EPA/635/R-ll/001Fb
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)
September 2013
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
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 vii
APPENDIX A. SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC COMMENTS, AND
DISPOSITION A-l
A.I. EXTERNAL PEER REVIEW PANEL COMMENTS A-l
A. 1.1. "Toxicokinetics and PBPK Modeling" A-2
A.1.2. Charge B: "Inhalation Reference Concentration (RfC) forMethanol" A-19
A.I.3. Charge C: "Oral Reference Dose (RID) for Methanol" A-39
A.1.4. Charge D: "General Charge Questions" A-42
A.2. PUBLIC COMMENTS A-55
A.2.1. April 18, 2011 to July 6, 2011 Public Comment Period A-55
A.2.2. May 3, 2013 to June 17, 2013 Public Comment Period A-64
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.2.1. Model Structure B-2
B.2.2. Model Parameters B-6
B.2.3. Rat Model Calibration B-8
B.2.4. Rat Model Sensitivity Analysis B-15
B.2.5. Rat Model Simulations B-18
B.2.6. Human Model Calibration B-19
B.2.7. Discussion and Sensitivity Analysis of Human Model B-27
B.2.8. Inhalation Route HECs and Oral Route HEDs B-32
B.3. MONKEY PK DATA AND MODEL ANALYSIS B-35
B.4. CONCLUSIONS AND DISCUSSION B-40
APPENDIX C. HUMAN CASE STUDIES C-l
APPENDIX D. RfC DERIVATION OPTIONS D-l
D.I. BENCHMARK DOSE MODELING SUMMARY D-l
D. 1.1. Evaluation of Model Fit D-l
D.1.2. Model Selection D-2
D.2. RFC DERIVATIONS USING THE NEDO METHANOL REPORT (NEDO, 1987) D-2
D.2.1. Decreased Brain Weight in Male Rats Exposed throughout Gestation and into the Postnatal Period D-2
D.2.2. Decreased Brain Weight in Male Rats Exposed During Gestation Only (GD7-GD17) D-14
D.2.3. C.l.2.2. BMD Approach with a BMR of 0.05 Change Relative to Control Mean (GD7-GD17) D-19
D.3. RFC DERIVATIONS USING ROGERS ETAL. (1993B) D-23
D.3.1. BMD Approach with a BMR of 0.10 Extra Risk D-24
D.3.2. BMD Approach with a BMR of 0.05 Extra Risk D-32
D.4. RFC-DERIVATIONS USING BURBACHER ETAL. (1999 A; 1999B) D-40
-------
APPENDIX E. DOCUMENTATION OF IMPLEMENTATION OF THE 2011 NATIONAL RESEARCH
COUNCIL RECOMMENDATIONS E-l
References R-l
LIST OF TABLES
Table A-l Summary of PODs for critical endpoints, application of UFs and conversion to candidate RfCs
using PBPK modeling A-38
Table A-2 Summary of PODs for critical endpoints, application of UFs and conversion to candidate RfDs
using PBPK modeling A-38
Table B-l Parameters used in the rat and human PBPK models B-7
Table B-2 Sensitivity of rat model dose metrics to fitted parameters B-16
Table B-3 Primate Km values reported in the literature B-28
Table B-4 Human PBPK model sensitivity analysis for steady-state inhalation exposure B-30
Table B-5 PBPK model sensitivity analysis for oral exposure B-32
Table B-6PBPK model predicted Cmax (Css) and 24-hour AUC for humans exposed to Methanol B-34
Table B-7 Monkey group exposure characteristics for Burbacher et al. (1999a) B-39
Table B-8 Monkey group exposure characteristics ofrNEDO (1987).a B-39
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-15
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-19
Table D-7 Methanol blood levels (Cmax above background) in mice following inhalation exposures D-24
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-25
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-33
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-41
IV
-------
Table D-l 1 Comparison of BMD modeling results for VDR in female monkeys using Cmax above
background of blood methanol as the dose metric D-42
Table E-1. National Research Council recommendation that EPA is implementing in the short-term E-2
Table E-2. National Research Council recommendations that the EPA is generally implementing in the
long term E-8
LIST OF FIGURES
Figure A-l Relationship of monkey blood levels associated with effects of uncertain adversity with
projected impact of daily peak alternative RfC and RfD exposures [derived using aUFD of 1]
on sample background methanol blood levels (mg MeOH/Liter [mg/L] blood) in humans A-39
Figure B-l Schematic of the PBPK model used to describe the inhalation, oral, and i.v. route
pharmacokinetics of methanol B-4
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-9
Figure B-3 Rat i.v.-route methanol blood kinetics B-ll
Figure B-4 Model fits to data sets from inhalation exposures in female Sprague-Dawley rats B-12
Figure B-5 Model simulations compared to 100 (squares) or 2,500 (diamonds) mg/kg oral methanol data
infemale Sprague-Dawley rat (expanded scale in panel B) B-14
Figure B-6 Simulated Sprague-Dawley rat inhalation exposures to 500, 1,000, or 2,000 ppm methanol B-19
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-21
Figure B-8 Urinary methanol elimination concentration (upper panel) and cumulative amount (lower
panel), following inhalation exposures to methanol in human volunteers B-23
Figure B-9 Blood methanol concentrations in subjects exposed for 30 min, 1 hr, or 2 hr at 800 ppm B-24
Figure B-10 Blood methanol concentrations in control (0 ppm) and methanol exposed (200 ppm) subjects B-24
Figure B-ll Oral exposure (10 mg/kg) to methanol in human volunteers (points) B-25
Figure B-12 Inhalation exposures to methanol in human volunteers B-26
Figure B-13 Intravenous exposure (10 mg/kg) to methanol in human volunteers (points) B-27
Figure B-14 Predicted human blood concentrations (increase above background) from total daily exposures
to 10 mg/kg-day methanol, consumed in a series of 6 boluses. Time is from the first bolus of
the day. See text for further details B-31
Figure B-15 Blood methanol concentration data from NP and pregnant monkeys B-36
Figure B-16 Chamber concentration profiles for monkey methanol exposures B-38
Figure B-17 PBPK model predictions of changes in blood methanol levels in humans for exposures at the
RfC and RfD B-42
Figure D-l Hill model, BMR of 1 Control Mean SD - 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, FI generation inhalational study D-13
-------
Figure D-3 Exponential model 4, BMR of 1 control mean SD - Decreased brain weight in male rats at 8
weeks of age versus Cmax above background, gestation only inhalational study D-18
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 -22
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-32
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-40
Figure D-7 Third (3rd) degree Polynomial model, BMR of 1 control mean SD - VDR in female monkeys
using Cmax above background of blood methanol as the dose metric D-45
VI
-------
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
VII
-------
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
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
NIEHS
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
apparent Michaelis-Menten constant;
substrate concentration at half the enzyme
maximum velocity (Vmax)
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)
National Institute of Environmental Health
Sciences
VIM
-------
NIOSH National Institute for 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
S-D Sprague-Dawley rat strain
SD 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; T,/2, t/2 time; 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 pseudo-maximal velocity of metabolism
VmaxC multiplier for allometric scaling of Vmax
VDR visually directed reaching test
VitC vitamin C
VPR ventilation perfusion ratio
v/v volume of solute/volume of solution
VYS visceral yolk sac
WBC white blood cell
WOE weight of evidence
w/v weight (mass of solute)/volume of solution
X2 chi square
IX
-------
APPENDIX A. SUMMARY OF EXTERNAL PEER
REVIEW AND PUBLIC COMMENTS, AND
DISPOSITION
A.1. External Peer Review Panel Comments
The draft toxicological review of methanol (U.S. EPA, 2011 a, c) has undergone a formal
external peer review performed by scientists in accordance with EPA guidance on peer review
(U.S. EPA, 2006b). An external peer-review meeting was held July 22, 2011. There were seven
external peer reviewers. The external peer reviewers were tasked with providing written answers
to general questions on the overall assessment and on chemical-specific questions in areas of
scientific controversy or uncertainty. At the workshop, they discussed their responses to each of
the charge questions and consensus was not sought. A summary of significant comments made
by the external reviewers and EPA's responses to these comments arranged by charge question
follow.
A subsequent follow-up peer review was completed in July 2013 to obtain feedback from
members of the original 2011 peer review panel on the 2013 revised draft methanol (noncancer)
toxicological review and EPA's response to the 2011 peer review comments. The follow-up
comments from these peer reviewers and EPA responses are presented in this section, with
general comments at the beginning of the section and charge specific comments at the end of
each charge question. Two other members of the original 2011 peer review panel submitted
written public comments, which are addressed in the public comment section (Section A.2) of
this appendix.
The summary of the peer review comments quotes the reviewer comments extensively,
but synthesizes and paraphrases in some cases for the sake of clarity and conciseness.
Additionally, the reviewers made a number of editorial suggestions to clarify specific portions of
the text. These changes were incorporated in the document as appropriate and are not discussed
further.
EPA received comments from the public on the 2011 and 2013 draft toxicological
reviews prior to the 2011 peer review and the 2013 follow-up peer review, which were
distributed to both peer review panels for their consideration. Public comments are posted to the
federal docket at www.regulations.gov: search for docket ID No. EPA-HQ-OJAD-2009-0398.[1] A
summary of these public comments and EPA's responses are included in Section A.2 of this
appendix.
[1] Public comments on the draft methanol (noncancer) Toxicological Review posted to www.regulations.gov can be
found at the following URL: http://www.regulations.gov/#!docketDetail:D=EPA-HQ-ORD-2009-0398.
A-1
-------
General Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer stated that "The revised (May 2013) version of
'lexicological Review of Methanol (noncancer)' has been improved significantly in comparison
to its external peer-review draft (U.S. EPA, 2011 a, c) version. It addressed the key
recommendations, comments, and suggestions provided in my Post-Meeting Comments of
7/31/2011."
Response: EPA appreciates the affirmation of sufficient revisions to the Toxicological
Review in response to previous peer-review comments.
Comment 2: One reviewer commented that "The EPA and the authors of this review of
the non-cancer effects of methanol are to be commended for this latest version" and added that
"[t]he overall document is much more concise and direct in detailing the key features of the risk
assessment that has been conducted." The reviewer stated that".. .a number of edits that
responded well to the comments of previous reviewers as well as the public.. .include the
utilization of background methanol levels in the PBPK model, the discussion of the relevance of
the blood levels resulting from the RfC/RfD numbers in comparison to endogenous methanol
levels, as well as a better explanation of various parameters in the PBPK models (such as using
only the Sprague-Dawley rat data, not the F344 rat, and adding more human data to the
validation)."
Response: EPA appreciates the affirmation of sufficient revisions to the Toxicological
Review in response to previous peer-review comments.
Comment 3: One reviewer requested that "a short statement that the CNS damage seen
in the acute overdose exposures most likely results from the acidosis and not from methanol per
se" be added to the beginning of Appendix C.
Response: A short statement has been added to the beginning of Appendix C, as
requested, to indicate that CNS damage seen in the acute overdose exposures most likely results
from the acidosis and not from methanol per se.
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.
Summary of Comments: In general, four reviewers stated that the PBPK model structure
was sound for the purposes of this assessment, a fifth reviewer stated that they noticed no
A-2
-------
obvious flaws but could not comment on a technical level due to a lack of expertise, and two
reviewers did not explicitly state whether the PBPK models were sound but provided comments.
Several reviewers commented that the models were comprehensively documented and that stated
assumptions are justified for the purposes of this assessment. One reviewer commended the
Agency for "developing a consistent model framework and sets of species-specific parameters
which have been validated across several somewhat diverse data sets. "Another reviewer
commented that "the Sprague-Dawley (S-D) rat PBPK model is inappropriately parameterized
(or insufficiently validated) for the inhalation route " and provided specific suggestions to
improve or validate the rat, mouse and human PBPK models developed by EPA for the purposes
of this assessment. Specific comments or suggestions made by the reviewers with respect to this
charge are described below, along with EPA responses.
Comment 1: One reviewer asked for "clarification of the process for evaluating the
usefulness of each model for the assessment and why the nonhuman primate model was not
included."
Response: EPA has described a framework (Chiu et al., 2007; U.S. EPA, 2006a) useful to
evaluate models for inclusion in an IRIS assessment. This framework includes review of the
model purpose, model structure, mathematical representation, parameter estimation, computer
implementation, predictive capacity, and statistical analyses. Currently, there is no specific EPA
policy or criteria for PBPK model use in IRIS assessments; consequently, the usefulness of a
PBPK model for a given species and assessment is a matter of scientific judgment and a number
of EPA PBPK experts are involved in making this judgment. Specific criteria used in evaluating
methanol models are presented in Section 3.4.1.2.
The ability of the model to fit a wide range of experimental data with a single set of
parameters is one of the critical considerations. When a chemical-specific (e.g., methanol) model
is able to predict experimental data for a range of doses or exposure conditions, there is
confidence that the model can predict chemical-specific (e.g., methanol) pharmacokinetics under
exposure conditions for which one does not have data. Confidence that one or more animal
species are properly represented by the model increases EPA's confidence that the models can be
used to extrapolate test animal exposures to human exposures.
Regarding the nonhuman primate model, EPA had incorrectly stated [in Appendix C,
Section C.3 of the draft assessment (U.S. EPA, 2011b)1 that external concentrations were used
for dose-response modeling for the monkey. However, a nonhuman primate classical PK model
(not a PBPK model) was adapted for use in the draft assessment and is used in the final
assessment to estimate internal doses (blood methanol Cmax values) for derivation of internal
A-3
-------
dose BMDLs associated with the Burbacher et al. (2004b; 1999b) monkey study. See
Appendix D, Section D.4, and Table D-10 in the final assessment.
Comment 2: One reviewer suggested that "the use [of] a bladder compartment is atypical
[thus] the EPA should consider receding the model to include a kidney/renal compartment that
considers excretion of methanol by the kidney."
Response: Urine passes through the bladder, which serves as a storage reservoir between
urine voids, so it is biologically realistic to include a compartment that represents this part of the
elimination pathway. Human urinary data are sufficient to identify a bladder residence-time
constant, but similar time-course data are not available for rats; therefore, the compartment only
impacts the human PBPK model. While only a small fraction of ingested/absorbed methanol is
excreted via the urinary elimination pathway, inclusion of a bladder compartment is significant in
that it allows a more precise fit to the human urine time-course data, which show a slight
nonlinearity in the human dosimetry.
Additionally, since the kidney glomeruli filter the blood directly, it is biologically realistic
to describe renal excretion as elimination directly from the blood compartment at a rate
proportional to the methanol concentration in blood. Inclusion of an explicit kidney compartment
of the form typically used for PBPK models would be most beneficial if the kidney was a target
tissue for toxicity. Development of a model of glomerular filtration for methanol would require
more extensive research, time, and resources. Addition of this type of compartment is not
expected to significantly impact the PBPK model predictions that are currently well predicted
and validated using the model structure applied in this assessment.
Comment 3: One reviewer questioned using a bladder component for only the human
model and suggested that the impact of a bladder component on the rodent models should be
tested.
Response: The bladder compartment is present in the model code used for both rats and
humans. The bladder compartment time constant (kbi) was identified for humans, and urinary
excretion rates were plotted and compared with existing human urine data (Figure B-7). Similar
urinary excretion data are not available for rats; hence, the bladder time-constant (ky) cannot be
identified for that species. The use of the bladder compartment and ky has no impact on
predicted blood concentrations; and hence, no impact on any of the rat model predictions.
Comment 4: One reviewer suggested that descriptions (e.g., page 3-45 of the draft
assessment) of the two divergent models that were considered (Michaelis-Menten or not) are
confusing, and should be clarified.
A-4
-------
Response: All discussion of this and other considerations in the development and
calibration of the human PBPK model have been consolidated into Appendix B, Section B.2.6,
and revised for clarification.
Comment 5: With respect to the monkey PK model, one reviewer commented that "the
description of the chamber volume [page 3-49 of the draft assessment] should be expanded" to
clarify the equipment in question and whether there is any evidence that incomplete mixing
occurred.
Response: The monkey PK model description and analysis is now found in Appendix B,
Section B.3 of this final assessment. Information regarding the chamber volume adjustment for
the PK model was added to this section. Briefly, the chamber volume was fit to the chamber
concentration data to allow for a better fit to the "mixing time" in the chamber (mixing time is
the time it takes for the chamber concentration to rise or fall after the inlet concentration is
turned on or off) and to account for the volume filled by the monkey and other chamber
equipment. The "accessible [chamber] volume" reported by Burbacher et al. (1999b) was
1,380 L, and the model fitted volume was 1,220 L. A detailed description of the chamber set-up
is included in the original Burbacher et al. (1999b) report and was not included in Appendix B of
this assessment.
Since methanol chamber time-course data were available for this study, and model
predictions did not match the data when the assumption of perfect mixing was used (chamber
volume of 1,380 L), EPA considered it reasonable to use the available data to calibrate the
residence- or mixing-time for the model of the chamber concentration. The text in Appendix B,
Section B.3 has been modified to indicate why Vch was varied.
Comment 6: With respect to the human PBPK and the monkey PK models, one reviewer
stated that EPA "has not clearly articulated why two different fractional absorption values were
used based on the same data base (see pages 3-50 (60%) and 3-42 (86.5%) of the draft
assessment)."
Response: In the external peer review draft document the derivation of the value of
FRACIN used for humans was described in detail in Appendix B, p. B-29. The exact mean
absorption measured by Sedivec et al. (1981) was 57.7%, but this was based on total ventilation.
However the human PBPK model uses alveolar ventilation, assumed to be 2/3 of total
ventilation, with the remaining 1/3 of each breath assumed to not enter the gas exchange area.
Therefore, to yield the same net uptake as Sedivec et al. (1981), 57.7% was divided by 2/3 to
yield 86.55% for the human model parameter, FRACIN.
A-5
-------
In contrast the monkey model used total respiration rather than alveolar ventilation. This
detail was included at the bottom of p. 3-49, where the monkey model parameter RC was defined
as "allometric scaling factor for total monkey respiration" (emphasis added). However the
parameter symbol "F" was used for the monkey "fraction of inhaled" to further distinguish it
from the human FRACIN, given that they are applied to different portions of total ventilation.
Further, in the description of the monkey parameter "F" on p. 3-50 (immediately after the value
of 60% is given), the document noted that 60% was the "(rounded)" value from Sedivec et al.
(1981) and went on to state, "F and Vmk cannot be uniquely identified, given the model structure,
so F was set to the (approximate) human value to obtain a realistic estimate of Vmk. For example,
if both F and Vmk are increased by 50%, then also increasing the fitted Vmax by 50% would yield
identical model fits. Any positive value could be assigned to F and it would not affect the
resulting model fits. Given that the airways in a 2-4 kg monkey are much smaller than those in a
70 kg human, it is unlikely that the transport characteristics for which F and FRACIN account
are identical in the two species. But a realistic value was considered desirable for the monkey, so
the approximate total ventilation human value was assumed to be sufficient.
Since the revised monkey model was ultimately not used in deriving the RfC or RfD, the
detailed explanation given here was not considered necessary and was not included in the draft
document. The human FRACIN has since been revised to 75% using additional data as noted by
another peer reviewer and described in detail in Appendix B, Section B.2. The value used for the
monkey model was therefore also revised to 50%, maintaining the 2/3 factor vs. human, and a
brief description provided. Reference to the Sedivec et al. (1981) paper was removed from this
part of the document, since a different data set is used.
Comment 7: Two reviewers noted that the blood methanol levels predicted by the EPA
rat PBPK model are much lower than the levels reported for S-D rats on recently located
(supplied by industry at the peer review meeting) pages from the NEDO (1987) report and
Perkins et al. (1995). One of the reviewers further noted that if the fraction inhaled (FRACIN)
model parameter is changed from 20% to a value more consistent with the mouse (66.5%) and
human (86.6%) estimates, the 1,000 ppm blood prediction is in agreement with Perkins et al.
(1995).
Response: In the draft assessment, inhalation data for F344 rats were used. Since the
current noncancer assessment does not use any bioassay data from F344 rats, all PK data (and
PBPK modeling results) for F344 rats have now been removed. The PBPK model in the final
assessment uses the S-D rat inhalation data from Perkins et al. (1996a) for calibration, yielding a
more appropriate fitted value for FRACIN (81%) and model predictions that are more consistent
with the NEDO (1987) reported blood levels (See Sections B.2.2, B.2.4, and B.2.5).
A-6
-------
Comment 8: One reviewer noted that, with respect to the Burbacher et al. (1999a)
monkey data, EPA has not justified why the second trimester group is considered the most
representative.
Response: When the data and fits shown in Figure 3-14 were evaluated, EPA noted that
overall there appears to be no significant or systematic difference among the NP and pregnant
groups. The solid lines, in the figure, are model simulations calibrated to only the 2nd trimester
data (details below), but they just as adequately represent average concentrations for the NP and
3rd trimester data. Likewise, a PK model calibrated to the NP PK data adequately predicted the
maternal methanol concentrations in the pregnant monkeys (results not shown). Since any
maternal:fetal methanol differences are expected to be similar in experimental animals and
humans (with the maternal:fetal ratio being close to one due to methanol's high aqueous
solubility and relatively limited metabolism by the fetus), the predicted levels for the 2nd
trimester maternal blood are used in place of measured or predicted fetal concentrations.
Thus the primary justification for only showing the results for the 2nd trimester is that it
does not matter which stage one selects, since there is not a significant difference in either the
data or the model fits among the stages. While there is no clear effect of pregnancy on the PK in
monkeys, to the extent that there is some trend (for example, if the AUC decreases slightly with
the extent of pregnancy), the value of the metric during the 2nd trimester was expected to be in
between the values for the 1st and 3rd trimesters, hence closer to an overall average. In short,
because the physiological changes induced by pregnancy are at an intermediate stage in the 2nd
trimester relative to the 1st and 3rd, PK parameters were expected to also be intermediate and
therefore most representative of the average over all of pregnancy. However, had there been clear
time-dependence in the PK data, a quantitative analysis could have been used to incorporate that
trend.
Comment 9: One reviewer suggested that the Km values estimated by the rat and human
models "don't seem to reflect the true Michaelis values of the metabolic enzymes themselves."
Response: Since methanol is metabolized by multiple enzymes with differing Km values,
at best one would expect the empirical Km values identified here to represent an average of the
enzyme-specific values (weighted by the contribution of each enzyme to total metabolism).
Further, it is quite typical to find that in vivo PK data are not well-predicted when Km values
measured in vitro are used in a model, hence Km values estimated from in vivo data are not
expected to be identical to values measured in vitro. The Km values identified for the revised
PBPK models described here are 28 mg/L for rats and 36 mg/L for humans. Pollack and Brouwer
(1996) analyzed the kinetics of formaldehyde formation in vitro and estimated Km = 39.3 mg/L
using nonpregnant adult rat liver homogenates and Km = 35.5 mg/L with GD 20 homogenates:
A-7
-------
quite similar to the revised value for the rat used in the final assessment. Mani et al. (1970)
measured methanol kinetics with human liver ADH and obtained a Km of 48 mg/L. This is
likewise quite similar to the Km estimated here with the human PBPK model.
Comment 10: With respect to the human PBPK model, one reviewer noted that useful
human kinetic studies (Haffner et al., 1992; Schmutte et al., 1988) were overlooked, and that
these studies "are potentially quite valuable in model parameterization because they do not
involve the inhalation route."
Response: Previously only inhalation data was included for humans. These studies
[(Haffner et al., 1992) and (Schmutte et al., 1988)] provide i.v. and oral data, and they have been
added to the PBPK analysis (see Appendix B, Section B.2.6). Specifically, the oral data are now
used for model calibration, allowing identification of human oral absorption rate constant and
bioavailability. The i.v. data from Haffner et al. (1992) are from only four individuals; these data
were used to validate the model by comparing model predictions following an i.v. dose with the
experimental data (Figure B-13).
Comment 11: One reviewer recommended that EPA perform sensitivity analyses of the
rat and human PBPK modeling results under conditions approximating the BMDL, stating that
"at a minimum, EPA should assess whether or not the model they used in the risk assessment can
(adequately) simulate the additional human data identified herein and conduct and provide
human model sensitivity analyses at the RfC and RfD."
Response: A sensitivity analysis has been conducted and a detailed description of this
analysis is included in Appendix B, Section B.2.7. However, such an analysis can only partly
inform the question of model adequacy, which is addressed in more detail in the response to
Charge Al Comment 1 above.
Comment 12: With respect to the mouse PBPK model, one reviewer stated that "it seems
odd that, for oral dosing, the mouse blood levels are reported to be insensitive to any parameter
related to clearance (e.g., metabolism, blood flow to the liver) (pp B-16 and B-18 of the draft
review)," and requested clarification in the text regarding the type of oral dose that is being
simulated.
Response: Since direct measurements of mouse (CD-I) blood concentrations for
bioassay exposures are available (Rogers et al., 1993b) and used for the BMD analysis in this
final assessment, the mouse PBPK model is not utilized in the final assessment to estimate an
internal dose metric. Therefore the description, analysis, and discussion of the mouse model are
not included in the final assessment.
A-8
-------
Comment 13: One reviewer commented that "the runtime files that should reproduce
Figures B-2 and B-5 yield simulations that are slightly off." The reviewer also commented,
regarding Figures B-6, B-7, and B-8, that "these files do not accurately reproduce the figures in
the document."
Response: The figures were produced with the background turned on, while the PBPK
runtime files had the background turned off. The current version of the PBPK model, available
electronically from the EPAFtERO database (U.S. EPA, 2012b), includes runtime files which
will exactly reproduce the figures in the toxicological review, aside from legend placement,
which is dependent on acslX window sizes.
Comment 14: With respect to the mouse PBPK model sensitivity analysis, one reviewer
noted that "EPA does not provide files that fully recreate the sensitivity analyses—only those
parameters demonstrated in Figures B-6, B-7, and B-8." This reviewer commented that "the
sensitivity analysis does not appear to have been comprehensive," and cited FRACIN as an
example of a parameter that was not tested, yet seems to be a parameter to which the mouse
PBPK model is sensitive.
Response: As stated in the response to Charge Al Comment 12, the mouse model has
been removed from the final assessment; thus no sensitivity analysis is included for the mouse
PBPK model.
Comment 15: One reviewer commented that "it is not clear why two saturable metabolic
pathways are needed for the Sprague-Dawley rat and only one for the F344 rat."
Response: The liver metabolism in the S-D rat is now described using a single saturable
rate equation. As discussed in response to Charge Al Comment 7, the analyses of F344 rat PK
data has been removed from the toxicological review.
Charge Al Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[w]hile the revised PBPK model used in the
derivation of reference toxicity values of methanol still contains many simplifications and
shortcuts, it seems to be adequate for chemical-specific risk assessment - the purpose for which
EPA developed this model."
Response: The EPA agrees with the reviewer that although the Agency's PBPK model
contains necessary, but simplifying assumptions, it is adequate for the purposes of the methanol
(noncancer) assessment.
A-9
-------
Comment 2: One reviewer suggested that EPA "reconnect the urine clearance to arterial
blood" because, while it "seems to approximate realistically the quantitative clearance of
methanol.. .the assumption that the urine equilibrates with mixed venous blood may be
inappropriate for some other chemicals."
Response: The reviewer is correct that by not including an explicit kidney compartment,
where urinary clearance would be limited by the arterial blood concentration and flow to the
kidney, the model does not include the limitation to clearance that occurs because renal flow is
only a fraction of total cardiac output. This limitation would be a significant factor if total urinary
clearance was a significant fraction of renal blood flow, which likely occurs for some other
chemicals. However for the methanol model the clearance rate for this pathway in the rat is only
0.24% of renal blood flow [using a renal flow fraction of 0.141 from Brown et al. (1997)1 and in
the human is only 0.07% of renal blood flow [using a renal flow fraction of 0.175 from Brown et
al. (1997)]. Therefore including an explicit kidney compartment with its own flow rate would
have a negligible impact on the methanol model results reported here. These calculations and a
statement that the approach should only be used when renal clearance is a small fraction (<10%)
of renal blood flow have been added to Appendix B.
A distinction between arterial and venous blood concentration can also occur due to high
rates of gas exchange in the lung. For oral (and i.v. exposures) the rate of exhalation is very low
and the arterial and venous blood concentrations were virtually indistinguishable for both rats
and humans. For inhalation exposures a small difference occurred, but less than 1% for the rat
and 4% for humans. Thus, as suggested by the reviewer, the difference is not significant for
methanol, hence has little impact on model predictions.
Comment 3: One reviewer encouraged EPA to "[p]lease keep unchanged equation and
parameters for methanol metabolism in the PBPK model, but change in the text the explanation
of meaning of Vmax to 'pseudo-maximal velocity of metabolism' and KM to 'apparent Michaelis-
Menten constant of metabolism'" because "the metabolism of methanol is potentially saturable"
and, while the Michaelis and Menten terms "Vmax" and "KM" are appropriate for use in EPA's
PBPK model equation, ".. .Michaelis and Menten equation describes initial velocity in
homogenous enzymatic systems" and "the [EPA] PBPK model describes rate of metabolism,
measured over time in the whole organism."
Response: The EPA appreciates the reviewer's support for the proposed model structure
and has chosen to retain this description of metabolism as a saturable process in the PBPK model
for both rats and humans. The EPA agrees that there is not an exact correspondence between the
parameters, Vmax and Km, as used and calibrated in the model versus values obtained from initial
A-10
-------
rate experiments in vitro. Therefore the descriptions of these parameters in the text and glossary
have been adjusted as indicated.
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 noncancer risks.
Summary of Comments: EPA stated two key assumptions for this approach in the peer
review draft: "(1) endogenous levels do not contribute significantly to the adverse effects of
methanol or its metabolites; and (2) the exclusion of endogenous levels does not significantly
alter PBPK model predictions. " Most reviewers were in general agreement with the first
assumption, but expressed the need to better characterize background levels of methanol and
their relationship to the RfC/D. Three reviewers were concerned that the first assumption, and
the subtraction of methanol background levels, gives the impression that
endogenous/background methanol levels are not important. With respect to the second
assumption, none of the reviewers disagreed with EPA's determination that the exclusion of
endogenous/background levels does not significantly alter PBPK model predictions; however,
two reviewers advocated the use of a PBPK model that incorporates a background term and one
reviewer favored the use of the simpler PBPK model (without background levels). Specific
comments or suggestions made by the reviewers with respect to this charge are described below,
along with EPA responses.
Comment 1: Three reviewers expressed concerns over the first assumption, that
endogenous/background levels do not contribute significantly to adverse effects. One reviewer
stated that EPA was giving the impression that "cumulative exposures from different sources are
not important." A second reviewer indicated that the first assumption is not met because "the RfC
and RfD correspond to blood methanol concentrations in humans squarely in the range of normal
'background' levels."." The third reviewer asked "If endogenous levels of methanol do not
contribute to adverse effects and an exposure does not produce an increase above background
levels, how can that exposure lead to an adverse effect?"
