, with others reporting CNS symptoms, GI effects, visual
31 disturbances, and acidosis. Putaminal necrosis was also described in case reports from Iran
32 (Sefidbakht et al.. 2007). Of 634 forensic autopsies carried out in Turkey during 1992-2003,
33 18 deaths appeared to be related to methanol poisoning (Azmak, 2006). Brain edema and focal
34 necrosis of the optic nerve were among various sequelae noted. Dethlefs and colleagues (Naraqi
35 et al., 1979; Dethlefs and Naraqi, 1978) described permanent ocular damage in 8/24 males who
36 ingested methanol in Papua New Guinea.
May 2013 C-7 Draft - Do Not Cite or Quote
-------
1 In summary, most cases of accidental/intentional methanol poisoning reveal a common
2 set of symptoms, many of which are likely to be presented upon hospital admission. See Section
3 4.1.1 for a list of common symptoms.
May 2013 C-8 Draft - Do Not Cite or Quote
-------
APPENDIX D. RFC DERIVATION OPTIONS
D.1. Benchmark Dose Modeling Summary
1 This appendix provides technical detail on dose-response evaluation and determination of
2 points of departure (POD) for relevant lexicological endpoints. The endpoints were modeled
3 using the U.S. EPA's Benchmark Dose Software (BMDS, version 2.2). Sections D. 1.1 and D. 1.2
4 describe the common practices used in evaluating the model fit and selecting the appropriate
5 model for determining the POD, as outlined in the Benchmark Dose Technical Guidance
6 Document (U.S. EPA. 2012a).
D.1.1. Evaluation of Model Fit
7 For the nested dichotomous endpoint (cervical rib), BMDS nested dichotomous models
8 were fitted to the data using the maximum likelihood method. Each model was tested for
9 goodness-of-fit using a chi-square goodness-of-fit test (% p-va\ue < 0.10 indicates lack of fit).
10 Other factors were also used to assess model fit, such as scaled residuals, visual fit, and adequacy
11 of fit in the low-dose region and in the vicinity of the BMR.
12 For each continuous endpoint (brain weight and VDR), BMDS continuous models1 were
13 fitted to the data using the maximum likelihood method. Model fit was assessed by a series of
14 tests as follows. For each model, first the homogeneity of the variances was tested using a
15 likelihood ratio test (BMDS Test 2). If Test 2 was not rejected (%2/'-value > 0.10), the model was
16 fitted to the data assuming constant variance. If Test 2 was rejected (% p-va\ue < 0.10), the
17 variance was modeled as a power function of the mean, and the variance model was tested for
18 adequacy of fit using a likelihood ratio test (BMDS Test 3). For fitting models using either
19 constant variance or modeled variance, models for the mean response were tested for adequacy
20 of fit using a likelihood ratio test (BMDS Test 4, with % p-va\ue < 0.10 indicating inadequate
21 fit). Other factors were also used to assess the model fit, such as scaled residuals, visual fit, and
22 adequacy of fit in the low-dose region and in the vicinity of the BMR.
1 Unless otherwise specified, all available BMDS continuous models were fitted. The following parameter
restrictions were applied: for the polynomial models, restrict the coefficients bl and higher to be nonnegative or
nonpositive if the direction of the adverse effect is upward or downward, respectively; for the Hill, power, and
exponential models, restrict power > 1.
May 2013 D-1 Draft - Do Not Cite or Quote
-------
D.1.2. Model Selection
1 For each endpoint, the BMDL estimate (95% lower confidence limit on the BMD, as
2 estimated by the profile likelihood method) and AIC value were used to select a best-fit model
3 from among the models exhibiting adequate fit. If the BMDL estimates were "sufficiently close,"
4 that is, differed by at most 3-fold, the model selected was the one that yielded the lowest AIC
5 value. If the BMDL estimates were not sufficiently close, the lowest BMDL was selected as the
6 POD.
D.2. RfC Derivations Using the NEDO Methanol Report (NEDO, 1987)
7 In the application of the BMD approach, continuous models in EPA's BMDS, version 2.2
8 (U.S. EPA, 2011 a), were fit to data sets for decreased brain weight in male rats exposed
9 throughout gestation and the postnatal period to 6 weeks and male rats exposed during gestation
10 on days 7-17 only (NEDO, 1987). Although there remains uncertainty surrounding the
11 identification of the proximate teratogen of importance (methanol, formaldehyde, or formate),
12 the dose metrics chosen for the derivation of RfCs were based on blood methanol levels. This
13 decision was primarily based on evidence that the toxic moiety is not likely to be the formate
14 metabolite of methanol (CERHR, 2004) and evidence that levels of the formaldehyde metabolite
15 following methanol maternal and/or neonate exposure would be much lower in the fetus and
16 neonate than in adults. While recent in vitro evidence indicates that formaldehyde is more
17 embryotoxic than methanol and formate, the high reactivity of formaldehyde would significantly
18 limit its transport from maternal to fetal blood, and the capacity for the metabolism of methanol
19 to formaldehyde is lower in the fetus and neonate versus adults.
D.2.1. Decreased Brain Weight in Male Rats Exposed throughout Gestation
and into the Postnatal Period
20 As discussed in Section 5.1.2.1, brain weight is susceptible to both the level and duration
21 of exposure suggesting that a dose metric that incorporates a time component would be the most
22 appropriate metric to use. For these reasons and because it is more typically used in internal-
23 dose-based assessments and better reflects total exposure within a given day, daily AUC
24 (measured for 22-hour exposure/day) was chosen as the most appropriate dose metric for
25 modeling the effects of methanol exposure on brain weights in rats exposed throughout gestation
26 and continuing into the FI generation.
27 As is discussed in Section 5.1.3.2.2, the additional routes of exposure presented to the
28 pups in this study (lactation and inhalation) present uncertainties in that the average blood levels
May 2013 D-2 Draft - Do Not Cite or Quote
-------
1 in pups is likely to be greater than those of their dams. The assumption made in this assessment
2 is that, if such differences exist between human mothers and their offspring, they are not
3 significantly greater than that which has been postulated for rats. Assuming this is true, the
4 PBPK model-estimated adult blood methanol level is considered to be an appropriate dose metric
5 for the purpose of this analysis and the estimation of a human equivalent concentration (HEC).
6 The first step in the current analysis is to convert the inhalation doses, given as ppm
7 values from the studies, to an internal dose surrogate or dose metric using the EPA PBPK model
8 (see Appendix B). Predicted AUC values for methanol in the blood of rats are summarized in
9 Table D-l. The AUC values above background (AUC - control) are then used as the dose metric
10 for the BMD analysis of response data shown in Table D-l for decreased brain weight at 6 weeks
11 in male rats following gestational and postnatal exposure.2 Decreases in brain weight at 6 weeks
12 (gestational and postnatal exposure), rather than those seen at 3 and 8 weeks, were chosen as the
13 basis for the RfC derivation because they resulted in lower estimated BMDs and BMDLs. The
14 details of this analysis are reported below. More details concerning the PBPK modeling were
15 presented in Appendix B.
Table D-l EPA PBPK model estimates of methanol blood levels (AUC)a in rat dams
following inhalation exposures and reported brain weights of 6 week old
male pups
Exposure
level (ppm)
0
500
1,000
2,000
Blood methanol AUC
(mg-hr/L)a in rats
72
619
2,380
17,600
Blood methanol AUC -
control (mg-hr/L)a in rats
0
547
2,308
17,528
Mean male rat (Fi generation)
brain weight at 6 weeksb
1.78 ±0.07
1.74 ±0.09
1.69 ± 0.06 c
1.52±0.07d
N
12
12
11
14
aAUC values were obtained by simulating 22 hr/day exposures for 5 days and calculated for the last 24 hours of that period, with
a simulated background blood level of 3 mg/L. (See Appendix B for further details.)
bExposed throughout gestation and F! generation. Values are means ± S.D.
°p < 0.01, dp < 0.001, as calculated by the authors.
Source: NEDO (1987)
16 The EPA BMD technical guidance (U.S. EPA, 2012a) suggests that in the absence of
17 knowledge as to what level of response to consider adverse, a change in the mean equal to
18 1 control S.D. from the control mean can be used as a BMR for continuous endpoints. However,
19 it has been suggested that other BMRs, such as 5% change relative to estimated control mean,
20 are also appropriate when performing BMD analyses on fetal weight change as a developmental
21 endpoint (Kavlock et al., 1995). Therefore, in this assessment, both a 1 control mean S.D. change
2All BMD assessments in this review were performed using BMDS version 2.2 (U.S. EPA. 201^)
May 2013 D-3 Draft - Do Not Cite or Quote
-------
1 and a 5% change relative to estimated control mean were considered. All models were fit using
2 restrictions and option settings suggested in the EPABMD Technical Guidance Document (U.S.
3 EPA. 2012a).
D.2.1.1. BMD Approach with a BMR of 1 Control Mean S.D. - Decreased
Brain Weight in Male Rats Exposed throughout Gestation and into the
Postnatal Period (NEDO, 1987)
4 A summary of the results most relevant to the development of a POD using the BMD
5 approach (BMD, BMDL, and model fit statistics) for decreased brain weight at 6 weeks in male
6 rats exposed to methanol throughout gestation and continuing into the FI generation, with a
7 BMR of 1 control mean S.D. (NEDO. 1987). is provided in Table D-2. Model fit and was
8 determined by statistics (AIC and ^ residuals of individual dose groups) and visual inspection,
9 as recommended by EPA (U.S. EPA, 2012a). There is a 5.1-fold range of BMDL estimates from
10 adequately fitting models, indicating considerable model dependence. In addition, the fit of the
11 Hill and more complex Exponential models are better than the other models in the dose region of
12 interest as indicated by a lower scaled residual at the dose group closest to the BMD (0.18 and
13 0.16 versus -1.4) and by visual inspection. In accordance with EPA BMD Technical Guidance
14 (U.S. EPA. 2012a). the BMDL from the Hill model (bolded), is selected as the most appropriate
15 basis for an RfC derivation because it results in the lowest BMDL from among a broad range of
16 BMDLs and provides a superior fit in the low dose region nearest the BMD. The detailed results
17 of the Hill model run, including text and a plot (Figure D-l) are shown after Table D-2. The
18 BMDLiso was determined to be 858 mg-hr/L using the 95% lower confidence limit of the dose-
19 response curve expressed in terms of the AUC above background for methanol in blood.
May 2013 D-4 Draft - Do Not Cite or Quote
-------
Table D-2 Comparison of BMDisD results for decreased brain weight in male rats at
6 weeks of age using modeled AUC above background of methanol as a dose
metric
Model
Linear
2nd degree Polynomial
3rd degree Polynomial
Power
Hillb
Exponential 2
Exponential 3
Exponential 4
Exponential 5
BMD1SD (AUC,
mg-hr/L)a
5469.53
5469.53
5469.53
5469.53
1730.35
5159.24
5159.24
1802.01
1802.01
BMDL1SD
(AUC,
mg-hr/L)a
4410.68
4410.68
4410.68
4410.68
858.04
4118.16
4118.16
997.71
997.71
/7-value AICb
0.1385 -201.13
0.1385 -201.13
0.1385 -201.13
0.1385 -201.13
0.5920 -202.79
0.1573 -201.38
0.1573 -201.38
0.5513 -202.72
0.5513 -202.72
aThe BMDL is the 95% lower confidence limit on the AUC estimated to decrease brain weight by 1
BMDS 2.2 (U.S. EPA, 201 la) and model options and restrictions suggested bv EPA BMD technical
Scaled residual0
-1.39
-1.39
-1.39
-1.39
0.179
-1.336
-1.336
0.163
0.163
control mean S.D. using
guidance (US. EPA, 2012a).
bAIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
°%2d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals that exceed 2.0 in
absolute value should cause one to question model fit in this region.
Source: NEDO (1987).
May 2013
D-5
Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 Hill Model. (Version: 2.16; Date: 04/06/2011)
3 Input Data File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
4 ConstantVariance-BMRlStd-Restrict.(d)
5 Gnuplot Plotting File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
6 ConstantVariance-BMRlStd-Restrict.pit
7 Tue Mar 27 08:42:04 2012
8 ====================================================================
9
10 BMDS Model Run
11 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12
13 The form of the response function is:
14
15 Y[dose] = intercept + v*dose/xn/(k^n + dose^n)
16
17
18 Dependent variable = Mean
19 Independent variable = Dose
20 rho is set to 0
21 Power parameter restricted to be greater than 1
22 A constant variance model is fit
23
24 Total number of dose groups = 4
25 Total number of records with missing values = 0
26 Maximum number of iterations = 250
27 Relative Function Convergence has been set to: le-008
28 Parameter Convergence has been set to: le-008
29
30
31
32 Default Initial Parameter Values
33 alpha = 0.00539333
34 rho = 0 Specified
35 intercept =1.78
36 v = -0.26
37 n = 0.698151
38 k = 5889.18
39
40
41 Asymptotic Correlation Matrix of Parameter Estimates
42
43 (*** The model parameter(s) -rho -n have been estimated at a boundary
44 point, or have been specified by the user, and do not appear in the
45 correlation matrix )
46
47
48 alpha intercept v k
49
50 alpha 1 1.7e-008 2.5e-008 -4e-008
51
52 intercept 1.7e-008 1 0.24 -0.62
53
54 v 2.5e-008 0.24 1 -0.85
55
56 k -4e-008 -0.62 -0.85 1
57
58
59
60 Parameter Estimates
61
62 95.0% Wald Confidence Interval
63 Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
64 alpha 0.00498218 0.00100655 0.00300938 0.00695499
65 intercept 1.77449 0.0177456 1.73971 1.80927
May 2013 D-6 Draft - Do Not Cite or Quote
-------
1 v -0.3555 0.0666435 -0.486119 -0.224881
2 n 1 NA
3 k 6984.58 4505.13 -1845.31 15814.5
4
5 NA - Indicates that this parameter has hit a bound
6 implied by some inequality constraint and thus
7 has no standard error.
8
9
10
11 Table of Data and Estimated Values of Interest
12
13 Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Res.
14
15
16 0 12 1.78 1.77 0.07 0.0706 0.27
17 547 12 1.74 1.75 0.09 0.0706 -0.425
18 2308 11 1.69 1.69 0.06 0.0706 0.179
19 1.753e+004 14 1.52 1.52 0.07 0.0706 -0.0151
20
21
22
23 Model Descriptions for likelihoods calculated
24
25
26 Model Al: Yij = Mu(i) + e(ij)
27 Var{e(ij)} = Sigma/x2
28
29 Model A2: Yij = Mu(i) + e(ij)
30 Var{e(ij)} = Sigma(i)^2
31
32 Model A3: Yij = Mu(i) + e(ij)
33 Var{e(ij)} = Sigma/x2
34 Model A3 uses any fixed variance parameters that
35 were specified by the user
36
37 Model R: Yi = Mu + e(i)
38 Var{e(i)} = Sigma/x2
39
40
41 Likelihoods of Interest
42
43 Model Log(likelihood) # Param's AIC
44 Al 105.539862 5 -201.079724
45 A2 106.570724 8 -197.141449
46 A3 105.539862 5 -201.079724
47 fitted 105.396232 4 -202.792465
48 R 77.428662 2 -150.857324
49
50
51 Explanation of Tests
52
53 Test 1: Do responses and/or variances differ among Dose levels?
54 (A2 vs. R)
55 Test 2: Are Variances Homogeneous? (Al vs A2)
56 Test 3: Are variances adequately modeled? (A2 vs. A3)
57 Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
58 (Note: When rho=0 the results of Test 3 and Test 2 will be the same.)
59
60 Tests of Interest
61
62 Test -2*log(Likelihood Ratio) Test df p-value
63
64 Test 1 58.2841 6 <.0001
65 Test 2 2.06173 3 0.5597
May 2013 D-7 Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Test 3 2.06173 3 0.5597
Test 4 0.287259 1 0.592
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels
It seems appropriate to model the data
The p-value for Test 2 is greater than .1. A homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than .1. The modeled variance appears
to be appropriate here
The p-value for Test 4 is greater than .1. The model chosen seems
to adeguately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 1730.35
BMDL = 858.038
1.85
1.75
1.7
1.65
1.6
1.55
1.5
1.45
Hill Model with 0.95 Confidence Level
Hill
\
BMDL
BMD
0
2000 4000 6000 8000 10000 12000 14000 16000 18000
dose
10:01 11/10 2011
Source: NEDO (1987).
Figure D-l Hill model, BMR of 1 Control Mean S.D. - Decreased Brain weight in male rats
at 6 weeks age versus AUC above background, Fl Generation inhalational
study.
May 2013
D-8
Draft - Do Not Cite or Quote
-------
D.2.1.2. BMD approach with a BMR of 0.05, change relative to estimated
control mean - Decreased brain weight in male rats exposed throughout
gestation and into the postnatal period (NEDO, 1987).
1 A summary of the results most relevant to the development of a POD using the BMD
2 approach (BMD, BMDL, and model fit statistics) for decreased brain weight at 6 weeks in male
3 rats exposed to methanol throughout gestation and continuing into the FI generation, with a
4 BMR of 0.05 change relative to estimated control mean, is provided in Table D-3. Model fit was
5 determined by statistics (AIC and $ residuals of individual dose groups) and visual inspection,
6 as recommended by the EPA BMD Technical Guidance (U.S. EPA. 2012a). There is a 4.7-fold
7 range of BMDL estimates from adequately fitting models, indicating considerable model
8 dependence. In addition, the fit of the Hill and more complex Exponential models are better than
9 the other models in the dose region of interest as indicated by a lower scaled residual at the dose
10 group closest to the BMD (0.18 and 0.16 versus -1.4) and visual inspection. In accordance with
11 EPA BMD Technical Guidance (U.S. EPA. 2012a). the BMDL from the Hill model (bolded), is
12 selected as the most appropriate basis for an RfC derivation because it results in the lowest
13 BMDL from among a broad range of BMDLs and provides a superior fit in the low dose region
14 nearest the BMD. Output from the Hill model, including text and plot (Figure D-2), is shown
15 after Table D-3. The BMDLos was determined to be 1,183 mg-hr/L, using the 95% lower
16 confidence limit of the dose-response curve expressed in terms of the AUC above background for
17 methanol in blood.
May 2013 D-9 Draft - Do Not Cite or Quote
-------
Table D-3 Comparison of BMD0s results for decreased brain weight in male rats at
6 weeks of age using modeled AUC above background of methanol as a
dose metric
Model
Linearb
2nd degree Polynomial
3rd degree Polynomial
Power
Hill
Exponential 2
Exponential 3
Exponential 4
Exponential 5
BMDos (AUC,
mg-hr/L)a
6,537.04
6,537.04
6,537.04
6,537.04
2.322.94
6,212.5
6,212.5
2,367.26
2, 367.26
BMDLos (AUC,
mg-hr/L)a
5,614.56
5,614.56
5,614.56
5,614.56
1.182.99
5,270.18
5,270.18
1,334.02
1,334.02
/7-value
0.1385
0.1385
0.1385
0.1385
0.5920
0.1573
0.1573
0.5513
0.5513
AICb Scaled Residual0
-201.13
-201.13
-201.13
-201.13
-202.79
-201.38
-201.38
-202.72
-202.72
-1.39
-1.39
-1.39
-1.39
0.179
-1.34
-1.34
0.163
0.163
aThe BMDL is the 95% lower confidence limit on the AUC estimated to decrease brain weight by 5% using BMDS 2.2 (U.S.
EPA, 201 la) and model options and restrictions suggested bv EPA BMD Technical Guidance ("U.S. EPA, 2012a1.
bAIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
°%2d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals that exceed 2.0 in
absolute value should cause one to question model fit in this region.
Source: NEDO (1987)
May 2013 D-10 Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 Hill Model. (Version: 2.16; Date: 04/06/2011)
3 Input Data File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
4 ConstantVariance-BMROS-Restrict.(d)
5 Gnuplot Plotting File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
6 ConstantVariance-BMROS-Restrict.pit
7 Tue Mar 27 10:57:37 2012
8 ====================================================================
9
10 BMDS Model Run
11 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12
13 The form of the response function is:
14
15 Y[dose] = intercept + v*dose/xn/(k^n + dose^n)
16
17
18 Dependent variable = Mean
19 Independent variable = Dose
20 rho is set to 0
21 Power parameter restricted to be greater than 1
22 A constant variance model is fit
23
24 Total number of dose groups = 4
25 Total number of records with missing values = 0
26 Maximum number of iterations = 250
27 Relative Function Convergence has been set to: le-008
28 Parameter Convergence has been set to: le-008
29
30
31
32 Default Initial Parameter Values
33 alpha = 0.00539333
34 rho = 0 Specified
35 intercept =1.78
36 v = -0.26
37 n = 0.698151
38 k = 5889.18
39
40
41 Asymptotic Correlation Matrix of Parameter Estimates
42
43 (*** The model parameter(s) -rho -n have been estimated at a boundary
44 point, or have been specified by the user, and do not appear in the
45 correlation matrix )
46
47
48 alpha intercept v k
49
50 alpha 1 1.7e-008 2.5e-008 -4e-008
51
52 intercept 1.7e-008 1 0.24 -0.62
53
54 v 2.5e-008 0.24 1 -0.85
55
56 k -4e-008 -0.62 -0.85 1
57
58
59
60 Parameter Estimates
61
62 95.0% Wald Confidence Interval
63 Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
64 alpha 0.00498218 0.00100655 0.00300938 0.00695499
65 intercept 1.77449 0.0177456 1.73971 1.80927
May 2013 D-11 Draft - Do Not Cite or Quote
-------
1 v -0.3555 0.0666435 -0.486119 -0.224881
2 n 1 NA
3 k 6984.58 4505.13 -1845.31 15814.5
4
5 NA - Indicates that this parameter has hit a bound
6 implied by some inequality constraint and thus
7 has no standard error.
8
9
10
11 Table of Data and Estimated Values of Interest
12
13 Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Res.
14
15
16 0 12 1.78 1.77 0.07 0.0706 0.27
17 547 12 1.74 1.75 0.09 0.0706 -0.425
18 2308 11 1.69 1.69 0.06 0.0706 0.179
19 1.753e+004 14 1.52 1.52 0.07 0.0706 -0.0151
20
21
22
23 Model Descriptions for likelihoods calculated
24
25
26 Model Al: Yij = Mu(i) + e(ij)
27 Var{e(ij)} = Sigma/x2
28
29 Model A2: Yij = Mu(i) + e(ij)
30 Var{e(ij)} = Sigma(i)^2
31
32 Model A3: Yij = Mu(i) + e(ij)
33 Var{e(ij)} = Sigma/x2
34 Model A3 uses any fixed variance parameters that
35 were specified by the user
36
37 Model R: Yi = Mu + e(i)
38 Var{e(i)} = Sigma/x2
39
40
41 Likelihoods of Interest
42
43 Model Log(likelihood) # Param's AIC
44 Al 105.539862 5 -201.079724
45 A2 106.570724 8 -197.141449
46 A3 105.539862 5 -201.079724
47 fitted 105.396232 4 -202.792465
48 R 77.428662 2 -150.857324
49
50
51 Explanation of Tests
52
53 Test 1: Do responses and/or variances differ among Dose levels?
54 (A2 vs. R)
55 Test 2: Are Variances Homogeneous? (Al vs A2)
56 Test 3: Are variances adequately modeled? (A2 vs. A3)
57 Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
58 (Note: When rho=0 the results of Test 3 and Test 2 will be the same.)
59
60 Tests of Interest
61
62 Test -2*log(Likelihood Ratio) Test df p-value
63
64 Test 1 58.2841 6 <.0001
65 Test 2 2.06173 3 0.5597
May 2013 D-12 Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Test 3 2.06173 3 0.5597
Test 4 0.287259 1 0.592
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels
It seems appropriate to model the data
The p-value for Test 2 is greater than .1. A homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than .1. The modeled variance appears
to be appropriate here
The p-value for Test 4 is greater than .1. The model chosen seems
to adeguately describe the data
Benchmark Dose Computation
Specified effect =0.05
Risk Type = Relative risk
Confidence level = 0.95
BMD = 2322.94
BMDL = 1182.99
1.85
1.75
1.7
1.65
1.6
1.55
1.5
1.45
Hill Model with 0.95 Confidence Level
Hill
BMDL
BMD
0
2000 4000 6000 8000 10000 12000 14000 16000 18000
dose
10:03 11/10 2011
Source: NEDO (1987).
Figure D-2 Hill model, BMR of 0.05 relative risk - decreased brain weight in male rats at
6 weeks age versus AUC above background of methanol, FI generation
inhalational study.
May 2013
D-13
Draft - Do Not Cite or Quote
-------
D.2.2. Decreased Brain Weight in Male Rats Exposed During Gestation Only
(GD7-GD17)
1 As discussed in Section 5.1.2.1, Cmax, as calculated by EPA's PBPK model, was selected
2 as the dose metric for this exposure scenario. Exposures occurred only during the major period of
3 organogenesis, during which the level of exposure is believed to be more important than the
4 duration of exposure.
5 The first step in the current analysis is to convert the inhalation doses, given as ppm
6 values from the studies, to an internal dose surrogate or dose metric using the EPA PBPK model
7 (see Appendix B). Predicted Cmax values for methanol in the blood of rats, with and without
8 background methanol levels, are summarized in Table D-4.
Table D-4 EPA PBPK model estimates of methanol blood levels (Cmax) in rat pups at 8
weeks following inhalation exposures during gestation
Exposure
level (ppm)
0
200
1,000
5,000
Blood methanol Cmax
(mg /L)a in rats
3
10.41
117.6
2,989
Blood methanol Cmax -
control (mg/L)a in rats
0
7.41
114.6
2,986
Mean male rat brain weight at 8
weeksb
2.00 ± 0.047
2.01 ±0.075
1.99 ±0.072
1.81±0.161C
N
11
11
12
10
aCmax values were obtained by simulating 22 hr/day exposures with a simulated background blood level of 3 mg/L. (See
Appendix B for further details).
bExposed throughout gestation. Values are means ± S.D.
°p < 0.01, as calculated by the authors.
Source: NEDO (1987).
9 The BMD technical guidance (U.S. EPA, 2012a) suggests that in the absence of
10 knowledge as to what level of response to consider adverse, a change in the mean equal to
11 1 control S.D. from the control mean can be used as a BMR for continuous endpoints. However,
12 it has been suggested that other BMRs, such as 5% change relative to estimated control mean,
13 are also appropriate when performing BMD analyses on fetal weight change as a developmental
14 endpoint (Kavlock et al., 1995). Therefore, in this assessment, both a 1 control mean S.D. change
15 and a 5% change relative to estimated control mean were considered. All models were fit using
16 restrictions and option settings suggested in the EPA's BMD Technical Guidance Document
17 (U.S. EPA. 2012a).
D.2.2.1. BMD Approach with a BMR of 1 Control Mean S.D. (GD7-GD17)
18 A summary of the results most relevant to the development of a POD using the BMD
19 approach (BMD, BMDL, and model fit statistics) (NEDO, 1987) for decreased brain weight at 8
May 2013 D-14 Draft - Do Not Cite or Quote
-------
1 weeks in male rats exposed to methanol during gestation from days 7-17, with a BMR of 1
2 control mean S.D, is provided in Table D-5. Male brain weight responses were chosen because
3 they resulted in lower BMD and BMDL estimates than female responses (data not shown).
4 Model fit was determined by statistics (AIC and $ residuals of individual dose groups) and
5 visual inspection, as recommended by EPA (U.S. EPA, 2012a). The Polynomial and Power
6 models reduced to Linear model and returned identical modeling results. There is a greater than
7 5-fold range of BMDL estimates from adequately fitting models, indicating considerable model
8 dependence. In addition, the fit of the Hill and Exponential 4 and 5 models are better than the
9 other models in the dose region of interest as indicated by a lower scaled residual at the dose
10 group closest to the BMD (-0.09 versus —0.3) and visual inspection. In accordance with EPA
11 BMD Technical Guidance (U.S. EPA, 2012a), the BMDL from the Exponential 4 and 5 models
12 (bolded), is selected as the most appropriate basis for an RfC derivation because it results in the
13 lowest BMDL from among a broad range of BMDLs and provides a superior fit in the low dose
14 region nearest the BMD. Output from the Exponential 4 model, including text and plot
15 (Figure D-3), is shown after Table D-5. The BMDLiso was determined to be 115 mg/L, using the
16 95% lower confidence limit of the dose-response curve expressed in terms of the Cmax above
17 background for methanol in blood.
May 2013 D-15 Draft - Do Not Cite or Quote
-------
Table D-5 Comparison of BMDisD results for decreased brain weight in male rats at
8 weeks of age using modeled Cmax above background of methanol as a dose
metric
Model
BMD1SD
(Cmax,mg/L)a
BMDL1SD
(Cmax,mg/L)a
/7-value
AICb
Scaled
residual0
Linear
960.78
626.64
0.8837
-173.347015
-0.28
2nd degree Polynomial
960.78
626.64
0.8837
-173.347015
-0.28
3rd degree Polynomial
960.78
626.64
0.8837
-173.347015
-0.28
Power
960.78
626.64
0.8837
-173.347015
-0.28
Hillb
449.28
115.97
0.9272
-171.586011
0.0944
Exponential 2
925.82
589.97
0.8910
-173.3635
-0.2674
Exponential:
925.92
589.97
0.8910
-173.3635
-0.2674
Exponential 4
433.46
114.86
0.9266
-171.5859
0.09421
Exponential 5
433.46
114.86
0.9266
-171.5859
0.09421
aThe BMDL is the 95% lower confidence limit on the Cmax estimated to decrease brain weight by 1 control mean S.D. using
BMDS 2.1.1 (U.S. EPA, 2009a) and model options and restrictions suggested by EPA BMD technical guidance (U.S. EPA.
2Q12a).
bAIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
°%2d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals that exceed 2.0 in
absolute value should cause one to question model fit in this region.
Source: NEDO (1987)
May 2013
D-16
Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 Exponential Model. (Version: 1.7; Date: 12/10/2009)
3 Input Data File: C:/USEPA/BMDS220/Data/Methanol/exp_NEDOrat-Gest-Cmax-Std_Exp-
4 ModelVariance-BMRlStd-Down.(d)
5 Gnuplot Plotting File:
6 Tue Mar 27 12:45:12 2012
7 ====================================================================
8
9 BMDS Model Run
10 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11
12 The form of the response function by Model:
13 Model 2: Y[dose] = a * expfsign * b * dose}
14 Model 3: Y[dose] = a * expfsign * (b * dose)^d}
15 Model 4: Y[dose] = a * [c-(c-l) * exp{-b * dose}]
16 Model 5: Y[dose] = a * [c-(c-l) * exp{-(b * dosej^d}]
17
18 Note: Y[dose] is the median response for exposure = dose;
19 sign = +1 for increasing trend in data;
20 sign = -1 for decreasing trend.
21
22 Model 2 is nested within Models 3 and 4.
23 Model 3 is nested within Model 5.
24 Model 4 is nested within Model 5.
25
26
27 Dependent variable = Mean
28 Independent variable = Dose
29 Data are assumed to be distributed: normally
30 Variance Model: exp(lnalpha +rho *ln(Y[dose]))
31 The variance is to be modeled as Var(i) = exp(lalpha + log(mean(i)) * rho)
32
33 Total number of dose groups = 4
34 Total number of records with missing values = 0
35 Maximum number of iterations = 250
36 Relative Function Convergence has been set to: le-008
37 Parameter Convergence has been set to: le-008
38
39 MLE solution provided: Exact
40
41
42 Initial Parameter Values
43
44 Variable Model 4
45
46 Inalpha 7.32457
47 rho -18.5236
48 a 2.1105
49 b 0.000507001
50 c 0.816778
51 d 1
52
53
54
55 Parameter Estimates
56
57 Variable Model 4
58
59 Inalpha 6.99305
60 rho -18.0776
61 a 2.00632
62 b 0.000758964
63 c 0.891583
64 d 1
65
May 2013 D-17 Draft - Do Not Cite or Quote
-------
1
2 Table of Stats From Input Data
o
J
4 Dose N Obs Mean Obs Std Dev
5
6 0 11 2 0.047
7 7.41 11 2.01 0.075
8 114.6 12 1.99 0.072
9 2986 10 1.81 0.161
10
11
12 Estimated Values of Interest
13
14 Dose Est Mean Est Std Scaled Residual
15
16 0 2.006 0.06098 -0.3437
17 7.41 2.005 0.06132 0.2651
18 114.6 1.988 0.06619 0.09421
19 2986 1.811 0.1536 -0.02792
20
21
22
23 Other models for which likelihoods are calculated:
24
25 Model Al: Yij = Mu(i) + e(ij)
26 Var{e(ij)} = SigmaA2
27
28 Model A2: Yij = Mu(i) + e(ij)
29 Var{e(ij)} = Sigma(i)^2
30
31 Model A3: Yij = Mu(i) + e(ij)
32 Var{e(ij)} = exp(lalpha + log(mean(i)) * rho)
33
34 Model R: Yij = Mu + e(i)
35 Var{e(ij)} = Sigma/x2
36
37
38 Likelihoods of Interest
39
40 Model Log(likelihood) DF AIC
41
42 Al 83.20596 5 -156.4119
43 A2 92.06049 8 -168.121
44 A3 90.61606 6 -169.2321
45 R 70.76186 2 -137.5237
46 4 90.79294 5 -171.5859
47
48
49 Additive constant for all log-likelihoods = -40.43. This constant added to the
50 above values gives the log-likelihood including the term that does not
51 depend on the model parameters.
52
53
54 Explanation of Tests
55
56 Test 1: Does response and/or variances differ among Dose levels? (A2 vs. R)
57 Test 2: Are Variances Homogeneous? (A2 vs. Al)
58 Test 3: Are variances adeguately modeled? (A2 vs. A3)
59
60 Test 6a: Does Model 4 fit the data? (A3 vs 4)
61
62
63
64
65
May 2013 D-18 Draft - Do Not Cite or Quote
-------
1 Tests of Interest
2
3 Test -2*log(Likelihood Ratio) D. F. p-value
4
5 Test 1 42.6 6 < 0.0001
6 Test 2 17.71 3 0.000505
7 Test 3 2.889 2 0.2359
8 Test 6a -0.3538 1 N/A
9
10
11 The p-value for Test 1 is less than .05. There appears to be a
12 difference between response and/or variances among the dose
13 levels, it seems appropriate to model the data.
14
15 The p-value for Test 2 is less than .1. A non-homogeneous
16 variance model appears to be appropriate.
17
18 The p-value for Test 3 is greater than .1. The modeled
19 variance appears to be appropriate here.
20
21 The p-value for Test 6a is less than .1. Model 4 may not adeguately
22 describe the data; you may want to consider another model.
23
24
25 Benchmark Dose Computations:
26
27 Specified Effect = 1.000000
28
29 Risk Type = Estimated standard deviations from control
30
31 Confidence Level = 0.950000
32
33 BMD = 433.456
34
35 BMDL = 114.856
May 2013 D-19 Draft - Do Not Cite or Quote
-------
Exponential Model 4 with 0.95 Confidence Level
2.05
8-
&
1.7
11:45 03/27 201 2
3000
Source: NEDO (1987).
Figure D-3 Exponential model 4, BMR of 1 control mean S.D. - Decreased brain weight in
male rats at 8 weeks of age versus Cmax above background, gestation only
inhalational study.