Response: The language in the draft assessment may have confused this issue and has
been clarified in Section 3.4.3.2. EPA acknowledges that endogenous/background methanol
concentrations can be a contributing factor in health effects that are associated with exogenous
methanol exposure. As indicated in response to Comment 2 below, for the sake of obtaining
more accurate and reliable toxicokinetic estimates, the PBPK models used in the final assessment
incorporate background/endogenous concentrations of methanol. Background estimates on
which the models were calibrated are described in Appendix B, Sections B.2.3 (rats), B.2.8
A-11
-------
(humans) and B.3 (monkeys). However, for BMD modeling of laboratory animal dose-response
data, the species-specific background estimates were subtracted from the dose metric predicted
under the relevant bioassay conditions. This approach takes into account the impact of
endogenous/background levels on the toxicokinetics of methanol, and allows for the derivation
of an RfC (or RfD) that is, by definition, a population level estimate (including sensitive
populations) of the amount of a substance that a person can inhale (or ingest) every day over the
course of a lifetime [above endogenous/background levels] without an appreciable risk of harm.
As pointed out by the 2nd and 3rd reviewers, the relationship between the RfD and RfC
and endogenous/background blood levels is an important consideration. Measured blood
concentrations of methanol in humans range between 0.25 mg/L and 5.2 mg/L (see Table 3-1).
As described in a new Section 5.3.6, PBPK model estimates of maximum blood level increases
of 0.44 mg/L and 0.41 mg/L associated with an RfD or RfC, respectively, are within the 0.7 mg/L
standard deviation estimated for the average methanol blood levels (1.5 ± 0.7 mg/L) in humans.
From this analysis EPA concludes that the estimated increase in blood levels of methanol from
exogenous exposures at the level of the RfD or the RfC (or from the RfC + RfD) are
distinguishable from natural background variation.
Comment 2: Two reviewers advocated the use of a PBPK model that incorporates a
background term and one reviewer favored the use of the simpler PBPK model. One of these
reviewers indicated that use of a background term would be "more rigorous and appropriate for
use in this assessment." The latter reviewer warned that "Including the background levels in the
models necessarily increases the model complexity and like any model enhancement may
increase the uncertainty in the final result, especially when as in this case it may be difficult to
design a test of its validity."
Response: As described in the response to Comment 1 above, EPA re-calibrated the
PBPK models to account for species-specific estimates of background/endogenous production of
methanol. For humans the model was tuned to have an average background level of 1.5 mg/L
determined from the corresponding human data in Table 3-1; for rats the model was tuned to a
background level of 3 mg/L from the corresponding (control) rat data in Table 3-5. These revised
PBPK models were used in estimation of internal dose metrics for the derivation of the RfD and
RfC. This addition did increase the model complexity by including an additional term (Robg, a
zero-order endogenous/background production rate, see Appendix B, Section B.2.1) for
endogenous/background production of methanol; however, this term was estimated using human
data for background blood methanol concentrations (Table 3-1). Since the background term was
tuned to match average observed background levels in rats and humans for the corresponding
models, there should be minimal systematic error or bias due to the incorporation of the
A-12
-------
background term; i.e., the average background level is neither under- nor over-predicted by the
model. Moreover, adding the term resulted in only minor changes, less than 20%, in model-
predicted blood levels at higher exposure levels (i.e., in the range of the bioassays for rats or the
HEC and HED values estimated for humans). Hence model predictions are not sensitive to the
inclusion of these average background levels vs. no background at all, and so the effect of and
uncertainty due to possible small changes (or errors) in the background term will be minimal.
Comment 3: One reviewer stated that "the upper bound on background concentrations of
methanol in target tissue should be carefully evaluated" and that "the lack of determination of the
upper statistical bound on normal physiological concentrations of methanol in relevant species,
including humans, can be considered to be a major deficiency of the reviewed document."
Response: Statistical bounds on normal physiological concentrations of methanol cannot
be determined for all tissues and species. The most complete dataset exists for blood levels of
methanol in humans. A discussion of endogenous/background levels of methanol and their
relationship to the RfC and RfD has been added to Chapter 5 (Section 5.3.6) and elsewhere in the
toxicological review. There is a scarcity of data for endogenous/background methanol levels in
the general population. Also, the existing data (Table 3-1) is from populations with various (e.g.,
age, gender, cultural) characteristics that were asked to adhere to a variety of diets, generally
restricted of food and drink that contain or convert to methanol. Measured values have been
documented as low as 0.25 mg/L and as high as 5.2 mg/L. From the data gathered for this
document (Table 3-1), EPA has estimated a mean background methanol level of 1.5 mg/L with an
approximated standard deviation of 0.7 mg/L (see Section 5.3.6).
Comment 4: One reviewer noted that, "in the simulations whose results are listed in the
Table B-5, a background level of 2 mg/L has been set to model human internal concentration
from inhalation (page B-92; line 29) but not from the oral exposure (page B-92; line 55)."
Response: This inconsistency was corrected and the values in Table B-5 have been
updated for both inhalation and oral exposures to reflect concentrations above
endogenous/b ackground.
Comment 5: One reviewer noted that EPA did not adequately explain the modest
differences in HED and HEC predictions from the PBPK models when background levels of
methanol were included or excluded.
Response: This comment was made in relation to a discussion of why including
background in the PBPK models might not be necessary. That discussion was removed from the
assessment because, as described above in response to Charge A2 Comment 2, the final versions
of the PBPK models do include background.
A-13
-------
Charge A2 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[i]n responding to Charge Question A2 for
the original report, I expressed concern about treating endogenous and exogenous methanol
differently lexicologically when both contribute to the internal dose" and that".. .this issue has
been addressed by including species-specific background/endogenous methanol in the PBPK
models." The reviewer added that "[t]his approach makes a great deal more sense, in my opinion,
and is consistent with the concept that risk is a function of internal dose, and that both
endogenous and exogenous sources contribute to that dose" and that "I consider this change
responsive to my comment."
Response: EPA appreciates the affirmation of sufficient revisions to the Toxicological
Review in response to previous peer-review comments and has retained the species-specific
estimates of background/endogenous methanol in the PBPK models.
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.
Summary of Comments: All reviewers agreed that the existing literature supports the
assumption of similar pharmacokinetics between pregnant and nonpregnant animals. Specific
comments or suggestions made by the reviewers with respect to this charge are described below,
along with EPA responses.
Comment 1: One reviewer stated they understood the rationale for omitting a fetal
compartment in the PBPK model, but felt that "for PBPK modeling to be effective, a fetal
compartment will ultimately be needed." This reviewer noted that "PBPK modeling is most
useful when the proximate form of the toxicant and mode of action are known, which is
unfortunately not the case with developmental effects of methanol."
Response: EPA agrees that a PBPK model with a fetal compartment would be ideal, and
that more mode of action information, including the identification of the proximate toxicant,
would be helpful. However, studies have shown, and reviewers have agreed, that methanol
pharmacokinetics between pregnant and nonpregnant animals are similar and, absent additional
information, provide a reasonable justification for extrapolation based on a PBPK model for non-
pregnant adults. If there are studies to the contrary, or studies that provide insight into fetal
metabolism or the embryotoxic moiety of methanol, a fetal compartment may be considered in
the future.
A-14
-------
Comment 2: One reviewer expressed concern over the model's ability to predict neonatal
blood levels, stating that "this issue is important since the critical study used by EPA to derive an
RfC involved combined gestational and lactational (inhalational) exposure of neonates" and that
"the use of an adult-based PBPK model could under predict potentially 'toxic' blood methanol
concentrations."
Response: It is recognized that neonatal blood levels will likely be higher than maternal
blood levels of methanol. Therefore, the ratio of blood concentrations between a human infant
and its mother is not expected to be significantly greater than the approximate 2-fold difference
that has been observed between rat pups and dams. Further, as stated in the final version of
Section 5.1.3.2.2, "the health-effects data indicate that most of the effects of concern are due to
fetal exposure, with a relatively small influence due to postnatal exposures." For these reasons
and because EPA has confidence in the ability of the PBPK model to accurately predict adult
blood levels of methanol, the maternal blood methanol levels for the estimation of FtECs from
the NEDO (1987) study were used as the dose metric.
Charge A3 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer stated that "[elven though [using methanol concentration in
maternal blood as surrogate dose metric for evaluating postnatal changes] seems to be
technically acceptable, without the understanding of exact mechanism of action (MOA) the
selection of such a surrogate dose metric remains somewhat speculative."
Response: EPA agrees that there is uncertainty regarding the decision to use methanol
concentration in maternal blood as a dose surrogate for evaluating postnatal changes. As
discussed in Section 4.7, the decision to model blood methanol concentration as opposed to one
of its metabolites was primarily based on a determination that (1) the toxic moiety for
developmental effects from methanol exposure is not likely to be the formate metabolite and (2)
methanol is an adequate dose metric, even if formaldehyde or ROS are determined to have a
significant role in the teratogenicity of methanol. The former determination has been endorsed by
other organizations (NTP-CERHR, 2004) and is supported by evidence that formate blood levels
do not correlate well with the developmental toxicity observed following methanol exposure.
The latter determination is based on evidence that (1) methanol can be metabolized to
formaldehye in situ by multiple organ systems, (2) the high reactivity of formaldehyde would
limit its unbound and unaltered transport as free formaldehyde and (3) the hypothesized ROS
MOA would require the presence of methanol to alter embryonic catalase activity (see further
discussion in Sections 4.7.1, 4.7.3 and 4.7.5).
A-15
-------
As described in Section 5.1.3.2.2, the decision to used maternal blood methanol as a
surrogate for neonatal blood levels was based on EPA's confidence in the PBPK models to
accurately predict maternal and fetal blood levels and an assumption that the ratio of the
difference in blood concentrations between a human infant and mother would be similar to and
not significantly greater than the difference in blood concentrations between a rat pup and their
rat dam. Further, the health-effects data indicate that most of the effects of concern are due to in
utero exposure, with a relatively small influence due to postnatal exposures.
Several research studies recommended by the reviewers in response to Charge D3
(below), including "dual labeled material" studies to "confirm fetal exposure" and "resolve
whether formaldehyde is involved in the developmental effects following perinatal methanol
exposure" and studies "to confirm the low activity of methanol metabolism in fetal tissues,"
could help to resolve the "principal sources of uncertainty" referred to by the reviewer. However,
as indicated in response to the Charge D3 comments, and consistent with the follow-up
comments to EPA's Charge D3 responses, EPA is not planning to delay the completion of this
assessment pending the completion of future studies.
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.
Summary of Comments: All reviewers agreed that this is a reasonable assumption given
the limited data available. Specific comments or suggestions made by the reviewers with respect
to this charge are described below, along with EPA responses.
Comment 1: Two reviewers thought that the assumption of limited methanol metabolism
in the fetus was valid based on the methanol pharmacokinetic data, but one of the reviewers
noted that embryotoxicity from methanol may be influenced by fetal catalase in mice as
demonstrated by a recent study (Miller and Wells, 2011). This reviewer further stated that this
and another study (Sweeting et al., 2011) suggest that fetal methanol concentrations in rodents
may not be a "good predictor of teratogenic responses in different species.".
Response: While these studies provide insights into the fetal metabolism of methanol, it
is unknown if fetal catalase is a controlling factor for methanol teratogenicity in mice.
Furthermore, a recent in vivo study (Siu et al., 2013) suggests that high catalase activity does not
protect against methanol teratogenicity in the strains of mice tested. EPA evaluated these studies
A-16
-------
and, as described in Section 5.3.5 "Choice of Species/Gender," concluded that the available
evidence related to fetal catalase and methanol's teratogenicity in mice is contradictory and
inadequate to suggest that rodent effects should not be used in an assessment of methanol's
potential to cause developmental effects in humans. Also, because the critical gestational window
for developmental effects could be different for rabbits versus mice, the claim that rabbits are
resistant to teratogenic effects of methanol needs to be verified over several gestational days, as
has been done for mice.
Comment 2: One reviewer commented that, "the assumption of limited methanol
metabolism in the fetus is probably justified based on the existing studies showing low levels of
ADH and catalase in fetal tissues" but added that "these studies have technically measured these
proteins using indirect measures such as immunoblotting showing protein amounts or activity
measures with ethanol as the substrate."
Response: EPA agrees with this comment. An activity measurement using methanol as
the substrate would be ideal. However, lacking such studies, it is reasonable to assume low
activity of methanol metabolism in fetal tissues from relevant, indirect studies.
Charge A4 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: No additional comments were received regarding this charge question.
Response: EPA has not changed the approach taken regarding fetal metabolism.
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.
Summary of Comments: Four reviewers agreed that a reasonable approach was taken
given the data available, though one of these reviewers reiterated that issues identified in Charge
Al with respect to the rat and human models need to be addressed. A fifth reviewer reiterated
comments made in response to Charge Al regarding the need for "clarification of the process for
evaluating the usefulness of each model for the assessment and why the nonhuman primate
model was not included" and noted that the use of the NEDO rat studies which included
neonatal exposures is "problematic, given the lack of data on lactational and early postnatal
inhalation exposure to methanol. " A sixth reviewer suggested that the model should be modified
to include gestational and lactational components and expressed concern over the use of rodent
data for estimating human risk from developmental effects. A seventh reviewer suggested that
A-17
-------
EPA's assumption that rats and humans would have similar maternal/offspring methanol
concentration ratios is a significant source of uncertainty.
Response: EPA agrees with the majority of the reviewers that the extrapolation approach
employed is justified given the available data. The reviewer concerns regarding the model
evaluation process and the perceived lack of a nonhuman primate model are addressed in
response to Charge Al Comment 1. The lack of data on lactational and early postnatal inhalation
exposure to methanol is a recognized data gap that led to the current approach. As discussed in
response to comments under Charge A3 above, gestational and lactational compartment may be
considered in a future assessment, but they are not necessary at this time for the purposes of this
toxicological review. Concerns over use of rodent studies stem from the Sweeting et al. (2011)
study. The relevance of the Sweeting et al. (2011) to these concerns is discussed in Section 5.3.5
and elsewhere in the toxicological review and in response to Comment 1 of Charge A4 and
Comment 1 of Charge D2. As discussed in response to Charge A3 Comment 2 and Section
5.1.3.2.2 of the toxicological review, the uncertainty surrounding the assumption of similar
maternal/offspring methanol concentration ratios between rats and humans is recognized, but the
ratio of blood concentrations between a human infant and its mother is not expected to be
significantly greater than the approximate 2-fold difference that has been observed between rat
pups and dams. Clarifications have been added in this regard to Sections 5.1.3.2.2 and 5.3.5.
Charge AS Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer stated that EPA "... continue[s] to state in the current
assessment, and reiterate in the response to panel comments from the original review, that the
ratio of blood methanol concentrations between a human infant and its mother is not expected to
be significantly different than the approximately 2-fold difference seen between rat pups and
dams." This reviewer noted that "[t]he main point made in that section [Section 5.1.3.2.2]
regarding this issue seems to be that this assumption isn't particularly important because most of
the effects of methanol occur in utero" and added that "[t]o the extent that it matters, the
assumption that maternal/offspring methanol concentration ratios are similar in both humans and
rats continues to be poorly justified in my opinion."
Response: The comment reflects the rationale provided, suggesting that the issue is not
lack of clarity in the rationale, but that there is not a strong justification, as noted by the reviewer.
The unfortunate fact is that methanol dosimetry data are not available for rat pups, human
infants, lactating rat dams, nor lactating human mothers (particularly, amounts in breast milk).
Given the high aqueous solubility of methanol, it may be reasonable to assume that
concentrations expressed in breast milk equal those in maternal blood. However dosimetry in the
A-18
-------
developing infant would depend on when and to what extent metabolic capacity develops in rat
pups versus human infants. So while it would be possible to extrapolate the existing adult models
to those life-stages, such extrapolations, for the infant in particular, would be quite speculative
and uncertain. In response to this comment, clarification has been added to Section 5.1.3.2.2 to
indicate why EPA does not believe that adding additional analyses would substantially reduce the
uncertainty and hence improve the justification around this assumption.
A.1.2. Charge B: "Inhalation Reference Concentration (RfC) for
Methancl"
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., 2004b;
Burbacher et al., 1999b) 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.
Summary of Comments: Two reviewers indicated that selection of NEDO (1987) as the
principal study was scientifically justified. Two reviewers stated that choice of the NEDO rat
study was based on "practical/technical grounds" or "policy" (i.e., use of the lowest RfC),
rather than scientific considerations. One reviewer did not explicitly state whether the use of the
NEDO rat study was scientifically justified, but stated that selection of the principal study is
contingent on the determination of the HEC/HED after the implementation of suggested model
revisions (e.g., Charge Al Comment 7). Two reviewers suggested that the NEDO rat
developmental study was not the most appropriate study for RfC derivation. Specific comments
or suggestions made by the reviewers with respect to the advantages and limitations of each of
the three studies addressed in the charge are described below, along with EPA responses.
Comment 1: Two reviewers indicated that the selection of NEDO (1987) as the principal
study was scientifically justified and noted the following scientific advantages:
• "The nearly continual exposure (20-22 hours per day depending on the study) represents the
types of exposures relative to the RfC/RfD (i.e. the daily exposure over the lifetime)."
• Selection of the NEDO study "is in accordance with the usual guidelines recommending use
of the study with the best data quality (including, in this case, availability of a validated
PBPK model) and greatest sensitivity."
A-19
-------
Two reviewers did not think that the NEDO study was the most appropriate choice and
noted the following concerns:
• The prior, EPA-sponsored peer review of the NEDO study questioned "procedures used in
the NEDO study (in utero and postnatal exposures, litter effects, etc) that make it difficult to
evaluate the study for RfC derivation."
• "The discussion on page 5-10 regarding the complications that arise from using the NEDO
study where exposure was both gestational and postnatal postulates a number of assumptions
that are supported by little or no data."
• "Data on lactational transfer and early postnatal inhalation exposures are limited."
• The neonatal brain weight response has not been replicated in other studies.
• "The analysis provided by the NEDO authors showed a gender difference (effects seen in
males but not female rats)."
• "The NEDO study relied on multiple t-tests as opposed to a more appropriate use of an
ANOVA to evaluate gender and treatment responses."
• There were no "corroborating clinical or pathological observations of depressed CNS activity
noted in the rats in the NEDO study."
The remaining three reviewers did not address the scientific merits of the NEDO study.
However, two of these reviewers suggested that its selection was based on it resulting in the
lowest RfC, and one stated that selection of the principal study is contingent on the determination
of the HEC/HED and suggested that (due to possible error in the PBPK model) the HEC/HED
value for the NEDO rat study "could be on the order of 6-fold too low."
Response: In addition to the advantages of the NEDO (1987) developmental rat study
noted by several reviewers (e.g., relevant exposure route and duration, validated PBPK model
estimates of internal dose, and a sensitive response endpoint), the NEDO study offers other
advantages (described further below) such as the identification of an endpoint that (1) is
biologically significant, (2) is observed at a sensitive developmental stage, (3) has been
replicated in adult rats and (4) is in an organ system for which suggestive pathology has been
observed in adult primates that received acute and chronic exposure to methanol via the same
exposure route. While reviewer comments on EPA's choice of the NEDO (1987) developmental
rat study were mixed, only two of seven reviewers indicated that its selection was not justified.
Though EPA recognizes that the NEDO (1987) study has limitations, these limitations do not
preclude its use as the principal study for RfC derivation (see Section 5.3.1 and below).
A-20
-------
EPA agrees that data on lactational transfer and early postnatal inhalation exposures are
limited, and this is largely the reason that maternal blood levels were used as a dose metric in the
analysis of the neonatal brain weight endpoint. The related discussion that was on page 5-10 of
the draft assessment has been revised to clarify the Agency's justification and address reviewer
concerns regarding the use of maternal (versus neonatal) blood levels of methanol as a basis for
the benchmark dose analysis of these data. In essence, the ratio of the difference in blood
concentrations between a human infant and mother is assumed to be similar to the approximate
two-fold difference that has been observed in rats. Further, while rat studies indicate that
postnatal exposure to methanol can impact brain weight, fetal exposure has been shown to have
the greatest influence on this endpoint. For these reasons and because EPA has confidence in the
ability of the PBPK model to accurately predict adult blood levels of methanol, the maternal
blood methanol levels for the estimation of FtECs from the NEDO (1987) study were used as the
dose metric. EPA has added text to Sections 4 and 5 to further clarify and discuss the limitations
of NEDO (1987).
NEDO (1987) observed brain weight reductions in the Fl and F2 generations of their two
generation study, in the Fl generation of the supplementary developmental study to the two
generation study and in a separate teratogenicity study. They also observed potentially adverse
histopathology (astrocytes) in the brains of monkeys receiving acute, subchronic and chronic
exposure to methanol (see further discussion in Section 4.4.2). While brain weight reduction has
not been observed in developmental bioassays of other laboratories, it has been observed in adult
rats exposed to methanol (TRL, 1986). Also, brain weight reduction is not an endpoint that has
been extensively measured or focused on in other developmental studies of methanol, such as the
Rogers et al. (1993b) mouse studies.
EPA agrees that the multiple t-tests applied in the NEDO study are not optimal for the
evaluation of the dose-response data from this study. For this reason, EPA did not rely on this
information and, instead, relied on the results of the more definitive benchmark dose analysis of
this data (Appendix D), as described in Section 5 of the final methanol toxicological review.
With respect to the use of absolute brain weight change without clinical or pathological
corroboration, the Agency's neurotoxicity guidelines (U.S. EPA, 1998a) states that a "change in
brain weight is considered to be a biologically significant effect," and further states that "it is
inappropriate to express brain weight changes as a ratio of body weight and thereby dismiss
changes in absolute brain weight" and that "changes in [absolute] brain weight are a more
reliable indicator of alteration in brain structure than are measurements of length or width in
fresh brain, because there is little historical data in the toxicology literature."
A-21
-------
With respect to the basis for EPA's study choice, while it is true that EPA guidelines
generally promote use of the more sensitive endpoint, the relative strengths of candidate studies
are not to be ignored (U.S. EPA, 2002, 1994). As discussed above and in Chapter 5 of the
methanol toxicological review, the NEDO (1987) study limitations were considered, but do not
preclude its use for the derivation of a candidate RfC/D. On the other hand, questions concerning
the Burbacher et al. (2004b; 1999b) monkey study dose-response are considered serious enough
to not use this study for RfC/D derivation, despite the possibility that a lower BMDL POD would
have been derived from this study (see Section 5.3.1 and Appendix D).
Comment 2: Regarding the Burbacher et al. (2004b; 1999b) monkey study, four
reviewers had no comment on its potential for use as the principal study and two reviewers stated
the following reasons why it should not be used as the principal study:
• "The lack of a dose-response function for the major effects."
• No "convincing evidence of an effect, given the inconsistencies in dose-response, multiple
comparisons, and the potential for unreliable identification of 'effects' in small studies."
However, one reviewer suggested that it would be a better choice than the NEDO rat
study because it "uses the most appropriate species (monkey) and examined a wide range of
reproductive and neurotoxicological endpoints and significant pharmacokinetic data," and two
reviewers suggested that the following limitations noted in the toxicological review were
overstated:
• Inclusion of wild-caught monkeys
• Influence of C-sections on results
• Not being relevant to persons who are folate deficient
• Lack of a dose-response for VDR in the male monkeys
Response: EPA agrees with reviewer comments regarding the significant difficulties of
assessing the dose-response data from the Burbacher et al. (2004b; 1999b) monkey study. These
concerns are addressed in Section 5.3.1 of the final review. In response to reviewer concerns,
EPA's attempt at performing a benchmark dose analysis of the Visually Directed Reaching
(VDR) endpoint from this study (described in Appendix D) are no longer presented in Chapter 5
(i.e., in Table 5-4 or 5-6) alongside the benchmark dose analyses of critical effects from the
candidate principal mouse and rat studies. With respect to the concerns that limitations in this
study were overstated, EPA has taken the following action:
• Inclusion of wild-caught or feral-born monkeys - One section of the draft toxicological
review inadvertently referred to monkeys from this study as being "wild" and this statement
A-22
-------
has been removed. In two sections, they were referred to as "a mixture of feral-born and
colony-bred animals." Since the Burbacher et al. (2004b; 1999b) study investigated for and
found no effects that were dependent on origin, EPA agrees that this statement is unnecessary
and it has been removed from the review.
• Influence ofC-sections on results - EPA agrees with the reviewers and the toxicological
review has been edited to reflect that Cesarean section (C-section) deliveries performed in
the methanol exposure groups did not impact the "decreased length of pregnancy" finding
(decreased length of pregnancy was observed in vaginally delivered animals).
• Not being relevant to persons who arefolate deficient - EPA agrees that this statement could
be made about most of the methanol studies reviewed. Hence the statement has been
removed.
• Lack of a dose-response for VDR in the male monkeys - While the ANOVA test in the male
monkeys suggests a statistical significant VDR change at 600 ppm (p = 0.007), there was no
significant difference between responses and/or variances (indicating lack of a dose-response
trend) among the dose levels for males only (p = 0.321), even when the high dose group is
excluded (p = 0.182). However, there was a significant dose-response trend for females only
(p = 0.0265). This is largely because the females had a larger overall sample size across dose
groups than males (21 females versus 13 males). Hence, only the VDR response for females
only exhibited a dose-response that could be adequately modeled (see Appendix D).
Comment 3: Regarding the Rogers et al. (1993b) mouse study, two reviewers supported
the use of this study over the NEDO rat study and noted the following advantages:
• "The study is scientifically sound and robust."
• "Exposures are limited to the prenatal period and the outcomes are clear."
• "The Rodgers (sic) study has undergone independent peer review, documents responses
reported by other laboratories, and has quite robust group sizes."
Response: EPA agrees with reviewer comments regarding the advantages of the Rogers
et al. (1993b) mouse developmental study. EPA also agrees with reviewer comments (see Charge
Bl Comment 1 above) regarding the advantages of the NEDO rat study, including the use of a
continuous, nearly full day exposure regimen and the adequacy of the reported response data for
dose-response analysis. As a result, EPA decided to treat both the Rogers and NEDO studies as
candidate principal studies and derived candidate RfCs and RfDs for the most sensitive endpoint
from each study (see Section 5.1.1.2).
A-23
-------
Charge Bl Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[i]n the current assessment and in responses
to comments, the U.S. EPA has more clearly and thoroughly described what it perceives as the
strengths of the NEDO study, making a better case for its selection as the principal study" and
that the Burbacher study is now more appropriately described. However, this and another
reviewer were not fully satisfied that EPAs use of BMD modeling fully addressed the suggestion
to reanalyze the brain weight data using more appropriate tests for statistical significance. Both
reviewers commented that without appropriate tests to determine if the NEDO study results
represent a statistically significant change, it is not clear that benchmark dose analysis is
warranted or reliable. One commenter stated that "Are-analysis using ANOVA would be easily
done and should be done, just to show validity of the NEDO data prior to using it for the BMD
analysis."
Response: EPA appreciates the affirmation from the reviewers of the revised study
characterizations. With respect to tests for statistical significance, EPA believes that trend tests
that use all of the dose-response data, in this case summarized continuous responses (means and
SDs), are more appropriate indicators of significance then pair-wise testing of the responses
within individual dose groups. With respect to whether the BMD modeling needs to be preceded
by other tests for statistical significance, EPA generally considers the results of the BMDS model
output to be sufficient to determine whether there is a significant increasing or decreasing trend
to the dose-response data. In the case of continuous data such as decreasing brain weight, the
results of four BMDS test results (described in detail in the BMDS Help manual available from
http://epa.gov/ncea/bmds) are considered. Test 1 is used to determine whether there are
significant differences among the means and, to some extent the variances, across dose groups.
Test 2 directly tests for homogeneity among the response variances. Test 3 is a test to determine
whether variances can be modeled as a power function of the mean. Test 4 is used to determine
whether the model adequately fits the mean responses. In general, Test 1, Test 4 and either Test 2
or Test 3 must pass for a set of dose-response data to be considered adequate for derivation of a
BMD. While none of the individual test results are suitable for determining whether there is a
significant overall (upward or downward) trend in the data, the combined test results are deemed
to be adequate for this purpose.
Nevertheless, it is true that none of the BMDS tests described above constitute a
traditional Analysis of variance (ANOVA) or trend test. Hence, in response to this comment EPA
has applied an ANOVA analysis for summarized response data (Larson, 1992) to all twelve of the
neonatal rat brain weight responses reported on page 202 of the NEDO (1987) report. A highly
A-24
-------
significant decreasing dose-response trend (p < 0.000001) was observed for all but the male and
female olfactory bulb weights, which were highly insignificant (p > 0.3).
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.
Summary of Comments: Four reviewers indicated that the use of brain weight change
was justified, but one of these and one other reviewer questioned the use of the 6-week time
point. Two reviewers suggest using the cervical rib endpoint from the Rogers et al. (1993b)
mouse study. One reviewer expressed a preference for endpoints from the Burbacher et al.
(2004b_; 1999b) monkey study or Rogers et al (1993b) mouse study over the NEDO rat study.
Specific comments or suggestions made by the reviewers with respect to this charge are
described below, along with EPA responses.
Comment 1: Two reviewers suggested that the increased incidence of cervical ribs
should serve as the critical effect for RfC derivation, with one stating that "the increases in
cervical ribs and supernumerary ribs observed in this Rogers et al. (1993b) study could be
considered a more scientifically justified critical effect."
Response: EPA agrees that the Rogers et al. (1993b) study is of high quality. In the final
assessment, it is considered a candidate principal study. The preference of these two reviewers
for the cervical rib endpoint seems to be based in part on perceived problems with the brain
weight change endpoint in rats (NEDO, 1987). The reviewer who stated that the cervical rib
endpoint "could be considered a more scientifically justified critical effect" pointed out that the
NEDO (1987) developmental rat study did not note "abnormal brain histopathology or functional
deficits" and a statistical analysis of the brain weight changes was performed that was questioned
in a separate peer review of this study that was conducted for EPA (ERG, 2009). As discussed in
the response to Charge Bl Comment 1, EPA neurotoxicity guidelines allow for the treatment of
absolute brain weight change as an adverse neurological effect regardless of the existence of
corroborating histopathological or functional observations, and EPA used benchmark dose
analyses in lieu of the statistical test results reported by the authors for this endpoint.
Comment 2: One reviewer commented that there is a "general lack of transparency"
regarding the basis for the selection of the critical effect and stated that EPA chose "the one that
led to the lowest RfC," without regard to the limitations of the NEDO study, including
A-25
-------
inappropriate use of statistical methods as described in a 2009 EPA-sponsored external peer
reviewer of the NEDO study (ERG, 2009). Another reviewer also commented that EPA had not
acknowledged errors in the NEDO statistical analysis and recommended that EPA conduct its
own analysis of variance (ANOVA) to determine "if there were an overall effect on brain
weight" and "which time frame and which methanol level are used in the BMD analysis."