D.2.3. C.1.2.2. BMD Approach with a BMR of 0.05 Change Relative to
Control Mean (GD7-GD17)
1 A summary of the results most relevant to the development of a POD using the BMD
2 approach (BMD, BMDL, and model fit statistics) for decreased brain weight at 8 weeks in male
3 rats exposed to methanol during gestation from days 7 to 17, with a BMR of 0.05 change relative
4 to estimated control mean, is provided in Table D-6. Model fit was determined by statistics (AIC
5 and % residuals of individual dose groups) and visual inspection, as recommended by EPA
6 (20J_2a). Modeling considerations and uncertainties for this data set were discussed in C. 1.2.1
7 and, as was done for the BMR of 1 S.D., the lowest BMDL was chosen for use in the RfC
8 derivation (NEDO, 1987), which in this case was the BMDLos of 119.51 mg methanol/L in blood
9 estimated by the Exponential 5 model. Results from the Exponential 5 model, including text and
10 plot (see Figure D-4), are shown after Table D-6.
May 2013
D-20
Draft - Do Not Cite or Quote
-------
Table D-6 Comparison of BMD0s modeling results for decreased brain weight in male rats
at 8 weeks of age using modeled Cmax above background of methanol as a
common dose metric
Model
Linearb
2nd degreePolynomial
3rd degree Polynomial
Power
Hillb
Exponential 2
Exponential 3
Exponential 4
Exponential 5
DlVlWoS v^max?
mg/L)a
1,542.49
1,542.49
1,542.49
1,542.49
871.996
1,502.61
1,502.61
814.76
814.76
BMDL05 (Cmax,
mg/L)a /7-value
1,061.91 0.8837
1,061.91 0.8837
1,061.91 0.8837
1,061.91 0.8837
Not Reported 0.9272
1,009.52 0.8910
1,009.52 0.8910
233.33 0.9266
119.51 0.9266
AICC
-173.347015
-173.347015
-173.347015
-173.347015
-171.586011
-173.3635
-173.3635
-171.5859
-171.5859
aThe BMDL is the 95% lower confidence limit on the Cmax estimated to decrease brain weight by 5% using
EPA, 20 11 a) and model options and restrictions suggested by EPA BMD Technical Guidance (2012a).
Scaled residuald
-0.28
-0.28
-0.28
-0.28
0.0944
-0.2674
-0.2674
0.09421
0.09421
BMDS 2.2 (U.S.
°AIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
d^d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals that exceed 2.0 in
absolute value should cause one to question model fit in this region.
Source: NEDO (1987).
May 2013
D-21
Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 Exponential Model. (Version: 1.7; Date: 12/10/2009)
3 Input Data File: C:/USEPA/BMDS220/Data/Methanol/exp_NEDOrat-Gest-Cmax-Std_Exp-
4 ModelVariance-BMR05-Down.(d)
5 Gnuplot Plotting File:
6 Tue Mar 27 15:30:45 2012
7 ====================================================================
8
9 BMDS Model Run
10 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
11
12 The form of the response function by Model:
13 Model 2: Y[dose] = a * expfsign * b * dose}
14 Model 3: Y[dose] = a * expfsign * (b * dose)^d}
15 Model 4: Y[dose] = a * [c-(c-l) * exp{-b * dose}]
16 Model 5: Y[dose] = a * [c-(c-l) * exp{-(b * dosej^d}]
17
18 Note: Y[dose] is the median response for exposure = dose;
19 sign = +1 for increasing trend in data;
20 sign = -1 for decreasing trend.
21
22 Model 2 is nested within Models 3 and 4.
23 Model 3 is nested within Model 5.
24 Model 4 is nested within Model 5.
25
26
27 Dependent variable = Mean
28 Independent variable = Dose
29 Data are assumed to be distributed: normally
30 Variance Model: exp(lnalpha +rho *ln(Y[dose]))
31 The variance is to be modeled as Var(i) = exp(lalpha + log(mean(i)) * rho)
32
33 Total number of dose groups = 4
34 Total number of records with missing values = 0
35 Maximum number of iterations = 250
36 Relative Function Convergence has been set to: le-008
37 Parameter Convergence has been set to: le-008
38
39 MLE solution provided: Exact
40
41
42 Initial Parameter Values
43
44 Variable Model 5
45
46 Inalpha 7.32457
47 rho -18.5236
48 a 2.1105
49 b 0.000507001
50 c 0.816778
51 d 1
52
53
54
55 Parameter Estimates
56
57 Variable Model 5
58
59 Inalpha 6.99305
60 rho -18.0776
61 a 2.00632
62 b 0.000758964
63 c 0.891583
64 d 1
65
May 2013 D-22 Draft - Do Not Cite or Quote
-------
1
2 Table of Stats From Input Data
o
J
4 Dose N Obs Mean Obs Std Dev
5
6 0 11 2 0.047
7 7.41 11 2.01 0.075
8 114.6 12 1.99 0.072
9 2986 10 1.81 0.161
10
11
12 Estimated Values of Interest
13
14 Dose Est Mean Est Std Scaled Residual
15
16 0 2.006 0.06098 -0.3437
17 7.41 2.005 0.06132 0.2651
18 114.6 1.988 0.06619 0.09421
19 2986 1.811 0.1536 -0.02792
20
21
22
23 Other models for which likelihoods are calculated:
24
25 Model Al: Yij = Mu(i) + e(ij)
26 Var{e(ij)} = SigmaA2
27
28 Model A2: Yij = Mu(i) + e(ij)
29 Var{e(ij)} = Sigma(i)^2
30
31 Model A3: Yij = Mu(i) + e(ij)
32 Var{e(ij)} = exp(lalpha + log(mean(i)) * rho)
33
34 Model R: Yij = Mu + e(i)
35 Var{e(ij)} = Sigma/x2
36
37
38 Likelihoods of Interest
39
40 Model Log(likelihood) DF AIC
41
42 Al 83.20596 5 -156.4119
43 A2 92.06049 8 -168.121
44 A3 90.61606 6 -169.2321
45 R 70.76186 2 -137.5237
46 5 90.79294 5 -171.5859
47
48
49 Additive constant for all log-likelihoods = -40.43. This constant added to the
50 above values gives the log-likelihood including the term that does not
51 depend on the model parameters.
52
53
54 Explanation of Tests
55
56 Test 1: Does response and/or variances differ among Dose levels? (A2 vs. R)
57 Test 2: Are Variances Homogeneous? (A2 vs. Al)
58 Test 3: Are variances adeguately modeled? (A2 vs. A3)
59
60 Test 7a: Does Model 5 fit the data? (A3 vs 5)
61
62
63 Tests of Interest
64
65 Test -2*log(Likelihood Ratio) D. F. p-value
May 2013 D-23 Draft - Do Not Cite or Quote
-------
1
2 Test 1 42.6 6 < 0.0001
3 Test 2 17.71 3 0.000505
4 Test 3 2.889 2 0.2359
5 Test 7a -0.3538 1 N/A
6
7
8 The p-value for Test 1 is less than .05. There appears to be a
9 difference between response and/or variances among the dose
10 levels, it seems appropriate to model the data.
11
12 The p-value for Test 2 is less than .1. A non-homogeneous
13 variance model appears to be appropriate.
14
15 The p-value for Test 3 is greater than .1. The modeled
16 variance appears to be appropriate here.
17
18 The p-value for Test 7a is less than .1. Model 5 may not adeguately
19 describe the data; you may want to consider another model.
20
21
22 Benchmark Dose Computations:
23
24 Specified Effect = 0.050000
25
26 Risk Type = Relative deviation
27
28 Confidence Level = 0.950000
29
30 BMD = 814.763
31
32 BMDL = 119.505
May 2013 D-24 Draft - Do Not Cite or Quote
-------
Exponential Model 5 with 0.95 Confidence Level
2.05
8-
&
1.7
14:3003/272012
3000
Source: NEDO (1987).
Figure D-4 Exponential model 4, BMR of 0.05 relative risk - Decreased brain weight in
male rats at 8 weeks age versus Cmax above background, gestation only
inhalational study.
D.3. RfC Derivations Using Rogers et al. (1993b)
1 For the purposes of deriving an RfC for methanol from developmental endpoints using
2 the BMD method and mouse data, cervical rib incidence data were evaluated from Rogers et al.
3 (1993b). In this paper, Rogers et al. (1993b) also utilized a BMD methodology, examining the
4 dosimetric threshold for cervical ribs and other developmental impacts by applying a log-logistic
5 maximum likelihood model to the dose-response data. Using air exposure concentrations (ppm)
6 as their dose metric, a value for the lower 95% confidence limit on the benchmark dose for 5%
7 additional risk in mice was 305 ppm (400 mg/m3), using the log-logistic model. Although the
8 teratology portion of the NEDO study (1987) also reported increases in cervical rib incidence in
9 Sprague-Dawley rats, the Rogers et al. (1993b) study was chosen for dose-response modeling
10 because effects were seen at lower doses, it was peer-reviewed and published in the open
11 literature, and data on individual animals were available for a more statistically robust analysis
12 utilizing nested models available in HMDS 2.2 (U.S. EPA. 201 la).
13 As described in Section 5.1.2.1, because exposure was during gestation only and due to
14 the small critical gestational window for cervical rib abnormalities, Cmax of methanol in blood
15 (mg/L) is chosen as the appropriate internal dose metric. Because the critical window for
May 2013
D-25
Draft - Do Not Cite or Quote
-------
1 methanol induction of cervical rib malformations in CD-I mice is between GD6 and GD7
2 (Rogers and Mole, 1997; Rogers et al., 1993a), the measured Cmax plasma methanol levels for
3 gestation day 6 from the Rogers study are used with background levels (1.6 g/L) subtracted. Cmax
4 values for methanol in the blood of mice are summarized in Table D-7. These Cmax values are
5 then used as the dose metric for the BMD analysis of the litter-specific cervical rib response. The
6 overall cervical rib/litter (%) reported by Rogers et al. (1993b) is shown in Table D-7, but litter-
7 specific response data from this study (170 litters) obtained from John Rogers (personal
8 communication) was used for the nested BMD analysis. Due to high mortality, the high
9 (15,000 ppm) dose group (5 litters) was excluded from this analysis. The individual animal
10 response data for the four dose groups are displayed below in the text output files for the
11 NLogistic model.
Table D-7 Methanol blood levels (Cmax above background) in mice following inhalation
exposures
Exposure (ppm)
0
1,000
2,000
5,000
Methanol in blood Cmax (mg/L)a in mice
0
61.4
485.4
2,124.4
Cervical Rib/Litter (%)
28
33.6
49.6
74.4
aReported Cmax background levels of 1.6 mg/L were subtracted from reported Cmax values.
Source: Rogers et al. (1993b)
12 A 10% BMR level is the value typically calculated for comparisons across chemicals and
13 endpoints for dichotomous responses because this level is near the low end of the observable
14 range for many types of toxicity studies. However, reproductive and developmental studies
15 having a nested design often have a greater sensitivity, and a 5% BMR is typically appropriate
16 for determination of a POD (U.S. EPA. 2012a: Allen etal.. 1994a). Rogers et al. Q993b) utilized
17 a 5% added risk for the BMR in the original study. This assessment utilizes both a 10% and 5%
18 extra risk level as a BMR for the determination of a POD.3 The nested suite of models available
19 in BMDS 2.2 (U.S. EPA, 2011 a) was used to model the cervical rib data. In general, data from
20 developmental toxicity studies are best modeled using nested models, as these models account
21 for any intralitter correlation (i.e., the tendency of littermates to respond similarly to one another
3 Starr and Festa (2003) have argued that the Rogers, et al. (1993b) study's experimental design lacked the statistical
power to detect a 5% risk and that a 5% level lay below the observable response data. However, EPA's BMD
guidance (U.S. EPA. 2012a) does not preclude the use of a BMR that is below observable response data and EPA
has deemed that Rogers et al. (1993b) is adequate for the consideration of a 5% BMR.
May 2013 D-26 Draft - Do Not Cite or Quote
-------
1 relative to other litters in a dose group). All models were fit using restrictions and option settings
2 suggested in the EPA's BMD Technical Guidance Document (U.S. EPA, 2012a).
D.3.1. BMD Approach with a BMR of 0.10 Extra Risk
3 A summary of the results most relevant to the development of a POD using the BMD
4 approach (BMD, BMDL, and model fit statistics) for increased incidence of cervical rib in mice
5 exposed to methanol during gestation from days 6 to 15, with a BMR of 0.10 extra risk, is
6 provided in Table D-8. Model fit was determined by statistics (AIC and ^ residuals of individual
7 dose groups) and visual inspection, as recommended by U.S. EPA (U.S. EPA, 2012a). The best
8 model fit to these data (from visual inspection and comparison of AIC values) was obtained
9 using the Nested Logistic (NLogistic) model. The textual and graphic (see Figure D-5) output
10 from this model follows Table D-8. The BMDLio was determined to be 90.9972 mg/L using the
11 95% lower confidence limit of the dose-response curve expressed in terms of the Cmaxfor
12 methanol in blood (Rogers et al., 1993b).
Table D-8 Comparison of BMD modeling results for 10% cervical rib incidence in mice
using modeled Cmax above background of methanol as a common dose metric
Model
NLogisticb
BMD10
(Cmax,mg/L)a
140.75
BMDL10(Cmax,
mg/L)a
91.00
/7-value
0.3359
AIC
1047.37
Scaled residual"1
0.5395
NCTR 223.55 111.78 0.2705 1050.32 0.5640
RaiandVanRyzin 233.61 116.81 0.2625 1052.14 0.6043
aCmax values are the blood levels of the dams on GD6 with background subtracted; the BMDL is the 95% lower confidence limit
on the Cmax for 10% extra risk (dichotomous endpoints) estimated by the model using the likelihood profile method (U.S. EPA,
2Q12a).
bModel choice based on adequate/) value (> 0.1), visual inspection, low AIC, and low (absolute) scaled residual.
°AIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
^d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals exceeding 2.0 in
absolute value should cause one to question model fit in this region.
Source: Rogers et al. (1993b).
May 2013 D-27 Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 NLogistic Model. (Version: 2.15; Date: 10/28/2009)
3 Input Data File: C:/Documents and
4 Settings/llowe/Desktop/ROGERS_CMAX_BMD/ROGERS_CMAX_BMD10/NLog_CR_10. (d)
5 Fri Dec 16 10:48:13 2011
6 ====================================================================
7
8 BMDS Model Run
Q ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10
11 The probability function is:
12
13
14 Prob. = alpha + thetal*Rij + [1 - alpha - thetal*Rij]/
15
16 [1+exp (-beta-theta2*Rij-rho*log (Dose))],
17
18 where Rij is the litter specific covariate.
19
20 Restrict Power rho >= 1.
21
22
23
24 Total number of observations =166
25 Total number of records with missing values = 0
26 Total number of parameters in model = 9
27 Total number of specified parameters = 0
28
29
30 Maximum number of iterations = 250
31 Relative Function Convergence has been set to: le-008
32 Parameter Convergence has been set to: le-008
33
34
35
36 Default Initial Parameter Values
37 alpha = 0.302379
38 beta = -7.2579
39 thetal = 0
40 theta2 = 0
41 rho = 1
42 phil = 0.214334
43 phi2 = 0.304943
44 phi3 = 0.220179
45 phi4 = 0.370088
46
47
48
49 Parameter Estimates
50
51 Variable Estimate Std. Err.
52 alpha 0.127131 *
53 beta -4.62736 *
54 thetal 0.0297845 *
55 theta2 -0.467856 *
56 rho 1 *
57 phil 0.203691 *
58 phi2 0.305429 *
59 phi3 0.212663 *
60 phi4 0.363199 *
61
62 * - Indicates that this value is not calculated.
63
64 Log-likelihood: -515.686AIC: 1047.37
65
May 2013 D-28 Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Litter Data
Lit. -Spec. Litter Scaled
Dose Cov. Est. Prob. Size Expected Observed Residual
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
1
1
2
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
7
8
8
8
8
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
,157
,157
,187
,187
,187
,187
,187
,216
,216
,216
,216
,246
,246
,246
,246
,276
,276
,276
,276
,276
,276
,276
,276
,276
,276
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,365
,365
,365
,365
1
1
2
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
7
8
8
8
8
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
,157
,157
,373
,373
,373
,373
,373
, 649
, 649
,649
,649
,985
,985
,985
,985
,380
,380
,380
,380
,380
,380
,380
,380
,380
,380
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,923
,923
,923
,923
0
0
0
2
0
0
1
1
0
1
0
0
1
0
1
0
1
1
3
0
0
0
1
1
1
0
5
3
3
0
3
5
3
1
3
2
2
1
6
0
0
1
5
1
2
1
2
5
2
3
3
0
0
2
4
2
8
3
-0.4314
-0.4314
-0.6176
2. 6904
-0.6176
-0.6176
1.0364
0.4142
-0.7674
0.4142
-0.7674
-0.9007
0.0136
-0.9007
0.0136
-1.0250
-0.2824
-0.2824
1.2028
-1.0250
-1.0250
-1.0250
-0.2824
-0.2824
-0.2824
-1.1444
1.9738
0.7265
0.7265
-1.1444
0.7265
1.9738
0.7265
-0.5208
0.7265
0.1029
0.1029
-0.5208
2.5975
-1.1444
-1.1444
-0.7245
1.4233
-0.7245
-0.1876
-0.7245
-0.1876
1.4233
-0.1876
0.3494
0.3494
-1.2615
-1.2615
-0.1876
0.5076
-0.4352
2.3932
0.0362
May 2013
D-29
Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
0.0000 8
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.0000 0
1.
1.
2.
2.
3.
3.
3.
3.
4.
4.
5.
5.
5.
5.
5.
6.
6.
7.
7.
7.
7.
7.
7.
8.
8.
8.
2
3
4
4
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.365 8
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
387
387
342
342
317
317
317
317
310
310
316
316
316
316
316
330
330
350
350
350
350
350
350
374
374
374
.716
.638
.564
.564
.503
.503
.503
.503
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.437
.437
.437
.437
.437
.437
.437
.437
.437
.437
.437
.437
2
1
1
2
2
3
3
3
3
4
4
5
5
5
5
5
6
6
7
7
7
7
7
7
8
8
8
2
3
4
4
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
.923 1 -
0.
0.
0.
0.
0.
0.
0.
0.
1.
1.
1.
1.