Response: The basis for the selection of the candidate principal studies and effects are
primarily described in Section 5.1.1, Choice of Principal Study and Critical Effect(s). The critical
effects considered for the derivation of an RfC and RfD were chosen because they were reported
in studies of adequate quality, are considered relevant to humans, evidence a clear dose-response
and are sensitive indicators of alterations in important organ systems. The dose-response data for
the effects that meet these criteria (in this case the mouse cervical ribs and rat brain weight
effects) were considered for the derivation of the RfC and RfD. If EPA had based its selection on
the effect that "led to the lowest RfC" an endpoint in the Burbacher et al. (2004b; 1999b) or
NEDO (1987) monkey studies might have been chosen as some of the endpoints in these studies
suggested a lower NOAEL or BMDL. However, as described in Sections 4.2.2.3, 4.4.2, and
5.1.1, these monkey studies did not meet all of the criteria necessary for an effect to be
considered a critical effect. As discussed in response to Charge Bl Comment 1, the limitations of
the NEDO rat developmental study, including the inappropriate use of statistical methods, are not
serious enough to preclude its consideration as a candidate principal study. There is no need for
the Agency to perform an ANOVA analysis because a benchmark dose analysis was performed in
accordance with EPA guidelines (U.S. EPA, 2012a) for all postnatal time frames (3, 6 and 8
weeks).
Comment 3: Two reviewers were concerned that EPA did not consider other postnatal
time points besides 6 weeks, with one stating that this approach "weakens the potential statistical
power for a response that appears stable over a wide range of time points (3 to 8 weeks)."
Response: A benchmark dose analysis of brain weight reductions in male and female rats
was performed for all postnatal time frames (3, 6 and 8 weeks). In accordance with EPA
guidelines (U.S. EPA, 2002, 1998a), the most sensitive developmental time point in the most
sensitive gender was used as the basis for the RfC/D. In order to achieve an increase in statistical
power by combining data together from separate ages, the animals must be exchangeable
(required for Bayesian statistics) or represent the same population (i.e., the brain weights from 3-
8 week old S-D rats would have to represent the same population; required of frequentist
statistics). The data for the more sensitive gender (males) suggests that each age represents a
separate subpopulation (3 wks (mean ± standard deviation): 1.45g ± 0.06; 6 wks: 1.78g ± 0.07; 8
wks: 1.99g ± 0.06). Randomly permuting the individuals across the groups would not yield the
A-26
-------
same conclusions, proving a lack of exchangeability. Thus, combining these samples together
would in all likelihood violate the exchangeability and independent and identically distributed
(i.i.d) assumptions required of Bayesian and frequentist methods, respectively.
Comment 4: Two reviewers were concerned over the "lack of histological or functional
follow-up for this [brain weight] response."
Response: See response to Charge Bl Comment 1 regarding brain weight.
Comment 5: Two reviewers noted that EPA may need to reevaluate the endpoint
selection if modification of the PBPK analysis for S-D rats significantly alters the relative
sensitivity (based on HECs) of the rat, mouse and monkey studies.
Response: EPA agrees with this comment. Consideration has been given to whether the
modified PBPK model results warrant a change in the critical effect. While the final candidate
RfDs and RfCs from S-D rat brain weight response (NEDO, 1987) and the CD-I mouse cervical
rib response (Rogers et al., 1993b) are similar, the PBPK model modifications do change the
relative sensitivities such that the mouse study now serves as the basis for the methanol RfD. The
RfC is still based on the brain weight changes observed in the rat study.
Charge B2 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[i]n the current assessment and in responses
to comments, the U.S. EPA has more clearly and thoroughly described what it perceives as the
strengths of the NEDO study, making a better case for its selection as the principal study."
Response: EPA appreciates the reviewer's affirmation of the revised NEDO study
characterization. As indicate in response to the Charge B1 Follow-up Peer Review Comments,
EPA agrees that the EPA analysis of the NEDO rat brain weight data would benefit from, and has
therefore applied and described in the assessment, a more reliable trend test to confirm that the
positive dose-response trend is significant.
A-27
-------
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.
Summary of Comments: Four reviewers indicated that the BMD analysis was
appropriate and appropriately applied. Three reviewers said that this was not their area of
expertise. Five reviewers accepted the choice of AUC as the dose metric, but noted limitations in
the data available (MOA and empirical information) for making that choice. One reviewer
preferred Cmax over A UC as the dose metric and one reviewer did not comment on the selected
dose metric. Specific comments or suggestions made by the reviewers with respect to this charge
are described below, along with EPA responses.
Comment 1: With respect to the selection of AUC as the dose metric for the BMD
analyses of the brain weight endpoint from the NEDO developmental study in which rats were
exposed gestationally and postnatally, one reviewer stated that "Without understanding of the
exact mechanism of action of the chemical, selection of any surrogate dose metric is somehow
speculative." A second reviewer commented that "The Agency has not adequately explained its
rationale for the use of AUC rather than Cmax (e.g., see literature related to methanol and 2-
methoxyethanol)." A third reviewer noted that justification for the AUC is "tenuous" because
"brain weight does not differ between the 3, 6 and 8 week periods."
Response: When performing BMD analyses, it is important to choose a reliably
measured or estimated dose metric that has a close relationship to the health effects under
consideration. For the BMD analyses of the mouse cervical rib endpoint, which has been shown
to result from just one day of gestational exposure, it is assumed that the level of exposure is
more important than duration. Internal methanol blood concentrations reported by Rogers et al.
(1993b) for the dams of each dose group at day 6 of gestation were assumed to be approximately
equivalent to Cmax levels and were used as the modeled dose metric. For the BMD analyses of
the rat brain weight endpoint following gestational and lactational exposure, PBPK model
estimates of AUC methanol in blood for the dams of each dose group were used as the modeled
dose metric. As described in Section 5.1.2.1, the decision to use AUC as the dose metric for the
gestationally and postnatally exposed rats was made because under this exposure regimen, brain
weight is susceptible to both the level and duration of exposure. It is true that the results of
A-28
-------
NEDO (1987), described in Section 4.4.2 and shown in Table 4-13, indicate that there is not an
obvious cumulative effect of ongoing exposure on brain-weight decrements in rats exposed
postnatally for 3, 6 and 8 weeks. However, there is a greater brain-weight effect in rats exposed
postnatally versus only during organogenesis (GD7-GD17). Further, brain weight reductions
have been observed in adult rats that were exposed for 90 days beginning no earlier than 30 days
of age (TRL, 1986). That brain weight is susceptible to continued exposure beyond gestation
suggests that a dose metric that incorporates a time component would be more appropriate. For
this reason, and because it is more typically used in internal-dose-based assessments and better
reflects total exposure within a given day, daily AUC (measured for 22 hours exposure/day) was
chosen as the most appropriate dose metric for modeling the effects of methanol exposure on
brain weights in rats exposed throughout gestation and continuing into the Fl generation.
Comment 2: One reviewer asked why, for the purposes of Table 5-2 and the PBPK
estimation of AUC methanol in rat dam blood, the AUC was calculated with a 5 day 22 hr/day
simulation.
Response: The full text of the subject footnote is "AUC values were obtained by
simulating 22 hr/day exposures for 5 days and calculated for the last 24 hours of that period."
Simulations were run for 5 days as this was sufficient to reach "periodicity" when the daily time-
course is the same from one day of exposure to the next. From Figure B-13 it can be seen that
model predictions for the second day and beyond are essentially identical, but because the blood
level does not drop to zero during the 2 hour "off period, the AUC is higher on the 2nd day and
beyond than the first day. More importantly, the AUC was calculated for a single day of
simulated exposure, which happened to be the fifth day. With the PK parameters used for those
simulations, the same results would have been obtained if the simulation had only been run for
3 days, or for 30 days.
Comment 3: With respect to the alternative hypothesis that formaldehyde is the
teratogenic moiety and that increased effects of methanol in GSH-depleted animals are due to
decreased formaldehyde elimination, one reviewer noted that GSH depletion does not necessarily
imply formaldehyde involvement because "depletion of GSH, as the major cellular antioxidant,
will also increase the accumulation of reactive oxygen species (ROS)."
Response: The toxicological review has been revised (Section 4.7) to reflect that the
impact of GSH depletion can support both formaldehyde and ROS involvement in the
teratogenic effects of methanol. However, this reviewer and another reviewer agreed with the
Agency's position that methanol would play a key transport role in either case, with the latter
reviewer stating that "even if the metabolism-related formation of ROS or formaldehyde are
important contributors to the observed toxic effects, a methanol-based dose metric is applicable
A-29
-------
when the downstream metabolic processes such as removal of ROS or formaldehyde are much
faster than the rate-limiting oxidation of methanol."
Comment 4: One reviewer commented that "Neither the 5% nor the 10% BMR have any
particular a priori justification for continuous data: the default assumption in this case is the
BMR of 1 standard deviation of the control dataset (as preferred here). In any case the data need
to be examined to determine an appropriate BMR representing a minimal detection level or
threshold of biologically significant response: this especially applies for continuous data."
Response: The reviewer's comments with respect to the selection of a BMR are correct
and consistent with EPABMD Technical Guidance (U.S. EPA, 2012a). In the case of the
methanol toxicological review, all BMR levels considered for RfC or RfD derivation lie well
within the range of the dose-response observations. As indicated in the EPABMD guidance (U.S.
EPA, 2012a), a series of papers (Allen etal., 1994a, b; Faustman et al., 1994) suggest that a 5%
BMR is appropriate for dichotomous response data from well designed nested developmental
studies such as the Rogers et al. (1993b). For continuous response data, EPA guidance (U.S.
EPA, 2012a) suggests that "if there is an accepted level of change in the endpoint that is
considered to be biologically significant then that amount of change is the BMR." For
continuous response data from developmental studies, comparisons with the NOAEL showed
that several cutoff values, including a 5% change in mean fetal weight, could be used to give
values similar to the NOAEL (Kavlock et al., 1995). If a 5% change in fetal weight is considered
biologically significant, it is reasonable to assume that a 5% brain weight change should also be
considered biologically significant. However, in a recent report on the statistical power in the
analyses of brain weight measures in pesticide neurotoxicity testing, Weichenthal et al. (2010)
state that "if toxicological experts ultimately decide that brain weight changes in the range of 5%
are physiologically meaningful, a larger [than 10 per dose group] sample size will be needed to
consistently achieve reasonable power to detect this magnitude of effect." EPABMD guidance
(U.S. EPA, 2012a) states that "in the absence of any other idea of what level of response to
consider adverse, a change in the mean equal to one control SD from the control mean can be
used." Because there is no clear biological basis for choosing one over the other, both are
considered and deference is given to the BMR that results in the lower RfC or RfD (see Tables 5-
4 and 5-6).
Charge B3 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: No additional comments were made regarding this charge question.
However, a related comment regarding the use of methanol concentration in maternal blood as
A-30
-------
surrogate dose metric for evaluating postnatal changes is addressed under the Charge A3 Follow-
up Peer Review Comments in Section A. 1.1, "Toxicokinetics and PBPK Modeling."
Response: EPA has responded to the comment regarding use of methanol concentration
in maternal blood as surrogate dose metric for evaluating postnatal changes under Charge A3
Follow-up Peer Review Comments in Section A. 1.1, "Toxicokinetics and PBPK Modeling."
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.
Summary of Comments: In general, four reviewers indicated that the selected UFs are
adequate and consistent with EPA policy and three reviewers did not agree with certain UFs. Of
the four reviewers that generally agreed with EPA's proposed UFs, one suggested that some
further examination and discussion of the UFn would be helpful and another noted that a strong
argument could be made for eliminating the UFr>. Of the three reviewers that expressed
disagreement, all three stated that the 3-fold UFD was not necessary, and one suggested that the
UFn of 10 is not warranted. With respect to where the UFs are applied, one reviewer supported
the Agency's practice of applying UFs to the HEC and one advocated application of UFs to
BMDLs (before HEC derivation). Specific comments or suggestions made by the reviewers with
respect to this charge are described below, along with EPA responses.
Comment 1: Regarding the UFn, one reviewer stated that a full UFn of 10 is not
warranted because "at the level of the proposed RfC and RfD, intraspecies differences in
disposition of exogenous methanol in humans will likely have no meaningful impact on the body
burden of'total' methanol." This reviewer recommended that EPA perform a sensitivity analysis
of the human PBPK model to identify variability and/or uncertainty in parameters which have an
impact on methanol levels predicted. Also, this and another reviewer stated that the UFn does not
need to account for uncertainty regarding the sensitivity of children because the critical study is
in neonates, two-generation study exists, and "no particular developmental susceptibility of
humans versus test species is expected." Another reviewer questioned whether a UFn of 10 was
"sufficient in the general case" and recommended further examination and discussion to establish
"the limits of the data available to inform the decision on the value for UFH."
Response: A sensitivity analysis of the human PBPK model has been performed (see
Appendix B), and the results suggest that parameter variability is not likely to result in methanol
A-31
-------
blood level estimates that vary more than 3-fold, the toxicokinetic portion of the 10-fold UFH.
However, one needs to also consider the variation in endogenous/background levels of methanol
(Table 3-1), and variation in toxicodynamics, because both may affect the impact of an
exogenous methanol exposure. Overall, the extent of human interindividual variation in
(endogenous and exogenous) methanol toxicokinetics and toxicodynamics would be very
difficult to quantify given the significant uncertainties that exist regarding background levels and
methanol's mode of action.
Toxicodynamic variability can only be discussed qualitatively. As discussed in Section
4.9, there are a number of issues that may lead to sensitive human subpopulations. Potentially
sensitive subpopulations would include individuals with polymorphisms in the enzymes involved
in the metabolism of methanol and individuals with significant folate deficiencies. The effects
used to derive the candidate RfCs are observed in a potentially susceptible and sensitive
fetal/neonatal subpopulation. However, there is also variability across fetuses and neonates that
need to be taken into account. Children vary in their ability to metabolize and eliminate methanol
and in their sensitivity to methanol's toxic developmental effects. Consequently, there exists
considerable uncertainty pertaining to human population variability in methanol metabolism,
which provides justification for the 10-fold intraspecies UF used to derive the RfC and RfD.
Comment 2: Regarding the UFA, one reviewer stated that "it is surprising that the EPA
used the same interspecies UFA for rodent and nonhuman primate studies - given the fact that
significant species difference exist between rodents and humans and less so between monkeys
and people (use UF = 1)."
Response: As discussed in response to Charge Bl Comment 2, due to uncertainties in the
dose-response data for the monkey studies, EPA has removed the alternative RfC derivation for
monkeys from the toxicological review.
Comment 3: Regarding the UFD, three reviewers provided the following reasons why a
3-fold UFD was not needed:
• Methanol has a "very rich toxicology database."
• "There is never enough data to be certain regarding a risk 'assessment' (that is why it is
called risk assessment not a risk determination)."
• Conservative assumptions are "always used," including:
o use of a single SD for BMDL rather than a 4 or 10% changes as commonly used
in some noncancer risk assessments (e.g., see page 5-23),"
o "PBPK assumes the most conservative scenarios,"
A-32
-------
o "BMD analysis itself favors the conservative numbers" and
o "when given the choice of alternative BMD numbers such as those obtained from
the 3 versus 6 versus 8 week data, the lowest (i.e., most conservative) number is
chosen."
• "The key endpoint is developmental toxicity, which has been evaluated in multiple species,
including primate, and special endpoints such as neurotoxicity and immunotoxicity have
been evaluated."
• "There is no need to have a UF because 'there is uncertainty regarding which test species is
most relevant to humans'—the lowest, high-quality point of departure was used."
• "There is also no need to have a UFo for "dose spacing" because the BMD analysis counters
this potential design deficiency.
Response: The database uncertainty factor accounts for the potential to underestimate
noncancer hazard as a result of data gaps. EPA agrees that the database for methanol toxicity is
quite extensive: there are chronic and developmental toxicity studies in rats, mice, and monkeys,
a two-generation reproductive toxicity study in rats, and neurotoxicity and immunotoxicity
studies. However, as discussed in Section 5.1.1.1, chronic and developmental studies in
monkeys, the species most likely to best represent the potential for developmental effects in
humans, were considered inadequate or inferior to the candidate principal rodent studies for the
purposes of RfC/D derivation. As discussed in Sections 5.1.3.2.3 and 5.3.6, the lack of a
quantifiable monkey study is an important data gap given the potential relevance to humans and
the uncertainties raised by existing monkey studies regarding this species sensitivity to
reproductive effects (e.g., shortened pregnancies discussed in Section 4.3.2), CNS degeneration
(e.g., stellate cell fibrosis described in Section 4.4.2) and delayed neurobehavioral development
(e.g., VDR response described in Section 4.4.2) from methanol exposure. In addition, a full
developmental neurotoxicity test (DNT) in rodents has not been performed and is warranted
given the critical effect of decreased brain weight in rats and the suggestive (but quantitatively
inconclusive) DNT results in monkeys. For these reasons, an UF of 3 was applied to account for
deficiencies in the database.
Comment 4: Regarding the application of all UFs, one reviewer stated that EPA should
"apply the uncertainty factors to the internal dose point of departure, prior to interspecies
extrapolation with the pharmacokinetic model to account for non-linearities in external versus
internal dose relationships." This reviewer suggested that EPA should discuss their choice of
applying UFs to the HEC/D rather than the BMDL. The reviewer estimated that if UFs are
applied first to the mouse cervical rib BMDL0s, then converted to the candidate RfC using the
A-33
-------
PBPK model, the candidate RfC would increase by more than 2-fold. Another reviewer noted
that the application of UFs to the HEC/D values is the standard procedure and is "preferred to
alternative suggestions that the UFs be applied to intermediate measures such as blood
concentrations or AUCs."
Response: The first reviewer is correct in that, after modifications were made to the rat
PBPK model (see Response to Charge Al Comment 7), BMDL estimations from both the rat and
mouse candidate principal studies are not within the linear range of EPA's PBPK model
predictions. EPA has reevaluated the analysis and applied the UFs prior to HEC/D derivation as
suggested. This approach results in more scientifically reliable model predictions by lowering the
BMDLs to within the more linear, calibrated range of the human PBPK model. Clarifying text
has been added to the Sections 5.1.3.2 and 5.2.2.3.
The concern expressed by the second reviewer regarding departure from EPA practice is
recognized, given the uncertainty associated with dividing internal dose BMDLs by UFs that are
at least partially based on empirical analyses of ratios of NOAELs obtained from external oral
exposures (U.S. EPA, 1994; Dourson and Stara, 1983). In the methanol (noncancer) assessment,
the general EPA practice of applying the human PBPK model to derive HEC/D values prior to
applying UFs (U.S. EPA. 2002. 1994) would result in RfC/Ds lower than if the PBPK model was
used to derive HEC/D estimates after dividing the BMDL internal doses by UFs. However, this
general practice if applied to methanol would result in greater model uncertainty because the
HECs (1,042 to 1,604 mg/m3) and HEDs (133 to 220 mg/kg-day) estimated from the BMDLs by
the revised PBPK model are well above the inhalation concentrations (655 mg/m3) and oral
exposures (50 mg/kg-day) for which there are human data to calibrate the PBPK model (see
Appendix B, Section B.2.7, Table B-6).
Charge B4 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[t]he authors of the document should be
commended for breaking off with almost ten-year-old, scientifically indefensible U.S. EPA
tradition of applying all uncertainty factors (UF) to the external human equivalent concentration
(HEC estimate)." This reviewer stated that "[e]ven if, within the validated range of a PBPK
model, the pharmacokinetics of a chemical appears to be almost linear, by the virtue of
potentially saturable mechanisms of absorption, metabolism and excretion - it is still prudent to
apply UFA, UFn, or perhaps in this case UFo, to the internal dose (paraphrasing Paracelsus: 'this
is the internal dose that makes a poison')."
Response: EPA appreciates the reviewer's affirmation of the Agency's decision to apply
UFs to the internal dose BMDL PODs for methanol. This approach has been retained in this
A-34
-------
assessment, but it should be recognized that there are chemical-specific circumstances
warranting this approach for the methanol (noncancer) assessment that may not pertain to other
chemical assessments.
Comment 2: Two reviewers thought that the lack of a quantifiable primate study was
used to justify both the UFA and UFD. One of these reviewers agreed with the public comment
that "[t]he lack of primate data should only be applied to one or the other, as this 'double
counting' adds yet another forced conservatism to the assessment" and suggested that EPA
should ".. .use the incomplete primate data once - to justify only UFA" and "[c]onsider using lack
of information on MO A to justify UFo." The other reviewer commented that "It seems to me that
the uncertainty contributed by limitations in knowledge about species sensitivity (per the monkey
studies) has in effect been double counted in the overall UF."
Response: In the revised Section 5.1.3.2.3, EPAhas clarified that the UFD is based on
deficiencies in the methanol toxicological database, particularly with respect to the interpretation
of the importance and relevance reproductive, developmental neurotoxicity and chronic CNS
effects observed in monkeys. The Agency has determined that a 3-fold UFD is necessary to
account for the possibility that a lower RfD/RfC might have been derived if additional data were
available. This is consistent with EPA (2002) guidance, which states that:
"The database UF is intended to account for the potential for deriving an underprotective
RfD/RfC as a result of an incomplete characterization of the chemical's toxicity. In
addition to identifying toxicity information that is lacking, review of existing data may
also suggest that a lower reference value might result if additional data were available.
Consequently, in deciding to apply this factor to account for deficiencies in the available
data set and in identifying its magnitude, the assessor should consider both the data
lacking and the data available for particular organ systems as well as life stages."
Additional studies to inform the MOA for the reproductive, developmental, and
neurological effects of methanol would be helpful and, as discussed in Section 5.1.3.2.3, were
suggested by peer reviewers of the monkey studies. However, uncertainty regarding the MOA is
generally considered for the purposes of determining the UFA.
Comment 3: Two reviewers questioned the need for a 3-fold UFD. One of these
reviewers commented that the EPA acknowledges that the database for methanol toxicity is
'quite extensive...'," ".. .the uncertainty contributed by limitations in knowledge about species
sensitivity (per the monkey studies) has in effect been double counted in the overall UF," "[t]he
absence of a full DNT test is rodents is stated to be important in part because of the critical effect
of decreased brain weight in rats, but I still have reservations about the strength of that finding,"
A-35
-------
and ".. .the report is responsive in terms of providing a clearer case for a database uncertainty
factor of 3, but I still question whether it is needed." The other reviewer stated that "[t]he
existing database on the developmental toxicity of methanol is sufficiently robust as to set the
UFd at 1," "... [the monkey studies] do corroborate the principal study and hence they offer
robust data that obviates the need for [a] UFd," and ".. .it isn't logical to consider these studies as
being necessary [to corroborate the principal study] and simultaneously insufficient [for deriving
andRfC]"
Response: EPAhas clarified in the revised Section 5.1.3.2.3 that while the database for
methanol toxicity is extensive in terms of the laboratory species and study design coverage,
consisting of chronic and developmental toxicity studies in rats, mice, and monkeys, a two-
generation reproductive toxicity study in rats, and neurotoxicity and immunotoxicity studies, it
leaves considerable uncertainty with respect to the importance and relevance of reproductive,
developmental and chronic effects observed in monkeys. As discussed in Section 5.1.1.1, the
available monkey studies are considered inadequate or inferior to the candidate principal rodent
studies for the purposes of RfC/D derivation. EPA agrees that this deficiency in the dose-
response data would not normally warrant a UFD given the scope of the existing database and the
qualitative value of the chronic and developmental monkey studies for hazard identification.
However, this deficiency is of particular concern for methanol given (1) metabolic similarities
that suggest monkeys should most closely represent the potential for effects in humans (see
Section 3.1) and (2) uncertainties regarding the importance and relevance of the monkey effects
(see Section 5.1.3.2.3).
The UFD does not have the same basis as the UFA, and was not "double counted" in the
overall UF. As stated above, consistent with EPA (2002) guidance , the Agency has determined
that a 3-fold UFo is necessary to account for the possibility that a lower RfD/RfC might have
been derived if better or additional data were available. EPA guidance places particular emphasis
in this regard on database deficiencies in the area of developmental toxicity, the primary focus of
the methanol (noncancer) assessment, stating that "[i]f data from the available toxicology studies
raise suspicions of developmental toxicity and signal the need for developmental data on specific
organ systems (e.g., detailed nervous system, immune system, carcinogenesis, or endocrine
system), then the database factor should take into account whether or not these data are available
and used in the assessment and their potential to affect the POD for the particular duration RfD
or RfC under development." As described in Section 5.1.3.2.3, NTP-CERHR (2004. 2003) and
HEI (Burbacher et al.. 2004a: 2004b: 1999a: 1999b) peer reviews of the monkey
reproductive/developmental studies and an EPA-sponsored peer review (ERG, 2009) of the
NEDO (1987) acute and chronic monkey studies were uncertain about the relevance of the
A-36
-------
effects observed, but all of the reviews signaled that the observed effects should not be ignored
and suggested additional research that might help resolve some of the uncertainty.
The reviewer's concern regarding the strength of the brain weight finding in the NEDO
rat study has been addressed in response to Charge Bl Follow-up Peer Review Comment 1.
Further, With respect to the developmental neurotoxicity (DNT) from methanol inhalation
exposure, Table 5-5 of Section 5.1.3.2.3 indicates that methanol blood levels associated with
DNT effects are a 12-fold higher in rodents versus primates. Some of this dissimilarity may be
due to differences in species sensitivity, for which the UFA of 3-fold is intended to account, but
some of the difference may be due to other factors, including whether appropriate and
comparable endpoints were examined and whether appropriate study designs and quality control
measures were used. To account for these additional uncertainties, a 3-fold UFD is applied.
Finally, for comparison purposes, EPA has performed an analysis of the alternative RfD
and RfC that would have been derived if a UFo of 1 had been applied instead of a UFo of 3.
Tables A-l and A-2 correspond to Tables 5-4 and 5-6 of the assessment and demonstrate that a
UFD of 1 would have resulted in an RfC of 60 mg/m3 and an RfD of 6 mg/kg-day (after rounding
to single digit significance). The EPA has decided against this approach because it believes there
is ample evidence that additional data from appropriate studies could result in the derivation of a
reference values considerably lower than 60 mg/m3 and a 6 mg/kg-day. With respect to DNT
effects, in addition to the 12-fold higher methanol blood LOAELs in rodents versus primates
noted above, a BMD analysis of the VDR DNT effect reported by Burbacher et al. (2004b;
1999b) resulted in a methanol blood Cmax BMDL of 19.59 mg/L, less than half the methanol
blood level PODs shown in Table A-l and A-2 for rodents. Potential chronic neurotoxicity
(fibrosis of "responsive stellate cells") were reported by NEDO (1987) at a 100 ppm exposure
level that EPA's monkey PK model estimates corresponds to a methanol blood level of 3 mg/L.
While Figure 5-4 of the assessment illustrates that a RfD or RfC exposure would not increase the
methanol blood levels of anyone with an background methanol blood level below 2.5 mg/L to
above 3 mg/L (see discussion in Section 5.3.6), Figure A-l shows that a 6 mg/kg-day alternative
RfD or a 60 mg/m3 alternative RfC exposure would result in blood levels above 3 m/L for a
substantial percentage (-25%) of those who started out with background blood levels below 2.5
mg/L. These analyses suggest that an alternative RfD of 6 mg/kg-day and an alternative RfC of
60 mg/m3 associated with a UFD of 1 are potentially under protective with respect to the DNT
and/or chronic neurotoxicity of methanol.
A-37
-------
Table A-l Summary of PODs for critical endpoints, application of UFs and conversion to
candidate RfCs using PBPK modeling.
BMDL = PODinternal
RfCinternal = PODinternal/UFsa
|RfC (mg/m3)b
Rogers et al. (1993b)
mouse cervical rib Cmax
10% BMR 5% BMR
90.9 mg/L 43.1 mg/L
3.03mg/L 1.44 mg/L
134.5 65.7
NEDO (1987)
rat brain weight ADC
5% BMR 1 SD BMR
1,183mg-hr/L 858 mg-hr/L
39.4 mg-hr/L 28.6 mg-hr/L
74.6 54.6
aUFA =3; UFD = 1; UFH = 10; UFS = 1; UFL = 1; product of all UFs = 30.
bEach candidate RfC is the inhalation exposure concentration predicted to yield a blood concentration equal to its corresponding
RfCintemai, using the human PBPK model with an background blood concentration of 2.5 mg/L, which corresponds to the estimated
maximum background exposure rate of 1,600 mg/day (COT. 2011) in a 70-kg person (see Section 5.3.6); the final RfC is rounded to
one significant figure.
Table A-2 Summary of PODs for critical endpoints, application of UFs and conversion to
candidate RfDs using PBPK modeling.
Rogers et al. (1993b)
(mouse cervical rib Cmax)
10% BMR 5% BMR
BMDL = PODmtemai 90.9 mg/L 43.1 mg/L
RfDinternai = PODintemai/UFsa 3.03 mg/L 1 .44 mg/L
|RfD (mg/kg/day)b 12.7 6.2
NEDO (1987)
(rat brain wt.AUC)
5% BMR 1 SD BMR
1,1 83 mg-hr/L 858 mg-hr/L
39.4 mg-hr/L 28.6mg-hr/L
16.4 12.1
aUFA=3; UFD = 3; UFH = 10; UFS = 1; UFL = 1; product of all UFs= 100; see Section 5.1.3.2 below for details.
bEach candidate RfC is the inhalation exposure concentration predicted to yield a blood concentration equal to its corresponding
RfCintemai, using the human PBPK model with an background blood concentration of 2.5 mg/L, which corresponds to the estimated
maximum background exposure rate of 1,600 mg/day (COT, 2011) in a 70-kg person (see Section 5.3.6); the final RfC is rounded
to one significant figure.
A-38
-------
Estimated high end
of uncontaminated
background blood
levels
Sample Background Blood Level Distribution
Sample Background + Peak RfD
Sample Background + RfC
Nfep (1987)
Chr&hic monkey study; 100 ppm;
21 hrs'fcjay; Minimal fibrosisof
"response stellate cells;" possibly
astrocytesSfcERG 2009)
Burbacher et al. (2004)
Pregnant monkeys;
200pprm; 2 hrs/day;
Shortened pregnancy
duration
2 2.5 3 3.5 4
mg Methanol/Liter Blood (mg/L)
Note: References in this figure refer to the NEDO (1987) and HEI (Burbacher et al.. 2004a: 2004b: 1999a: 1999b) reports and the
EPA sponsored external peer review (ERG. 2009) of the NEDO (1987) report.
Figure A-l Relationship of monkey blood levels associated with effects of uncertain
adversity with projected impact of daily peak alternative RfC and RfD
exposures [derived using alIFD of 1] on sample background methanol blood
levels (mg MeOH/Liter [mg/L] blood) in humans.
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.
Summary of Comments: Six reviewers indicated that the approach taken by EPA was
appropriate and one reviewer did not comment due to a lack of expertise. Specific comments or
suggestions made by the reviewers with respect to this charge are described below, along with
EPA responses.
A-39
-------
Comment 1: One reviewer recommended that EPA "provide alternative RfC estimates
that would be derived using traditional approaches."
Response: Since this comment was made in response to the oral RfD charge, it is
assumed that the reviewer is requesting that EPA provide traditional RfD estimates and not
"RfC" estimates. None of the oral studies provided sufficient dose-response data for a dose-
response analysis and none of the developmental toxicity studies identified a NOAEL for use in a
traditional RfD estimate. The only NOAEL identified was 500 mg/kg-day from the subchronic
oral study in adult rats CTRL, 1986). As discussed in Section 5.2.4, the previous IRIS assessment
of methanol divided this NOAEL by a 1,000-fold uncertainty factor to obtain an RfD of
0.5 mg/kg-day. This value is lower than the current proposed RfD of 2 mg/kg-day, largely
because a 10-fold higher uncertainty factor was employed in the previous assessment.
Comment 2: One reviewer stated that "Human model validation using the oral data of
Schmutte et al. (1988) (see Charge D2) could further strengthen confidence in the route-to-route
extrapolation."
Response: EPA agrees and as discussed in response to Charge Al Comment 11, this oral
study has been added to EPA's PBPK analysis and used in the validation of the oral human PBPK
model.
Charge Cl Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: No additional comments were received regarding this charge question.
Response: EPA has retained the route-to-route extrapolation approach in the assessment.
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.
Summary of Comments: Five reviewers stated that the approach taken by EPA was
appropriate and a sixth reviewer did not comment due to a lack of expertise. A seventh reviewer
cited the lack of gestational and lactational components as a weakness in the EPA approach. All
A-40
-------
reviewers either referred to or repeated previous comments on the PBPK model and the RfC
derivation approach.
Summary Response: The reviewers did not offer any new comments in response to this
charge question that were not covered in response to previous charge questions.
Charge C2 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: No additional comments were received regarding this charge question.
Response: EPA has retained the route-to-route extrapolation approach in the assessment.
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.
Summary of Comments: All but one reviewer agreed with the use of the same UFs for
the RfD as for the RfC. One reviewer stated that this was "unexpected" because the database for
oral and inhalation are very different.
Summary Response: The critical effects were systemic, developmental effects that are
assumed to be dependent on blood concentrations of methanol. EPA was able to use methanol
blood concentrations in its benchmark dose analyses of the critical effects in the candidate
principal studies because blood levels were either reported in the study or could be estimated
using a validated PBPK model. After application of UFs, a validated human PBPK model was
then used to convert the adjusted benchmark dose estimates to an RfD and RfC. For these
reasons, EPA was able to derive the oral RfD and inhalation RfC with a similar degree of
confidence using the same data set, endpoint, BMD methods and PBPK model.
Charge C3 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: No additional comments were received regarding this charge question.
Response: EPA will continue to apply the same UFs for the RfC and RfD.
A-41
-------
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?
Summary of Comments: In general, reviewers commented that the Toxicological review
was logical, comprehensive and clear, but not concise. One reviewer stated that the review "is
thorough and well written, and takes care to provide descriptions of the available evidence in a
clear, complete, and unbiased form" and "presents a careful and well justified synthesis of these
data. " However, five of the seven reviewers criticized the repetitive or redundant nature of the
review and four reviewers were critical of the review format, with one reviewer stating that "a
different format could be much more effective in conveying critical information, interpretations,
and decisions regarding available, relevant toxicological literature. " Specific major (non-
editorial) suggestions made by the reviewers with respect to this charge are described below,
along with EPA responses.
Comment 1: One reviewer suggested that EPA add a "decision tree" to make choices for
major decisions more transparent.
Response: In light of this concern, an Executive Summary has been added to the
beginning of the toxicological review which makes the choices for major decisions more readily
apparent and transparent. Exposure-Response arrays for oral and inhalation toxicity studies were
also added as Figures 4-1 and 4-2 to better depict the relationship of NOAELs and LOAELs in
the overall database of studies.
Comment 2: Two reviewers suggested that EPA make edits consistent with recent NRC
(2011) recommendations for the draft EPA formaldehyde assessment to:
a. Reduce text volume, narrative approach, redundancies and inconsistencies,
b. rely more heavily on tables and not repeat individual study descriptions,
c. include "inclusion and exclusion criteria" for cited references, and
d. reduce "extraneous information" contained in Appendices.
Response: In response to this comment EPA has made the following format and text edits
to the toxicological review:
a. In response to the suggestion to "reduce text volume, narrative approach,
redundancies and inconsistencies," EPA has extensively condensed Section 3.4
(e.g., abbreviating the discussion of model structure and deleting detailed
A-42
-------
discussion of model parameter and model calibration in deference to Appendix
B), combined Section 4.6 "MECHANISTIC DATA AND OTHER STUDIES IN
SUPPORT OF THE MO A" with Section 4.8 "NONCANCER MOA
INFORMATION" into a new Section 4.7, deleted or merged redundant portions
of Section 5.3 "UNCERTAINTIES IN THE INHALATION RFC AND ORAL
RFD" with Section 5.1.3.2 "Application of UFs," removed portions of Section 5
that were unnecessarily redundant with Appendix D, revised and consolidated
portions Section 5 related to the justification of the dose metric (AUC versus
Cmax versus total metabolites) employed for the BMD analyses of candidate
principal studies and removed Section 6 (in lieu of a new Executive Summary).
b. In response to the suggestion to "rely more heavily on tables and not repeat
individual study descriptions," EPA has created new tables for whole embryo
studies described in Section "4.3.3 Other Reproductive and Developmental
Studies" and the i.p. studies described in Section "4.4.3 Neurotoxicity Studies
Employing In Vitro and I.P. Methanol Exposures" and has edited all Sections to
reduce unnecessary repetition of individual study descriptions. EPA has also
added an exposure-response array (Figure 4-1) to the assessment.
c. In response to the suggestion to include 'inclusion and exclusion criteria' for cited
references," EPA has added text to the Preface of the methanol toxicological
review that describes how EPA evaluates the quality of studies.
d. In response to the suggestion to "reduce 'extraneous information' contained in
Appendices," EPA has removed Appendix D, having determined it to be
extraneous due to the stronger biological basis for the choice of a Cmax dose
metric described in Section 5.1.2.1, and removed the source code text from
Appendix B. The source code text is posted on the EPA HERO database (U.S.
EPA, 2012b), as part of a Windows zip file containing a complete package of
acslX code necessary to run all of the developed models.
Comment 3: One reviewer suggested that "Table 3-3 should include the Dorman
cynomolgus monkey study with a clear indication that it involved lung only exposure of
anesthetized monkeys."
Response: Table 3-3 has been revised to include the Dorman cynomolgus monkey study
in response to this comment.
A-43
-------
Comment 4: One reviewer asked whether EPA considers the Fagan test results from the
Burbacher et al. (2004b; 1999b) monkey study to be "biologically significant despite the lack of
a statistically significant response?"
Response: There is uncertainty regarding both the biological and statistical significance
of the Fagan test results from the Burbacher et al. (2004b; 1999b) monkey study. As explained in
Section 4.4.2,
"Unlike the VDR results discussed previously, results of this test did not appear to be
gender specific and were neither statistically significant (ANOVA p = 0.38) nor related to
exposure concentration. The findings indicated a cohort effect which appeared to reduce
the statistical power of this analysis. The authors' exploratory analysis of differences in
outcomes between the 2 cohorts indicated an effect of exposure in the second cohort and
not the first cohort due to higher mean performance in controls of cohort 2 (70% + 5%
versus 55% ± 4% for cohort 1). In addition, this latter finding could reflect the inherent
constraints of this endpoint. If the control group performs at the 60% level and the most
impaired subjects perform at approximately the 50% chance level (worse than chance
performance would not be expected), the range over which a concentration-response
relationship can be expressed is limited."
However, the Fagan test results cannot be ignored and, as described in Section 5.1.1.2.2:
"Although not statistically significant and not quantifiable, the results of this test
need to be considered, in conjunction with VDR test results and brain weight
changes noted in the NEDO (1987) rat study, as a possible indication of CNS
effects."
Comment 5: One reviewer found Section 3.4.2.4 confusing, and suggested that other
models that have been developed with inhaled manganese (Schroeter et al., 2011; Yoon et al.,
2011; Yoon et al., 2009a, b) "could form the basis for a gestational and lactational model."
Response: To reduce text volume in response to Charge Dl Comment 2a, Section 3.2.4
has been removed from the toxicological review. It contained an unnecessary discussion of the
rat and human isopropanol models described by Gentry et al. (2003; 2002) and Clewell et al.
(2001). It was originally included because it was thought to be a possible guide had EPA decided
to develop a more complex gestational and lactational model. The reviewer is right in that, had
EPA decided to take this approach, other gestational and lactational models, such as the one
developed for manganese, could have been considered. However, EPA has determined, and the
peer reviewers generally agreed (see "Summary of Comments" under Charges A3 and A5), that
such a model was unnecessary for the purposes of the methanol toxicological review.
A-44
-------
Comment 6: One reviewer stated that the "the discussion of a two compartment stomach
(page 3-28 and elsewhere) for rodents need additional justification (squamous and epithelial
portions?)" and questioned whether this structure is "appropriate for people (as indicated on page
3-51)."
Response: EPA agrees with the reviewer and, in response, has simplified the GI
absorption model and revised the associated text in the toxicological review. In particular, the GI
model for humans has been reduced to a single, first order compartment and rate (see Appendix
B, Section B.2.6).
Comment 7: One reviewer commented that the use of "terms that describe model fits as
'quite poor' (e.g., see page 3-40 and elsewhere)" need to be "better clarified (visual inspection,
goodness of fit, other?)."
Response: Except where numerical measures of fit are given, all such references to
model fit reflect visual inspection. This has been clarified in the toxicological review.
Comment 8: One reviewer requested that EPA "pick one set of units (ppm would be
preferred until calculation of the actual RfC value)."
Response: In general, both units are given, with mg/m3 values provided parenthetically
after the ppm values, except for RfC/D and point of departure (e.g., BMDL) values discussed in
Section 5.
Comment 9: One reviewer requested a discussion of the use of alcohol dehydrogenase
inhibitors as a clinical 'antidote' on "page 4-7 (and possibly elsewhere)."
Response: Explanatory text has been added on page 4-1 and 4-4 to explain that infusion
of ADH1 inhibitors such as ethanol or fomepizole (4-methylpyrazole) can serve as treatment for
methanol poisoning.
Comment 10: One reviewer asked whether the folate deficiency described on page 4-40
affects methanol concentrations significantly, and "which data support this conclusion?"
Response: Folate is the coenzyme of tetrahydrofolate synthetase, an enzyme that is rate
limiting in the removal of formate. However, there is limited evidence regarding how folate
deficiency would impact methanol and formaldehyde levels. Hence, the statement on page 4-40
of the draft assessment, that "Folate deficiency would be expected to cause potentially toxic
levels of methanol, formaldehyde, and formate to be retained" has been revised in the final
assessment (Section 4.3.2), to read" "Folate deficiency would be expected to cause potentially
toxic levels of formate to be retained."
A-45
-------
Comment 11: One reviewer recommended that EPA remove the Section 4.1 discussion
of the CNS effects produced by acute methanol overdosing because it could be perceived as an
inappropriate and "biased way to validate the subsequent choice of the NEDO study (decrease in
brain weights suggesting a methanol-induced CNS effect)."
Response: Because of the limited usefulness of human case study information to this
assessment, this portion of Section 4.1 was moved to a new Appendix C. However, the remainder
of Section 4.1 is retained because it contains important information relevant to the acute toxicity
of methanol and is one of the only sections in the toxicological review for which human data are
available. It is recognized that the CNS effects from acute exposure to methanol are likely the
result of a different mode of action than methanol's developmental effects. This is discussed in
several places in the toxicological review, particularly Section 4.7 on the MOAfor noncancer
effects.
Comment 12: One reviewer suggested that EPA needs to improve the synthesis of S-D
rat toxicokinetic data for purposes of PBPK model development.
Response: This has been done and Appendix B has been revised accordingly.
Comment 13: One reviewer suggested that EPA correct inconsistencies between the
toxicokinetics section of Section 3 and Appendix B.
Response: To avoid redundancy and address inconsistencies, the PBPK discussions in
Section 3 have been removed and the reader is referred to Appendix B for technical details.
Comment 14: One reviewer noted that the "clarity of the document is hampered by the
lack of a clear synthesis of evidence regarding plausible modes of action for developmental
toxicity."
Response: The mode of action discussions previously divided between Section 4.6 and
4.8 have been revised for clarity and consolidated into Section 4.7.
Charge Dl Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer stated that "[t]he overall document is much more concise and
direct in presenting the key features of the risk assessment that has been conducted." Another
reviewer stated that "The organization and presentation of the information is greatly improved,
creating a much more readable document" and that "[t]he information seems to flow better,
redundancies are minimized, and tables are used to more effectively summarize information."
Another reviewer stated that "[t]he revised (May, 2013) version of 'Toxicological Review of Methanol
A-46
-------
(non-cancer)' has been improved significantly in comparison to its external peer-review draft (2011)
version."
Response: EPA appreciates the affirmation of the Agency's revisions to make the
methanol (noncancer) toxicological review more clear and concise. The revised format, including
the Executive Summary and increased use of tables and appendices, has been retained.
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 noncancer health effects of methanol.
Summary of Comments: Three reviewers identified seven additional studies for EPA to
consider (Miller and Wells, 2011; Leavens et al, 2006; Dorman et al, 1995; Bolon et al, 1994;
Bolon et al, 1993; Haffner et al., 1992; Schmutte et al., 1988). Specific comments or suggestions
made by the reviewers with respect to this charge are described below, along with EPA
responses.
Summary Response: The identified papers were evaluated and are now discussed and
referenced in the final assessment. As discussed in response to Charge Al Comment 10, oral data
from two of these studies (Haffner et al., 1992) and (Schmutte et al., 1988)1 are now used for
model calibration, allowing identification of human oral absorption rate constant and
bioavailability. The most informative of the remaining studies may be the in-vitro study of Miller
and Wells (2011) which demonstrated that methanol-induced developmental effects are enhanced
in mouse embryos with low catalase activity and reduced in mouse embryos with high catalase
activity. The authors propose that this observation is related to methanol's impact on the ability
of catalase to control the damaging effects of reactive oxygen species (ROS) activity, which
would be greater in mouse embryos with low catalase activity. As discussed in Section 5.3.5,
there are several problems with this interpretation, including that in vivo results from the same
laboratory (Siu et al., 2013) do not support the Miller and Wells (2011) in vitro findings. Further,
these observations do not preclude alternative explanations that involve a more direct interaction
between methanol and the embryo.
Comment 1: One reviewer suggested that the University of Toronto rabbit studies
published by Sweeting and coworkers "were not considered in the EPA's consideration of inter-
species differences (i.e., are rat or mice studies appropriate)." Another reviewer commented that
the discussion of the University of Toronto studies, especially the "publication regarding the role
of ROS in mediating the effects of methanol," needs to be improved and included in the
"sections related to choice of POD, critical effect, etc."
A-47
-------
Response: EPA has added additional discussion of the University of Toronto (Miller and
Wells, 2011; Sweeting et al., 2011) research to the toxicological review. A detailed discussion of
the University of Toronto findings and hypotheses regarding species differences and the role of
ROS following methanol exposure has been added to Section 5.3 "UNCERTAINTIES IN THE
INHALATION RFC AND ORAL RFD" of the toxicological review (see Section 5.3.5 "Choice
of Species/Gender"). Miller and Wells (2011) have suggested that developmental studies in
rodents may not be suitable for assessing human risk, and Sweeting et al. (2011) have suggested
that rabbits would be a more appropriate test species than mice and that rabbits are resistant to
methanol teratogenicity. A developmental study in rabbits via an appropriate route of exposure
would be of interest, particularly if it involved an investigation of effects over a broad set of
gestational days. However, more research is needed before it can be definitively stated that rabbit
developmental study would be more relevant to humans than rodent studies and that rabbits are
resistant to methanol teratogenicity.
Comment 2: One reviewer stated that "there are also other studies, including work in
monkeys, with aspartame that may be supportive (e.g., Reynolds). Since Table 3-2 includes
results from aspartame exposure this does not seem to be a clear exclusion criterion."
Response: A review of the aspartame literature is beyond the scope of this toxicological
review. The aspartame exposure studies have been removed from Table 3-2.
Comment 3: One reviewer noted that "the ethanol teratology literature has been largely
ignored despite some similarities in teratogenic response" and that "this larger literature may
help inform the MOA discussions in the draft document and help guide whether formaldehyde
should be considered as the proximate teratogen."
Response: A review of the ethanol literature is beyond the scope of this toxicological
review.
Comment 4: One reviewer stated that "search terms and databases examined have been
poorly defined" and that "there is a lack of inclusion and exclusion criteria" for references.
Response: EPA has added text to the Preface of the methanol toxicological review that
describes how literature searches are performed and how studies are evaluated and selected.
Charge D2 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[i]n relation to this, on pages 4-2 and 4-3 of
the main document, the authors indirectly state that formaldehyde and not formate is the likely
cause of the ocular toxicity of methanol" and that "[t]his is clearly not true and needs to be
revised, as shown by Martin-Amat et al (Methanol poisoning: Ocular toxicity produced by
A-48
-------
formate. Toxicol. Appl. Pharmacol. 45: 201-208, 1978; also, McMartin et al, Lack of a role for
formaldehyde in methanol poisoning in the monkey. Biochem. Pharmacol. 28: 645-649, 1979)."
Response: Edits have been made to the subject paragraph to emphasize that formate is
the likely cause of ocular toxicity and the new reference provided by the reviewer has been cited.
A.l.4.3. Charge D3. Please discuss research likely to substantially increase
confidence in the database for future assessments of methanol.
Summary of Comments: Reviewers suggested the following research to increase
confidence in the database for a future assessment:
• A proper study should be performed "to confirm the low activity of methanol metabolism in
fetal tissues."
• "Future studies using different animal models from rodents to primates should focus on
outcomes related to reproductive function, early sensorimotor development and object
memory as well as changes in brain architecture and size.
• "Development of a PBPK model that considers gestation and lactational exposure. "
* Studies that "replicate the findings of the critical study used by NEDO including the
inclusion of additional neuropathological and neurobehavioral assessments " and using
"NEDO-type " exposures.
* "Although additional monkey studies could be considered the Burbacher study is extremely
robust and should receive more attention by EPA. "
* Studies using "dual labeled material to confirm fetal exposure " and "designed to resolve
whether formaldehyde is involved in the developmental effects following perinatal methanol
exposure."
• "Completion of surveys to examine blood methanol concentrations in the U.S. population. "
• "A study that fully characterizes methanol metabolism [including estimates the Km and Vmax]
in the intact fetus and the dam using the rat as model... (as opposed to the existing studies
that only assess protein levels or activities using ethanol as substrate). "
* "Studies of the role ofADH and catalase in the metabolism of methanol by F-344 and
Sprague-Dawley rats " to "clarify why there might be two saturable pathways in one strain
but only one in the other (as implied by the PBPK model) "
* An oral developmental study of methanol sufficient for use in the derivation of an RfD.
A-49
-------
• "Research to explain the basis for differences in species/strain developmental effects " and
determine "the proximate toxicant and mode of action for developmental toxicity. "
• "Further studies to illuminate the relative sensitivity of rodents and primates to chronic
methanol toxicity, especially with regard to developmental and neurotoxicity endpoints. "
• Studies to elicit better "inhalation kinetic data for Sprague-Dawley rats. "
• "Monkey studies with longer exposure durations and similar endpoints. "
• "Additional mode-of-action motivated [including in-vitro] studies "
Summary Response: EPA agrees that these suggested research studies could enhance the
methanol toxicological review. However, EPA is not planning to, and none of the reviewers
suggested that EPA should, delay the completion of this assessment pending the completion of
any of these future study suggestions.
Charge D3 Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: No additional comments were received regarding this charge question.
Response: EPA completed the methanol (noncancer) toxicological review using available
studies.
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.
Summary of Comments: The six reviewers that provided comments seem to be in
agreement that there needs to be more discussion of the relation of the RfD and RfC to existing
endogenous/background blood levels. Five of six reviewers suggested that the RfD and RfC
values were more conservative (lower) than necessary. One reviewer pointed out "the RfC and
RfD are specifically defined as levels at which the risk assessor can be reasonably confident that
adverse effects will not appear " and are "not threshold levels at which effects might start to
appear."
A-50
-------
Two reviewers suggested that estimates of the increased blood levels associated with the
RfD/C values should be compared with either an upper bound or a standard deviation for
existing or normal physiological background levels ofmethanol. However, another reviewer
warned that "in view of the uncertainties as to fetal metabolism, mode of action and contribution
of diet and individual metabolic or toxicodynamic differences which are identified in the report it
seems very unwise to conclude that high-end [of the distribution of background] exposures which
are apparently safe for some individuals are necessarily safe for all. "
One reviewer supported the NTP CERHR (2003) opinion that a blood methanol
concentration of< 10 mg/L would not be associated with adverse developmental effects. Another
reviewer cited the NTP CERHR (2003) report as indicating that "common exposures " are not a
concern for developmental toxicity, and suggested that this presents a credibility problem for the
proposed RfD andRfC values, which have been likened to common exposures such as a glass of
orange juice. Another reviewer expressed concern that the assumption that common exposures or
"current background levels " are safe has not been analytically investigated, and suggested that
the uncertainty factors applied are needed to reflect these concerns, "which therefore indicates
that the proposed values for RfC and RfD are not necessarily unreasonable. "
Summary Response: The RfC and RfD have increased by several-fold due to PBPK
model revisions made in response to the comments received during external peer review. The
final RfD of 2 mg/kg-day and RfC of 20 mg/m3 are not overly conservative because they (1) are
well above the levels associated with common exposures to methanol such as from a glass of
orange juice and (2) need to account for uncertainty regarding the sensitivity of primates to the
reproductive and developmental neurotoxic effects ofmethanol.
EPA addressed the recommendation of a reviewer that estimates of the increased blood
levels associated with the RfD and RfC be compared with a standard deviation for existing or
normal physiological (endogenous blood) background levels ofmethanol. As described in
Section 5.3.6, the methanol blood levels of individuals receiving both an RfC and RfD exposure
would increase by a daily maximum of 0.86 mg/L and a daily average of 0.59 mg/L. As shown in
Figures 5-3, 5-4 and B-17, these increases are comparable to the 0.7 mg/L standard deviation
estimated for the average methanol blood levels (1.5 ± 0.7 mg/L) in humans. Thus, the estimated
increase in blood levels ofmethanol from exogenous exposures at the level of the RfD or the
RfC (or from the RfC + RfD) are distinguishable from natural background variation. These RfC
and RfD methanol blood level increases are also more than 100-fold higher than the increase that
would be associated with a "common exposure" such as from a glass of pasteurized orange juice
and about 10-fold higher than the increase that would be associated with exposure to a glass of
unpasteurized orange juice (note that this is a relatively rare exposure and FDA requires warning
A-51
-------
labels on unpasteurized juice that state "This product has not been pasteurized and therefore may
contain harmful bacteria that can cause serious illness in children, the elderly, and persons with
weakened immune systems"). Hence there is consistency with NTP CERHR (2003) in this
regard.
However, there is uncertainty with respect to the NTP CERHR (2003) statement that
methanol blood levels below 10 mg/L would not be associated with adverse developmental
effects. As discussed in Sections 5.1.3.2.3 and 5.3.6, there is uncertainty as to whether rodents
are as sensitive as monkeys and humans to the reproductive and developmental neurotoxic
effects of methanol. The lack of a reliably quantifiable monkey study is an important data gap
given the potential relevance to humans and the uncertainties raised by existing monkey studies
regarding monkey sensitivity to reproductive effects (e.g., shortened pregnancies discussed in
Section 4.3.2), CNS degeneration (e.g., stellate cell fibrosis described in Section 4.4.2) and
delayed neurobehavioral development (e.g., VDR response described in Section 4.4.2) from
methanol exposure. In the Burbacher et al. (2004b; 1999b) study, statistically significant
shortened pregnancy duration was observed in monkeys exposed to 200 ppm and statistically
significant VDR delay was observed in male monkey infants exposed to 600 ppm methanol for
just 2 hours per day. EPA estimates that these exposures raised the methanol blood levels over
background methanol blood levels in these monkeys to peak values of just 3 and 10 mg/L,
respectively (see Appendix D, Table D-10), corresponding to total blood levels of 5 and 12 mg/L,
respectively. Also, NEDO (1987) observed potential signs of CNS degeneration in
histopathology reported for monkeys exposed chronically to 100 ppm for 21 hours per day,
which is estimated to be associated with an increase in methanol blood levels over background
levels of approximately 1 mg/L (based on EPA monkey model), corresponding to total methanol
blood levels of roughly 3 mg/L (assuming an background in these monkeys of 2 mg/L).
Regarding the comment warning that it should not be assumed that "high-end [of the
distribution of background] exposures which are apparently safe for some individuals are
necessarily safe for all", EPA agrees some individuals may have a high background level of
methanol and/or high susceptibility. However, for the purposes of this assessment, EPA assumes
that background blood levels of methanol in a human population with normal background
variation do not elicit adverse health effects. This greatly simplifies the derivation of an RfD and
an RfC which are, by definition, population level estimates (including sensitive populations) of
the amount of a substance that a person can inhale (or ingest) every day over the course of a
lifetime [above background levels] without an appreciable risk of harm.
As discussed in response to Charge A2 Comment 1, the discussion of the RfC and RfD
and their relation to endogenous/background blood levels has been clarified in the revised draft
A-52
-------
assessment (see Section 5.3.6). In summary, EPA does not feel that the RfD of 2 mg/kg-day and
RfC of 20 mg/m3 are overly conservative. They are well above the levels associated with
common exposures to methanol and they appropriately account for uncertainty regarding the
sensitivity of primates to the reproductive and developmental neurotoxic effects of methanol.
Bonus Charge Follow-up Peer Review Comments on 2013 Revised Draft Assessment
Comment 1: One reviewer commented that "[t]he new proposed RfC and RfD are higher
values, and the IRIS assessment addresses the issue of comparison of associated blood methanol
concentrations with background levels directly" and that "[t]his is a very important addition to
the document and helps place the RfC and RfD in perspective."
Response: EPA appreciates the affirmation of sufficient revisions to the Toxicological
Review in response to previous peer-review comments.
Comment 2: One reviewer commented that "Nearly all of the studies used to obtain data
in humans had restricted dietary intake of foods that might increase methanol levels," and noted
that".. .this perhaps provides data on methanol blood concentrations that can be expected from
endogenous metabolism, but hardly captures the range of blood methanol concentrations in
individuals consuming a normal diet." This reviewer also stated that "[i]f the statement in the
current assessment that typical blood methanol concentrations are assumed to be without adverse
effect (which I support), that presumably applies to the higher blood methanol concentrations
that would be expected without dietary restriction." Another reviewer commented that "[t]he
endogenous levels in the general population are likely higher than the EPA-derived 1.5 + 0.7
mg/L (1 SD) (from the special "meta-analaysis" of studies in Table 3-1)" and that "[t]he EPA
should search for and include data from studies without dietary restrictions."
Response: In response to these reviewer and public comments on this topic, EPA has
revised the methanol (noncancer) assessment, particularly Section 5.3.6, to more clearly reflect
the limitations of the data shown in Table 3-1 for estimating population blood levels of methanol
(restricted diets, small numbers of studies and individuals, differing results by study), and
supplemented its analysis with data and conclusions from the United Kingdom (U.K.) Food
Standards Agency "COT Statement on the effects of chronic dietary exposure to methanol"
(COT, 2011). In Section 5.3.6, EPA has derived a sample lognormal distribution of methanol
blood levels that is consistent with data from study groups in Table 3-1 that did not involve
dietary restrictions other than alcohol. EPA compares this sample distribution with data for
background (endogenous and dietary) "exposure" rates estimated by the U.K. (COT, 2011). The
EPA sample background distribution and U.K. background methanol estimates are consistent
with one another in that the methanol blood level predicted by EPA's PBPK model for the U.K.'s
A-53
-------
23 mg/kg-day maximum exposure rate estimate and the EPA sample background distribution's
mean + 2xSD are similar, 2.5 mg/L and 2.9 mg/L, respectively. EPA recognizes that some
individuals may have background methanol blood levels above 2.5 mg/L (-7% according to the
sample distribution), but believes that methanol blood level of 2.5 mg/L represents a reasonable
approximation of the upper end of the range of background blood levels associated with a diet
that includes fruits and vegetables (as discussed in Section 5.3.6).
Comment 3: One reviewer commented that".. .for a sizable fraction of the population,
exposure at these doses/concentrations would not result in blood methanol concentrations outside
the normal range, particularly considering the first point, above [Comment 1]." This reviewer
further commented that ".. .the report represents progress in dealing with the problem of
assessing risk from exogenous exposure to endogenous chemicals, but falls short of presenting a
compelling case why the toxicity values are not excessively conservative." Another reviewer
commented that Figures 5-3 and 5-4 were "persuasive visually" but suggested a "statistical
analysis of confidence limits/significance of RfC and/or RfD impact in relation to
background/endogenous levels of methanol. "Another reviewer commented that RfC/RfD
exposures would not be distinguishable from background because "[f]or example, if a person
with 0.8 added 0.86, you would be at 1.66 or almost right at the mean level!" This reviewer
stated that "Figures 5.3 and 5.4 clearly show how the added levels are NOT distinguishable from
background" and that".. .in slide #11 of the presentation [given by EPA at the June 26, 2013
follow-up peer review webinar], when the UFo happened to be set at 1, the two distributions of
blood methanol are readily distinguishable."
Response: By definition, RfDs and RfCs are intended to protect "sensitive subgroups"
from "appreciable risk." In the case of methanol, sensitive subgroups would include pregnant
females and the elderly (toxicodynamic susceptibility) with lifetime methanol blood levels that
are at the high end of the range of background methanol blood levels associated with a diet that
includes fruits and vegetables (toxicokinetic susceptibility), estimated to be approximately 2.5
mg/L. A determination of whether the daily exogenous exposures at a RfD or RfC are
"distinguishable" relative to endogenous methanol blood levels is not a primary consideration.
However, in response to this concern, EPA provides an example in Section 5.3.6 which illustrates
that the shift in a sample background methanol blood level distribution that would be associated
with daily exposures of the entire population to methanol at the RfC or the RfD is estimated to
increase the percentage of individuals with peak methanol blood levels at or above 2.5 mg/L
from -7% to -14%. As discussed in Section 5.3.6, these estimates are not precise and do not
account for interindividual variability. However, they suggest that the increase in individuals
with higher than 2.5 mg/L methanol blood levels (i.e., higher than the upper range of background
A-54
-------
methanol blood levels associated with a diet that includes fruits and vegetables) following a RfD
or RfC exposure would not be negligible.