1.
1.
1.
1.
1.
2.
2.
2.
2.
2.
2.
2.
2.
2.
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
387
387
684
684
952
952
952
952
240
240
578
578
578
578
578
981
981
453
453
453
453
453
453
994
994
994
.432
.915
.258
.258
.517
.517
.517
.517
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.057
.057
.057
.057
.057
.057
.057
.057
.057
.057
.057
.057
0
0
2
0
1
3
1
2
3
0
0
0
1
4
0
3
2
2
2
3
0
2
2
2
8
0
2
3
1
2
5
1
3
3
2
5
2
3
2
0
4
0
5
2
4
3
6
3
5
3
4
5
0
5
1
4
3
4
1
3
3
1
0.9066
-0.7951
-0.7951
1.7177
-0.8919
0.0472
2.0021
0.0472
1.0246
1.3743
-0.9685
-1.0189
-1.0189
-0.3733
1.5633
-1.0189
0.5566
0.0105
-0.2131
-0.2131
0.2577
-1.1545
-0.2131
-0.2131
-0.4101
2.0644
-1.2350
0.8091
1.0920
-0.9912
-0.2032
1.6327
-0.9972
0.3178
0.3178
-0.4350
1.2756
-0.4350
0.1352
-0.4350
-1.5754
0.7054
-1.5754
1.2756
-0.4350
0.7054
0.1352
1.8458
0.1352
1.2756
0.1352
0.4762
0.9813
-1.5443
0.9813
-1.0392
0.4762
-0.0289
0.4762
-1.0392
-0.0289
-0.0289
-1.0392
May 2013
D-30
Draft - Do Not Cite or Quote
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
•SQ
jy
60
61
62
63
64
65
485.
485.
485.
485.
485.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
8.0000
8.0000
8.0000
9.0000
9.0000
1.0000
1.0000
1.0000
2.0000
2.0000
3.0000
3.0000
3.0000
3.0000
4.0000
4.0000
4.0000
4.0000
4.0000
4.0000
4.0000
5.0000
5.0000
5.0000
5.0000
5.0000
5.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
7.0000
7.0000
7.0000
7.0000
7.0000
8.0000
Combine litters
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.430
.430
.430
.435
.435
.940
.940
.940
.911
.911
.872
.872
.872
.872
.820
.820
.820
.820
.820
.820
.820
.759
.759
.759
.759
.759
.759
. 692
.692
.692
. 692
. 692
.692
.692
. 692
. 692
.692
.692
. 692
. 628
.628
.628
. 628
. 628
.575
8
8
8
9
9
1
1
1
2
2
3
3
3
3
4
4
4
4
4
4
4
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
8
3
3
3
3
3
0
0
0
1
1
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
.436
.436
.436
.915
.915
.940
.940
.940
.822
.822
. 615
.615
.615
. 615
.282
.282
.282
.282
.282
.282
.282
.795
.795
.795
.795
.795
.795
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.396
.396
.396
.396
.396
.598
7
5
0
0
6
1
1
1
2
1
3
3
1
1
4
4
2
4
4
3
4
1
5
5
4
4
3
5
6
6
3
6
2
4
0
5
0
5
4
5
5
7
6
7
0
1.
0.
-1
-1
0.
0.
0.
0.
0.
-1
0.
0.
-2
-2
0.
0.
-1
0.
0.
-0
0.
-1
0.
0.
0.
0.
-0
0.
0.
0.
-0
0.
-1
-0
-2
0.
-2
0.
-0
0.
0.
1.
0.
1.
-1
6134
7079
.5558
.6016
8530
2530
2530
2530
3783
.7500
5058
5058
.1218
.1218
6473
6473
.1551
6473
6473
.2539
6473
.8656
8047
8047
1371
1371
.5305
4466
9736
9736
.6074
9736
.1344
.0804
.1885
4466
.1885
4466
.0804
2650
2650
1421
7036
1421
.7470
with adjacent levels of the litter-specific covariate
within dose groups
the fit of
the X
until
the expected count exceeds 3.0, to help improve
.A2 statistic to
chi
-sguare.
Grouped Data
Mean
Dose
Scaled
Lit.
0.0000 1.
0.0000 2.
0.0000 3.
0.0000 4.
0.0000 5.
0.0000 5.
-Spec.
0000 0
0000 1
0000 2
0000 3
0000 4
0000 4
Cov.
.314 0
.867 3
.598 2
.940 2
.141 2
.141 3
Expected
-0
0.
-0
-0
-0
-0
6101
Observed Residual
8381
.
.
.
.
3532
8870
9178
4891
May 2013
D-31
Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
0.0000 5.0000 4.141 2 -0.9178
0.0000 5.0000 1.380 1 -0.2824
0.0000 6.0000 3.670 5 0.5865
0.0000 6.0000 3.670 6 1.0275
0.0000 6.0000 3.670 3 -0.2955
0.0000 6.0000 3.670 8 1.9094
0.0000 6.0000 3.670 4 0.1455
0.0000 6.0000 3.670 4 0.1455
0.0000 6.0000 3.670 7 1.4685
0.0000 6.0000 3.670 0 -1.6184
0.0000 7.0000 4.699 6 0.4941
0.0000 7.0000 4.699 3 -0.6450
0.0000 7.0000 4.699 3 -0.6450
0.0000 7.0000 4.699 7 0.8738
0.0000 7.0000 4.699 6 0.4941
0.0000 7.0000 4.699 0 -1.7840
0.0000 7.0000 2.349 2 -0.1876
0.0000 8.0000 5.847 6 0.0512
0.0000 8.0000 5.847 11 1.7178
0.0000 8.0000 2.923 1 -0.9066
61.4000 1.0000 0.775 0 -1.1245
61.4000 2.0000 1.367 2 0.5840
61.4000 3.0000 3.807 7 1.5606
61.4000 4.0000 2.480 3 0.2870
61.4000 5.0000 3.157 0 -1.4409
61.4000 5.0000 3.157 5 0.8414
61.4000 5.0000 1.578 0 -1.0189
61.4000 6.0000 3.962 5 0.4010
61.4000 7.0000 4.905 4 -0.3013
61.4000 7.0000 4.905 3 -0.6342
61.4000 7.0000 4.905 4 -0.3013
61.4000 8.0000 5.989 10 1.1697
61.4000 8.0000 2.994 0 -1.2350
485.4000 2.0000 1.432 2 0.8091
485.4000 3.0000 1.915 3 1.0920
485.4000 4.0000 4.516 3 -0.8446
485.4000 5.0000 5.033 6 0.4494
485.4000 5.0000 5.033 6 0.4494
485.4000 6.0000 5.526 7 0.5944
485.4000 6.0000 5.526 5 -0.2120
485.4000 6.0000 5.526 2 -1.4216
485.4000 6.0000 5.526 4 -0.6152
485.4000 6.0000 5.526 7 0.5944
485.4000 6.0000 5.526 7 0.5944
485.4000 6.0000 5.526 9 1.4008
485.4000 6.0000 5.526 8 0.9976
485.4000 7.0000 3.057 4 0.4762
485.4000 7.0000 3.057 5 0.9813
485.4000 7.0000 3.057 0 -1.5443
485.4000 7.0000 3.057 5 0.9813
485.4000 7.0000 3.057 1 -1.0392
485.4000 7.0000 3.057 4 0.4762
485.4000 7.0000 3.057 3 -0.0289
485.4000 7.0000 3.057 4 0.4762
485.4000 7.0000 3.057 1 -1.0392
485.4000 7.0000 3.057 3 -0.0289
485.4000 7.0000 3.057 3 -0.0289
485.4000 7.0000 3.057 1 -1.0392
485.4000 8.0000 3.436 7 1.6134
485.4000 8.0000 3.436 5 0.7079
485.4000 8.0000 3.436 0 -1.5558
485.4000 9.0000 3.915 0 -1.6016
485.4000 9.0000 3.915 6 0.8530
May 2013
D-32
Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
1.
2.
3.
3.
4.
4.
4.
4.
4.
4.
4.
5.
5.
5.
5.
5.
5.
6.
6.
6.
6.
6.
6.
6.
6.
6.
6.
6.
6.
7.
7.
7.
7.
7.
8.
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
2
3
5
5
3
3
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
.820
. 645
.230
.230
.282
.282
.282
.282
.282
.282
.282
.795
.795
.795
.795
.795
.795
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.396
.396
.396
.396
.396
.598
3
3
6
2
4
4
2
4
4
3
4
1
5
5
4
4
3
5
6
6
3
6
2
4
0
5
0
5
4
5
5
7
6
7
0
0.
-0
0.
-3
0.
0.
-1
0.
0.
-0
0.
-1
0.
0.
0.
0.
-0
0.
0.
0.
-0
0.
-1
-0
-2
0.
-2
0.
-0
0.
0.
1.
0.
1.
-1
4382
.9699
7153
.0007
6473
6473
.1551
6473
6473
.2539
6473
.8656
8047
8047
1371
1371
.5305
4466
9736
9736
.6074
9736
.1344
.0804
.1885
4466
.1885
4466
.0804
2650
2650
1421
7036
1421
.7470
Chi-square = 101.30 DF = 96 P-value = 0.3359
To calculate the BMD and BMDL, the litter specific covariate is fixed
at the mean litter specific covariate of all the data: 5.379518
Benchmark Dose Computation
Specified effect =0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 140.749
BMDL = 90.9972
May 2013
D-33
Draft - Do Not Cite or Quote
-------
Nested Logistic Model with 0.95 Confidence Level
1
I
.1
o
0.8
0.7
0.6
0.5
0.4
0.3
0.2
10:48 12/16 2011
2000
Source: Rogers et al. Q993b).
Figure D-5 Nested logistic model, 0.1 extra risk - Incidence of cervical rib in mice versus
Cmax above background of methanol, GD6-GD15 inhalational study.
D.3.2. BMD Approach with a BMR of 0.05 Extra Risk
1 A summary of the results most relevant to the development of a POD using the BMD
2 approach (BMD, BMDL, and model fit statistics) for increased incidence of cervical rib in mice
3 exposed to methanol during gestation from days 6 to 15, with a BMR of 0.05 extra risk, is
4 provided in Table D-9. Model fit was determined by statistics (AIC and ^ residuals of individual
5 dose groups) and visual inspection, as recommended by U.S. EPA (2012a). The best model fit to
6 these data (from visual inspection and comparison of AIC values) was obtained using the
7 NLogistic model. The text and graphic (see Figure D-6) output from this model follow
8 Table D-6. The BMDL0s was determined to be 43.1039 mg/L using the 95% lower confidence
9 limit of the dose-response curve expressed in terms of the Cmax for methanol in blood (Rogers et
10 al.. 1993b).
May 2013
D-34
Draft - Do Not Cite or Quote
-------
Table D-9 Comparison of BMD modeling results for 5% cervical rib incidence in mice
using modeled Cmax above background of methanol as a common dose metric
Model
NLogisticb
NCTR
Rai and Van Ryzin
BMD05
(Cmax,mg/L)a
66.67
108.83
113.73
BMDL05
(Cmax,mg/L)a
43.10
54.42
56.87
/7-value
0.3359
0.2705
0.2625
AICC
1047.37
1050.32
1052.14
Scaled residual11
0.5395
0.5640
0.6043
aCmax are the blood levels of the dams on GD6 with background subtracted; the BMDL is the 95% lower confidence limit on the
Cmax for a 5% extra risk (dichotomous endpoints) estimated by the model using the likelihood profile method (U.S. EPA, 2012a).
bModel choice based on adequate/) value (> 0.1), visual inspection, low AIC, and low (absolute) scaled residual.
°AIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
d^d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals exceeding 2.0 in absolute
value should cause one to question model fit in this region.
Source: Rogers et al. (1993V).
May 2013 D-35 Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 NLogistic Model. (Version: 2.15; Date: 10/28/2009)
3 Input Data File: C:/Documents and
4 Settings/llowe/Desktop/ROGERS_CMAX_BMD/ROGERS_CMAX_BMD05/NLog_CR_5. (d)
5 Fri Dec 16 10:56:05 2011
6 ====================================================================
7
8 BMDS Model Run
Q ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
10
11 The probability function is:
12
13
14 Prob. = alpha + thetal*Rij + [1 - alpha - thetal*Rij]/
15
16 [1+exp (-beta-theta2*Rij-rho*log (Dose))],
17
18 where Rij is the litter specific covariate.
19
20 Restrict Power rho >= 1.
21
22
23
24 Total number of observations =166
25 Total number of records with missing values = 0
26 Total number of parameters in model = 9
27 Total number of specified parameters = 0
28
29
30 Maximum number of iterations = 250
31 Relative Function Convergence has been set to: le-008
32 Parameter Convergence has been set to: le-008
33
34
35
36 Default Initial Parameter Values
37 alpha = 0.302379
38 beta = -7.2579
39 thetal = 0
40 theta2 = 0
41 rho = 1
42 phil = 0.214334
43 phi2 = 0.304943
44 phi3 = 0.220179
45 phi4 = 0.370088
46
47
48
49 Parameter Estimates
50
51 Variable Estimate Std. Err.
52 alpha 0.127131 *
53 beta -4.62736 *
54 thetal 0.0297845 *
55 theta2 -0.467856 *
56 rho 1 *
57 phil 0.203691 *
58 phi2 0.305429 *
59 phi3 0.212663 *
60 phi4 0.363199 *
61
62 * - Indicates that this value is not calculated.
63
64 Log-likelihood: -515.686AIC: 1047.37
65
May 2013 D-36 Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Litter Data
Lit. -Spec. Litter Scaled
Dose Cov. Est. Prob. Size Expected Observed Residual
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
1
1
2
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
7
8
8
8
8
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
,157
,157
,187
,187
,187
,187
,187
,216
,216
,216
,216
,246
,246
,246
,246
,276
,276
,276
,276
,276
,276
,276
,276
,276
,276
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,306
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,336
,365
,365
,365
,365
1
1
2
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
7
8
8
8
8
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
2.
,157
,157
,373
,373
,373
,373
,373
, 649
, 649
,649
,649
,985
,985
,985
,985
,380
,380
,380
,380
,380
,380
,380
,380
,380
,380
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,835
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,349
,923
,923
,923
,923
0
0
0
2
0
0
1
1
0
1
0
0
1
0
1
0
1
1
3
0
0
0
1
1
1
0
5
3
3
0
3
5
3
1
3
2
2
1
6
0
0
1
5
1
2
1
2
5
2
3
3
0
0
2
4
2
8
3
-0.4314
-0.4314
-0.6176
2. 6904
-0.6176
-0.6176
1.0364
0.4142
-0.7674
0.4142
-0.7674
-0.9007
0.0136
-0.9007
0.0136
-1.0250
-0.2824
-0.2824
1.2028
-1.0250
-1.0250
-1.0250
-0.2824
-0.2824
-0.2824
-1.1444
1.9738
0.7265
0.7265
-1.1444
0.7265
1.9738
0.7265
-0.5208
0.7265
0.1029
0.1029
-0.5208
2.5975
-1.1444
-1.1444
-0.7245
1.4233
-0.7245
-0.1876
-0.7245
-0.1876
1.4233
-0.1876
0.3494
0.3494
-1.2615
-1.2615
-0.1876
0.5076
-0.4352
2.3932
0.0362
May 2013
D-37
Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
0.0000 8
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
61.
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
485
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.0000 0
1.
1.
2.
2.
3.
3.
3.
3.
4.
4.
5.
5.
5.
5.
5.
6.
6.
7.
7.
7.
7.
7.
7.
8.
8.
8.
2
3
4
4
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.365 8
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
387
387
342
342
317
317
317
317
310
310
316
316
316
316
316
330
330
350
350
350
350
350
350
374
374
374
.716
.638
.564
.564
.503
.503
.503
.503
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.460
.437
.437
.437
.437
.437
.437
.437
.437
.437
.437
.437
.437
2
1
1
2
2
3
3
3
3
4
4
5
5
5
5
5
6
6
7
7
7
7
7
7
8
8
8
2
3
4
4
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
.923 1 -
0.
0.
0.
0.
0.
0.
0.
0.
1.
1.
1.
1.
1.
1.
1.
1.
1.
2.
2.
2.
2.
2.
2.
2.
2.
2.
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
387
387
684
684
952
952
952
952
240
240
578
578
578
578
578
981
981
453
453
453
453
453
453
994
994
994
.432
.915
.258
.258
.517
.517
.517
.517
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.763
.057
.057
.057
.057
.057
.057
.057
.057
.057
.057
.057
.057
0
0
2
0
1
3
1
2
3
0
0
0
1
4
0
3
2
2
2
3
0
2
2
2
8
0
2
3
1
2
5
1
3
3
2
5
2
3
2
0
4
0
5
2
4
3
6
3
5
3
4
5
0
5
1
4
3
4
1
3
3
1
0.9066
-0.7951
-0.7951
1.7177
-0.8919
0.0472
2.0021
0.0472
1.0246
1.3743
-0.9685
-1.0189
-1.0189
-0.3733
1.5633
-1.0189
0.5566
0.0105
-0.2131
-0.2131
0.2577
-1.1545
-0.2131
-0.2131
-0.4101
2.0644
-1.2350
0.8091
1.0920
-0.9912
-0.2032
1.6327
-0.9972
0.3178
0.3178
-0.4350
1.2756
-0.4350
0.1352
-0.4350
-1.5754
0.7054
-1.5754
1.2756
-0.4350
0.7054
0.1352
1.8458
0.1352
1.2756
0.1352
0.4762
0.9813
-1.5443
0.9813
-1.0392
0.4762
-0.0289
0.4762
-1.0392
-0.0289
-0.0289
-1.0392
May 2013
D-38
Draft - Do Not Cite or Quote
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
•SQ
jy
60
61
62
63
64
65
485.
485.
485.
485.
485.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
2124.
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
4000
8.0000
8.0000
8.0000
9.0000
9.0000
1.0000
1.0000
1.0000
2.0000
2.0000
3.0000
3.0000
3.0000
3.0000
4.0000
4.0000
4.0000
4.0000
4.0000
4.0000
4.0000
5.0000
5.0000
5.0000
5.0000
5.0000
5.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
6.0000
7.0000
7.0000
7.0000
7.0000
7.0000
8.0000
Combine litters
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.430
.430
.430
.435
.435
.940
.940
.940
.911
.911
.872
.872
.872
.872
.820
.820
.820
.820
.820
.820
.820
.759
.759
.759
.759
.759
.759
. 692
.692
.692
. 692
. 692
.692
.692
. 692
. 692
.692
.692
. 692
. 628
.628
.628
. 628
. 628
.575
8
8
8
9
9
1
1
1
2
2
3
3
3
3
4
4
4
4
4
4
4
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
8
3
3
3
3
3
0
0
0
1
1
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
.436
.436
.436
.915
.915
.940
.940
.940
.822
.822
. 615
.615
.615
. 615
.282
.282
.282
.282
.282
.282
.282
.795
.795
.795
.795
.795
.795
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.153
.396
.396
.396
.396
.396
.598
7
5
0
0
6
1
1
1
2
1
3
3
1
1
4
4
2
4
4
3
4
1
5
5
4
4
3
5
6
6
3
6
2
4
0
5
0
5
4
5
5
7
6
7
0
1.
0.
-1
-1
0.
0.
0.
0.
0.
-1
0.
0.
-2
-2
0.
0.
-1
0.
0.
-0
0.
-1
0.
0.
0.
0.
-0
0.
0.
0.
-0
0.
-1
-0
-2
0.
-2
0.
-0
0.
0.
1.
0.
1.
-1
6134
7079
.5558
.6016
8530
2530
2530
2530
3783
.7500
5058
5058
.1218
.1218
6473
6473
.1551
6473
6473
.2539
6473
.8656
8047
8047
1371
1371
.5305
4466
9736
9736
.6074
9736
.1344
.0804
.1885
4466
.1885
4466
.0804
2650
2650
1421
7036
1421
.7470
with adjacent levels of the litter-specific covariate
within dose groups
the fit of
the X
until
the expected count exceeds 3.0, to help improve
.A2 statistic to
chi
-sguare.
Grouped Data
Mean
Dose
Scaled
Lit.
0.0000 1.
0.0000 2.
0.0000 3.
0.0000 4.
0.0000 5.
0.0000 5.
-Spec.
0000 0
0000 1
0000 2
0000 3
0000 4
0000 4
Cov.
.314 0
.867 3
.598 2
.940 2
.141 2
.141 3
Expected
-0
0.
-0
-0
-0
-0
6101
Observed Residual
8381
.
.
.
.
3532
8870
9178
4891
May 2013
D-39
Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
0.0000 5.0000 4.141 2 -0.9178
0.0000 5.0000 1.380 1 -0.2824
0.0000 6.0000 3.670 5 0.5865
0.0000 6.0000 3.670 6 1.0275
0.0000 6.0000 3.670 3 -0.2955
0.0000 6.0000 3.670 8 1.9094
0.0000 6.0000 3.670 4 0.1455
0.0000 6.0000 3.670 4 0.1455
0.0000 6.0000 3.670 7 1.4685
0.0000 6.0000 3.670 0 -1.6184
0.0000 7.0000 4.699 6 0.4941
0.0000 7.0000 4.699 3 -0.6450
0.0000 7.0000 4.699 3 -0.6450
0.0000 7.0000 4.699 7 0.8738
0.0000 7.0000 4.699 6 0.4941
0.0000 7.0000 4.699 0 -1.7840
0.0000 7.0000 2.349 2 -0.1876
0.0000 8.0000 5.847 6 0.0512
0.0000 8.0000 5.847 11 1.7178
0.0000 8.0000 2.923 1 -0.9066
61.4000 1.0000 0.775 0 -1.1245
61.4000 2.0000 1.367 2 0.5840
61.4000 3.0000 3.807 7 1.5606
61.4000 4.0000 2.480 3 0.2870
61.4000 5.0000 3.157 0 -1.4409
61.4000 5.0000 3.157 5 0.8414
61.4000 5.0000 1.578 0 -1.0189
61.4000 6.0000 3.962 5 0.4010
61.4000 7.0000 4.905 4 -0.3013
61.4000 7.0000 4.905 3 -0.6342
61.4000 7.0000 4.905 4 -0.3013
61.4000 8.0000 5.989 10 1.1697
61.4000 8.0000 2.994 0 -1.2350
485.4000 2.0000 1.432 2 0.8091
485.4000 3.0000 1.915 3 1.0920
485.4000 4.0000 4.516 3 -0.8446
485.4000 5.0000 5.033 6 0.4494
485.4000 5.0000 5.033 6 0.4494
485.4000 6.0000 5.526 7 0.5944
485.4000 6.0000 5.526 5 -0.2120
485.4000 6.0000 5.526 2 -1.4216
485.4000 6.0000 5.526 4 -0.6152
485.4000 6.0000 5.526 7 0.5944
485.4000 6.0000 5.526 7 0.5944
485.4000 6.0000 5.526 9 1.4008
485.4000 6.0000 5.526 8 0.9976
485.4000 7.0000 3.057 4 0.4762
485.4000 7.0000 3.057 5 0.9813
485.4000 7.0000 3.057 0 -1.5443
485.4000 7.0000 3.057 5 0.9813
485.4000 7.0000 3.057 1 -1.0392
485.4000 7.0000 3.057 4 0.4762
485.4000 7.0000 3.057 3 -0.0289
485.4000 7.0000 3.057 4 0.4762
485.4000 7.0000 3.057 1 -1.0392
485.4000 7.0000 3.057 3 -0.0289
485.4000 7.0000 3.057 3 -0.0289
485.4000 7.0000 3.057 1 -1.0392
485.4000 8.0000 3.436 7 1.6134
485.4000 8.0000 3.436 5 0.7079
485.4000 8.0000 3.436 0 -1.5558
485.4000 9.0000 3.915 0 -1.6016
485.4000 9.0000 3.915 6 0.8530
May 2013
D-40
Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
2124,
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
.4000
1.
2.
3.
3.
4.
4.
4.
4.
4.
4.
4.
5.
5.
5.
5.
5.
5.
6.
6.
6.
6.
6.
6.
6.
6.
6.
6.
6.
6.
7.
7.
7.
7.
7.
8.
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
,0000
2.820 3
3.645 3
5.230 6
5.230 2
3.282 4
3.282
3.282
3.282
3.282
3.795 1
3.795 5
3.795 5
3.795 4
3.795 4
3.795 3
4.153 5
4.153 6
4.153 6
4.153 3
4.153 6
4.153 2
4.153 4
4.153 0
4.153 5
4.153 0
4.153 5
4.153 4
4.396 5
4.396 5
4.396 7
4.396 6
4.396 7
4.598 0
0.4382
-0.9699
0.7153
-3.0007
0.6473
0.6473
-1.1551
0.6473
0.6473
-0.2539
0.6473
-1.8656
0.8047
0.8047
0.1371
0.1371
-0.5305
0.4466
0.9736
0.9736
-0.6074
0.9736
-1.1344
-0.0804
-2.1885
0.4466
-2.1885
0.4466
-0.0804
0.2650
0.2650
1.1421
0.7036
1.1421
-1.7470
Chi-square = 101.30 DF = 96 P-value = 0.3359
To calculate the BMD and BMDL, the litter specific covariate is fixed
at the mean litter specific covariate of all the data: 5.379518
Benchmark Dose Computation
Specified effect =0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 66.6706
BMDL = 43.1039
May 2013
D-41
Draft - Do Not Cite or Quote
-------
Nested Logistic Model with 0.95 Confidence Level
1
I
.1
o
0.8
0.7
0.6
0.5
0.4
0.3
0.2
10:56 12/16 2011
2000
Source: Rogers et al.(1993bX
Figure D-6 Nested logistic model, 0.05 extra risk - Incidence of cervical rib in mice versus
Cmax above background of methanol, GD6-GD15 inhalational study.
D.4. RfC-Derivations Using Burbacheretal. (1999a; 1999b)
1 The BMD approach was utilized in the derivation of potential chronic inhalation
2 reference values from effects seen in monkeys due to prenatal methanol exposure. Deficits in
3 VDR were evaluated from Burbacher et al. (1999b: 1999a). In the application of the BMD
4 approach, continuous models in EPA's BMDS 2.2 were fit to the data set for increased latency in
5 VDR in neonatal monkeys. The maximum blood methanol values (Cmax) above background
6 estimated using the PK model described in Appendix B were used as the dose metric.
7 The VDR test, which assesses time (from birth) it takes for an infant to grasp for a
8 brightly colored object containing an applesauce-covered nipple, is a measure of sensorimotor
9 development. Beginning at 2 weeks after birth, infants were tested 5 times/day, 4 days/week.
10 Performance on that test, measured as age from birth at achievement of test criterion (successful
11 object retrieval on 8/10 consecutive trials over 2 testing sessions), was reduced in all treated male
12 infants. The times (days after birth) to achieve the criteria for the VDR test in means ± S.D. were
13 23.7 ± 8.3 (n = 3), 32.4 ± 9.2 (n = 5), 42.7 ± 13.9 (n = 3), and 40.5 ± 17.7 (n = 2) days for males
14 and 34.2 ± 4.0 (n = 5), 33.0 ± 5.8 (n = 4), 27.6 ± 6.0 (n = 5), and 40.0 ± 10.6 (n = 7) days for
15 females in the control to 1,800 ppm groups, respectively. As discussed in Section 4.4.2, this type
16 of response data is sometimes adjusted to account for premature births by subtracting time (days)
May 2013
D-42
Draft - Do Not Cite or Quote
-------
1 premature from the time (days from birth) needed to meet the test criteria (Wilson and Cradock,
2 2004). When this type of adjustment is applied, the times (days after birth or, if shorter, days
3 after control mean gestation length) to achieve the criteria for VDR test in means ± S.D. were
4 22.0 ± 16.5 (n = 3), 26.2 ± 19.3 (n = 5), 33.3 ± 17.3 (n = 3), and 39.5 ± 23.1 (n = 2) days for
5 males and 32.0 ± 9.6 (n = 5), 21.8 ± 11.2 (n = 4), 24.0 ± 12.7 (n = 5), and 32.0 ± 39.2 (n =
6 7) days for females in the control to 1,800 ppm groups, respectively. When these data were
7 modeled within BMDS 2.1.1 (U.S. EPA, 2009a), there was no significant difference between
8 unadjusted responses and/or variances among the dose levels (indicating lack of a dose-response)
9 for males and females combined (p = 0.244), for males only (p = 0.321) and for males only with
10 the high-dose group excluded (p = 0.182), or for adjusted responses of males and females
11 combined (p = 0.12), males only (p = 0.448) and males only with the high-dose group excluded
12 (p = 0.586).4 The only data that offered a significant dose-response trend was that for unadjusted
13 (p = 0.0265) and adjusted (p = 0.009) female responses, largely because of the much larger
14 overall sample size across dose groups for females versus males (21 females versus 13 males).
15 However, the model fits for the adjusted female response data were unacceptable. Only the
16 unadjusted female VDR response data (Table D-10) offered both a dose-response trend and
17 acceptable model fits.
Table D-10 EPAPK model estimates of methanol blood levels (Cmax) above background in
monkeys following inhalation exposures and VDR test results for their
offspring
Exposure
concentration (ppm)a
0
206
610
1,822
Blood methanol Cmax above
background (mg/L)b
0
2.87
10.4
38.4
Days After Birth to Achieve VDR
Test Criteria"1
34.2 ±4.0
33.0 ±5.8
27.6 ±6.0
40.0 ±10.6
N
5
4
5
7
"From Burbacher, et al. (199%) and Burbacher, et al. [(1999al Table 2].
bEstimated from the two-compartment PK monkey model described in Appendix B.
"Data reported in means ± standard deviation.
18 The BMD technical guidance (U.S. EPA, 2012a) suggests that in the absence of
19 knowledge as to what level of response to consider adverse, a change in the mean equal to
20 1 control S.D. from the control mean can be used as a BMR for continuous endpoints. A
21 summary of the results most relevant to the development of a POD using the BMD approach
22 (BMD, BMDL, and model fit statistics) for increased latency of VDR in female neonatal
4 BMDS (U.S. EPA. 201^) continuous models contain a test for dose-response trend, test 1, which compares a
model that fits a distinct mean and variance for each dose group to a model that contains a single mean and variance.
The dose response is considered to be significant if this comparison returns ap value < 0.05.
May 2013 D-43 Draft - Do Not Cite or Quote
-------
1 monkeys exposed to methanol with a BMR of 1 control mean S.D. is provided in Table D-l 1.
2 Model fit was determined by statistics (AIC and ^ residuals of individual dose groups) and
3 visual inspection, as recommended by EPA (2012a). The Power model returned a lower AIC than
4 the other models.5 The text and graphic (see Figure D-7) output from this model follows
5 Table D-10. The BMDLiso was determined to be 19.59 mg/L, using the 95% lower confidence
6 limit of the dose-response curve expressed in terms of the ppm of external methanol
7 concentration.
Table D-ll Comparison of BMD modeling results for VDR in female monkeys using AUC
above background of blood methanol as the dose metric
Model
Linear
2nd degree Polynomial
3rd degree Polynomial
Powerb
Hill
Exponential 2
Exponential 3
Exponential 4
Exponential 5
BMD1SD
(Cmax, mg/L)a
38.92
32.27
33.53
37.50
36.90
36.54
37.32
38.92
37.19
BMDL1SD
(Cmax, mg/L)a
15.19
17.59
18.94
19.59
Not Reported
16.22
20.00
15.18
10.81
p-value
0.13
0.2166
0.2646
0.2862
0.1137
0.133
0.1137
0.0433
Not Reported
AIC
110.51
109.49
109.09
108.93
110.93
110.46
110.93
112.51
112.93
Scaled residual11
0.746
0.177
0.0461
7.35E-08
7.65E-07
0.6748
-2.28E-07
0.7457
1.71E-07
aCmax was estimated using the monkey PK model described in Appendix B of the methanol toxicological review; the BMDL is
the 95% lower confidence limit on the AUC of a decrease of 1 control mean S.D. estimated by the model using the likelihood
profile method (U.S. EPA, 2012a).
bModel choice based on adequate/) value (> 0.1), visual inspection, low AIC, and low (absolute) scaled residual.
°AIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the
parameters, and P is the number of modeled degrees of freedom (usually the number of parameters estimated).
^d residual (measure of how model-predicted responses deviate from the actual data) for the dose group closest to the BMD
scaled by an estimate of its S.D. Provides a comparative measure of model fit near the BMD. Residuals that exceed 2.0 in
absolute value should cause one to question model fit in this region.
Source: Burbacher et al. (199%).
5 A detailed analysis of this dose response revealed that modeling results, particularly the BMDL estimation, are
very sensitive to the high-dose response. There is no data to inform the shape of the curve between the mid- and
high-exposure levels, making the derivation of a BMDL very uncertain. The data were analyzed without the high
dose to determine if the downward trend in the low- and mid-exposure groups is significant. It was not, so
nonnegative restriction on the (3 coefficients of the polynomial models was retained.
May 2013
D-44
Draft - Do Not Cite or Quote
-------
1 ====================================================================
2 Power Model. (Version: 2.16; Date: 10/28/2009)
3 Input Data File: C:/USEPA/BMDS220/Data/pow_monkey_Pow-ModelVariance-BMRlStd-
4 Restrict, (d)
5 Gnuplot Plotting File: C:/USEPA/BMDS220/Data/pow_monkey_Pow-ModelVariance-BMRlStd-
6 Restrict.pit
7 Wed Nov 16 11:02:04 2011
8 ====================================================================
9
10 BMDS Model Run
11 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12
13 The form of the response function is:
14
15 Y[dose] = control + slope * dose^power
16
17
18 Dependent variable = Mean
19 Independent variable = Dose
20 The power is restricted to be greater than or egual to 1
21 The variance is to be modeled as Var (i) = exp (lalpha + log (mean (i)) * rho)
22
23 Total number of dose groups = 4
24 Total number of records with missing values = 0
25 Maximum number of iterations = 250
26 Relative Function Convergence has been set to: le-008
27 Parameter Convergence has been set to: le-008
28
29
30
31 Default Initial Parameter Values
32 lalpha = 4.05748
33 rho = 0
34 control =27.6
35 slope = 3.85161
36 power = 0.320501
37
38
39 Asymptotic Correlation Matrix of Parameter Estimates
40
41 ( *** The model parameter (s) -power
42 have been estimated at a boundary point, or have been specified by the user,
43 and do not appear in the correlation matrix )
44
45 lalpha rho control slope
46
47 lalpha 1 -1 -0.29 0.6
48
49 rho -1 1 0.27 -0.6
50
51 control -0.29 0.27 1 -0.37
52
53 slope 0.6 -0.6 -0.37 1
54
55
56
57 Parameter Estimates
58
59 95.0% Wald Confidence Interval
60 Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
61 lalpha -13.0645 12.1112 -36.8021 10.673
62 rho 4.77979 3.44065 -1.96376 11.5233
63 control 31.5 1.4819 28.5955 34.4045
64 slope 2.57903e-028 1.21193e-028 2.03691e-029 4.95437e-028
65 power 18 NA
May 2013 D-45 Draft - Do Not Cite or Quote
-------
1
2 NA - Indicates that this parameter has hit a bound
3 implied by some inequality constraint and thus
4 has no standard error.
5
6
7
8 Table of Data and Estimated Values of Interest
9
10 Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Res.