EPA also believes that the RfD and RfC are not "excessively conservative." As stated in
response to the Charge B4 Follow-up Peer Review Comment 3, EPA guidance places particular
emphasis on database deficiencies in the area of developmental toxicity, the primary focus of the
methanol (noncancer) assessment, stating that "[i]f data from the available toxicology studies
raise suspicions of developmental toxicity and signal the need for developmental data on specific
organ systems (e.g., detailed nervous system, immune system, carcinogenesis, or endocrine
system), then the database factor should take into account whether or not these data are available
and used in the assessment and their potential to affect the POD for the particular duration RfD
or RfC under development." As discussed in Section 5.1.3.2.3, a database uncertainty factor of 3-
fold has been applied to account for deficiencies in the methanol database, particularly with
respect to deficiencies that do not allow the Agency to full assess methanol's potential to cause
developmental neurotoxicity. As discussed in Section 5.3.7, uncertain, but potentially adverse
effects have been observed in monkeys at blood levels as low as 3 mg/L. As discussed in Section
5.3.6, a RfC or RfD exposure is expected to raise the methanol blood level of an individual with
a high end background blood level of approximately 2.5 mg/L to just below 3 mg/L. A higher
exposure would raise background levels above the lowest methanol blood level that has been
associated with uncertain, but potentially adverse effects.
A.2. Public Comments
A.2.1. April 18, 2011 to July 6, 2011 Public Comment Period
The following public comments and EPA responses pertain to comments received on the
2011 draft of the methanol (noncancer) toxicological review during the April 18, 2011 to July 6,
2011 public comment period.
Comment 1: "EPA has arbitrarily decided to establish these reference levels to identify
risks ONLY for exposure to methanol that increases the body burden of methanol or its
metabolites."
Response: As discussed above in response to external peer review Comment 1 of Charge
A2 and comments associated with the Bonus Charge Question, the decision to base the RfC/D on
exposures that increase the body burden of an individual above their naturally occurring
endogenous/background blood levels was not arbitrary. The PBPK models used in the final
A-55
-------
assessment incorporate endogenous/background concentrations of methanol; however, for BMD
modeling, the PBPK model estimate of background concentration is subtracted from the
predicted dose metric under bioassay conditions. This approach for dealing with
endogenous/background concentrations of methanol and it metabolites avoids the issue of
whether or not individuals experience health effects from endogenous/background concentrations
of methanol or its metabolites because only the risk due to exposures above background is
thereby evaluated. This greatly simplifies the dose-response assessment for methanol and the
derivation of an RfC (or RfD), which is by definition a population level estimate (including
sensitive populations) of the amount of a substance that a person can inhale (or ingest) every day
over the course of a lifetime [above background levels] without an appreciable risk of harm.
Comment 2: "The recommended reference levels represent a very small addition to the
average person's body burden of methanol and by implication suggest that half the population is
at risk from their own background level of methanol."
Response: The approach taken by EPA in deriving the RfC and the RfD assumes that
endogenous/background blood levels of methanol in a human population with normal
background variation do not elicit adverse health effects. There is currently little evidence,
epidemiological or otherwise, to challenge this assumption. Given this assumption and lack of
evidence to the contrary, if the 2 mg/kg-day RfD or 2x 101 mg/m3 RfC were so low that the
resulting (predicted) change in methanol blood levels was only a small fraction of the normal
variation in background levels (e.g., 1% of one standard deviation), one could argue that this
would be indistinguishable from natural variation and lexicologically irrelevant. Therefore, a
comparison of the expected increase in methanol levels in blood resulting from exposure to
methanol at the level of the RfC or RfD to the variation in endogenous/background levels of
methanol observed in humans is provided in Section 5.3.6 to determine if this might be the case.
As shown in Figures 5-3, 5-4 and B 17, the estimated increase in blood levels of methanol
resulting from exposure to methanol at the RfC alone, at the RfD alone, or at the RfC + RfD
combined is comparable to the variability (represented as one standard deviation) observed in the
average estimated methanol blood levels (1.5 ± 0.7 mg/L) in humans (see Table 3-1 and Section
5.3.6). This then demonstrates that the estimated increase in levels of methanol from the RfD or
the RfC (or from the RfC + RfD) are distinguishable from natural background variation, but the
overall derivation of the RfD and RfC ensures that these increases will not significantly increase
adverse health outcomes.
Comment 3: "EPA incorrectly supports its decision to ignore the naturally-occurring
background levels of methanol in human blood by citing the results of its PBPK modeling."
A-56
-------
Response: As explained above in response to external peer review Comment 2 of Charge
A2, EPA has re-calibrated the PBPK models to account for background levels and has used them
to derive the revised RfD and RfC. Hence, the justification for not including a background term
in the PBPK models has been removed from the toxicological review.
Comment 4: "While the increment of a reference level dose of methanol is a small
percentage of the average background blood level in humans, the intake of certain common
foods can easily exceed EPAs recommended reference level."
Response: Due to changes in the rat PBPK model, the RfC and RfD are several-fold
higher than the previously proposed values and, as explained above in response to comments
related to the external peer review Bonus Charge Question, the increase in an individual's
methanol blood levels after an exposure equivalent to the final RfC or RfD is expected to be well
in excess of the increase that would be associated with a "common exposure" such as from a
glass of orange juice.
Comment 5: "Application of a physiologically based pharmacokinetic (PBPK) model to
these study data that is inappropriate for modeling exposures to pregnant animals, neonates, and
weanling rats, and that is based on a data set that severely underestimates the likely exposures in
both studies."
Response: As explained above in response to external peer review Comment 2 of Charge
A2, EPA recognizes that neonatal blood levels will likely be higher, approximately 2-fold higher
for rats, than maternal blood levels of methanol. However, the ratio of blood concentrations
between a human infant and its mother is not expected to be significantly greater than the
approximate 2-fold difference that has been observed between rat pups and dams. Further, the
health-effects data indicate that most of the effects of concern are due to fetal exposure, with
only a small influence due to postnatal exposures. As stated in Section 5.1.3.2.2, for these
reasons and because EPA has confidence in the ability of the PBPK model to accurately predict
adult blood levels of methanol, the maternal blood methanol levels for the estimation of HECs
from the NEDO (1987) study were used as the dose metric.
Comment 6: "Failure to confirm the results of EPAs PBPK modeling against blood
methanol concentration data collected in both [Rogers et al. (1993b) and NEDO (1987)] studies."
Response: The mouse PBPK model has now been removed from the assessment and the
blood concentration data from Rogers et al. (1993b) used directly for the benchmark dose
analysis. For example, the BMDLio for the mouse cervical rib effect based on the blood Cmax
metric has thereby changed from 94.3 mg/L to 90.9 mg/L (both representing concentration
increases above background). Data were not available for validating the rat PBPK model
A-57
-------
predictions prior to the methanol external peer review. Subsequent to the methanol external peer
review, EPA received blood measurements from the NEDO (1987) rat study and has validated
model predictions against them (Appendix B, Sections B.2.3, B.2.4 and B.2.5).
Comment 7: "Recent research by Dr. Peter Wells of the University of Toronto, which we
detail in these comments, raises serious questions about the use of rodent models for hazard
assessment of methanol in humans because rodents and humans metabolize methanol very
differently."
Response: As explained above in response to comments made under external peer review
Charge D2, a detailed discussion of the University of Toronto findings have been added to
Section 5.3 "UNCERTAINTIES IN THE INHALATION RfC AND ORAL RfD" of the
toxicological review (see Section "5.3.5 Choice of Species/Gender"). The in-vitro study of Miller
and Wells (2011) demonstrated that methanol-induced developmental effects are enhanced in
mouse embryos with low catalase activity and reduced in mouse embryos with high catalase
activity. The authors propose that this observation is related to methanol's impact on the ability
of catalase to control the damaging effects of reactive oxygen species (ROS) activity, which
would be greater in mouse embryos with low catalase activity. However, as discussed in Section
5.3.5, there are several problems with this interpretation, including that in vivo results from the
same laboratory (Siu et al., 2013) do not support the Miller and Wells (2011) in vitro findings.
The University of Toronto studies are informative, but do not demonstrate conclusively that
rodent developmental studies are irrelevant to humans.
Comment 8: "Severe reporting deficiencies in the two-generation reproductive study,
including a lack of mean and individual animal data in the main study and the absence of details
regarding methods and data related to maternal or gestational outcomes in the supplementary
study."
Response: As described in Section 5.1.2.2, the supplementary study to the NEDO two
generation study provides sufficient dose and response information for a benchmark dose
analysis. Uncertainties associated with this study as they relate to the benchmark dose analysis,
including the absence of a detailed reporting of methods and maternal or gestational outcomes,
are discussed in Section "5.3.1 Choice of Study/Endpoint." Though the methods for this
supplementary study are not described, the methods for the parent two-generation study are
adequately described and it is reasonable to assume that the supplementary study was performed
under the same protocol starting with a number of FO females appropriate for a one-generation
developmental study (see response to public Comment 11 below). While data related to maternal
or gestational outcomes in the supplementary study are not given, signs of overt maternal
toxicity were not reported in the two-generation study at similar exposure levels and it is
A-58
-------
reasonable to assume that they did not occur, and would have been reported had they been
observed, in the supplementary study. While this supplementary study no longer forms the basis
of the RfD, it does form the basis for the RfC because its limitations are not considered serious
enough to preclude its consideration as a candidate principal study and because it documents a
clear dose-response for a relevant endpoint for a critical organ system, brain weight reduction,
which is consistent with its parent two-generation study and with other teratogenicity (NEDO,
1987) and subchronic CTRL, 1986) study findings with respect to the effect of methanol exposure
on brain weight.
Comment 9: "The lack of utility of the NEDO (1987) reproductive study for the purpose
of human health risk assessment, as judged by other authoritative bodies."
Response: By "authoritative bodies" the commenter refers specifically to a 2003 report
from the National Toxicology Program's Center for the Evaluation of Risks to Human
Reproduction (NTP-CERHR, 2003) on the reproductive and developmental toxicity of methanol.
The methanol toxicological review cites the subsequently peer reviewed and published version of
this report (NTP-CERHR, 2004). In this more recent published version of the report, the NTP-
CERHR panel states that "a summary of a two-generation rat reproductive toxicity study done by
the Japanese NEDO was received, but data were not available in sufficient detail for Expert
Panel review." The NEDO summary reviewed by the NTP-CERHR panel did not contain the
more detailed supplementary study data, with pup brain weight means and standard deviations
that EPA evaluated in its benchmark dose analysis (Appendix D). This information, as well as
supplemental methanol blood measurements, was obtained by EPA after the NTP-CERHR panel
completed its report. In addition, EPA sponsored an external peer review (ERG, 2009) of the
NEDO (1987) report that contained this information, along with two other NEDO chronic rat and
mouse studies (NEDO, 1985a, b). This expert peer review panel of five scientists was asked
specifically to "Describe the reliability of the subject NEDO studies for consideration in the
derivation of EPA IRIS quantitative health benchmarks." With respect to this charge and the two-
generation study, including the supplementary study, the main concerns expressed by the peer
reviewers were that they "may be useful for RfD derivation if brain weight changes persist when
normalized by body weight" and that the authors "should have used ANOVA plus multiple
comparison tests to analyze these data." With respect to the former concern, NEDO only reported
means and standard deviations for absolute brain weight change and did not report body weight
data for the supplementary study. However, body weight data reported for the parent, two-
generation study did not indicate a body weight effect in the exposed Fl or F2 generation pups.
Further, the absolute brain weights reported by NEDO are an appropriate basis for a dose-
response assessment. With respect to the concern over the lack of appropriate statistical testing,
A-59
-------
EPA did not rely on the NEDO statistical determinations, but performed its own more definitive
benchmark dose analysis of the data (see response to external peer review Comment 1 of Charge
Bl). Hence, while the NEDO report of the two-generation supplementary study results has
limitations, particularly with respect to reporting of methods (see discussion in Section 5.3.1), it
was not evaluated by the NTP-CERHR (2004) expert panel, and it was not deemed inadequate,
for the purposes of RfC derivation, by a panel of expert peer reviewers (ERG, 2009).
Comment 10: "The use of exposure regimens in both the reproductive study and the
24-month rat study that confound the estimates of exposure."
Response: The stated basis for this comment is the concern over (1) "consumption of
[methanol] contaminated feed," (2) "ingestion of methanol during the act of preening,"
(3) dermal absorption in adult rats, and (4) "increased dermal absorption of neonatal animals
[over adults]" because "the epidermal layers of neonatal rats are thinner than those of adult
animals, they lack fur for the first week after birth." With respect to the first concern, EPA
estimates that data on a rodent breeder diet (labdiet #5013) indicates 10% moisture content. If
one assumes that chamber methanol concentrations equilibrate with this moisture content, using
the blood:air PC for methanol, and uses a typical pregnancy food consumption of 30 g/day in
rats, then the amount of methanol ingested in the chow would be about 3% of that inhaled during
a 22 hr/day exposure. This is likely an upper bound since it would take some time for methanol
to diffuse into and through a container of chow (equilibration with the chow would take time),
and fresh chow is provided each day. Thus the amount ingested by this route is not considered
significant and dosimetry calculations have not been adjusted to reflect that possibility.
With respect to the second, third and fourth concerns, methanol is not known to adhere to
or be absorbed by rat dermal surfaces in amounts that would significantly impact model
predictions. According to Perkins et al. (1996b page 160):
"The method [flow-through chamber exposure], like all whole-body methods, exposes
the animal to the vapor at all dermal surfaces. For the very water-soluble vapors, such as
methanol, dermal exposure is not significant; indeed, when taring the chamber by
inserting a dead rat versus just opening an empty chamber for the same length of time, no
difference in methanol loss was noted except at 20,000 ppm, in which case the steady-
state loss was 27% higher with a dead rat than an empty chamber. This higher steady-
state loss at 20,000 ppm methanol may be related to physical properties of the compound;
at the high vapor level, somewhat more methanol may condense and become adsorbed to
the fur. Further experimentation is required to clarify-the significance of this
observation."
A-60
-------
Since the concentrations used in the NEDO rat studies were well below the 20,000 ppm
level at which Perkins et al. (1996b) observed a difference in methanol loss in the chamber,
methanol dermal absorption is not expected to significantly impact model predictions of
methanol blood levels in the NEDO rat studies [also, EPA presumes that the empty-chamber loss
rate, of which the dead rat caused a 27% increase, was fairly small. The actual loss rate was not
reported by Perkins et al.(1996b)]. Though the Perkins et al. (1996b) conclusion was based on an
adult rat, there is no scientific basis for the belief that neonatal rats would absorb a significantly
greater amount of methanol.
Comment 11: "Use of an insufficient number of parental animals in the supplementary
reproductive study (from which EPA derives its RfC) to support proper statistical evaluation."
Response: The basis provided by the commenter for this statement is that "EPA and
Organisation for Economic Cooperation and Development (OECD) guidelines recommend
evaluation of at least 20 litters per group in a two-generation reproduction toxicity test, in order
to ensure sufficient statistical power in the study (OECD. 2001: U.S. EPA. 1998b)." The number
of FO parental animals used in the NEDO two-generation study (30 males and 30 females per
dose group) was appropriate and in accordance with both EPA and OECD two-generation
reproduction toxicity test guidelines. The supplementary study performed by NEDO does not fall
under these guidelines because it was not a two-generation reproduction study. According to
NEDO (1987 page 201 ), the purpose of the supplementary study was "to confirm its [decreased
brain weight] relationship with the treatment and to know from what period after birth such
changes would appear" and, therefore, the test rats were only exposed "from Day 0 of gestation
throughout the Fl generation." This type of study and purpose would more appropriately fall
under the Agency's developmental neurotoxicity guidelines (U.S. EPA, 1998b), which state that
"on postnatal day 11, either 1 male or 1 female pup from each litter (total of 10 males and 10
females per dose group) should be sacrificed" and that "brain weights should be measured in all
of these pups." The number of FO parental animals included per group in the supplemental
experiment was not reported. However, the number of pups per dose group was reported and it is
reasonable to assume that, consistent with the standard culling protocol used for both the Fl and
F2 generations of the two-generation study (NEDO, 1987 pages 185 and 189 ), each dose group
pup came from a different litter (to avoid problems associated with litter correlation). Hence, by
examining more than 10 male and 10 female litter-specific pups per dose group at three time
points (3, 6 and 8 weeks), the NEDO supplementary study actually went well beyond EPA
recommendations for this type of study.
Comment 12: "Use of statistical methods in both the reproductive study and the 24-
month rat study that, by today's standards, are considered inadequate."
A-61
-------
Response: As mentioned above in response to public Comment 9, EPA did not rely on
the NEDO statistical determinations, but performed its own more definitive benchmark dose
analysis of the NEDO (1987) rat two-generation an teratogenicity data (see response to external
peer review Comment 1 of Charge Bl).
Comment 13: "Derivation of an RfC based on absolute brain weight data without
considering the significance of other gestational outcome data (including body-weight data) that
would put these data in proper context for risk assessment purposes."
Response: As mentioned above in response to public Comment 9, the absolute brain
weights reported by NEDO in a supplementary developmental study are an appropriate basis for
a dose-response assessment (also see response to external peer review Comment 1 of Charge
Bl). Other gestational outcome data, including body weight data, were not provided for the
supplementary developmental study. However, body weight data reported for the parent, two-
generation study did not indicate a body weight effect in the exposed Fl or F2 generation pups.
The commenter argued that relative brain weights are important for neonates. While it would
have been helpful to have the body weight information for the neonates from the supplementary
study, the two-generation data indicate that methanol does not significantly impact pup body
weight at the exposure levels of concern. Further, because brain weights are conserved in both
neonates and adults, a dose-related reduction in absolute brain weight is an important
consideration for both neonates and adults.
Comment 14: "Lack of proper consideration of species differences in sensitivity to
developmental toxicity due to methanol exposures in the RfC derivation."
Response: The commenter cites differences in breathing rates, minute volumes and
metabolism (i.e., the preference for metabolism via catalase over ADH that is unique to rodents)
as factors that are not properly considered. The first two factors are accounted for by the
Agency's rat and human PBPK models. The latter factor is considered extensively in the
toxicological review (e.g., Section 5.3.5) and is discussed above in response to external peer
review Comment 1 of Charge A4, Comment 1 of Charge D2, and public Comment 7. There is
currently not enough known about methanol's teratological mode of action to conclude that
rodent developmental studies are not relevant to humans.
Comment 15: "Failure of EPA to consider more robust developmental toxicity data in
derivation of an RfC value." Specifically, the commenter suggests that the Rogers et al. (1993b)
study would be the more appropriate study on which to base an assessment of the developmental
toxicity of methanol.
A-62
-------
Response: EPA agrees that the Rogers et al. (1993b) study is an appropriate study, and
the final RfC and RfD are derived from quantitative analyses of the Rogers et al. (1993b) and
NEDO (1987) studies.
Comment 16: "Sweeting et al. demonstrated a large difference in developmental toxicity
between mice and rabbits; minor differences in number of stillbirths or postpartum mortality do
not equal developmental effects and the EPA's reliance on them is not appropriate."
Response: EPA is not relying on the Sweeting et al. (2011) study results as evidence of
teratogenic effects in rabbits, but simply points out that their claim that rabbits are resistant
relative to mice to the teratogenic effects of methanol needs to be verified over several
gestational days, as has been done for mice, because the critical gestational window for
developmental effects could be different for rabbits versus mice. Under different study
conditions, the observed increase in postpartum lethality (11% versus 5% in controls) and
stillbirths (4% versus 0% in controls) may prove significant given that postpartum lethality
("wasting syndrome") and a shortened gestational period were possible adverse outcomes
observed in methanol exposed monkeys (see discussion of Burbacher et al., (2004b; 1999b) in
Section 4.3.2).
Comment 17: "The draft assessment states that Sweeting et al. (2011) suggests that low
ADH activity in mouse embryos could lead to a 'greater depletion of'catalase.' The assessment
further states (line 7,8) 'If ROS accumulation due to this catalase consumption...' Sweeting et al.
do not postulate a depletion or consumption of catalase."
Response: The text in the final assessment has been clarified.
Comment 18: "The [draft] IRIS Assessment should note here [page 9, paragraph 1] that
Sweeting et al. postulated that methanol and/or its metabolites may enhance the embryonic
production of ROS (by mechanisms that do not involve catalase)."
Response: The text in the final assessment has been clarified.
Comment 19: "The use of the citation from the Tran et al. study to suggest that embryos
are in danger of development effects of methanol draws overly broad conclusions from a very
limited study."
Response: EPA is not making a broad conclusion regarding the danger of methanol to
human fetuses based on the Tran et al. (2007) study. The Tran et al. (2007) study lends
uncertainty to the hypothesis presented by others, including Sweeting et al. (2011), that
developmental studies in mice are not relevant to humans because human infants do not rely on
A-63
-------
catalase to metabolize methanol as do mice. The Iran et al. (2007) study provides limited
evidence that catalase may play a role in the metabolism of alcohols in neonates.
Comment 20: "The embryo culture model [used in the Miller and Wells (2011) study]
removes the confounding effects of maternal catalase activity, and specifically the maternal
peroxidative activity of catalase responsible for metabolizing methanol."
Response: This is offered by the commenter as an explanation for why the in-vivo
studies of Siu et al. (2013) did not observe the enhanced embryopathies in aCat (catalase-
deficient) mice that were reported in the in-vitro studies of Miller and Wells (2011). As discussed
in Section 5.3.5, Miller and Wells (2011) acknowledge that aCat mice in the in-vivo study of Siu
et al. (2013) "appeared resistant to methanol teratogenicity." However, they suggest that the in-
vivo results were confounded by "maternal factors, including the metabolism of methanol and its
formic acid metabolite by maternal catalase (Dorman et al., 1995)," which would presumably
reduce the methanol body burden to levels that do not competitively inhibit embryonic catalase
antioxidant activity. Alternatively, maternal factors could be protecting the embryo from a more
direct interaction with methanol, the compound which this assessment assumes to be the toxic
agent.
A.2.2. May 3, 2013 to June 17, 2013 Public Comment Period
The following public comments and EPA responses pertain to comments received from
seven individuals/organizations on the 2013 draft of the methanol (noncancer) toxicological
review during the May 3, 2013 to June 17, 2013 public comment period.
Comment 1: Compliments/Affirmations:
• "We commend EPA for revising the March 2011 draft toxicological review in order to
address previous public comments, peer reviewer concerns and to improve the
scientific basis for the derivation of reference values."
• "The May 2013 revised draft assessment addresses the significant concerns that we
raised during reviews of the initial draft, and does increase the inhalation reference
concentration (RfC) and oral reference dose (RfD)."
• "EPA is to be commended for basing its BMD analyses of the Rogers et al. (1993b)
cervical rib malformation data on gestation day 6 blood methanol data that were
collected in the very same study, rather than on simulated blood methanol levels that
were predicted with an EPA mouse PBPK model."
A-64
-------
"EPA is also to be commended for eliminating the mouse PBPK model altogether
from its updated assessment."
"EPA is to be commended for their responsiveness to concerns raised by the external
peer review panel and the public regarding the previous external peer review draft
(2011)."
"I would also like to note that I am very pleased that EPA is applying the uncertainty
factors to the internal dose prior to using the human PBPK model for conversion of
the internal dose to an acceptable external dose (Section 5.1.3.2, pages 5-15- and
5-16)."
".. .the revised draft assessment demonstrates a clear intent to address the significant
concerns that were raised during reviews of the initial draft, and does significantly
raise the inhalation reference concentration (RfC) and oral reference dose (RfD)."
"Given the availability (subsequent to the original draft review) of the toxicokinetic
data for S-D rats which are the strain used in the critical experiments for the RfC, it
makes sense to use these data as the basis for rat PBPK modeling, and apparently this
has been successful in that blood levels reported in the NEDO experiments are
matched by the model."
"It appears to me that it is perfectly reasonable to apply the dose metric conversion
after the UFs as has been done in the revised toxicity review. This ensures that the
dose metric conversion is made at the actual concentration of interest, i.e. the
RfC/RfD. This would also usually be done for a cancer risk assessment, where the
dose metric conversion would be done at an observed or estimated concentration or a
risk specific level."
"The change in critical study for the RfD to the Rogers et al. (1993b) is a natural
response to the change in relative sensitivities of this and the NEDO studies based on
the revised toxicokinetic model. This study is of adequate quality and is reported in
detail. Although the data used are from the inhalation study (the oral experiment
being much more limited) it has the advantage of having actual blood level
measurements which are useable as the internal dose metric, reducing the
uncertainties associated with application of the animal PBPK model (although the
human PBPK model is still needed to convert to an external oral dose metric for the
RfD.)"
A-65
-------
"The Executive Summary is thorough but readable and lays out the basis of the RfC
and RfD derivations clearly."
"The Exposure Response Arrays clearly lay out the comparison between available
endpoints which was the basis for selection of the critical studies."
"The increased use of tabular presentations for study data and comparisons is helpful,
generally making the narrative clearer as well as more compact."
"The move of model details etc. to appendices is also an improvement."
"The additional materials identified at earlier stages of the review and comment
process have been included and incorporated into the report satisfactorily. The most
important change obviously is the incorporation of the S-D rat toxicokinetic data and
the resulting recalibration of the PBPK model."
"This additional discussion [of the relation of RfD and RfC to existing endogenous
blood levels] addresses the questions raised earlier in the review process and is
helpful in clarifying the situation. It is important to note that even without this
clarification the RfC and RfD proposals are reasonable."
"It isn't necessarily incumbent on EPA to show that the predicted exposure increases
are 'distinguishable from endogenous background' if a clear hazard is identified at the
BMDL exposure level. However, the ability to do so as shown here certainly adds
confidence and provides an answer to critics who are naturally disposed to weaken
the health protective standards if they can."
".. .the application of the S-D rat toxicokinetic data, which has had the effect of
significantly raising the values of both the RfC and RfD and improving the
confidence in these values, has significantly eased the task [of demonstrating that
RfD and RfC exposures are "distinguishable" from background] by increasing the
gap between the endogenous background and the dose-related level at the RfC/RfD."
"The selection of 1 sd as the BMR [for NEDO brain weight data] is justified in the
review since this gives the lower value for the BMDL. However, it is additionally
justified as it is based on a generally accepted statistical criterion of what constitutes a
clearly observable change in a lexicologically significant parameter."
"For the Rogers et al. cervical rib incidence data a BMR or 5% is selected and
justified based on established U.S. EPA guidance for dichotomous responses in
nested-design developmental studies."
A-66
-------
• "Many of the comments raised in the previous round of peer review have been
addressed as discussed above, or incidentally as a result of the revised basis of the rat
PBPK model and selection of Rogers et al. (1993b) as the critical study for the RfD.
Other points raised in the earlier peer review discussion appear to have been
responded to thoroughly and appropriately."
Response: EPA appreciates these comments and will continue to work towards
developing assessments that are responsive to peer reviewer and public comments.
Comment 2: One commenter noted that "EPA's revised draft assessment provides a brief
discussion of some potential exposure pathways (e.g., foodstuff or commercial products) but it
does not provide information on the specific levels of methanol that humans are likely to be
exposed to on a daily basis" and that "[t]he United Kingdom (U.K.) Food Standards Agency has
stated that endogenous methanol production ranges from 300 to 600 mg/day and that up to 1,000
mg/day methanol can be consumed in food, particularly fruit and vegetables."
Response: EPA has added the U.K. estimates of endogenous methanol production and
consumption in food, particularly fruits and vegetables, to Section 2 of the assessment. Also,
discussion of how these estimates compare to EPA's estimate of background blood levels has
been added to section 5.3.6. The U.K. report referred to by the commenter (COT, 2011) is now
used to support the EPA's estimate of the upper end of the range of methanol blood levels
associated with a diet that includes fruits and vegetables.
Comment 3: One commenter noted that "[t]he [ten] studies used to derive the estimated
endogenous background methanol blood level involved fasting and or some forms of dietary
restrictions." This commenter stated that one of these studies, Woo et al. (2005), involved "... no
food intake from the time subjects [18 males] went to sleep the previous night until after the 8:00
a.m. blood sampling" and therefore "... provides data on background blood methanol that may
be more representative of endogenous levels with little or no contribution from foods." This
commenter further suggested that the RfD would not be lexicologically relevant had EPA used
the Woo et al. (2005) study to estimate endogenous methanol blood levels because "the mean
incremental blood level of 0.41 mg/L [associated with the RfD] is well within the background
level of variation" of 2.62 ± 1.33 mg/L reported by Woo et al. (2005).
Response: In the revised methanol (noncancer) assessment, EPA has clarified that the
methanol (noncancer) Toxicological Review ".. .provides scientific support and rationale for a
hazard identification and dose-response assessment of the noncancer effects associated with
chronic exposures to exogenous sources of methanol that add to background levels of methanol
derived from a diet that includes fruits and vegetables (see further discussion in Section 5.3.6).
A-67
-------
Thus, studies that substantially restricted the consumption of fruits and vegetables (Ernstgard et
al.. 2005: Osterloh et al.. 1996: Cooketal.. 1991: Davolietal.. 1986) are considered
inappropriate under this definition. The remaining studies, Batterman and Franzblau (1997),
Batterman et al. (1998). Lee et al. (1992). Sarkola and Eriksson (2001). Turner et al.(2006) and
Woo et al. (2005), are considered appropriate for the purpose of this analysis as they did not
involve substantial fasting (i.e., only two involved fasting, one overnight and one for 4 hours) or
dietary restrictions (i.e., only one involved a minimal dietary restriction, no aspartame-containing
cereals and no juice on the morning of the test). The analysis of these six studies (see details in
Section 5.3.6), after weighting them in accordance with the extent to which they represent the
U.S. population (see footnote 61), yields an estimate for the mean and SD for endogenous
background methanol blood of 1.3 mg/L and 0.8 mg/L, respectively. As discussed in Section
5.3.6 and in response to Bonus Charge Follow-up Peer Review Comment 1, these estimates are
consistent with endogenous methanol production plus dietary exposure ranges reported by the
U.K. Food Standards Agency (COT. 2011).
The Woo et al. (2005) alone is not considered an appropriate basis for the estimation of a
sample background methanol blood distribution that would be representative of the general U.S.
population. The Woo et al. (2005) study subjects were all male with a mean age 23.7 years (range
20 -29 years), the sample size of 18 is small and likely biased somewhat towards subjects who
regularly consume alcohol; while the subjects scored low on an alcoholism screening test, they
agreed to self-induce an "alcohol hangover state" and all 18 participants "had experienced
hangover at least once." Further, since the study was performed in Korea, the subjects are
presumed to be Korean, a population prone to having more than one variant of the genes coding
for alcohol and aldehyde dehydrogenase (Eng et al., 2007). This suggests marked differences in
their alcohol metabolism relative to other ethnicities (see discussion in Section 3.3).
Comment 4: One commenter stated that "EPAs approach to endogenous methanol may
set a dangerous precedent" because "... [tjoxic levels of methanol are proposed [by EPA] to be
those which fall outside 1 standard deviation [associated with a RfD + RfC exposure] of the
population mean of endogenous methanol." Another commenter stated that".. .well over one-
fifth of the population will have a background level of methanol (without exposure to external
methanol) that is above the level deemed safe by EPA (1.5 mg/L average background + 0.4 mg/L
RfD exposure) and asked ".. .how does EPA distinguish between endogenous and exogenous
methanol" and "[h]ow can one be "safe" and smaller levels of the other be "unsafe"?"
Response: To address these comments it is helpful to reiterate the definition of an RfD.