11
12
13 05 34.2 31.5 4 5.54 1.09
14 2.87 4 33 31.5 5.8 5.54 0.541
15 10.4 5 27.6 31.5 6 5.54 -1.57
16 38.4 7 40 40 10.6 9.81 7.35e-008
17
18
19
20 Model Descriptions for likelihoods calculated
21
22
23 Model Al: Yij = Mu (i) + e (ij)
24 Var {e (ij)} = SigmaA2
25
26 Model A2: Yij = Mu (i) + e (ij)
27 Var {e (ij)} = Sigma (i)A2
28
29 Model A3: Yij = Mu (i) + e (ij)
30 Var {e (ij)} = exp (lalpha + rho*ln (Mu (i)))
31 Model A3 uses any fixed variance parameters that
32 were specified by the user
33
34 Model R: Yi = Mu + e (i)
35 Var {e (i) } = SigmaA2
36
37
38 Likelihoods of Interest
39
40 Model Log (likelihood) # Param's AIC
41 Al -50.884765 5 111.769529
42 A2 -47.717070 8 111.434139
43 A3 -49.215263 6 110.430526
44 fitted -50.466380 4 108.932759
45 R -54.905426 2 113.810852
46
47
48 Explanation of Tests
49
50 Test 1: Do responses and/or variances differ among Dose levels?
51 (A2 vs. R)
52 Test 2: Are Variances Homogeneous? (Al vs A2)
53 Test 3: Are variances adequately modeled? (A2 vs. A3)
54 Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
55 (Note: When rho=0 the results of Test 3 and Test 2 will be the same.)
56
57 Tests of Interest
58
59 Test -2*log (Likelihood Ratio) Test df p-value
60
61 Test 1 14.3767 6 0.0257
62 Test 2 6.33539 3 0.09639
63 Test 3 2.99639 2 0.2235
64 Test 4 2.50223 2 0.2862
65
May 2013 D-46 Draft - Do Not Cite or Quote
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels
It seems appropriate to model the data
The p-value for Test 2 is less than .1. A non-homogeneous variance
model appears to be appropriate
The p-value for Test 3 is greater than .1. The modeled variance appears
to be appropriate here
The p-value for Test 4 is greater than .1. The model chosen seems
to adeguately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 37.4993
BMDL = 19.5918
27
g.
(/}
o>
on
50
45
40
35
30
25
20
Power Model with 0.95 Confidence Level
Power
BMDL
BMD:
10
15
20
dose
25
30
35
40
11:02 11/16 2011
Source: Burbacher et al. (1999b; 1999a)
Figure D-7 Third (3rd) degree Polynomial model, BMR of 1 control mean S.D. - VDR in
female monkeys using Cmax above background of blood methanol as the dose
metric.
May 2013
D-47
Draft - Do Not Cite or Quote
-------
APPENDIX E. DOCUMENTATION OF
IMPLEMENTATION OF THE 2011 NATIONAL
RESEARCH COUNCIL RECOMMENDATIONS
Documentation of the IRIS Program's Implementation of the 2011 NRC Recommendations in the Draft
Toxicological Review of Methanol (May 2013)
1 Background: On December 23, 2011, The Consolidated Appropriations Act, 2012, was signed into law6.
2 The report language included direction to EPA for the IRIS Program related to recommendations
3 provided by the National Research Council (NRC) in their review of EPA's draft IRIS assessment of
4 formaldehyde7. The report language included the following:
5 "The Agency shall incorporate, as appropriate, based on chemical-specific datasets and
6 biological effects, the recommendations of Chapter 7 of the National Research Council's
7 Review of the Environmental Protection Agency's Draft IRIS Assessment of Formaldehyde
8 into the IRIS process...For draft assessments released in fiscal year 2012, the Agency shall
9 include documentation describing how the Chapter 7 recommendations of the National
10 Academy of Sciences (NAS) have been implemented or addressed, including an explanation
11 for why certain recommendations were not incorporated."
12 The NRC's recommendations, provided in Chapter 7 of their review report, offered suggestions to EPA
13 for improving the development of IRIS assessments. Consistent with the direction provided by Congress,
14 documentation of how the recommendations from Chapter 7 of the NRC report have been implemented
15 in this assessment is provided in the table below. Where necessary, the documentation includes an
16 explanation for why certain recommendations were not incorporated.
17 The IRIS Program's implementation of the NRC recommendations is following a phased approach that is
18 consistent with the NRC's "Roadmapfor Revision" as described in Chapter 7 of the formaldehyde review
19 report. The NRC stated that "the committee recognizes that the changes suggested would involve a
20 multi-year process and extensive effort by the staff at the National Center for Environmental
21 Assessment and input and review by the EPA Science Advisory Board and others."
22 The IRIS methanol (noncancer) assessment is in Phase 1 of implementation, which focuses on a subset
23 of the short-term recommendations, such as editing and streamlining documents, increasing
24 transparency and clarity, and using more tables, figures, and appendices to present information and
25 data in assessments. Phase 1 also focuses on assessments near the end of the development process and
26 close to final posting. Chemical assessments in Phase 2 of implementation will address all of the short-
27 term recommendations from Table E-l. Chemical assessments in Phase 3 of implementation will
28 incorporate the longer-term recommendations made by the NRC as outlined below in Table E-2,
29 including the development of a standardized approach to describe the strength of evidence for
30 noncancer effects . On May 16, 2012, EPA announced8 that as a part of a review of the IRIS Program's
31 assessment development process, the NRC will also review current methods for weight-of-evidence
32 analyses and recommend approaches for weighing scientific evidence for chemical hazard identification.
33 This effort is included in Phase 3 of EPA's implementation plan.
34
6Pub. L No. 112-74, Consolidated Appropriations Act, 2012.
7National Research Council, 2011. Review of the Environmental Protection Agency's Draft IRIS Assessment of
Formaldehyde.
8EPA Announces NAS' Review of IRIS Assessment Development Process (www.epa.gov/iris)
May 2013 E-1 Draft - Do Not Cite or Quote
-------
Table E-l National Research Council recommendations that EPA is implementing in the
short term
NATIONAL RESEARCH COUNCIL (NRC)
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE SHORT TERM
IMPLEMENTATION STATUS
General recommendations for completing the IRIS formaldehyde assessment that EPA will adopt
for all IRIS assessments (p. 152 of the NRC report)
1. To enhance the clarity of the document, the
draft IRIS assessment needs rigorous editing to
reduce the volume of text substantially and
address redundancies and inconsistencies. Long
descriptions of particular studies should be
replaced with informative evidence tables. When
study details are appropriate, they could be
provided in appendices.
Partially Implemented. Methanol is a post-peer review,
Phase 1 chemical; as such, implementation has focused
on a subset of the short-term recommendations, such as
editing and streamlining, increasing transparency and
clarity, and using more tables, figures, and appendices to
present information and data. For example:
• details of EPA PBPK models were moved from
Chapter 3 to Appendix B,
• descriptions of human case studies were moved
to Appendix C,
• tables were added to Chapter 4, replacing
textual descriptions, and
• details of benchmark dose analyses were moved
to Appendix D.
2. Chapter 1 needs to be expanded to describe
more fully the methods of the assessment,
including a description of search strategies used
to identify studies with the exclusion and
inclusion criteria articulated and a better
description of the outcomes of the searches and
clear descriptions of the weight-of-evidence
approaches used for the various noncancer
outcomes. The committee emphasizes that it is
not recommending the addition of long
descriptions of EPA guidelines to the
introduction, but rather clear concise statements
of criteria used to exclude, include, and advance
studies for derivation of the RfCs and unit risk
estimates.
Partially Implemented. Text in Chapter 1 has been added
that describes the literature search and study evaluation
process in greater detail. This section also provides a link
to EPA's Health and Environmental Research Online
(HERO) database (www.epa.gov/hero) that contains the
references that were cited in the document, along with
those that were considered but not cited. As indicated in
the comment for recommendation #1, methanol is a post-
peer review, phase 1 chemical. Consequently, literature
search and study evaluation processes were not
substantially revised.
May 2013
E-2
Draft - Do Not Cite or Quote
-------
NATIONAL RESEARCH COUNCIL (NRC)
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE SHORT TERM
IMPLEMENTATION STATUS
3. Standardized evidence tables for all health
outcomes need to be developed. If there were
appropriates tables, long text descriptions of
studies could be moved to an appendix of
deleted.
Partially Implemented. The methanol (noncancer)
assessment contains evidence tables for relevant study
types, including oral, inhalation, i.p., in vitro study
designs. Additional tables with study specific health
outcomes have been added to Chapter 4 in response to
this recommendation. However, because methanol is a
post-peer review, phase 1 chemical, the format of the
existing tables was not substantially changed.
4. All critical studies need to be thoroughly
evaluated with standardized approaches that are
clearly formulated and based on the type of
research, for example, observational
epidemiologic or animal bioassays. The findings
of the reviews might be presented in tables to
ensure transparency.
Implemented. All critical studies are thoroughly
evaluated; Study design, results, and limitations are
described in Chapter 4, and the basis for their selection,
along with uncertainties, are discussed in Chapter 5.
5. The rationales for the selection of the studies
that are advanced for consideration in
calculating the RfCs and unit risks need to be
expanded. All candidate RfCs should be
evaluated together with the aid of graphic
displays that incorporate selected information
on attributes relevant to the database.
Implemented. The Dose-Response Analysis section of the
methanol (noncancer) assessment provides a clear
explanation of the rationale used and uncertainties
considered in selecting and advancing studies that were
considered for calculating toxicity values. Rationales for
the selection of studies advanced for reference value
derivation are informed by the weight-of-evidence for
hazard identification. In support of the RfC and RfD
derivations, exposure-response arrays were included that
compare effect levels for several toxicological effects
following oral (Figure 4-1) and inhalation (Figure 4-2)
exposure. The exposure-response arrays provide a visual
representation of points of departure for various effects
resulting from exposure to methanol. The arrays inform
the identification of doses associated with specific effects,
and the choice of principal studies and critical effects. In
the case of methanol, the database supported
development of multiple candidate RfCs and RfDs. The
candidate RfCs and RfDs are presented in Tables 5-1, 5-3
and 5-4. Uncertainties with the RfD and RfC derivations
are summarized in Table 5-6.
May 2013
E-3
Draft - Do Not Cite or Quote
-------
NATIONAL RESEARCH COUNCIL (NRC)
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE SHORT TERM
6. Strengthened, more integrative and more
transparent discussions of weight-of-evidence
are needed. The discussions would benefit from
more rigorous and systematic coverage of the
various determinants of weight-of-evidence,
such as consistency.
IMPLEMENTATION STATUS
Partially implemented. Weight-of-evidence
considerations were added or revised based on peer
review comments (see Appendix A). Table 5-6 summarizes
considerations and uncertainties in the assessment and
their potential impact on the RfC/RfD. Additional
discussion of approaches to ensure systematic coverage
of the various determinants of weight-of-evidence will be
added to phase 2 chemicals, but were not added to the
methanol (noncancer) assessment because it is a post-
peer review phase 1 chemical.
Other specific recommendations
General Guidance for the Overall Process (p. 164)
7. Elaborate an overall, documented, and quality-
controlled process for IRIS assessments.
8. Ensure standardization of review and evaluation
approaches among contributors and teams of
contributors; for example, include standard
approaches for reviews of various types of
studies to ensure uniformity.
9. Assess disciplinary structure of teams needed to
conduct the assessments.
Partially Implemented. A team approach has been
utilized for the development of the methanol (noncancer)
assessment to help ensure that the necessary disciplinary
expertise is available for assessment development and
review, to provide a forum for identifying and addressing
key issues. Because methanol is a post-peer review, phase
one chemical, the methanol team did not have access to
the "overall, documented, and quality-controlled process"
that is now being developed in response to the NRC
recommendations.
May 2013
E-4
Draft - Do Not Cite or Quote
-------
NATIONAL RESEARCH COUNCIL (NRC)
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE SHORT TERM
IMPLEMENTATION STATUS
Evidence Identification: Literature Collection and
Collation Phase (p. 164)
10. Select outcomes on the basis of available
evidence and understanding of mode of action.
11. Establish standard protocols for evidence
identification.
12. Develop a template for description of the search
approach.
13. Use a database, such as the Health and
Environmental Research Online (HERO)
database, to capture study information and
relevant quantitative data.
Partially Implemented. More detailed information on the
literature search strategy used for the methanol
(noncancer) assessment has been added to Chapter 1.
Information on how studies were selected to be included
in the document is presented, along with a link to an
external database (www.epa.gov/hero) that contains the
references that were cited in the document, along with
those that were considered but not cited. Each citation in
the Toxicological Review is linked to HERO such that the
public can access the references and abstracts to the
scientific studies used in the assessment.
Outcomes have been selected on the basis of available
evidence and understanding mode of action in
accordance EPA guidelines (U.S. EPA, 2002,1994b).
Uncertainties associated with the available evidence are
described in Section 5.3. Available evidence played an
important role in the selection of candidate studies and
endpoints. For example, questions concerning the
Burbacher et al. (2004a; 1999a) monkey study endpoint
and dose-response are considered serious enough to
preclude its use for RfC/D derivation, despite the
possibility that a lower BMDL POD would have been
derived (Section 5.3.1 and Appendix D).
The study evaluation processes was not revised because
methanol is a phase 1 chemical. The methanol
(noncancer) assessment was developed using standard
protocols for evidence identification that are provided in
existing EPA guidance.
Evidence Evaluation: Hazard Identification and Dose-
Response Modeling (p. 165)
14. Standardize the presentation of reviewed studies
in tabular or graphic form to capture the key
dimensions of study characteristics, weight-of-
evidence, and utility as a basis for deriving
reference values and unit risks.
Partially Implemented. Standardized tables have been
developed that provide summaries of key study design
information and results by health effect. In addition,
exposure-response arrays are utilized in the assessment
to provide a graphical representation of points of
departure for various effects resulting from exposure to
TMB. The exposure-response arrays inform the
identification of doses associated with specific effects and
the weight-of- evidence for those effects.
15. Develop templates for evidence tables, forest
plots, or other displays.
Partially Implemented. Graphic displays in the form of
exposure arrays have been added to Chapter 4 (Figures 4-
1 and 4-2).
May 2013
E-5
Draft - Do Not Cite or Quote
-------
NATIONAL RESEARCH COUNCIL (NRC)
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE SHORT TERM
IMPLEMENTATION STATUS
16. Establish protocols for review of major types of
studies, such as epidemiologic and bioassay.
Partially Implemented. The study evaluation processes
was not revised because methanol is a phase 1 chemical.
However, the methanol (noncancer) assessment was
developed using standard protocols for evidence
identification that are provided in existing EPA guidance.
Selection of Studies for Derivation of Reference
Values and Unit Risks (p. 165)
17. Establish clear guidelines for study selection.
a. Balance strengths and weaknesses.
b. Weigh human vs. experimental evidence.
c. Determine whether combining estimates
among studies is warranted.
Partially Implemented. The basis for study selection is
described in Sections 5.1.1 (RfC) and 5.2.1 (RfD). Existing
EPA guidelines for study selection were applied to inform
the evaluation of the weight-of-evidence across health
effects and the strengths and weaknesses of individual
studies. Sections 5.1.2, 5.1.3, and 5.2.2 discuss
uncertainties that are addressed quantitatively via
uncertainty factors.
Section 5.3 provides an additional discussion of the
uncertainties associated with the RfC and RfD derivation.
A summary of these uncertainties is presented in Table 5-
6. Section 5.3.1 specifically addresses the uncertainties
associated with the choice of study and endpoint. Other
aspects besides the choice of study and endpoint that can
impact RfC/D derivation that are discussed include dose-
response modeling (5.3.2), route-to-route extrapolation
(5.3.3), statistical uncertainty at the POD (5.3.4), choice of
species and gender (5.3.5) and the relationship of the RfC
and RfD with endogenous methanol blood Levels (5.3.6).
In the case of methanol, the database did not support the
combination of estimates across studies. In future
assessments, combining estimates across studies will be
routinely considered.
Calculation of Reference Values and Unit Risks (pp.
165-166)
18. Describe and justify assumptions and models
used. This step includes review of dosimetry
models and the implications of the models for
uncertainty factors; determination of
appropriate points of departure (such as
benchmark dose, no-observed-adverse-effect
level, and lowest observed-adverse-effect level),
and assessment of the analyses that underlie the
points of departure.
Implemented. Appendix B documents EPA's PBPK model.
Appendix D documents the benchmark dose modeling
analyses used to derive candidate points of departure.
The implications of the models for uncertainty factors are
described in Sections 5.1.3 and 5.2.2, and the impact of
model choices are further described in section 5.3.
May 2013
E-6
Draft - Do Not Cite or Quote
-------
NATIONAL RESEARCH COUNCIL (NRC)
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE SHORT TERM
IMPLEMENTATION STATUS
19. Provide explanation of the risk-estimation
modeling processes (for example, a statistical or
biologic model fit to the data) that are used to
develop a unit risk estimate.
Not applicable. A cancer unit risk estimate was not
derived in this assessment.
20. Provide adequate documentation for conclusions
and estimation of reference values and unit risks.
As noted by the committee throughout the
present report, sufficient support for conclusions
in the formaldehyde draft IRIS assessment is
often lacking. Given that the development of
specific IRIS assessments and their conclusions
are of interest to many stakeholders, it is
important that they provide sufficient references
and supporting documentation for their
conclusions. Detailed appendixes, which might
be made available only electronically, should be
provided when appropriate.
Implemented. Chapter 5 documents the approach taken
for the estimation of reference values and provides
support for the conclusions drawn. As recommended,
supplementary information is provided in the
accompanying appendices. Appendix D documents the
benchmark dose modeling analyses used to derive
candidate points of departure.
May 2013
E-7
Draft - Do Not Cite or Quote
-------
Table E-2 National Research Council recommendations that EPA is implementing in the
long-term
NATIONAL RESEARCH COUNCIL
RECOMMENDATIONS THAT EPA IS
IMPLEMENTING IN THE LONG-TERM
IMPLEMENTATION STATUS
Weight-of-Evidence Evaluation: Synthesis of Evidence for
Hazard Identification (p. 165)
1. Review use of existing weight-of-evidence
guidelines.
2. Standardize approach to using weight-of-evidence
guidelines.
3. Conduct agency workshops on approaches to
implementing weight-of-evidence guidelines.
4. Develop uniform language to describe strength of
evidence on noncancer effects.
5. Expand and harmonize the approach for
characterizing uncertainty and variability.
6. To the extent possible, unify consideration of
outcomes around common modes of action rather
than considering multiple outcomes separately.
As indicated above, Phase 3 of EPA's implementation
plan will incorporate the longer-term
recommendations made by the NRC, including the
development of a standardized approach to describe
the strength of evidence for noncancer effects. On
May 16, 2012, EPA announced9 that as a part of a
review of the IRIS Program's assessment
development process, the NRC will also review
current methods for weight-of-evidence analyses and
recommend approaches for weighing scientific
evidence for chemical hazard identification. In
addition, EPA may hold additional workshops on
issues related to weight-of-evidence to inform future
assessments.
Calculation of Reference Values and Unit Risks (pp. 165-
166)
7. Assess the sensitivity of derived estimates to model
assumptions and end points selected. This step
should include appropriate tabular and graphic
displays to illustrate the range of the estimates and
the effect of uncertainty factors on the estimates.
Partially Implemented. Chapter 5 describes the
derivation of candidate RfCsand RfDsfrom data for
multiple endpoints in multiple species. In addition, a
sensitivity analysis on model parameters used in the
rat and human PBPK models has been conducted and
results are tabulated in Appendix B, Sections B.2.4
and B.2.7. However, such analyses can only partly
inform the question of model adequacy, which is
addressed in more detail in the response to Charge
Al Comment 1 of Appendix A.
EPA Announces NAS' Review of IRIS Assessment Development Process (www.epa.gov/iris)
May 2013
E-8
Draft - Do Not Cite or Quote
-------
REFERENCES
Adanir. J: Ozkalkanti. MY: Aksun. M. (2005). Percutaneous methanol intoxication: Case report [Abstract]. Eur
JAnaesthesiol22: 560-561.
Airas. L: Paavilainen. T: Marttila. RJ: Rinne. J. (2008). Methanol intoxication-induced nigrostriatal
dysfunction detected using 6-[18F]fluoro-L-dopa PET. Neurotoxicology 29: 671-674.
http://dx.doi.0rg/10.1016/j.neuro.2008.03.010
Allen. BC: Kavlock. RJ: Kimmel CA: Faustman. EM. (1994a). Dose-response assessment for developmental
toxicity II: Comparison of generic benchmark dose estimates with no observed adverse effect levels.
Fundam Appl Toxicol 23: 487-495. http://dx.doi.org/10.1006/faat. 1994.1133
Allen. BC: Kavlock. RJ: Kimmel CA: Faustman. EM. (1994b). Dose-response assessment for developmental
toxicity III: statistical models. Fundam Appl Toxicol 23: 496-509. http://dx.doi.org/10.1006/faat. 1994.1134
Andresen. H: Schmoldt H: Matschke. J: Flachskampfc. FA: Ee. T. (2008). Fatal methanol intoxication with
different survival times -morphological findings and postmortem methanol distribution. Forensic Sci Int
179: 206-210. http://dx.doi.0rg/10.1016/j.forsciint.2008.05.014
Andrews. JE: Ebron-Mccov. M: Logsdon. TR: Mole. LM: Kavlock. RJ: Rogers. JM. (1993). Developmental
toxicity of methanol in whole embryo culture: a comparative study with mouse and rat embryos.
Toxicology 81:205-215.
Arora. V: Nijjar. IBS: Multani. AS: Singh. JP: Abrol R: Chopra. R: Attri. R. (2007). MRI findings in methanol
intoxication: a report of two cases. Br J Radiol 80: 243-246. http://dx.doi.org/10.1259/bjr/40137535
Arora. V: Nijjar. IS: Thukral H: Roopa. (2005). Bilateral putaminal necrosis caused by methanol intoxication-
a case report. Neuroradiology 15: 341-342.
Aufderheide. TP: White. SM: Brady. WJ: Stueven. HA. (1993). Inhalational and percutaneous methanol
toxicity in two firefighters. Ann Emerg Med 22: 1916-1918.
Azmak. D. (2006). Methanol related deaths in Edirne. Leg Med 8: 39-42.
http://dx.doi.0rg/10.1016/j.legalmed.2005.07.002
Barceloux. DG: Bond. GR: Krenzelok. EP: Cooper. H: Vale. JA. (2002). American academy of clinical
toxicology practice guidelines on the treatment of methanol poisoning [Review]. Clin Toxicol 40: 415-446.
Batterman. SA: Franzblau. A: D'Arcv. JB: Sargent NE: Gross. KB: Schreck. RM. (1998). Breath, urine, and
blood measurements as biological exposure indices of short-term inhalation exposure to methanol. Int Arch
Occup Environ Health 71: 325-335.
Bebarta. VS: Heard. K: Dart. RC. (2006). Inhalational abuse of methanol products: elevated methanol and
formate levels without vision loss. Am J Emerg Med 24: 725-728.
http://dx.doi.0rg/10.1016/j.aiem.2006.03.004
Bennett. IL. Jr: Gary. FH: Mitchell. GL. Jr: Cooper. MN. (1953). Acute methyl alcohol poisoning: A review
based on experiences in an outbreak of 323 cases [Review]. Medicine (Baltimore) 32: 431-463.
Bessell-Browne. RJ: Bynevelt M. (2007). Two cases of methanol poisoning: CT and MRI features.
Australasian Radiology 51: 175-178. http://dx.doi.0rg/10.llll/i.1440-1673.2007.01691.x
Bhatia. KP: Marsden. CD. (1994). The behavioural and motor consequences of focal lesions of the basal
ganglia in man. Brain 117: 859-876. http://dx.doi.org/10.1093/brain/l 17.4.859
Blanco. M: Casado. R: Vazquez. F: Pumar. JM. (2006). CT and MR imaging findings in methanol intoxication.
AJNR Am J Neuroradiol 27: 452-454.
Bolon. B: Dorman. DC: Janszen. D: Morgan. KT: Welsch. F. (1993). Phase-specific developmental toxicity in
mice following maternal methanol inhalation. Toxicol Sci 21: 508-516.
May 2013 R-1 Draft - Do Not Cite or Quote
-------
Bolon. B: Welsch. F: Morgan. KT. (1994). Methanol-induced neural tube defects in mice: Pathogenesis during
neurulation. Teratology 49: 497-517. http://dx.doi.org/10.1002/tera.1420490610
Bouchard. M: Brunei RC: Droz. PO: Carrier. G. (2001). A biologically based dynamic model for predicting
the disposition of methanol and its metabolites in animals and humans. Toxicol Sci 64: 169-184.
Braden. GL: Strayhorn. CH: Germain. MJ: Mulhern. JG: Skutches. CL. (1993). Increased osmolal gap in
alcoholic acidosis. Arch Intern Med 153: 2377-2380.
Brahmj N: Blel. Y: Abidi N: Kouraichi. N: Thabet H: Hedhili. A: Amamou. M. (2007). Methanol poisoning
in Tunisia: Report of 16 cases. Clin Toxicol 45: 717-720. http://dx.doi.org/10.1080/15563650701502600
Brent. J: Lucas. M: Kulig. K: Rumack. BH. (1991). Methanol poisoning in a six-week-old infant. J Paediatr
Child Health 118: 644-646.
Brown. RP: Delp. MD: Lindstedt SL: Rhomberg. LR: Beliles. RP. (1997). Physiological parameter values for
physiologically based pharmacokinetic models [Review]. Toxicol Ind Health 13: 407-484.
http://dx.doi.org/10.1177/074823379701300401
Burbacher. TM: Grant K: Shen. D: Damian. D: Ellis. S: Liberate. N. (1999a). Reproductive and offspring
developmental effects following maternal inhalation exposure to methanol in nonhuman primates Part II:
developmental effects in infants exposed prenatally to methanol. Cambridge, MA: Health Effects Institute.
Burbacher. TM: Grant KS: Shen. DP: Sheppard. L: Damian. D: Ellis. S: Liberate. N. (2004a). Chronic
maternal methanol inhalation in nonhuman primates (Macaca fascicularis): reproductive performance and
birth outcome. Neurotoxicol Teratol 26: 639-650. http://dx.doi.0rg/10.1016/i.ntt.2004.06.001
Burbacher. TM: Shen. D: Grant. K: Sheppard. L: Damian. D: Ellis. S: Liberate. N. (1999b). Reproductive and
offspring developmental effects following maternal inhalation exposure to methanol in nonhuman primates
Part I: methanol disposition and reproductive toxicity in adult females. Cambridge, MA: Health Effects
Institute.
Burbacher. TM: Shen. DP: Lalovic. B: Grant KS: Sheppard. L: Damian. D: Ellis. S: Liberate. N. (2004b).
Chronic maternal methanol inhalation in nonhuman primates (Macaca fascicularis): exposure and
toxicokinetics prior to and during pregnancy. Neurotoxicol Teratol 26: 201-221.
http://dx.doi.0rg/10.1016/j.ntt.2003.10.003
CERHR (NTP Center for the Evaluation of Risks to Human Reproduction). (2003). NTP-CERHR monograph
on the potential human reproductive and developmental effects of methanol.
CERHR (NTP Center for the Evaluation of Risks to Human Reproduction). (2004). NTP-CERHR expert panel
report on the reproductive and developmental toxicity of methanol [Review]. Reprod Toxicol 18: 303-390.
http://dx.doi.org/10.1016/j.reprotox.20Q3.10.013
Chen. JC: Schneiderman. JF: Wortzman. G. (1991). Methanol poisoning: Bilateral putaminal and cerebellar
cortical lesions on CT and MR. J Comput Assist Tomogr 15: 522-524.
Chiu. WA: Barton. HA: Dewoskin. RS: Schlosser. P: Thompson. CM: Sonawane. B: Lipscomb. JC: Krishnan.
K. (2007). Evaluation of physiologically based pharmacokinetic models for use in risk assessment
[Review]. J Appl Toxicol 27: 218 - 237. http://dx.doi.org/10.1002/iat. 1225
Chiu. WA: White. P. (2006). Steady-state solutions to PBPK models and their applications to risk assessment I:
Route-to-route extrapolation of volatile chemicals. Risk Anal 26: 769-780. http://dx.doi.org/10. Ill 1/i. 1539-
6924.2006.00762.x
Clary. JJ. (2003). Methanol, is it a developmental risk to humans? [Review]. Regul Toxicol Pharmacol 37: 83-
91.http://dx.doi.org/10.1016/S0273-2300(02)00031-4
Clewell HJ. Ill: Gentry. PR: Gearhart. JM: Covington. TR: Banton. MI: Andersen. ME. (2001). Development
of a physiologically based pharmacokinetic model of isopropanol and its metabolite acetone. Toxicol Sci
63: 160-172.
Corlev. RA: Bormett GA: Ghanavem. BI. (1994). Physiologically-based pharmacokinetics of 2-butoxyethanol
and its major metabolite 2-butoxyacetic acid, in rats and humans. Toxicol Appl Pharmacol 129: 61-79.
http://dx.doi.org/10.1006/taap.1994.1229
May 2013 R-2 Draft - Do Not Cite or Quote
-------
De Brabander. N: Wojciechowskj M: De Decker. K: De Weerdt A: Jorens. PG. (2005). Fomepizole as a
therapeutic strategy in paediatric methanol poisoning: a case report and review of the literature [Review].
Eur JPediatr 164: 158-161. http://dx.doi.org/10.1007/s00431-004-1588-5
Degitz. SJ: Rogers. JM: Zucker. RM: Hunter. ES. III. (2004a). Developmental toxicity of methanol:
pathogenesis in CD-I and C57BL/6J mice exposed in whole embryo culture. Birth Defects Res A Clin Mol
Teratol 70: 179-184. http://dx.doi.org/10.1002/bdra.20009
Degitz. SJ: Zucker. RM: Kawanishi. CY: Massenburg. GS: Rogers. JM. (2004b). Pathogenesis of methanol-
induced craniofacial defects in C57BL/6J mice. Birth Defects Res A Clin Mol Teratol 70: 172-178.
http://dx.doi.org/10.1002/bdra.20010
Dethlefs. R: Naraqi S. (1978). Ocular manifestations and complications of acute methyl alcohol intoxication.
MedJAust 2: 483-485.
Dorman. DC: Bolon. B: Starve. MF: Laperle. KMD: Wong. BA: Elswick. B: Welsch. F. (1995). Role of
formate in methanol-induced exencephaly in CD-I mice. Teratology 52: 30-40.
http://dx.doi.org/10.1002/tera.1420520105
Dorman. DC: Welsch. F. (1996). Developmental toxicity of methanol in rodents. CUT Activities 16: 40360.
Dourson. ML: Stara. JF. (1983). Regulatory history and experimental support of uncertainty (safety) factors.
Regul Toxicol Pharmacol 3: 224-238.
Downie. A: Khattab. TM: Malik. MI: Samara. IN. (1992). A case of percutaneous industrial methanol toxicity.
Occup Med (Lond) 42: 47-49.
Dutkiewicz. B: Konczalik. J: Karwacki. W. (1980). Skin absorption and per os administration of methanol in
men. Int Arch Occup Environ Health 47: 81-88.
ERG (Eastern Research Group Inc.). (2009). External letter peer review of reports documenting methanol
studies in monkeys, rats and mice performed by the New Energy Development Organization (NEDO).
Lexington, MA.
Ernstgard. L. (2005). E-mail correspondance from Lena Ernstgard to Torka Poet. Available online
Ernstgard. L: Shibata. E: Johanson. G. (2005). Uptake and disposition of inhaled methanol vapor in humans.
Toxicol Sci 88: 30-38. http://dx.doi.org/10.1093/toxsci/kfi281
Faustman. EM: Allen. BC: Kavlock. RJ: Kimmel CA. (1994). Dose-response assessment for developmental
toxicity: I characterization of data base and determination of no observed adverse effect levels. Fundam
Appl Toxicol 23: 478-486.
Feany. MB: Anthony. DC: Frosch. MP: Zane. W: De Girolami. U. (2001). August 2000: two cases with
necrosis and hemorrhage in the putamen and white matter. Brain Pathol 11: 125.
Finkelstein. Y: Vardi J. (2002). Progressive parkinsonism in a young experimental physicist following long-
term exposure to methanol. Neurotoxicology 23: 521-525.
Fiserova-Bergerova. V: Diaz. ML. (1986). Determination and prediction of tissue-gas partition coefficients. Int
Arch Occup Environ Health 58: 75-87. http://dx.doi.org/10.1007/BF00378543
Fisher. JW: Dorman. DC: Medinsky. MA: Welsch. F: Conolly. RB. (2000). Analysis of respiratory exchange of
methanol in the lung of the monkey using a physiological model. Toxicol Sci 53: 185-193.
Fontenot AP: Pelak. VS. (2002). Development of neurologic symptoms in a 26-year-old woman following
recovery from methanol intoxication. Chest 122: 1436-1439.
Fu. SS: Sakanashj TM: Rogers. JM: Hong. (1996). Influence of dietary folic acid on the developmental
toxicity of methanol and the frequency of chromosomal breakage in the CD-I mouse. Reprod Toxicol 10:
455-463.
Gaul HP: Wallace. CJ: Auer. RN: Chen Fong. T. (1995). MRI findings in methanol intoxication. AJNR Am J
Neuroradiol 16: 1783-1786.
May 2013 R-3 Draft - Do Not Cite or Quote
-------
Gentry. PR: Covington. TR: Andersen. ME: Clewell HJ. III. (2002). Application of a physiologically based
pharmacokinetic model for isopropanol in the derivation of a reference dose and reference concentration
[Review]. Regul Toxicol Pharmacol 36: 51-68.
Gentry. PR: Covington. TR: Clewell HJ. III. (2003). Evaluation of the potential impact of pharmacokinetic
differences on tissue dosimetry in offspring during pregnancy and lactation. Regul Toxicol Pharmacol 38:
1-16. http://dx.doi.org/10.1016/S0273-2300(03)00047-3
Guggenheim. MA: Couch. JR: Weinberg. W. (1971). Motor dysfunction as a permanent complication of
methanol ingestion. Arch Neurol 24: 550-554.
Haffner. HT: Wehner. HP: Schevtt KD: Besserer. K. (1992). The elimination kinetics of methanol and the
influence of ethanol. Int J Legal Med 105: 111-114.
Hantson. P: Duprez. T: Mahieu. P. (1997a). Neurotoxicity to the basal ganglia shown by magnetic resonance
imaging (MRI) following poisoning by methanol and other substances. J Toxicol Clin Toxicol 35: 151-161.
Hantson. P: Lambermont JY: Mahieu. P. (1997b). Methanol poisoning during late pregnancy. Clin Toxicol 35:
187-191.
Hassanian-Moghaddam. H: Paioumand. A: Dadgar. SM: Shadnia. SH. (2007). Prognostic factors in methanol
poisoning. HumExp Toxicol 26: 583-586. http://dx.doi.org/10.1177/0960327106080077
Henderson. WR: Brubacher. J. (2002). Methanol and ethylene glycol poisoning: A case study and review of
current literature [Review]. C J E M 4: 34-40.
Horton. VL: Higuchi. MA: Rickert DE. (1992). Physiologically based pharmacokinetic model for methanol in
rats, monkeys, and humans. Toxicol Appl Pharmacol 117: 26-36.
Hovda. KE: Hunderi. OH: Tafjord. AB: Dunlop. O: Rudberg. N: Jacobsen. D. (2005). Methanol outbreak in
Norway 2002-2004: epidemiology, clinical features and prognostic signs. J Intern Med 258: 181-190.
http://dx.doi.0rg/10.llll/i.1365-2796.2005.01521.x
Hovda. KE: Mundal H: Urdal P: Mcmartin. K: Jacobsen. D. (2007). Extremely slow formate elimination in
severe methanol poisoning: A fatal case report. Clin Toxicol 45: 516-521.
http://dx.doi.org/10.1080/15563650701354150
Hsu. HH: Chen. CY: Chen. FH: Lee. CC: Chou. TY: Zimmerman. RA. (1997). Optic atrophy and cerebral
infarcts caused by methanol intoxication: MRI. Neuroradiology 39: 192-194.
http://dx.doi.org/10.1007/s002340050391
Hunderi. OH: Hovda. KE: Jacobsen. D. (2006). Use of the osmolal gap to guide the start and duration of
dialysis in methanol poisoning. Scand J Urol Nephrol 40: 70-74.
http://dx.doi.org/10.1080/00365590500190755
Johanson. G: Kronborg. H: Naslund. PH: Nordqvist MB. (1986). Toxicokinetics of inhaled 2-butoxyethanol
(ethylene glycol monobutyl ether) in man. Scand J Work Environ Health 12: 594-602.