As stated in the introduction to the methanol (noncancer) toxicological review, "[t]he RfD
(expressed in units of milligrams per kilogram per day [mg/kg-day]) is defined as an estimate
A-68
-------
(with uncertainty spanning perhaps an order of magnitude) of a daily exposure to the human
population (including sensitive subgroups) that is likely to be without an appreciable risk of
deleterious effects during a lifetime." Thus, the addition of the RfD to the mean of EPA's
estimated sample distribution is not meaningful and the RfD is not an estimate of the level above
which all are effected. The RfD is the level at or below which there is likely to be no appreciable
risk, even to sensitive subgroups. In the case of methanol, sensitive subgroups would include
pregnant females (toxicodynamic susceptibility) with relatively high endogenous methanol blood
levels (toxicokinetic susceptibility). Thus, it is more appropriate to consider the impact of the
RfD on sensitive individuals with methanol blood levels at the high end of the range of
background blood levels of methanol associated with a diet that includes fruits and vegetables.
As described in Section 5.3.6, EPA has defined the upper end of this background methanol blood
levels as 2.5 mg/L. In response to the last question in this comment, smaller exogenous levels
can be "unsafe" when they are added to a subgroup's background blood methanol that are
already susceptible from a toxicodynamic (e.g., pregnancy) and/or toxicokinetic (e.g., high
endogenous methanol blood levels) perspective. This is because, as indicated in the Executive
Summary and other sections of the assessment, the RfD is intended to protect sensitive
subgroups and the combination of endogenous background levels plus exogenous exposure can
lead to toxicity.
Figure 5-4 illustrates how methanol blood level distributions for RfD and RfC exposures
to the EPA sample background distribution compares with the blood levels that have been
associated with these uncertain, but potentially adverse effects in monkeys. As discussed in the
previous section, a RfC or RfD exposure is expected to raise the methanol blood level of an
individual with a high end background methanol blood level of 2.5 mg/L to just under 3 mg/L,
the lowest methanol blood level that has been associated with these uncertain, but potentially
adverse effects.
Comment 5: One commenter noted that "It is quite possible that levels of exposure to
some chemicals at background levels are in fact hazardous to human health, and if the
technology exists now or at some time in the future to control that exposure, it is incumbent on
the Agency to inform the public of the potential risk of these background exposures." This
commenter further stated that "In many cases, the question also arises, if the general public is
being exposed to this hazardous substance on a routine basis because of background levels,
whether we are seeing any adverse health effects resulting from this exposure." Another
commenter stated that".. .it cannot be assumed automatically that any reported endogenous level
is safe... [T]here are a number of examples in the toxicological literature where some individuals,
as a result of idiosyncrasies of metabolism, diet, exposure etc., show actual toxicity or at least
A-69
-------
substantially increased susceptibility from 'background' exposures." This commenter added that
"[tjhese considerations are adequately addressed in the revised toxicity review."
Response: EPA agrees that it is possible that background levels of methanol are
hazardous to human health. In the revised assessment, EPA does not state that background levels
are without health risk, but states that the "lexicological Review provides scientific support and
rationale for a hazard identification and dose-response assessment of the noncancer effects
associated with chronic exposures to exogenous sources of methanol that add to background
levels of methanol derived from a diet that includes fruits and vegetables (see further discussion
in Section 5.3.6)." In Section 5.3.6 EPA discusses the relationship of the RfD and RfC to
background blood levels and derives a sample background distribution of methanol blood levels.
EPA sample background distribution and the U.K. (COT, 2011) background methanol estimates
are consistent with one another. The upper bound of the combined endogenous and dietary
exposures estimated in the U.K. is 23 mg/kg-day. The methanol blood level predicted by EPAs
PBPK model for this 23 mg/kg-day maximum exposure rate of 2.5 mg/L is slightly below EPAs
sample background distribution estimated mean + 2xSD of 2.9 mg/L. A small percentage (-7%)
of the EPA sample background population is predicted to have methanol blood levels above 2.5
mg/L.
Comment 6: One commenter stated that "EPA has indicated that it believes these
increases [from a RfC, RfD or RfC + RfD exposure] would be distinguishable from background,
but we fail to see how this is possible given that the combined RfC and RfD methanol values are
well within the range of background blood methanol levels." Another commenter stated that
"[t]his additional discussion [of the relation of RfD and RfC to existing endogenous blood levels]
addresses the questions raised earlier in the review process and is helpful in clarifying the
situation." This commenter also stated that "[i]t isn't necessarily incumbent on EPA to show that
the predicted exposure increases are 'distinguishable from endogenous background' if a clear
hazard is identified at the BMDL exposure level."
Response: As indicated in response to the Bonus Charge Follow-up Peer Review
Comments and Comment 4 above, the primary consideration in deriving the methanol RfD and
RfC is whether they represent daily exogenous exposures that, when added to background levels
of methanol associated with a diet that includes fruits and vegetables (resulting in methanol
blood levels of up to 2.5 mg/L; see discussion in Section 5.3.6), are not likely to result in an
appreciable health risk, even to sensitive subgroups. Consistent with the view of the second
commenter above, a determination of whether the daily exogenous exposures at a RfD or RfC
are "distinguishable" relative to background methanol blood levels is not a primary consideration
for an IRIS risk assessment. However, reviewers of the 2011 draft methanol (noncancer)
A-70
-------
toxicological review (U.S. EPA, 2011 a, c) asked EPA to investigate this topic and EPA responded
by adding Section 5.3.6 to the assessment. In Section 5.3.6, EPA provides an example which
illustrates that an RfC or RfD exposure is estimated to increase the percentage of individuals
with peak methanol blood levels at or above 2.5 mg/L from -7% to -14%. These estimates are
not precise and do not account for interindividual variability. However, they suggest that the
increase in individuals with higher than 2.5 mg/L methanol blood levels (i.e., higher than the
upper range of background methanol blood levels associated with a diet that includes fruits and
vegetables) following a RfD or RfC exposure would not be negligible.
Comment 7: One commenter stated that".. .EPA has not provided a discussion of the
threshold blood methanol level for adverse effects" and noted that".. .in 2003 the National
Toxicology Program (NTP) issued a monograph which reviewed the potential human
reproductive and developmental effects of methanol and found minimal concern that adverse
health effects would be associated with <10 mg/L blood methanol concentrations." This
commenter suggested that EPAs stated uncertainty with respect to NTP's assumption,
"particularly whether rodents are as sensitive as monkeys and humans to the reproductive and
developmental effects of methanol exposure... should be further elucidated as well as the
threshold blood methanol levels necessary to illicit adverse effects."
Another commenter stated that "EPA's revised draft assessment does not appear to
adequately address the plausible adverse health risks associated with levels above background
exposures." More specifically, the commenter contends that "[t]he Agency has failed to show
how this "measurable" variation from external exposure to [a RfD and/or RfC of] methanol
increases any health hazard" and that".. .the real question that still needs to be asked which is
'where's the risk?'"
Response: As was discussed in response to the Bonus Charge Follow-up Peer Review
and 2013 Public Comments 4 and 6 above, the primary consideration in deriving the methanol
RfD and RfC is whether they represent daily exogenous exposures that, when added to
background levels of methanol associated with a diet that includes fruits and vegetables (as
defined in Section 5.3.6), are not likely to result in an appreciable health risk, even to sensitive
subgroups. The RfD and RfC are not estimates of exposures that are health hazards, and the risk
associated with exposures above the RfD will vary from individual to individual, with the
greatest risk experienced by sensitive subgroups. Nevertheless, the revised assessment includes
an expanded discussion of the relationship of the RfD and RfC to methanol blood levels that
have been associated with effects in monkeys and humans. This discussion has been moved to a
new Section 5.3.7 titled "The Relationship of the RfC and RfD to Methanol Blood Levels In
Monkeys Associated with Unquantifiable Effects of Uncertain Adversity." Section 5.3.7
A-71
-------
discusses the reasons EPA believes that blood levels of methanol below 10 mg/L, but above 3
mg/L, could pose an uncertain, but potential health risk. EPA's conclusion differs from the NTP-
CERHR (2003) conclusion partly because of an evaluation of the methanol blood levels
corresponding to effects observed in the Burbacher et al. (2004b; 1999b) reproductive and
developmental monkey study using the EPA monkey PK model, and partly because the NTP-
CERHR (2003) report focused on the reproductive and developmental effects of methanol, and
did not assess the potential for effects from chronic exposure. As discussed in Section 5.3.6, a
RfC or RfD exposure is not expected to raise the methanol blood level of an individual with a
high end background methanol blood level of 2.5 mg/L to more than 3 mg/L, the lowest
methanol blood level that has been associated with uncertain, but potentially adverse effects in
monkeys.
Comment 8: Concerning all IRIS chemicals, one commenter stated that".. .to date, EPA
has not indicated which substances, under review by the IRIS program, will benefit from
implementation of the NRC recommendations... [n]or has EPA provided an updated timeline of
when it anticipates having each phase completed and fully implemented." The commenter
suggests that "EPA should provide this information as soon as possible and expeditiously move
forward with fully implementing all of the NRC's recommendations, regardless of the phase"
that has been assigned by EPA to each assessment for progressive implementation of the NRC
recommendations. Concerning the methanol (noncancer) assessment, the commenter noted that
"[t]he revised draft assessment only provides a general overview of the process utilized in EPA's
literature search strategy" and suggests that "[i]t would have been more useful for EPA to have
included the search terms it used to select appropriate studies for inclusion in the literature search
as well as listing the specific inclusion/exclusion criteria." The commenter further suggested that
"EPA should be consistent and clear in identifying which elements it will focus on with regard to
'partial implementation' [of the NRC recommendations]" and noted that ".. .the necessary, major
substantive changes that are needed, such as more fully considering an integrated weight of
evidence approach, that includes mode of action, as part of IRIS, remain largely unaddressed."
Response: In April 2011, the National Research Council (NRC), in their report Review
of the Environmental Protection Agency's Draft IRIS Assessment of Formaldehyde (NRC,
2011), made several recommendations to EPA for improving IRIS assessments and the IRIS
Program. The NRC's recommendations were focused on Step 1 of the IRIS process, the
development of draft assessments. All IRIS assessments currently under review by the IRIS
Program will benefit from implementation of the NRC recommendations, and consistent with the
advice of the NRC, the IRIS Program is implementing these recommendations using a phased
approach and is making the most extensive changes to assessments that are in the earlier stages
A-72
-------
of the IRIS process. The methanol (noncancer) assessment is in Phase 1 of implementation of the
NRC recommendations. Phase 1 focuses on a subset of the shorter-term recommendations for
assessments near the end of the document development process or close to final posting.
Consistent with the focus of this phase of implementation of NRC recommendations, the
Toxicological Review was edited to be more concise, the rationales for decisions have been made
more transparent, and the description of the literature search strategy has been augmented for
clarity. Additionally, consistent with direction provided by Congress in The Consolidated
Appropriations Act of 2012, the Agency will include documentation in the final methanol
(noncancer) IRIS assessment describing how the recommendations of the NRC have been
implemented or addressed in the methanol (noncancer) assessment, including an explanation for
why certain recommendations were not incorporated.
Comment 9: One commenter noted that".. .there does not appear to be a full discussion
in the document regarding what study quality criteria were applied to each study in determining
its 'acceptable quality.'"
Response: As indicated in Appendix E, because the methanol (noncancer) toxicological
review is near the end of the development process and close to final posting, it is a post-peer
review, Phase 1 assessment. This means the methanol (noncancer) assessment focuses on a
subset of the short-term recommendations, such as editing and streamlining documents,
increasing transparency and clarity, and using more tables, figures, and appendices to present
information and data in assessments. Literature search and study evaluation processes were not
substantially revised.
Comment 10: One commenter stated that, for a dose-response analysis of the Rogers et
al. (1993b) mouse study, ".. .average values [see Table 1 in Starr and Festa (2003)1 are expected
to provide a more accurate and precise characterization of mean blood methanol levels
throughout the temporal window of vulnerability for mice from gestation day 6 to 15." The
commenter noted that "[t]here were no temporal trends in these data, and EPA has not provided
any justification for its selection only of the data from day 6, which were collected from just 3
mice per treatment group." The commenter suggested that "[a]t a minimum, EPA should evaluate
what difference it makes in the final results of their BMD computations to employ the full blood
methanol data from Rogers et al. (1993b) as opposed to just the data from day 6."
Response: In response to the 2011 peer review comments to streamline the assessment,
EPA moved some of the PBPK and dose-response modeling details to Appendices B and D.
EPAs rationale for using GD6 blood levels is described in Appendix D, Section D.3. Subsequent
to the Rogers et al. (1993b) study, their laboratory narrowed the temporal window of
vulnerability for developmental effects in mice to GDs 6 and 7 (Rogers and Mole, 1997; Rogers
A-73
-------
et al., 1993a). Therefore the scientific justification for not using measurements of blood
concentrations later in gestation is that they are not directly relevant to the endpoint being
evaluated. However, EPA agrees that, given that methanol inhalation dosimetry appears to be not
significantly affected by the stage of pregnancy, data from the later gestation days could be
viewed simply as additional measurements in female CD1 mice. Therefore the BMD modeling
results of using weighted concentration averages for all three gestation days measured were
compared with EPA's primary approach (using only the GD6 data). The results are not
substantially different, and the model fits were not as good as the model fits to the data using the
GD6 blood levels. Thus, EPA has decided that the use of the GD6 data as the dose metric is
appropriate for this analysis. Text has been added to Section 5.1.2.3 to describe this analysis of
the dose metric options.
Comment 11: One commenter noted that ".. .in Table B-l, only central estimates of these
two parameters [Vmaxc and Km] are presented for both the rat and human PBPK models."
Response: While the software used for PBPK modeling, acslX, nominally reports
measures of certainty in the estimated parameters, these estimates are based on a statistical
model which assumes that each data point is an independent measurement from a separate
experiment. The model does not account for data in the form of group means (and SDs), nor for
repeated measurement sampling as occurs with urine collection over time or repeated blood
samples from the same individual. While one could potentially build the correct statistical model
as a computational script, that would require the individual subject data rather than summary
statistics, and those data are not available for some, likely most of the studies used for parameter
estimation. Thus the measures of parameter variance reported by acslX are incorrect and the data
needed to correctly estimate the variance are not available. The EPA believes it is preferable to
simply not report measures of variance, rather than providing incorrect values.
Comment 12: One commenter stated that ".. .there is little, if any, data showing that the
human metabolism of methanol begins to saturate at blood methanol levels as low as 36 mg/L,"
that "EPA should make a concerted effort to quantify the uncertainty inherent in its estimates of
the human Vmax and Km parameters, including the strong positive covariance that is expected to
occur between these two parameter estimates" and that "EPA should also implement a purely
linear form of their human PBPK model to determine what differences in final RfC values could
arise as a consequence of the large degree of residual uncertainty that remains regarding exactly
where on the blood methanol scale human metabolism actually does begin to saturate."
Response: It should first be noted that the values of Km reported previously for monkeys
and one human subject by Perkins et al. (1995) were estimated using a classical (non-
physiological) model. For one of the studies analyzed by Perkins et al. (1995), Makar et al.
A-74
-------
(1968), blood concentrations in the monkey were not measured but were estimated from the total
dose and clearance was estimated based on exhaled CC>2 and urinary excretion. For all monkey
studies analyzed the initial doses ranged from 1-6 g/kg and for the human subject the initial
blood level was estimated to be 1.4 g/L. None of the blood data used by Perkins et al. (1995) to
estimate Km values fell below -100 mg/L. Thus none of those other data sets were particularly
suited to identifying a potential Km below 100 mg/L, even if that exists. By contrast, the analysis
of the much lower, more relevant monkey exposure data from Burbacher et al. (2004b; 2004a)
shown in B.3 yielded an estimated Km of 14.4 mg/L, a value which none of the commenters has
called into question. (This result was also obtained with a classical PK model structure.)
Therefore, EPA concluded that the prior Km results from a single published analysis show that
the value obtained for humans is likely incorrect. The model as parameterized is clearly
consistent with (fits well) the existing human data. The difference between the previously
reported values for monkeys and EPA's value for humans is explained by the use of a different
data set which covers a much lower range of internal concentrations, in humans.
Comment 13: One commenter noted that EPA's human PBPK model is "seriously
flawed" because "...predicted human levels are actually markedly larger than the corresponding
mouse levels at the effect levels in the Rogers et al. (1993b) study" and "... [f]or many years, the
conventional toxicological wisdom regarding interspecies differences in methanol toxicity has
been that the human is the least sensitive of the three species (human, rat, mouse), which is
followed by the rat, and finally by the mouse, which is consistently the most sensitive, and this
hypersensitivity of the mouse is due at least in part to the fact that the mouse's metabolism of
methanol saturates at far lower blood methanol levels than do either the rat or human metabolism
of methanol."
Response: First, as discussed in response to Comment 7 above and in more detail in
Sections 5.1.3.2.3 and 5.3.7 of the methanol (noncancer) toxicological review, there is
considerable uncertainty regarding the potential adversity of low level exposures in humans
(e.g., resulting in blood concentrations -10 mg/L). Second, the EPA notes that the "conventional
toxicological wisdom" which the commenter states exists for a greater sensitivity to the
toxicological effects of methanol of rodents versus human and non-human primates is a
hypothesis (i.e., lacks supporting data). While it may have been believed that humans are less
sensitive than mice to high exposure levels for certain developmental effects, there are no dose-
response data in humans or primates to support this assumption. In fact, there is evidence to
indicate the opposite relationship for effects associated with acute (e.g., ocular) and chronic
(e.g., CNS) methanol exposures. Third, sensitivity is the result of both pharmacokinetic and
pharmacodynamic factors. Hence it is possible that humans experience higher internal doses of
A-75
-------
methanol but are less sensitive due to differences in pharmacodynamics. Fourth, the
commenter's statement about sensitivity assumes that it is parent methanol that is responsible for
methanol's developmental effects. While the EPA has chosen to operate under that assumption
for the purpose of this assessment, given the lack of MO A information, if in fact the toxicity is
due to methanol metabolites then a low Km in humans would in fact be consistent with lower
sensitivity to high exposure levels. Hence the estimation of Km = 36 mg/L by the EPA is not
inconsistent with that assumption for other dose metrics that might be considered, so it is
effectively neutral relative to that assumption.
The EPA's analysis of the human data does indicate that these data do not clearly
establish a concentration level at which human metabolism saturates, as reflected by the caveats
placed on use of the human model. But the EPA also does not believe that "the fact that the
mouse's metabolism of methanol saturates at far lower blood methanol levels than ... human
metabolism of methanol," is supported by the existing data. The exact value of the Km in humans
is uncertain but the results of EPA's modeling clearly show that the human data are consistent
with Km = 36 mg/L.
Comment 14: Two commenters stated that the lack of usable primate studies has been
cited as the justification for retaining a database UF of 3 and an interspecies UF of 10, "thereby
using the same source of uncertainty twice" or "double-counting." One of these two commenters
also stated that".. .the lack of a developmental neurotoxicity study should not trigger a database
UF of 3" because "[i]t is unlikely that neurotoxic effects would be seen in a developmental
neurotoxicity study at exposures lower than those that caused reduced brain weights in a
reproduction study." Another of these two commenters stated that "The UF for database gaps,
even though set at 3 rather than 10 is not needed based upon the available data and the choice of
the most conservative decision at several points is the derivation of the reference values (See
Comment 3 to Charge Question B.4, P. A-26)." Another commenter stated that "the continued
use of a database UF of 3 is highly questionable given the well-recognized large data set that
already exists for methanol. This commenter noted that".. .EPA was able to use not one but two
studies with two different toxicological endpoints as a basis for developing a common RfC" and
asks "... how much data are enough?" Another commenter stated they were "... not convinced
that both a UFD of 3 and UFH of 10 are necessary." Another commenter noted that ".. .the wide
variability of 'endogenous' levels in humans, and the uncertainty in the primate sensitivity data
definitely leave grounds for concern. This certainly justifies the inclusion of an additional
database uncertainty factor."
Response: The inter-species (animal-human) uncertainty factor, UFA was not set to 10,
but was set to 3, the default value for pharmacodynamic differences between animals and
A-76
-------
humans (i.e., allowing that humans may be more sensitive to a given internal dose). The UFA is 3
instead of 10 because the PBPK model is used to capture inter-species differences in
pharmacokinetics. As discussed in response to Charge B4 Follow-up Peer Review Comments 2
and 3, in the revised Section 5.1.3.2.3, EPAhas clarified that the UFD is based on deficiencies in
the methanol toxicological database, particularly with respect to the interpretation of the
importance and relevance reproductive, developmental neurotoxicity and chronic CNS effects
observed in monkeys. Thus, the UFD does not have the same basis as the UFA, and was not
"double counted."
With respect to the need for an additional developmental neurotoxicity (DNT) study, EPA
(2002) guidance places particular emphasis in this regard on database deficiencies in the area of
developmental toxicity, the primary focus of the methanol (noncancer) assessment, stating that
"[i]f data from the available toxicology studies raise suspicions of developmental toxicity and
signal the need for developmental data on specific organ systems (e.g., detailed nervous system,
immune system, carcinogenesis, or endocrine system), then the database factor should take into
account whether or not these data are available and used in the assessment and their potential to
affect the POD for the particular duration RfD or RfC under development." EPA believes that,
with respect to the methanol database, the available toxicology studies in monkeys "raise
suspicions of developmental toxicity and signal the need for developmental data," particularly
with respect to DNT. Table 5-5 of Section 5.1.3.2.3 indicates that methanol blood levels
associated with DNT effects are a 12-fold higher in rodents versus primates. Some of this
dissimilarity may be due to differences in species sensitivity, for which the UFA of 3-fold is
intended to account, but some of the difference may be due to other factors, including whether
appropriate and comparable endpoints were examined and whether appropriate study designs and
quality control measures were used. To account for these additional uncertainties, a 3-fold UFD is
applied.
With regard to the sufficiency of the existing database, EPA has clarified in the revised
Section 5.1.3.2.3 that while the database for methanol toxicity is extensive in terms of the
laboratory species and study design coverage, consisting of chronic and developmental toxicity
studies in rats, mice, and monkeys, a two-generation reproductive toxicity study in rats, and
neurotoxicity and immunotoxicity studies, it leaves considerable uncertainty with respect to the
importance and relevance of reproductive, developmental and chronic effects observed in
monkeys.
EPAs response to Comment 15 below addresses the general concern regarding EPA's
choice of ".. .the most conservative decision at several points." It should be noted here, however,
that in consideration of the credibility of the available scientific information, EPA did not derive
A-77
-------
lower RfD and RfC values using the monkey dose-response data. For instance, as can be seen
from Table 5-5, blood levels associated with DNT effects in monkeys were 12-fold lower than
blood levels that caused DNT in rats. Thus, a POD derived from the developmental monkey
study would have likely been substantially lower than the POD derived from the rat
developmental study. Thus, the UFo is not an additional "conservative decision" but is intended
to account for the possibility that deficiencies in the methanol database are causing EPA to derive
a RfD/RfC that may not be as health protective as required (i.e., by using the rodent studies
instead of the monkey studies). As stated in response to Charge B4 Follow-up Peer Review
Comment 2 above, this use of the UFD is consistent with EPA (2002) guidelines which state that
"[t]he database UF is intended to account for the potential for deriving an under protective
RfD/RfC as a result of an incomplete characterization of the chemical's toxicity."
EPA agrees with the commenter that stated that".. .the wide variability of 'endogenous'
levels in humans, and the uncertainty in the primate sensitivity data definitely leave grounds for
concern." As stated above, the UFo accounts for deficiencies in the database that limit the
Agency's ability to interpret the overall findings, particularly the primate studies, and derive
sufficiently protective RfD and RfC values.
Comment 15: One commenter stated that ".. .EPAhas unnecessarily relied upon
automatic use of the choice that will result in the lowest reference value that is scientifically
defensible." Another commenter recommended that".. .whenever the Agency calculates a
reference concentration below background, it should cause the Agency to pause and ask whether
the proposed reference concentration calculation is being driven by the best available science or
by assumptions about uncertainty and by the choice of a particular model that may be too
conservative in this particular case." This commenter noted that "[although EPA has developed
and published methodology for performing Benchmark Dose analysis (BMD), the choice of
which of the ten or so models to use [for derivation of the RfC POD] is not based on any
understanding of mode of action" but".. .is the one that gives the lowest BMD value without
indication of bad fit to the data. According to this commenter, "[t]he Agency ignored models that
resulted in a 10-fold higher BMD, including the linear model" but".. .in other venues, EPA has
been asserting that all toxicological risk is linear with dose." This commenter further stated that
"[t]he lower the set values, the greater the cost of regulations and clean-ups for manufacturers
and consumers, but the benefit to human health must still be determined" and "[t]hus, RfC and
RfD values ought to be set at the highest reasonable health-protective values, not the lowest
"justifiable" or "measureable" values, and ought to be driven by science, not by arbitrary
Agency-made rules."
A-78
-------
Response: Consistent with EPA guidelines for the development of RfDs and RfCs (U.S.
EPA, 2002, 1994), use of the best available, sound science has been a key focus of the methanol
assessment. The credibility of the available science was an important consideration at every step
of the RfD/RfC derivation process. EPA's commitment to use the best scientific and most
credible toxicological approach available resulted in several choices that do not represent "the
lowest reference value that is scientifically defensible," including the use of:
• rodent studies instead of the more uncertain, but potentially more sensitive
monkey studies,
• developmental endpoints instead of more sensitive endpoints of uncertain
adversity such as the "reduction in the size of thyroid follicles" and "transient
reduction in plasma testosterone levels" endpoints illustrated in Figure 4-2,
• PBPK modeling to derive the HEC in lieu of the 3-fold toxicokinetic portion of
the UFA, which would have resulted in a 10-fold lower RfC (BMDL from NEDO
rat developmental study of 670 mg/m3 +• 300 = 2 mg/m3 versus 2X101 mg/m3)
• PBPK modeling to derive the HED from the Rogers et al. (1993b) inhalation
study in lieu of using the oral subchronic study CTRL, 1986) that was used to
derive the old, 4-fold lower RfD of 0.5 mg/kg-day (see Section 5.2.3),
• BMD modeling in lieu of a NOAEL approach which would have resulted in a
2-fold reduction in the RfC from 2x 101 mg/m3 to 1 x 101 mg/m3 (using a POD of
547 mg-hr/L AUC at the 500 ppm NOAEL of the NEDO (1987) rat
developmental study [see Appendix D, Table D-l] instead of the 858 mg-hr/L
BMDL [see Table B-4]);
• BMDL/UFsfor HEC and HED derivations (i.e., applying UFs to internal dose
BMDLs) instead of using the BMDLs directly for the HEC and HED derivations,
which would have resulted in ~2-fold lower, but less reliable (i.e., less
scientifically credible) reference values (see discussion Section 5.1.3.2).
With respect to the suggestion that the Agency has calculated "a reference concentration
below background," as described in numerous places in the methanol (noncancer) toxicological
review, including the Executive Summary, the RfD and RfC are exposures above background
blood levels of methanol associated with a diet that includes fruits and vegetables, which are
estimated to be below approximately 2.5 mg/L (see Section 5.3.6). As discussed in the response
to the Bonus Charge Follow-up Peer Review Comments and Comments 4 and 6 above, the
primary consideration in deriving the methanol RfD and RfC is whether they represent daily
A-79
-------
exogenous exposures that, when added to background levels of methanol associated with a diet
that includes fruits and vegetables, are not likely to result in an appreciable health risk, even to
sensitive subgroups. Consistent with the view of the second commenter above, a determination
of whether the daily exogenous exposures at a RfD or RfC are "distinguishable" relative to
background methanol blood levels is not a primary consideration for an IRIS risk assessment but
has been accounted for in this assessment.
With respect to the Agency's choice of the Hill BMD model for derivation of the RfC
POD, the Hill model was the proper choice in accordance with established BMD technical
guidance (U.S. EPA, 2012a). However, it also provided a substantially better fit overall (as
indicated by a 4-fold higher p-value for model fit to the response means) and in the area of the
BMD (as indicated by an 8-fold higher scaled residual for model fit at the dose group closest to
the BMD) over other models, including the linear model (See Appendix D, Table D-2). Further,
as mentioned in the 5th bullet above, had BMD modeling not been performed the RfC would
have been reduced by 2-fold. Finally, it is assumed that the "other venues" the commenter is
referring to involve the analysis of dichotomous cancer dose-response data using EPAs
multistage cancer model, which is not the same, and bears little relation to, the linear model EPA
uses to evaluate continuous noncancer data.
In summary, the methanol (noncancer) reference values that have been established are
consistent with EPA guidelines for BMD analysis and the development of RfDs and RfCs (U.S.
EPA, 2012a, 2002, 1994) and are supported by the best available, sound science. They represent
estimates of exposures over background levels of methanol that are not arbitrary and are not
necessarily the lowest "justifiable" or "measureable" values, but are derived by applying
scientifically sound methods to the best available science.
Comment 16 - How to take into account endogenous levels of compounds is a key
issue that goes beyond methanol: One commenter stated that "[t]he question of how to take
into account endogenous levels of compounds for which regulatory exposure values are being
developed is an issue that goes far beyond methanol" and ".. .needs to be discussed in a much
larger arena than methanol or any one chemical alone." Another commenter stated that
"[b]ecause methanol is not the only substance where there are natural levels of the substance in
the human body, absent exogenous exposure, this larger issue of how to conduct hazard
assessments of such chemicals needs to be addressed squarely by the Agency, and the methanol
non-cancer assessment would be the place to start."
Response: The Agency is considering the cross-cutting issues associated with chemicals
for which there are natural, endogenous levels in the human body.
A-80
-------
Comment 17 - EPA has not made its PBPK model publically available: One
commenter stated that, as of June 12, 2013, they were unable to locate the PBPK model source
code files, which EPA states will be "available electronically on the IRIS website
[www.epa.gov/iris}" in one place (page A-9) and "will be posted on the EPA IRIS website, along
with the final methanol (noncancer) assessment" (Page A-32).
Response: An error was made in the link for the EPA model code on the pages noted by
the commenter. The correct link to the EPA HERO database record for the model code is
provided here in the following citation, (U.S. EPA, 2012b), and was available in the citation on
page B-34 and in the reference section of the 2011 draft Appendices. The correct link is given
throughout the Appendices of the current version of the methanol (noncancer) toxicological
review.
Comment 18 - EPA implemented nonphysiological urinary clearance in its PBPK
model: One commenter stated that they ".. .find the approach of modeling clearance as occurring
from mixed venous blood to be lacking with respect to physiological realism" and that "[t]he
more appropriate location for urinary clearance would be from the fraction of arterial blood flow
directed to the kidney (Corley RA, Bartels MJ, Carney EW, Weitz KK, Soelberg JJ, Gies RA,
Thrall KD. Development of a physiologically based pharmacokinetic model for ethylene glycol
and its metabolite, glycolic acid, in rats and humans. Toxicol Sci. 2005 May; 85(1):476-90)."