Kahn. A: Blum. D. (1979). Methyl alcohol poisoning in an 8-month-old boy: An unusual route of intoxication.
JPediatr94:841-843.
Kavet R: Nauss. KM. (1990). Thetoxicity of inhaled methanol vapors [Review]. Crit Rev Toxicol 21: 21-50.
http://dx.doi.org/10.3109/10408449009089872
Kavlock. RJ: Allen. BC: Faustman. EM: Kimmel CA. (1995). Dose-response assessments for developmental
toxicity. IV. Benchmark doses for fetal weight changes. Toxicol Sci 26: 211-222.
Keles. GT: Orguc. S: Toprak. B: Ozaslan. S: Sakarya. M. (2007). Methanol poisoning with necrosis corpus
callosum. Clin Toxicol 45: 307-308.
Kuteifan. K: Oesterle. H: Taiahmadv. T: Gutbub. AM: Laplatte. G. (1998). Necrosis and haemorrhage of the
putamen in methanol poisoning shown on MRI. Neuroradiology 40: 158-160.
Leavens. TL: Parkinson. CU: James. RA: House. D: Elswick. B: Dorman. DC. (2006). Respiration in Sprague-
Dawley rats during pregnancy. Inhal Toxicol 18: 305-312. http://dx.doi.org/10.1080/08958370500444361
Lev. CO: Gall FG. (1983). Parkinsonian syndrome after methanol intoxication. Eur Neurol 22: 405-409.
May 2013 R-4 Draft - Do Not Cite or Quote
-------
Liu. JJ: Dava. MR: Carrasquillo. O: Kales. SN. (1998). Prognostic factors in patients with methanol poisoning.
ClinToxicol36: 175-181.
Loccisano. AE: Campbell JL. Jr: Andersen. ME: Clewell HJ. III. (2011). Evaluation and prediction of
pharmacokinetics of PFOA and PFOS in the monkey and human using a PBPK model. Regul Toxicol
Pharmacol 59: 157-175. http://dx.doi.0rg/10.1016/j.vrtph.2010.12.004
Mahieu. P: Hassoun. A: Lauwerys. R. (1989). Predictors of methanol intoxication with unfavourable outcome.
Hum Toxicol 8: 135-137.
Mani. J: Pietruszko. R: Theorell H. (1970). Methanol activity of alcohol dehydrogenases from human liver,
horse liver, and yeast. Arch Biochem Biophys 140: 52-59. http://dx.doi.org/10.1016/0003-9861(70)90009-3
Meyer. RJ: Beard. ME: Ardagh. MW: Henderson. S. (2000). Methanol poisoning. N Z Med 1113:11-13.
Miller. L: Wells. PG. (2011). Altered methanol embryopathies in embryo culture with mutant catalase-deficient
mice and transgenic mice expressing human catalase. Toxicol Appl Pharmacol 252: 55-61.
http://dx.doi.0rg/10.1016/i.taap.2011.01.019
Moonev. SM: Miller. MW. (2001). Episodic exposure to ethanol during development differentially affects
brainstem nuclei in the macaque. J Neurocytol 30: 973-982. http://dx.doi.Org/10.1023/A:1021832522701
Naraqi S: Dethlefs. RF: Slobodniuk. RA: Sairere. JS. (1979). An outbreak of acute methyl alcohol
intoxication. Aust N Z J Med 9: 65-68. http://dx.doi.0rg/10.llll/i.1445-5994.1979.tb04116.x
NEDO (New Energy Development Organization). (1985a). Test report: 18-month inhalation carcinogenicity
study on methanol inB6C3Fl mice (test no. 4A-223). Tokyo, Japan: Mitsubishi Kasei Institute of
Toxicology and Environmental Sciences.
NEDO (New Energy Development Organization). (1985b). Test report: 24-month inhalation carcinogenicity
study on methanol in Fischer rats (Test No. 5A-268). Toyko, Japan Mitsubishi Kasei Institute of Toxicology
and Environmental Sciences.
NEDO (New Energy Development Organization). (1987). Toxicological research of methanol as a fuel for
power station: summary report on tests with monkeys, rats and mice. Tokyo, Japan.
NRC (National Research Council). (2011). Review of the Environmental Protection Agency's draft IRIS
assessment of formaldehyde. Washington, DC: National Academies Press.
http://www.nap.edu/catalog/13142.html
OECD (Organisation for Economic Co-operation and Development). (2001). OECD guideline for testing of
chemicals: Two-generation reproduction toxicity study.
Osterloh. JD: D'Alessandro. A: Chuwers. P: Mogadeddi. H: Kelly. TJ. (1996). Serum concentrations of
methanol after inhalation at 200 ppm. J Occup Environ Med 38: 571-576.
Pasting. GM: Conolly. RB. (2000). Application of a physiologically based pharmacokinetic model to estimate
the bioavailability of ethanol in male rats: distinction between gastric and hepatic pathways of metabolic
clearance. Toxicol Sci 55: 256-265.
Patankar. T: Bichile. L: Karnad. D: Prasad. S: Rathod. K. (1999). Methanol poisoning: Brain computed
tomography scan findings in four patients. Australasian Radiology 43: 526-528.
http://dx.doi.0rg/10.1046/i.1440-1673.1999.00723.x
Pelletier. J: Habib. MH: Khalil. R: Salamon. G: Bartoli. D: Jean. P. (1992). Putaminal necrosis after menhanol
intoxication [letter] [Letter]. J Neurol Neurosurg Psychiatry 55: 234-235.
Perkins. RA: Ward. KW: Pollack. GM. (1995a). Comparative toxicokinetics of inhaled methanol in the female
CD-I mouse and Sprague-Dawley rat. Toxicol Sci 28: 245-254.
Perkins. RA: Ward. KW: Pollack. GM. (1996a). Methanol inhalation: site and other factors influencing
absorption, and an inhalation toxicokinetic model for the rat. Pharm Res 13: 749-755.
http://dx.doi.Org/10.1023/A:1016055701736
Perkins. RA: Ward. KW: Pollack. GM. (1996b). Ventilation of ambulatory rats exposed to methanol vapor in a
flow-through exposure system: Measurement and input to toxicokinetic models. Inhal Toxicol 8: 143-162.
May 2013 R-5 Draft - Do Not Cite or Quote
-------
Phang. PT: Passerini. L: Mielke. B: Berendt R: King. EG. (1988). Brain hemorrhage associated with methanol
poisoning. Crit Care Med 16: 137-140.
Pollack. GM: Brouwer. KLR. (1996). Maternal-fetal pharmacokinetics of methanol (pp. 63 pp). (74). Boston,
MA: Health Effects Institute.
Prabhakaran. V: Ettler. H: Mills. A. (1993). Methanol poisoning: Two cases with similar plasma concentrations
but different outcomes [Review]. CanMedAssoc J 148: 981-984.
Ramsey. JC: Andersen. ME. (1984). A physiologically based description of the inhalation pharmacokinetics of
styrene in rats and humans. Toxicol Appl Pharmacol 73: 159-175. http://dx.doi.org/10.1016/0041-
008X(84)90064-4
Riegel. H: Wolf. G. (1966). Schwere neurologische Ausfalle als Folge einer Methylakoholvergiftung. Fortschr
Neurol Psychiatr 34: 346-351.
Rogers. JM: Barbee. BD: Rehnberg. BF. (1993a). Critical periods of sensitivity for the developmental toxicity
of inhaled methanol [Abstract]. Teratology 47: 395.
Rogers. JM: Brannen. KC: Barbee. BD: Zucker. RM: Degitz. SJ. (2004). Methanol exposure during
gastrulation causes holoprosencephaly, facial dysgenesis, and cervical vertebral malformations in
C57BL/6J mice. Birth Defects Res B Dev Reprod Toxicol 71: 80-88. http://dx.doi.org/10.1002/bdrb.20003
Rogers. JM: Mole. ML. (1997). Critical periods of sensitivity to the developmental toxicity of inhaled
methanol in the CD-I mouse. Teratology 55: 364-372. http://dx.doi.org/10.1002/(SICI) 1096-
9926(199706)55:6<:364::AID-TERA2>:3.0.CO:2-Y
Rogers. JM: Mole. ML: Chernoff. N: Barbee. BD: Turner. CI: Logsdon. TR: Kavlock. RJ. (1993b). The
developmental toxicity of inhaled methanol in the CD-I mouse, with quantitative dose-response modeling
for estimation of benchmark doses. Teratology 47: 175-188. http://dx.doi.org/10.1002/tera. 1420470302
Rubinstein. D: Escott E: Kelly. JP. (1995). Methanol intoxication with putaminal and white matter necrosis:
MR and CT findings. AJNR Am J Neuroradiol 16: 1492-1494.
Schaefer. PW: Grant PE: Gonzalez. RG. (2000). Diffusion-weighted MR imaging of the brain. Radiology 217:
331-345.
Schmutte. P: Bilzer. N: Penners. BM. (1988). Zur nuchternkinetik der begleitalkohole methanol und propanol-
l.Blutalkohol25: 137-142.
Schroeter. JD: Nong. A: Yoon. M: Taylor. MD: Dorman. DC: Andersen. ME: Clewell HJ. III. (2011). Analysis
of manganese tracer kinetics and target tissue dosimetry in monkeys and humans with multi-route
physiologically based pharmacokinetic models. Toxicol Sci 120: 481-498.
http://dx.doi.org/10.1093/toxsci/kfq389
Sedivec. V: Mraz. M: Flek. J. (1981). Biological monitoring of persons exposed to methanol vapours. Int Arch
Occup Environ Health 48: 257-271.
Sefidbakht S: Rasekhi. AR: Kamali K: Borhani. HA: Salooti. A: Meshksar. A: Abbasi. HR: Moghadami. M:
Nabavizadeh. SA. (2007). Methanol poisoning: acute MR and CT findings in nine patients. Neuroradiology
49: 427-435. http://dx.doi.org/10.1007/s00234-007-0210-8
Siu. MT: Wiley. MJ: Wells. PG. (2013). Methanol teratogenicity in mutant mice with deficient catalase activity
and transgenic mice expressing human catalase. Reprod Toxicol 36: 33-39.
http://dx.doi.org/10.1016/j.reprotox.2Q12.11.006
Starr. TB: Festa. JL. (2003). A proposed inhalation reference concentration for methanol [Review]. Regul
Toxicol Pharmacol 38: 224-231.
Sultatos. LG: Pasting. GM: Rosenfeld. CA: Flynn. EJ. (2004). Incorporation of the genetic control of alcohol
dehydrogenase into a physiologically based pharmacokinetic model for ethanol in humans. Toxicol Sci 78:
20-31. http://dx.doi.org/10.1093/toxsci/kfh057
Sweeting. JN: Siu. M: Wiley. MJ: Wells. PG. (2011). Species- and strain-dependent teratogenicity of methanol
in rabbits and mice. Reprod Toxicol 31: 50-58. http://dx.doi.0rg/10.1016/i.reprotox.2010.09.014
May 2013 R-6 Draft - Do Not Cite or Quote
-------
Iran. MN: Wu. AH: Hill. DW. (2007). Alcohol dehydrogenase and catalase content in perinatal infant and adult
livers: potential influence on neonatal alcohol metabolism. Toxicol Lett 169: 245-252.
http://dx.doi.0rg/10.1016/i.toxlet.2007.01.012
TRL (Toxicity Research Laboratories). (1986). Rat oral subchronic toxicity study with methanol. (TRL No.
032-005). Muskegon, MI: Research Triangle Institute.
U.S. EPA (U.S. Environmental Protection Agency). (1994b). Methods for derivation of inhalation reference
concentrations and application of inhalation dosimetry [EPA Report]. (EPA/600/8-90/066F). Research
Triangle Park, NC. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=71993
U.S. EPA (U.S. Environmental Protection Agency). (1998a). Guidelines for neurotoxicity risk assessment [EPA
Report]. (EPA/630/R-95/001F). Washington, DC.
http://www.epa.gov/raf/publications/pdfs/NEUROTOX.PDF
U.S. EPA (U.S. Environmental Protection Agency). (1998b). Health effects test guidelines OPPTS 870.6300
developmental neurotoxicity study [EPAReport]. (EPA712C98239). Washington, DC: U.S.
Environmentall Protection Agency, http://www.regulations.gov/#!documentDetail:D=EPA-HQ-OPPT-
2009-0156-0042
U.S. EPA (U.S. Environmental Protection Agency). (2000a). IRIS summary for vinyl chloride. Available online
at http://www.epa.gov/ncea/iris/subst/1001.htm (accessed luly 13, 2010).
U.S. EPA (U.S. Environmental Protection Agency). (2002). A review of the reference dose and reference
concentration processes [EPA Report]. (EPA/630/P-02/002F). Washington, DC.
http://cfpub.epa.gov/ncea/cfm/recordisplav. cfm?deid=51717
U.S. EPA (U.S. Environmental Protection Agency). (2006a). Approaches for the application of physiologically
based pharmacokinetic (PBPK) models and supporting data in risk assessment (Final Report) [EPA Report].
(EPA/600/R-05/043F). Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). (2006c). Peer review handbook (3rd edition) [EPAReport].
(EPA/100/B-06/002). Washington, DC. http://www.epa.gov/peerreview/
U.S. EPA (U.S. Environmental Protection Agency). (2009a). Benchmark dose software (BMDS). Available
online at http://www.epa.gov/NCEA/bmds (accessed December 19, 2009).
U.S. EPA (U.S. Environmental Protection Agency). (2011a). Benchmark Dose Software (BMDS) (Version 2.2
R65 [Build: 04/13/2011]) [Computer Pro gram]. Research Triangle Park, NC: National Center for
Environmental Assessment. Retrieved from http://www.epa. gov/NCEA/bmds/index.html
U.S. EPA (U.S. Environmental Protection Agency). (2011b). Toxicological Review of Methanol (Non-Cancer)
(External Review Draft) [EPAReport]. (EPA/635/R-11/001A). Washington, D.C.
http ://cfpub. epa. gov/ncea/cfm/recordisplav. cfm?deid=23 3 771
U.S. EPA (U.S. Environmental Protection Agency). (2012a). Benchmark dose technical guidance. (EPA/100/R-
12/001). Washington, DC. http://www.epa.gov/raf/publications/pdfs/benchmark dose guidance.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2012b). Methanol PBPK Model [PBPK].
Vara-Castrodeza. A: Perez-Castrillon. IL: Duenas-Laita. A. (2007). Magnetic resonance imaging in methanol
poisoning. Clin Toxicol 45: 429-430. http://dx.doi.org/10.1080/15563650701285313
Ward. KW. (1995) A comprehensive description of the toxicokinetics of methanol in non-pregnant and
pregnant rats and mice. (Doctoral Dissertation). University of North Carolina at Chapel Hill, Chapel Hill,
NC.
Ward. KW: Blumenthal GM: Welsch. F: Pollack. GM. (1997). Development of a physiologically based
pharmacokinetic model to describe the disposition of methanol in pregnant rats and mice. Toxicol Appl
Pharmacol 145: 311-322. http://dx.doi.org/10.1006/taap.1997.8170
Ward. KW: Perkins. RA: Kawagoe. JL: Pollack. GM. (1995). Comparative toxicokinetics of methanol in the
female mouse and rat. Toxicol Sci 26: 258-264.
May 2013 R-7 Draft - Do Not Cite or Quote
-------
Weichenthal S: Hancock. S: Raffaele. K. (2010). Statistical power in the analyses of brain weight measures in
pesticide neurotoxicity testing and the relationship between brain and body weight. Regul Toxicol
Pharmacol 57: 235-240. http://dx.doi.0rg/10.1016/j.vrtph.2010.03.001
Wilson. SL: Cradock. MM. (2004). Review: Accounting for prematurity in developmental assessment and the
use of age-adjusted scores [Review]. J Pediatr Psychol 29: 641-649.
http://dx.doi.org/10.1093/jpepsv/jsh067
Wu. AHB: Kelly. T: Mckav. C: Ostheimer. D: Forte. E: Hill. D. (1995). Definitive identification of an
exceptionally high methanol concentration in an intoxication of a surviving infant: Methanol metabolism
by first-order elimination kinetics. J Forensic Sci 40: 315-320.
Yoon. M: Nong. A: Clewell HJ. Ill: Taylor. MD: Dorman. DC: Andersen. ME. (2009a). Evaluating placenta!
transfer and tissue concentrations of manganese in the pregnant rat and fetuses after inhalation exposures
with a PBPK model. Toxicol Sci 112: 44-58. http://dx.doi.org/10.1093/toxsci/kfpl98
Yoon. M: Nong. A: Clewell HJ. Ill: Taylor. MD: Dorman. DC: Andersen. ME. (2009b). Lactational transfer of
manganese in rats: Predicting manganese tissue concentration in the dam and pups from inhalation
exposure with a pharmacokinetic model. Toxicol Sci 112: 23-43. http://dx.doi.org/10.1093/toxsci/kfpl97
Yoon. M: Schroeter. JD: Nong. A: Taylor. MD: Dorman. DC: Andersen. ME: Clewell HJ. III. (2011).
Physiologically based pharmacokinetic modeling of fetal and neonatal manganese exposure in humans:
Describing manganese homeostasis during development. Toxicol Sci 122: 297-316.
http://dx.doi.org/10.1093/toxsci/kfrl41
May 2013 R-8 Draft - Do Not Cite or Quote
------- |