The commenter further state that".. .the greatest danger is that such a model structure as EPA
used could inadvertently lead to an optimized rate of 'urinary' clearance that exceeds blood flow
to the kidney." This commenter notes that "[t]he implications are of lesser concern for methanol,
since parent compound concentrations are used in the risk assessment, but could have
implications if total metabolism, or levels of a metabolite were a key consideration" and
concludes that they ".. .would not recommend that such a structure be used for other chemicals
without better justification than EPA has provided in this document."
Response: As indicated above in response to the Charge Al Follow-up Peer Review
Comments, the commenters are correct in that the lack of an explicit kidney compartment would
be a significant factor if total urinary clearance was a significant fraction of renal blood flow,
which likely occurs for some other chemicals. However for the methanol model the clearance
rate for this pathway in the rat is only 0.24% of renal blood flow [using a renal flow fraction of
0.141 from Brown et al. (1997)] and in the human is only 0.07% of renal blood flow [using a
renal flow fraction of 0.175 from Brown et al. (1997)1. Therefore including an explicit kidney
compartment with its own flow rate would have a negligible impact on the methanol model
results reported here. These calculations and a statement that the approach should only be used
A-81
-------
when renal clearance is a small fraction (< 10%) of renal blood flow have been added to
appendix B.
Comment 19 - The drinking water scenario used by EPA for derivation of the RfD
lacks explanation/justification: One commenter stated the following concerns and suggestions
regarding the drinking water scenario that EPA applied to derive the RfD:
• "No references/precedents are cited to justify this scenario."
• "Probabilistic descriptions should be considered."
• "... EPA has not provided any analysis that indicates which of the assumptions
embedded in this scenario had an impact on the resulting RfD, and to what
degree."
• "My concern is primarily the precedent that may be set (or continued?...) with
this assessment;"
• "EPA should better justify this departure from established practice."
Response: The commenter is correct that the drinking water pattern used has not been
extensively evaluated, but it was previously used in the posted dichloromethane toxicological
review (which underwent extensive peer review) and has been described in the recent peer-
reviewed paper by Sasso et al. (2013) for chloroform. While a probabilistic analysis as suggested
by the commenter would be ideal, the EPA is not aware of an available published statistical
model for water/fluid ingestion in a given day. Variability in total water imbibed from one day to
another is likely available, but a description of the detailed drinking pattern within a day (e.g.,
probabilities of ingestion in a given time increment) would need to be generated, the analysis
conducted, and the results subject to peer-review.
As stated in the draft review, the pattern is meant to be representative, rather than exact.
Except for individuals receiving medical treatment, an assumption of continuous ingestion over
the entire 24 hours of each day, prior standard practice, is clearly unrealistic. At the other
extreme, to assume that all water was ingested in a single daily bolus would be equally
unrealistic. The proposed scenario is most certainly between those two unrealistic extremes.
A detailed analysis of the uncertainty associated with the assumed pattern could be
insightful, but to determine the extent to which the pattern should be varied in such an analysis
would require development of the probabilistic model mentioned just above. Future research
may provide information to be considered beyond the current assessment. The EPA believes that
the assumed pattern is sufficiently more realistic than assuming continuous exposure (the
A-82
-------
previous standard approach), that this realism is effectively self-evident, and therefore that it can
be used and considered an improvement over prior practice.
Comment 20. One commenter stated that "I am not so content with moving the model
source code files to the HERO database since that database is not freely accessible to the public
(needing password access and an EPA account)."
Response: The model source code files are available in a publically accessible record of
the HERO database via the following citation link, (U.S. EPA, 2012b). This citation here, and
elsewhere in the assessment, links directly to the HERO record.
Comment 21: One commenter stated that they ".. .disagree with the comments made in
the Appendix (D26, Iinel2) that 10% is a typically justifiable BMR for non-developmental
dichotomous data, and that nested developmental studies are necessarily 'more sensitive' than
straightforward non-developmental studies." This commenter noted that "[t]he reason for the
nested design in developmental studies is to accommodate the additional complicating litter
effects in these data: it allows the analysis to deal with additional variability and bias in the data
(which do not exist in the non-developmental data) rather than providing more accuracy or
sensitivity." This commenter also noted that "[i]n order to retain comparability with existing
assessments, and to retain compatibility with the guidance on UFs, it is necessary to have a
BMDL which is at least in general properties similar to a NOAEL, and practical experience has
shown that in fact the BMDL05 best meets this criterion for dichotomous general toxicity data as
well as developmental studies."
Response: The statements in Appendix D referred to by the commenter are taken from
the recent EPABMD technical guidance (U.S. EPA. 2012aX which states that "[t]he 10%
response level has customarily been used for comparisons because it is at or near the limit of
sensitivity in most cancer bioassays and in noncancer bioassays of comparable size" and further
states that "[f]rom a statistical standpoint, most reproductive and developmental studies with
nested study designs easily support a BMR of 5%." While EPA agrees that the nested study
design of developmental studies does not necessarily makes them more sensitive, in many cases,
developmental studies have the advantage of a larger sample size, which can increase statistical
power and allow for the use of the lower BMR. Also, as the commenter suggests, a series of
papers have shown that when data are expressed as the proportion of affected fetuses per litter
(nested dichotomous data), the NOAEL was on average 0.7 times the BMDL for a 10% excess
probability of response and was approximately equal, on average, to the BMDL for a 5% excess
probability of response (U.S. EPA, 2012a). The text in Appendix D has been modified to remove
the suggestion that the nested study design of developmental studies justifies a BMR of 5%.
A-83
-------
APPENDIX B. DEVELOPMENT, CALIBRATION,
AND APPLICATION OF A METHANOL PBPK
MODEL
B.1. Summary
This appendix describes the development, calibration, and approach for application of
PBPK models for adult (non-pregnant) Sprague-Dawley (S-D) rats and humans to extrapolate rat
methanol inhalation-route internal dose metrics to human equivalent inhalation exposure
concentrations (HECs) or oral exposure doses (HEDs) that result in the same internal doses. This
model is a revision of the model reported by Ward et al. (1997), reflecting significant
simplifications (removal of compartments for placenta, embryo/fetus, and extraembryonic fluid)
and several elaborations (details follow), which allow the model to describe methanol blood
kinetics. The reasoning for removal of the pregnancy description is given in Section 3.4.1.2, so is
not reiterated here.
The model includes compartments for lung/blood methanol exchange, liver, fat, and the
rest of the body. A single set of parameters was identified for each species modeled, whereas
Ward et al. (1997) employed a number of data-set specific parameters. Fitting parameters to each
data set make it difficult at best to apply the model to bioassay conditions (i.e., to extrapolate the
model to exposure scenarios not used for model calibration). Other biokinetic methanol models
that were considered as starting points for the current model also used varied parameters by data
set to achieve model fits to the data. For example, the model of Bouchard et al. (2001) used
different respiratory rates and fractional inhalation absorbed for different human exposures.
Thus, model re-calibration using a single set of parameters was considered necessary for use in a
health assessment.
The model structure common to rats and humans is described in further detail in Section
B.2.1. Three model features are species-specific:
• (1) A term to account for observed decreases in respiration rate (and assumed
corresponding decrease in cardiac output) was used to match rat data for rats reported by
Pollack and Brouwer (1996). Human exposures used for model calibration, and for which
model application is expected, are assumed to be low enough that the term is inactive for the
human model.
• (2) A urinary bladder compartment is used to simulate urine excretion time-course data in
humans. Human urinary data are sufficient to identify a bladder residence-time constant, but
no such data are available for rats. Urinary elimination is included in the rat model, but the
kinetics of methanol appearance in rat urine were not analyzed.
B-1
-------
• (3) For rats, the body:blood PC had to be adjusted to match the short-time i.v. data (i.e., the
data indicated that the volume-of-distribution predicted by assuming that body:blood
partitioning was identical to muscle:blood was incorrect). But once this was done, the oral
data were well predicted with 100% bioavailability. However for humans, since there was a
single limited i.v. data set which could not be used to calibrate the body:blood PC, the value
for muscle was used, but it was then found that less than 100% oral bioavailability must be
used to match the oral PK data.
Further details of and justification for these features are given in corresponding Sections below.
Algebraic functions which approximate the full human PBPK model to within -1% are
also presented. These functions allow one to calculate human oral methanol doses (HEDs) and
inhalation concentrations (HECs) yielding internal dose(s) equal to specified maximum
concentrations (Cmax values) or area-under-the-curve (AUC) values, specifically to match
internal doses (internal PODs) determined from rat dose-response data.
B.2. Model Development
B.2.1. Model Structure
The model structure is shown in Figure B-l. A gas-exchange model for inhalation
exposure was added, with an adjustment factor (FRACIN) for methanol absorption/desorption in
the conducting airways, as was done by Fisher et al. (2000), to describe delivery of methanol to
blood as a function of ventilation, partitioning, and blood flow rather than the less standard
approach used by Ward et al. (1997). A second (non-physiological) GI compartment was added
to better describe oral uptake in rats. For humans the limited oral PK data were not sufficient to
identify the two additional parameters associated with the second GI compartment, but the data
were consistent with less than 100% bioavailability. (Rat data were consistent with 100%
bioavailability, with the second GI compartment included.) The kidney was lumped with the
body compartment because the blood:tissue partition coefficients for these tissues were similar
and recent practice in PBPK modeling is to treat urinary clearance as occurring from the blood
compartment rather than the kidney tissue [for example, (Loccisano et al., 2011)]. In particular
this reflects the biological reality that renal excretion is initiated by filtration of blood flowing
through the glomeruli, rather than a partitioning form kidney tissue into the nephrons. Where the
current model is not realistic is that it uses venous blood rather than arterial blood concentrations
and the rate of clearance is not limited by the fraction of blood flow to the kidney. The use of
venous vs. arterial blood could be significant where there is very high clearance or uptake in
B-2
-------
various tissues, but examining these two for the experimental studies used for model calibration
showed the difference was below 1% for rats and below 4% for humans, with the maximum
difference occurring during inhalation exposures. Where the kidney blood-flow limitation would
be significant is if the clearance in the kidney was a significant fraction (i.e., greater than 10%)
of the kidney blood flow. However for the methanol model the clearance rate for this pathway in
the rat is only 0.24% of renal blood flow [using a renal flow fraction of 0.141 from Brown et al.
(1997)1 and in the human is only 0.07% of renal blood flow [using a renal flow fraction of 0.175
from Brown et al. (1997)1. Therefore including an explicit kidney compartment with its own
flow rate would have a negligible impact on the methanol model results reported here.
A fat compartment was included because it is the only tissue with a tissue:blood
partitioning coefficient appreciably different than unity, and the liver is included because it is the
primary site of metabolism. Background levels of methanol are included through use of a zero-
order rate of infusion, R0bg. Equations in the model code allow R0bg to be calculated as a
function of other model parameters to match a user-specified background blood or urine
concentration.
B-3
-------
Inhalation Fracin
exposure
Exhaled
air
Bladder
(human only)
Urine
Metabolism
Endogenous
production
-(fStomacrO
•si ^-*- _-—^
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 Gl compartment; ksi, first-order
transfer between stomach and 2nd Gl; Vmax and apparent Km Michaelis-Menten rate constants for metabolism in liver; k-i, first-order
rate constant for urinary elimination; kb!, 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.
Methanol is well absorbed by the inhalation and oral routes, and is readily metabolized to
formaldehyde, which is rapidly converted to formate in both rodents and humans. Although the
primary enzymes responsible for metabolizing formaldehyde are different in rodents (CAT) and
adult humans (ADH); the metabolite, formate, is the same, and the metabolic rates are similar
(Clary, 2003). The published rodent kinetic models for methanol differ in how they describe the
metabolism of methanol (Bouchard et al.. 2001: Fisher et al.. 2000: Wardetal.. 1997: Hortonet
al., 1992). Ward et al. (1997) used one saturable and one first-order pathway for mice, and
Horton et al. (1992) applied two saturable pathways of metabolism to describe methanol
elimination in rats. Bouchard et al. (2001) employed one metabolic pathway and a second
pathway described as urinary elimination in rats and humans, both being first-order.
Since metabolic reactions are known to be saturable - the rate is ultimately limited by the
amount of enzyme present - and metabolism is known to be the primary route of elimination in
rats and humans, the starting point for both rats and humans was to assume the simplest model
B-4
-------
form consistent with this biochemistry: a single saturable pathway, described by Michaelis-
Menten kinetics. This model structure provided a reasonable fit to a range of data, with the first-
order urinary pathway included. If the human PK data, in particular, were completely linear, then
attempts to fit this structure would result in a lack of parameter convergence, with the saturation
constant (Km) approaching infinity, which did not occur. Parameter estimation converged to a
reasonable value for Km when the model was fit to the human data, and the resulting fit to the
data was slightly but clearly improved versus a forced first-order function (evaluated by both a
quantitative measure of fit and visual inspection). Further, when human model optimization was
begun with larger values of Km, which would make the model predictions more linear, the
optimization still converged to the value reported here, clearly indicating that the human data are
more consistent with a saturable metabolic description than a first-order description. The impact
of uncertainty in this model choice is discussed in the corresponding section of the review
(Section B.2.6.1). However, that uncertainty would not be reduced by assuming strictly first-
order metabolism, given the data available and results described here.
Inclusion of a second metabolic pathway was tested for the rat model, but was found to
create problems with parameter convergence and not found to significantly improve model fits.
For the rat, suppression of respiration rate at higher exposure levels was reported by
(Perkins et al., 1996a). Therefore, an empirical function was fit to the respiration rate vs. blood
data from Perkins et al. (1996a) and, assuming this indicates a parallel depression in both cardiac
output and ventilation, the function was applied to the rat cardiac output with ventilation-
perfusion-ratio fixed. Further details are given in Section B.2.3 on rat model calibration below.
While the PBPK model explicitly describes the concentration of methanol, it only
describes the rate of metabolism or conversion of methanol to its metabolites. Distribution and
metabolism of formaldehyde is not considered by the model, and this model does not track
formate or formaldehyde. The data needed to parameterize or validate a specific description of
either of these metabolites is not available. Since the metabolic conversion of formaldehyde to
formate is rapid (< 1 minute) in all species (Kavet and Nauss, 1990), the methanol metabolism
rate should approximate a formate production rate, though this has not been verified. Thus the
rate of methanol metabolism predicted by the model can be used as a dose metric for either or
both of these metabolites, but scaling of that metabolic rate metric to humans requires that the
rate be normalized to BW°'75, (i.e., scaled rate = mg/kg0'75 - time), to account for the general
expectation metabolic elimination of the metabolites scales as BW°'75, hence is slower in
humans. First-order rate constants were scaled as 1/BW0'25, since the resulting rate is also
multiplied by tissue volume which scales as BW1.
The model was initially coded in acslXtreme vl.4 and was subsequently updated in acslX
v 3.0.2.1 (The AEgis Technologies Group, Inc., Huntsville, AL). Most procedures used to
generate this report, except those for the optimization, may be run by executing the
B-5
-------
corresponding .m files. The model code (acslX .csl file) and supporting .m files are available
electronically on the EPA HERO database (U.S. EPA. 2012b). A key identifying .m files
associated with figures and tables in this report is also provided in the supporting materials.
B.2.2. Model Parameters
Physiological parameters such as tissue volumes, blood flows, and ventilation rates were
obtained from the open literature (Table B-l). Parameters for blood flow, ventilation, and
metabolic capacity were scaled as BW°'75, according to the methods of Ramsey and Andersen
(1984). Pulmonary air-flow (Qp) was coded as the product of cardiac output (Qc) and a
ventilation-perfusion ratio (VPR) in order to facilitate coding of changes in these quantities due
to exercise or respiratory depression in rats. In particular it was generally assumed that VPR
remained constant, so Qp and Qc varied in proportion to one another during such changes, unless
data specifically indicated otherwise.
As briefly described in the summary, when published partition coefficients (PCs) were
used for all body compartments for the rat; the predicted blood levels immediately following i.v.
doses were not well estimated. Since those blood levels only depend on the tissue partitioning
and the rest-of-body compartment is compromised of multiple tissues which have differing
partition coefficients, it was therefore decided to initially fit the body:blood PC to the i.v. data
and then to the total PK data set in global parameter estimation. This approach is validated by the
observation that the resulting fitted PC was in the range of those measured for other tissues and
the rat model was then consistent with 100% oral bioavailability. Rat PCs were taken as
measured for that species by Horton et al. (1992) for liverblood and blood:air. The "slow-to-
blood" PC (1.1) in rats reported by Horton et al. (1992), is inconsistent with the value for
fatblood (0.083) in mice from Ward et al. (1997), and that determined for rat fatblood (0.11)
partitioning of ethanol by Pastino and Conolly (2000): these other results indicate much lower
partitioning of alcohols into fat. Therefore the Ward et al. (1997) PC for mouse fatblood, was
used.
B-6
-------
Table B-l Parameters used in the rat and human PBPK models.
S-D Rat
Human
Data Source
Body weight (kg)
0.275a
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)
Calculated"
Flows: Total
Cardiac output (QCC; L/hr/kg°75)c
Ventilation-perfusion ratio (VPR)C
16.4
1
16.5
1.45
Brown et al. (1997): Perkins et al.
(1995): U.S. EPA (2000a)
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)
kiC (kg°25/hr)
21.4
29
0.153
41
36
0.034
Fitted, except rat kiC which is call
from Pollack and Bronwor (199P)
culated
Oral absorption
kasfhr'1)
ksifhr'1)
kaifhr'1)
Bav (fraction)
12.8
3.1
0.38
1
0.21
3.17
3.28
0.79
Rat: fitted, except Bav assumed -
[from (Sultatos et al., 2004)1: Bav
= 1
anol
fitted.
B-7
-------
Table B-l (Continued): Parameters used in the rat and human PBPK models.
S-D Rat
Human
Data Source
Partition coefficients
LiverBlood
Fat: Blood
Blood:Air
Body: Blood
Lung:Blood
Bladder time-constant (kbi, hr"1)f
Inhalation fractional availability
(FRACIN, %)
1.6
0.083
1,350
0.89
1
NA
0.81
0.5838
0.142
1,626
0.805
1.07
0.76
0.75
Human: Fiserova-Bergerova & Diaz
(1ysfc>) (human "body" assumed -
muscle)'
Rat: Morton et al. (1992): except rat
fatblood assumed equal to mouse
(Wdid el dl., 1997) body. blood Wdb III lo
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
bThe 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.
°ln the model cardiac output (QC; L/hr) was set as the primary constant, via the scaling constant QCC (QC/BW075), 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 a saturable metabolic process assumed to occur solely in the liver. Vmax used in the model =
Vmaxc (mg/kg075-hr)xBW°75. kiC is the first-order urinary elimination constant (from the blood compartment), ki used in the
model = kiC/BW° .
eHuman liverblood partition coefficient estimated by Fiserova-Bergerova and Diaz (1986) from correlation to measured fat:blood
partition coefficient, based on data from 27 other solvents.
kbi - 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
The S-D rat model was calibrated to fit blood concentration data from intravenous,
inhalation, and oral exposures. However, the urinary clearance constant and the respiratory
depression function were specified outside of the PBPK model, using separate data. Pollack and
Brouwer (1996) used linear regression of urine excretion rates versus blood concentration in
non-pregnant rats to obtain a clearance constant, ki = 0.00916 L/hr/kg. This was converted to the
equivalent kiC in the PBPK model by normalizing to the venous blood fraction of BW,
0.0443 L/kg, and multiplying by an average SD BW of 0.3 kg (raised to 0.25) to obtain the
allometric constant:
kiC = (0.00916 L/hr/kg) x (0.3 kg)0-25/ (0.0443 L/kg) = 0.153 kg°-25/hr.
As mentioned above, suppression of respiration rate in the rat at higher exposure levels
was reported by (Perkins et al., 1996a). An empirical function was therefore fit to the respiration
rate versus blood data from that source, shown in Figure B-2. It was assumed that cardiac output
decreased proportionately with ventilation, so the inhibition term (1+ ([MeOH])/3,940)5'5}"1, was
applied to the rat cardiac output with ventilation-perfusion-ratio fixed. However, when the
response was assumed to occur instantaneously due to changes in mixed venous blood
B-8
-------
concentration (i.e., the mixed venous blood concentration was used for [MeOH]), the model
predicted an unreasonable level of suppression immediately after i.v. dosing because of the short-
term spike in blood levels predicted to occur. If instead the concentration in venous blood exiting
the "body" compartment was used for [MeOH], reasonable model simulations resulted. Since
some (short) time is likely needed for methanol to interact with the neurons involved in
respiratory and cardiac control, and for neural processing of the resulting signal to the heart and
lungs, the use of this body-tissue-blood concentration, for which the methanol concentration
changes are slightly delayed and "smoothed" relative to the mixed venous blood, seems a
reasonable option.
25
.0
J2 20
o> 15
10 -
o
I 5
V0/[l + ([MeOH]/3940)A5.5]
data
0
1000 2000 3000
Blood methanol (mg/L)
4000
Source: Adapted with permission of Informa Healthcare; 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.
B-9
-------
All of the data available for S-D rats are reported with background levels subtracted.
These data are from the laboratory of Gary M. Pollack (then at the University of North Carolina,
Chapel Hill) and most are also presented in the thesis of Keith W. Ward (1995). The original
source reported only values with background subtracted, and neither Dr. Ward nor Dr. Pollack
have retained any other records of these experiments (personal communications). Therefore, the
methanol blood levels reported by NEDO (NEDO, 1987) in control animals, 3 mg/L, was
assumed for all PK experiments analyzed. Rather than adding this number to the reported data,
however, this background was subtracted from model simulation result obtained with this
background level set to match the reported data. Specifically, model simulations were run with a
zero-order endogenous production rate, Robg, set to produce a concentration in venous blood
(CVB) of 3 mg/L in the absence of exposure. This background level is denoted CVBbg and is a
constant in the model code. A secondary variable was defined in the code: CvBmb = CVB - CvBbg,
i.e., the concentration predicted including background, CVB, minus that constant background.
Since the rates of metabolism, including saturation, calculated in the model all used the total
concentration, which includes that produced from the zero-order term, this approach accounts for
background methanol in the animals to the extent possible, given the data, without adjusting the
data using an otherwise assumed background level. All of the plots which follow, demonstrating
model fits to various data, then show model predictions of CvBmb versus the data as reported in
the various publications. Total blood concentrations, CVB, are listed in tables of internal metrics
and show in plots depicting internal dosimetry under bioassay conditions.
Initial values Vmaxc, Km, and the body:blood partition coefficient (PR) were then obtained
by fitting the model to the 100 and 2,500 mg/kg i.v. data provided in the command file of Ward
et al. (1997) (holding other parameters constant). As mentioned previously, if PR was not also
adjusted, the predicted concentration immediately following the distribution phase, which are
only dependent on the partition coefficients, were discrepant from the data. Without adjusting
PR, this then created a bias in the metabolic parameters to correct for the error in the distribution
phase. Model predictions were also compared to 500 mg/kg i.v. data in the command file of
Ward et al. (1997), with additional early time-points reported by Pollack and Brouwer (1996).
With PR adjusted this way to fit the 100 and 2,500 mg/kg data, the model matched the initial
time points of the 500 mg/kg data quite well (see Figure B-3). However, the subsequent
clearance rate fit to these other two dose levels was inconsistent with the 500 mg/kg data. All
three data sets with globally-fit model parameters are shown in Figure B-3. If one compares the
clearance rate of the 500 mg/kg data at 20 hours and beyond, when the concentration range is the
same as the 100 mg/kg data, it is clear that the two data sets are discrepant. Thus no model with a
single set of parameters could simultaneously match both data sets. That the model does fit the
2,500 mg/kg data quite well indicates that the discrepancy is not due to a simple dose-
dependency. Since it is most important that the model describe the low-dose data well, in the
B-10
-------
range of the point-of-departure for toxicity extrapolation, while capturing as much of the high-
dose dependency as possible, the 500 mg/kg i.v. data were not used in subsequent model
calibration.
10,000
en
1,000
o 100
-------
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): Adapted with permission of Springer; Perkins et al. (1996a): (Panel B): Reprinted with the permission of the
Health Effects Institute, Boston, MA; from 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.
B-12
-------
Oral absorption parameters were first fit to the lower dose (100 mg/kg) oral absorption
data reported by Ward et al. (1997) (with other parameters held constant). The initial fit, with a
single GI compartment was not very good, even with the oral bioavailability adjusted at the same
time (dashed line in panel B of Figure B-4). Therefore, an empirical (non-physiological) second
GI compartment was considered, like that used by Sultatos et al. (2004) for ethanol. With
bioavailability fixed at 100%, use of this second compartment gave an excellent fit to the data
(solid line in panel B of Figure B-4). Therefore the two-compartment GI structure was used.
While the fit to the 100 mg/kg oral data was quite good, the fit to the 2,500 mg/kg data
exhibited a much faster and higher peak than shown by the data and under-predicted the data
between 24 and 31 hr, during the clearance phase (Figure B-5, panel A). Notably, the two-
compartment model reproduces these high-dose data much better than the single-compartment
model. Even when the model was fit to both the high- and low-concentration data
simultaneously, the fit to the high-concentration data could not be significantly improved without
completely degrading the low-concentration fit (not shown). Several variations of the GI
compartment rate equations were tested, in part reflecting data available from the ethanol
literature, but none could significantly improve the fit to the 2,500 mg/kg data without
introducing otherwise untested parameters and hypotheses. Since the primary concern is with
fitting low-dose data, which produce blood concentrations near the point of departure, it was
therefore decided to not use the 2,500 mg/kg data for parameter estimation, though comparisons
of model predictions to those data are still presented, and to use first-order kinetics with the
empirical, two-compartment GI model shown in Figure B-l.
A final set of fitted model parameters for the rat was obtained by allowing all of the
adjusted parameters (Vmaxc, Km, PR, kas, ks;, ka;, and FRACIN) to the data sets as described
above: 100 and 2,500 mg/kg i.v. doses [Figure B-3; (WardetaL 1997). squares]; 1,000 to
20,000 ppm inhalation time-course data [Figure B-4, panel A, (Perkins et al., 1996a)], and
100 mg/kg oral dose data [Figure B-5; (Ward et al., 1997)1. The resulting parameter values are
listed in Table B-l and the simulations with solid lines in Figures B-3 to B-5 all use this global
set. Although the model does not fit all of the data as well as one might like, particularly the
1,000 and 5,000 ppm data in Figure B-4, panel A, the overall quality of the fits is considered
good. The number of parameters adjusted is considered modest, since could reduce the number
of parameters by keeping PR at a value measured for muscle or using a one-compartment GI
model. But either of these choices significantly degrades the model fits (shown for GI model),
which indicates that the number and variety of data available are sufficient to inform these seven
fitted parameters. One can consider the urinary excretion constant, kiC, and the two parameters
used to define the level of respiratory depression, as additional adjusted parameters. These two
parameters were fit using additional data on urinary excretion and respiration rate, respectively.
So the total number of fitted parameters is considered to be well supported by the corresponding
B-13
-------
data used to determine their values, and hence a fairly good level of confidence should be held
for model predictions of bioassay dosimetry. To further elucidate the level of confidence one can
place in model predictions, evaluation of model sensitivity to these parameters was conducted as
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
0
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).
B-14
-------
B.2.4. Rat Model Sensitivity Analysis
An evaluation of the importance of selected parameters on rat model estimates of blood
methanol concentration was performed. Since the rat model was only used to evaluate internal
doses during inhalation exposures, the sensitivity to the oral uptake parameters was not
evaluated. The parameters which can affect inhalation dosimetry that were identified by
matching to PK (and respiratory response) data were Vmaxc, Km, kiC [estimated by Pollack and
Brouwer (1996)1, PR (body:blood partition coefficient), FRACIN, and kiv (respiratory/cardiac
depression constant). For the purpose of comparison, the blood:air partition coefficient (PB) was
also included. Sensitivity of the dose metrics, Cmax and AUC (both above background) was
estimated under conditions of the NEDO bioassay (NEDO, 1987), 22 hr/day inhalation exposure,
at the bounding levels of 200 and 5,000 ppm. The analysis was conducted by measuring the
change in each metric resulting from a ±1% change in a given model parameter when all other
parameters were held fixed. The normalized sensitivity coefficient is then:
SC = (A metric/metrico)/ (A p/po),
where metrico and po are the values of the metric and parameter, respectively with the unchanged
(as fitted) values and A metric and A p are the differences between the values obtained with p
increased by 1% and decreased by 1%.
A normalized sensitivity coefficient of 1 indicates that there is a one-to-one relationship
between the fractional change in the parameter and model output; values close to zero indicate a
small effect on model output. A positive value for the normalized sensitivity coefficient indicates
that the output and the corresponding model parameter are directly related while a negative value
indicates they are inversely related. Results are listed in Table B-2.
B-15
-------
Table B-2 Sensitivity of rat model dose metrics to fitted parameters.
Exposure level, metric
Parameter3
VmaxC
Km
PR
PB
k-iC
FRACIN
kiv
Cmax
-1.1
0.7
0.0
0.0
0.0
1.2
0
200 ppm
AUC
-1.0
0.7
0.0
0.0
0.0
1.2
0
5,000 ppm
Cmax
-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.
The sensitivity analysis results are mostly not surprising. At the lower concentration of
200 ppm, metabolic elimination has a significant influence, with both Vmaxc and Km having high
SCs. The SC for Vmaxc is negative since an increase in its value decrease blood concentration,
while Km is positive for the opposite reason. At 5,000 ppm VmaxC is only marginally significant
and Km not at all, but urinary elimination (kiC) becomes significant, though only slightly. The
one somewhat surprising result is that the body:blood partition coefficient, PR, has very little
influence on the inhalation dose predictions. However, the analysis was conducted on conditions
near steady-state with 22 hr/day exposure. As shown by Chiu and White (2006), the steady-state
level predicted in blood by a PBPK model depends on only a small number of parameters: those
affecting absorption, elimination (metabolic), and the blood:air partition coefficient (PB). For
this model, at 200 ppm the rate of absorption by inhalation is likely limited by respiration rate,
hence PB has little influence at that concentration, but it does significantly impact uptake at
5,000 ppm. More importantly, since PR has so little effect on these predictions means that any
uncertainty in its value is inconsequential to the outcome of this assessment. (PR is expected to
more strongly influence non-steady-state conditions, such as when oral ingestion occurs in
boluses.)
The fraction inhaled (FRACIN) is highly sensitive at both dose levels. The respiration
inhibition constant, k;v, has no influence at 200 ppm but is sensitive at 5,000 ppm. Since
increasing k;v decreases the level of inhibition - increases respiration - its coefficient is positive.
Differences in the sensitivities of the two metrics existed in the second decimal place, but
otherwise the two are closely correlated for this exposure scenario, hence the SCs are effectively
identical.
Thus, all of the adjusted parameters except PR have a significant influence on model
predictions over part of the relevant range of concentrations. Of these fitted parameters, kiC and
B-16
-------
k;v were fit to independent data sets, not used to fit any other parameters. Hence a good degree of
confidence can be given to their values. Because of the wide range of doses, particularly by the
i.v. route, used for the PK data, VmaxC and Km can also be considered fairly well identified.
However the model's inability to fit the 500 mg/kg i.v. data (Figure B-3) and 1,000 and
5,000 ppm inhalation data (Figure B-4, panel A) create some level of uncertainty in their values
and that of FRACIN. That the model fits rather well both the 100 and 2,500 mg/kg i.v. data,
makes it difficult to come up with a simple explanation for the lack-of-fit to the intermediate
dose. Since the clearance observations at 500 ppm go beyond 24 hr, it is possible that there is a
time-dependent process that reduces clearance in that time range. The 2,500 mg/kg i.v. dose
clearance was only measured to 43 hours, when it had just dropped to -100 mg/L, so one cannot
say if the clearance from then on would have been more like the 100 mg/kg data or the
500 mg/kg data.
For FRACIN, the poor fit to the lower two inhalation exposures (Figure B-4, panel A)
suggests a concentration-dependence; (i.e., FRACIN is higher at low concentrations. However,
even if FRACIN is set to 100%, the later time points for the 1,000 and 5,000 ppm concentration
curves are ^/'//under-predicted (results not shown). One hypothesis is that at low concentrations,
deposition in the conducting airways leads to a significant amount of absorption, not accounted
for in the standard gas-exchange model used here. Including such a mechanism would increase
model complexity significantly, and such a hypothesis should be tested by also comparing model
predictions to methanol gas uptake experiments, which would clearly show if methanol is being
taken up more efficiently at low concentrations versus an error in the model's description of
metabolic elimination or some other systemic process.
This consideration of possible model errors and potential future improvements (with
necessary data) should be balanced against the observation that the model-predicted blood level
at 8 hours from a 1,000 ppm exposure, 81 mg/L, almost exactly matches the measured
concentration reported in Pollack and Brouwer's (1996) (Table 16): 83 ± 15 mg/L. The
discrepancy between that result and the value obtained from a plot (digitized) in the same report
which also appears in Perkins et al. (1996a), -290 mg/L, can only be attributed to experimental
variability, which no model can fully describe. Since the model does fit the lower-concentration
8-hr data (Figure B-4, panel B) fairly well, it is considered adequate for use in the assessment as
is, without further complication and additional parameters, and FRACIN is assumed to provide a
reasonable adjustment to the internal doses with the value obtained here.
B-17
-------
B.2.5. Rat Model Simulations
A range of adverse developmental effects was noted in rat pups exposed to methanol
throughout embryogenesis (NEDO, 1987). In particular, model simulations were conducted for
S-D rats in utero over different periods of pregnancy and as neonates via inhalation. Inhalation
exposures to methanol were carried out for 18-22 hours, depending on the exposure group.
Simulations of predicted Cmaxand 22-hour exposures to 500, 1,000, and 2,000 ppm methanol are
shown in Figure B-6. Although the exposures in these studies are to rats over long periods and in
some cases exposures of the newborn pups, the model simulations are for NP adult rats only, and
do not take into account changes is body weight or composition. These simulated values are
presumed to be a better surrogate for and predictor of target-tissue concentrations in developing
rats, and the corresponding estimated human concentrations a better predictor of developmental
risk in humans than would be obtained using the applied concentration or dose and default
extrapolations. The logic here is simply that the ratio of actual target tissue concentration (in the
developing rat pup or human) to the simulated concentration in the NP adult is expected to be the
same in both species and hence, that proportionality drops out in calculating a HEC.
Figure B-6 depicts rat model simulations to determine internal doses for 22 hours/day
inhalation exposures at 500, 1,000, or 2,000 ppm. Atypical BW of 0.3 kg was used, since
predicted inhalation dosimetry is usually insensitive to the exact BW. Simulation results for
continuous inhalation exposures are shown for contrast. The simulations show that for all but the
highest dose (2,000 ppm) steady-state is reached within 22 hours, and that "periodicity," where
the concentration time course is the same for each subsequent day, is reached by the 3rd day of
exposure. At 2,000 ppm, however, steady state is not reached until after 8 days for the continuous
exposure. Therefore, the Cmax and 24-hour AUC were calculated by simulating 22 hours/day
exposures for 12 days, with the AUC calculated over the last day (24 hours) of that period. The
AUC values shown in Figure B-6 are calculated from the concentration increase above the
background or endogenous level; (i.e.,
AUC = I (C - Cbg)dt
Jo
where the integration is over 24 hours, C is the instantaneous blood concentration, and Cbg is the
endogenous/background level, set to 3 mg/L for the rat).
B-18
-------
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 BW 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/background 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
The rat model was scaled to human body weight (70 kg or study-specific average), using
human tissue compartment volumes and blood flows, and calibrated to fit the human inhalation-
exposure data available from the open literature, which comprised data from four publications
(Ernstgard et al.. 2005: Batterman et al.. 1998: Osterloh et al.. 1996: Sedivec et al.. 19811 and a
single data set for oral exposure (Schmutte et al., 1988). Model predictions are also compared to
an i.v. data set that was not used for parameter estimation (Haffner et al., 1992). Since the bulk of
the human data were from inhalation exposures, the approach to identifying parameters was to
first fit the metabolic (and endogenous level) parameters to those data sets. Initial estimates for
the oral uptake parameters were then obtained by fitting the oral PK data with other parameters
held constant. Lastly, a global fit over the inhalation and oral data sets combined, with all fitted
B-19
-------
parameters varied simultaneously, was performed to obtain final parameter values. The two key
differences in model structure and parameters adjusted are discussed below, followed by a more
detailed description of the calibration against specific data.
More specifically, the human model calibration differed from the rat calibration in two
ways. First, a bladder compartment was included (calibrated) to better describe the kinetics of
human urinary data, where both the rise and the drop in excretion rate is slower than the
predicted decline in blood and tissue methanol and hence rate of metabolite production. This
difference is shown in Figure B-7 for the 231 ppm exposure data of Sedivec et al. (1981). The
model-predicted venous blood and body tissue concentration curves show the pattern typical for
PBPK models which use the common venous-equilibration equations for tissue distribution (used
in this model) for fixed-duration inhalation exposures: an asymptotic rise in concentration during
the exposure period and then a sharp decline starting the moment that exposure ends. If urinary
excretion was assumed to be proportional to the body tissue concentration (which includes the
kidney tissue) or a separate kidney compartment was used wih the same venous-equilibration
equations, then the shape of the predicted time-course would simply mirror that of the tissue
level shown in Figure B-7, which is clearly a poor representation of the data. However, fitting the
one additional parameter introduced for the bladder compartment, the bladder clearance constant,
kbi, allows the model to reproduce the distinct kinetics of urinary excretion quite well. Thus this
addition is considered both biologically realistic and well justified.
The second difference from the rat calibration is that the body:blood partition coefficient
(PR) was not adjusted but the oral bioavailability (Bav) was adjusted. In particular, PR was not
adjusted because only limited i.v. dosing data were available (a single dose level with actual data
only available for one subject). Instead the value measured for muscle by Fiserova-Bergerova
and Diaz (1986) was used for PR without adjustment. However, when attempting to match the
model to the oral PK data, model predictions then significantly over-predicted those data (with
parameters otherwise consistent with the inhalation data). Therefore the oral bioavailability was
allowed to vary to less than 100% to fit the oral PK data.
In summary, the set of key parameters fit for the human model were the metabolic (Vmaxc
and Km) and urinary elimination (kiC and kbi) constants, the inhalation fraction (FRACIN), and
the oral bioavailability (Bav). In addition, the endogenous background concentration and an
increment in background over time were fit to control data from Osterloh et al. (1996). A detailed
description of each data set and the parameter(s) that it primarily informs follow. However, as
with the rat, the final set of parameters was obtained by global optimization: varying all
parameters while fitting all of data sets simultaneously. Other human parameters were set as
reported in Table B-l.
B-20
-------
10 H
Model urine from bladder
Venous blood
Body tissue
• Urine data
CD
12
Time (hr)
16
20
24
Source: Sedivec et a\. (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.
The first-order rate of clearance of methanol from the blood to urine, kiC, and first-order
bladder compartment time constant, kbi, were used to describe urinary methanol elimination. (See
Section B.2.1 on the reasoning for treating urinary elimination as occurring from the blood
compartment versus a kidney tissue compartment.) The inhalation-route urinary methanol kinetic
data described by Sedivec et al. (1981) (Figure B-8) were used to inform these parameters. The
urine methanol concentration data reported by the authors were converted to amount in urine by
assuming 0.5 mL/hr/kg total urinary output (Horton et al., 1992). Since the resulting values of
kiC and kM (Table B-l) are only calibrated using a small data set, they should be considered an
estimate. Urine is a minor route of methanol clearance in humans, with little impact on total
blood methanol concentration, but changes in urine levels are expected to closely reflect
corresponding changes in blood levels, hence the slight nonlinearity in the urine data also inform
the apparent metabolic saturation constant, Km. The potential for this information is lost,
however, if the kinetics of urinary elimination are not well matched; i.e., if the bladder
compartment is not used.
B-21
-------
To estimate both the Michaelis-Menten (hepatic) and first-order (urinary) clearance rates,
all human inhalation data under nonworking conditions were used (Batterman et al., 1998;
Osterloh et al.. 1996: Sedivec et al.. 1981).
• The initial urine concentrations from Sedivec et al. (1981) (reported at time = 0; see
Figure B-8) were assumed to represent endogenous background levels, and therefore were
used to set a (constant) endogenous level for each exposure level to match that urinary level
(i.e., the endogenous blood level that must exist to match the observed urine concentration,
given the urinary clearance constant, kiC). The endogenous blood concentrations so
estimated were 0.6-0.74 mg/L.
• Batterman et al. (1998) subtracted background levels before reporting their results, but also
included the exposure-specific background (pre-exposure) concentrations in a separate table.
Therefore those background levels were added back to the reported exposure-group values
and treated as actual blood concentrations. Results of model fits to the Batterman et al.
(1998) data are shown in Figure B-9.
• Osterloh et al. (1996) measured and reported (plotted) blood methanol in nonexposed
controls (data shown in Figure B-10). The data for Osterloh et al. (1996) clearly show a time-
dependent trend which is close to linear. Therefore, the endogenous methanol production rate
was assumed to increase at a constant rate over time when simulating the Osterloh et al.
(1996) data (both controls and methanol-exposed), with the rate of increase fit to the control
data set. The results shown in Figure B-10 (solid lines) include this increase. For comparison,
the thin dashed line shows results for the 200 ppm exposure if the endogenous production is
assumed to be constant.
B-22
-------
I
o
10
9-
8-
7
6-
5
4
3-
2
1
0
cs
c
Urine concentration
• 231 ppm
O 157 ppm
A 78 ppm
0
4
12
Time (hr)
16
24
Total urinary excretion
• 231 ppm
O 157 ppm
A 78 ppm
0
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.
B-23
-------
16
14
d 12
u>
'— 10
I
o
01
o
_o
m
• 2hr
4- 1 hr
A 30 min
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
as
I
0 3
o J
o
_o
m
0
• 200 ppm
^ 0 ppm
— Constant bgd
0
345
Tirne(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.
B-24
-------
Oral PK data from Schmutte et al. (1988) from a 10 mg/kg dose was used to set an oral
bioavailability for humans and to test the assumption that human oral absorption of methanol
could otherwise be described using the simple two-compartment GI model of Sultatos et al.
(2004). with parameters fit by Sultatos et al. (2004) to ethanol PK data. Sultatos et al. (2004)
included a rate of metabolism for ethanol in the stomach, which would reduce the systemic
bioavailability of that compound from 100%. Lacking the data to fit a specific rate constant for
methanol metabolism in the GI, the simulated dose was simply reduced using a bioavailability
constant (Bav), but the mechanism for less than 100% availability could also be metabolism in
the GI. A value of Bav = 0.79 was obtained and the simulation curve matches the data of
Schmutte et al. (1988) fairly well (Figure B-l 1). The initial condition was set to the reported pre-
exposure background by Schmutte et al. (1988) (1.1 mg/L). The model reproduces the data well,
considering that only one parameter is adjusted for the oral dose route. Data were only collected
for 1.5 hours: a longer sampling time would have provided a better evaluation of the model's
ability to predict longer-term kinetics from oral exposures.
,—, !°-
& 8-
6-
o
'•u
re
o
o
c
o
_'
•u
o
_o
CO
4-
2-
0-1
0.25 0.5 0.75 1
Time (hr)
1.25
1.5
Source: Modified with permission of; 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-1] were used.
Figure B-ll Oral exposure (10 mg/kg) to methanol in human volunteers (points).
The data from Ernstgard et al. (2005) was used to assess the use of the model parameters
with a data set collected under conditions of light work. Historical measures of VPR (2.023) and
QcC (26 L/hr/kg0'75) for individuals exposed under conditions of 50 W of work from that
laboratory (Ernstgard, 2005; Corley et al., 1994; Johanson et al., 1986) were used for the 2-hour
exposure period (Figure B-12). Also, a linear rate of increase in the endogenous production rate
B-25
-------
was fit to the control data set, as this set showed an increasing trend over time, like Osterloh
et al. (1996), and the initial background level was set to match the observed value at time = 0 for
each data set. Otherwise, there were no changes in the model parameters (no fitting to these
data). The results are remarkably good, given the lack of parameter adjustment to data collected
in a different laboratory, using different human subjects than those to which the model was
calibrated.
"
O
_O
CO
8
7^
6
5
4
3
2
1
0
Ernstgard et al. (2005)
n 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/kg075 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
B-26
-------
various body tissues and hence under-predicts blood concentrations in this time period; i.e., that
distribution to body tissues is diffusion-limited, with the effect being significant at shorter times.
The slope of the simulation line closely matches that of Subject B, indicating similar clearance
kinetics. Subjects C and D exhibited faster elimination kinetics than predicted by the model. The
authors report elimination rate constants of 0.259, 0.325, 0.406, and 0.475 hr"1 for Subjects A-D,
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.
Horton et al. (1992) employed two sets of metabolic rate constants to describe human
methanol disposition, but in vitro studies using monkey tissues with non-methanol substrates
were used as justification for this approach. Although Bouchard et al. (2001) described their
metabolism using Michaelis-Menten metabolism, Starr and Festa (2003) reduced that to an
effective first-order equation and showed adequate fits. Perkins et al. (1995) estimated a Km of
320 + 1,273 mg/L (mean + S.E.) by fitting a one-compartment model to data from a single oral
poisoning to an estimated dose. In addition to the extremely high standard error, the large
standard error for the associated Vmax (93 + 87 mg/kg/hr) indicates that the set of Michaelis-
Menten constants was not uniquely identifiable using this data. Other Michaelis-Menten
constants that have been used to describe methanol metabolism in various models for primates
are given in Table B-3. Because the Km calculated by Perkins et al. (1995) from the high-dose
B-27
-------
oral exposure is 320 mg/L, while the highest observed concentration in the data sets considered
here is 14 mg/L (Batterman et al., 1998), forcing the model to use this higher Km would simply
result in fits that are effectively indistinguishable from the linear model. The value obtained in
this analysis, 36 mg/L, allows the model to describe the slight nonlinearity that exists in the data.
For example, the peak urine concentration observed by Sedivec et al. (1981) after the 231 ppm
exposure was increased 8.66 mg/L above the time zero value, while that observed after the
78 ppm exposure was 2.63 mg/L above the time zero value, so a 3-fold increase in exposure lead
to a 3.3-fold increase (8.66/2.63) increase in the urinary excretion above background. It is
possible that a much higher Km pathway is also operant in humans, but is only significant at
much higher concentrations than evaluated here.
Table B-3 Primate Km values reported in
Km (mg/L)
320 + 1,273
716+489
278
252 + 116
33.9
0.66
36a
Note: The values from Perkins et
Reference
Perkins et al. (1995)
Perkins et al. (1995)
Perkins et al. (1995)
Perkins et al. (1995)
Morton et al. (1992)
Fisher et al. (2000)
(This analysis.)
al. (1995) are ±S.E.
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
"This Km was optimized while also varying Vmax, k-iC, kb!, 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.
Sedivec et al. (1981) estimated a fractional uptake of 57.7%, based on total amount
inhaled. Since the PBPK model uses alveolar rather than total ventilation and this is typically
assumed to be 2/3 of total ventilation, one might correct this value by dividing by 2/3 to obtain a
value for FRACIN of 0.8655. Ernstgard et al. (2005) also estimated a fractional uptake, 51% at
100 ppm and 49.3% at 200 ppm under light exercise. It is reasonable to expect uptake efficiency
to decrease with more rapid breathing due to exercise, since an inhaled volume element of air
spends less time in the respiratory tract, allowing less time for uptake, as respiration increases.
Also, while Ernstgard et al. (2005) based their calculation on estimated pulmonary ventilation,
they used the difference between inhaled air concentration and exhaled air concentration.
Exhaled air will be a mixture of air that was taken into the pulmonary airways and air that only
entered the conducting airways. Very little methanol would be absorbed from the later air and
hence the mixed exhaled concentration will be higher than that which exits the pulmonary region
and the resulting calculation will then under-estimate the fraction of methanol absorbed from
B-28
-------
pulmonary air. Thus the "fraction inhaled" estimated from a given data set will depend on which
flow rates and concentrations are being used in the calculation, or to which it might be applied;
i.e., the value depends on the model "context" in which it is used. Therefore, EPA decided to fit
FRACIN with the other parameters estimated in the context of the PBPK model used here, as
was done for the rat, and obtained a value of 0.75. This indicates that the concentration entering
the pulmonary space is reduced by 25% due to deposition in the conducting airways (with that
material assumed to desorb on exhalation), and is not the fraction removed in the pulmonary
space. At 200 ppm, for example, the model predicts that 99.9976% of the methanol entering the
pulmonary region is absorbed. The value is slightly less than estimated for the rat (rat FRACIN =
81%) which seems reasonable since the larger human airways would reduce uptake efficiency
somewhat. Assuming that 2/3 of inhaled air goes to the pulmonary region, the total rate of
inhalation would be 1.5*Qp*CONC (rate of inhalation through nose and mouth at air
concentration CONC), and the amount removed in the pulmonary region roughly
0.75*QP*CONC (using FRACIN = 0.75), so the fraction of each breath absorbed is predicted by
the model to be:
(0.75*QP*CONC)/ (1.5*Qp*CONC) = 50%,
which closely matches the estimates of Ernstgard et al. (2005). Considering that EPA did not fit
FRACIN to the Ernstgard et al. data, this appears to be a good validation of the value obtained
for this parameter.
Considering the model simulations versus the data of Haffner et al. (1992) (Figure B-13),
it is first evident that the model is not capturing the short-term kinetics shown for Subject A.
Since Haffner et al. (1992) did not indicate that the data for this subject were discrepant from the
other subjects in the first hour it is assumed that these data represent human distribution, hence
that the model does not describe well what happens immediately after such an exposure. Given
that i.v. exposures are not a route for which risk is estimated, model failure is not considered
critical here; however, it does suggest an area for future research and model improvement. The
model does track the longer-term clearance data for Subject A quite well. Since the model is
intended to represent an average adult human, it is also not alarming that it does not match so
well the individually-fitted clearance curves for Subjects B-D, which indicates a range of human
variability. In particular the results for those other subjects indicate that some people will clear
methanol more quickly than predicted by the model, which means that the model will somewhat
over-estimate internal doses and health effects for those individuals. Since other data to which
the model is fit are averages among individuals, and the model does not show a strong bias with
regard to those data (Figures B-8 to B-ll), neither does it appear that the model is systematically
under-predicting clearance for most of the population. Therefore, the model predictions are
expected to provide reasonably good estimates of average adult human methanol PK under long-
B-29
-------
term exposure scenarios. Caution is suggested, though, in potential use of the model to estimate
internal doses shortly after accidental exposures.
A sensitivity analysis for human model predictions to the primary fitted parameters was
conducted for continuous inhalation exposures, and results are shown in Table B-4. Normalized
sensitivity coefficients are calculated using the method described for the rat (see B.2.4). To
bracket the range of likely concern for human exposures, inhalation sensitivities were evaluated
at 10 and 200 ppm concentration. The bladder time constant, kbi, was not included in the analysis
since it has no influence on blood concentrations. The resulting coefficients (Table B-4) are not
surprising. Vmaxc and Km both strongly influence model predictions. At these exposure levels the
urinary pathway (kiC) has little effect on blood level. There is essentially a 1:1 correspondence
with FRACIN, which follows from the fact that close to 100% of what enters the gas-exchange
compartment is absorbed. That all of the sensitivities are slightly higher at 200 ppm than at
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
10 ppm
-0.81
0.74
-0.0024
1.00
Exposure level"
200 ppm
-0.93
0.75
-0.0029
1.11
formalized sensitivity coefficients for steady-state blood levels (increase above background) at the indicated concentrations.
For oral exposures ingestion is assumed to occur in a series of six boluses over the course
of the day, with the fraction of the total daily dose and respective times ingested being: 25% at
7 a.m., 10% at 10 a.m., 25% at 12 p.m., 10% at 3 p.m., 25% at 6 p.m., and 5% at 9 p.m. The
pattern is meant to be representative of human ingestion patterns, recognizing that this will vary
among the population. The impact of changing the pattern on estimated AUC values is fairly
small, since the total ingestion remains the same. However the pattern will clearly influence the
peak concentration, since an assumption of ingestion in a single bolus would lead to the highest
predicted daily peak, while assumption of continuous ingestion would lead to the minimal peak
possible, for a given total daily exposure. With the pattern used here, the blood concentration
profile predicted at 10 mg/kg-day is shown in Figure B-14 (time is in hours from first bolus). In
particular, the model predicts a following peak to occur -45-55 minutes after each bolus
(depending on dose size), with the overall daily peak occurring just before 1 p.m. (~6 and 30 hr
time-points in Figure B-14). While the boluses assumed to be ingested at 7 a.m. and 12 p.m. are
both 25%, because methanol is predicted to accumulate somewhat over the morning, the later
B-30
-------
bolus leads to a peak that is roughly 30% higher than the first of the day. At this exposure level
there is a very small residual blood level at 24 hr, about 1% of the mid-day peak. At higher
exposure levels more significant day-to-day accumulation would be predicted until a state of
"periodicity" is reached, when the day-to-day pattern no longer changes. For example, at
200 mg/kg-day the blood level just prior to the next day's ingestion is predicted to approach 3%
of the daily peak (-80 mg/L). At 500 mg/kg-day, the model predicts that it will take about 2
weeks to reach periodicity, where the peak during the first day is -350 mg/L, but this increases
to 740 mg/L after two weeks, and the end-of-day minimum is 460 mg/L.
The model sensitivities to the key fitted parameters at 0.2 and 10 mg/kg-day under this
oral exposure scenario are listed in Table B-5. As with the inhalation sensitivity analysis, these
exposure levels are selected to bracket the range of primary concern for this assessment. The
results are qualitatively the same as for inhalation exposure (Table B-4), with oral bioavailability
(Bav) having an effect essentially identical to that of FRACIN for inhalation. The metabolic
parameters have slightly less impact for these exposure levels; probably due to blood-flow and
oral-absorption limitation, and the increase in sensitivity from 0.2 to 10 mg/L is not as large as in
going from 10 to 200 ppm inhalation concentrations.
01
\-^
I
O
01
CD
3 6
12 15 18 21 24 27 30 33 36 39 42 45 48
Time (Kr)
Figure B-14 Predicted human blood concentrations (increase above background) from
total daily exposures to 10 mg/kg-day methanol, consumed in a series of
6 boluses. Time is from the first bolus of the day. See text for further details.
B-31
-------
Table B-5 PBPK model sensitivity analysis for oral exposure.
Exposure level, metric3
0.2 mg/kg-day 10 mg/kg-day
Parameter Cmax ADC Cmax ADC
Vmaxc -0.67 -0.98 -0.68 -1.01
Km 0.62 0.91 0.59 0.89
_kiC -0.0014 -0.0024 -0.0015 -0.0025
Bav 100 100 103 104
"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.
Considering the multiple data sets used for human model calibration and validation, there
is fairly high confidence in the fitted metabolic/clearance parameters, Vmaxc, Km, and kiC. Since
the pharmacokinetics are mostly linear in the range of interest, it is really Vmaxc/Km that is the
critical determinant of predicted internal doses, but as that is equally true of the model fits to the
data, this does not decrease confidence in model predictions. Where more uncertainty and
concern exists is with the oral bioavailability, since it is only estimated from a small data set
[4 individuals; (Schmutte et al., 1988)], with measurements only extending to 90 minutes after
ingestion. However, bioavailability can be no more than 100%, a 25% increase over the fitted
value (79%). Hence any under-prediction of human dosimetry after oral exposure should be no
greater than that factor, well within the general variability and uncertainty expected for human
dosimetry (for which the UFn of 10 is used).
B.2.8. Inhalation Route HECs and Oral Route HEDs
The atmospheric methanol concentration resulting in a human daily blood methanol AUC
(hr x mg/L) or Cmax (mg/L) equal to that occurring in experimental animals following exposure
at the POD concentration is termed the HEC. Similarly, the oral dose (rate) resulting in human
daily blood methanol AUC (hr x mg/L) equivalent to that occurring in an experimental animal at
the POD concentration is termed the HED. For humans these estimates are made using long-term
exposure patterns, after a steady-state is reached from continuous inhalation exposures, or
otherwise there is no longer a variation from day-to-day in the blood concentration profile, given
an assumed consistent exposure pattern, as indicated in Figure B-14. Internal concentration
PODs in mice were estimated by BMD analysis applied to measured (peak) blood concentrations
(Cmax values), as described in Section 5. For the rat, internal Cmax and AUC values were
estimated using the rat PBPK model as described in B.2.5 for bioassay exposures prior to BMD
analysis.
B-32
-------
To estimate the HEC for specific blood methanol Cmax and 24-hour AUC values,
continuous 1,000-hour exposures were simulated, to assure steady state was achieved, for which
the human Cmax was the steady state blood methanol concentration (Css) so predicted and the
AUC calculated from the last 24 hours of that period. (AUC = 24*CSS.). For oral exposure, the
daily ingestion pattern described in B.2.6.1 was used, simulations were again run for 1,000
hours, the Cmax selected as the maximum achieved over the resulting time-course, and the AUC
calculated over the last 24 hours. Results for selected exposure levels are given in Table B-6.
While the PBPK computational code was used to derive the HECs and HEDs used in this
assessment, using a computational script that will be described below, an alternative approach
was developed that provides an initial approximation, which also allows non-PBPK model users
to estimate methanol HECs and HEDs from BMDs in the form of Cmax (or Css) and AUC values.
This approach uses algebraic equations describing the relationship between predicted methanol
Cmax or 24-hour AUC and the inhalation exposure level (i.e., an HEC in ppm) (Equations 1 or 2
below) or oral exposure rate (i.e., an HED in mg/kg-day) (Equations 3 or 4 below). The
equations were derived by generating tables of exposure-dose values like Table B-6, but with
more entries to define the relationship, then selecting and fitting equations to interpolate among
the simulated points from that table. The resulting approximations match the exact PBPK model
results to within a few percent. To use the equations to derive an HEC or HED, the target human
Cmax or AUC is simply plugged into to the appropriate equation.
B-33
-------
Table B-6 PBPK model predicted Cr
Methanol.
(Css) and 24-hour AUC for humans exposed to
Concentration
(ppm)
1
5
10
50
100
200
500
1,000
2,000
5,000
Inhalation exposure3
AUC
(mg-hr/L)
0.65
3.27
6.56
33.5
68.7
145
437
1,380
15,400
115,000
C— p
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 exposure3
AUC (mg-hr/L)
0.21
2.14
22.2
130
320
984
15,000
80,600
216,000
625,000
Cmax
(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) = 0.554 xCss +
1734 x Css
45.73 + Css
NEC (ppm) = 0.02308 x A UC +
\734xAUC
IQ9& + AUC
Equation 1
Equation 2
//ŁZ)(mg/kg-day) = 0.1904x Cmax + 440-4xCmax Equations
109.9 + C max
419 0 x A UC
//ŁD(mg/kg-day) = 0.007257 x AUC + : Equation 4
J
In Equations 1-4 above, AUC, Css, and Cmax are above endogenous background. The endogenous
background blood concentration (Cmax or Css) was set to 1.5 mg/L, so the endogenous
background AUC = 1.5 (mg/L)x24 (hr) = 36 mg-hr/L. So to identify an HEC or HED that lead to
a total daily AUC of 50 mg-hr/L, for example, one would then plug 50 - 36 = 14 mg-hr/L into
Equation 2 or 4.
While the preceding equations approximate the PBPK model fairly well, an exact
solution is preferred if the full PBPK model can be run. For example, for Cmax = 20 mg/L (above
B-34
-------
background), Equation (1) estimates HEC = 538.6 ppm, but running the PBPK model at this
exposure level predicts a (peak) blood level of 20.2 mg/L. An exact HEC (to 4 significant
figures) is 535.6 ppm. Two .m file scripts were created as part of the acslX PBPK model
workspace for methanol, which calculate HEC and HED values through a simple search
algorithm (U.S. EPA, 2012b). These were used to generate all of the HEC and HED values
reported in Section 5 of this assessment.
B.3. Monkey PK Data and Model Analysis
In order to estimate internal doses (blood Cmax values) for the monkey health-effects
study of Burbacher et al. (1999b) and further elucidate the potential differences in methanol
pharmacokinetics between NP and pregnant individuals (2nd and 3rd trimester), a focused
reanalysis of the data of Burbacher et al. (1999a) was performed. The monkeys in this study were
exposed for 2.5 hours/day, with the methanol concentration raised to approximately the target
concentration for the first 2 hours of each exposure and the last 30 minutes providing a chamber
"wash-out" period, when the exposure chamber concentration was allowed to drop to 0. Blood
samples were taken and analyzed for methanol concentration at 30 minutes, 1, 2, 3, 4, and 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
active exposure). These data were analyzed to compare the PK in NP versus pregnant animals,
and fitted with a simple PK model to estimate blood Cmax values above background for each
exposure level. Dr. Burbacher graciously provided the original data, which were used in this
analysis.
Two cohorts of monkeys were examined, but the data (plots) did not indicate a systematic
difference between the two, so the data from the two cohorts were combined. The data from the
scatter plots of Burbacher et al. (1999a) for the NP (pre-pregnancy), first pregnancy (2nd
trimester), and second pregnancy (3rd trimester) studies are compared in Figure B-15, along with
model simulations (explained below). Since the pregnancy time points were from animals that
had been previously exposed for 87 days plus the duration of pregnancy to that time point, the
pre-exposed NP animals were used for comparison, rather than naive animals, with the
expectation that effects due to changes in enzyme expression (i.e., induction) from the
subchronic exposure would not be a distinguishing factor. Note that each exposure group
included a pre-exposure baseline or background measurement, also shown. To aid in
distinguishing the data visually, the NP data are plotted at times 5 minutes prior to the actual
blood draws and the 3rd trimester at 5 minutes after each blood draw.
B-35
-------
200 ppm
10.0 -,
-200
x
X
*r-
1
200
300
400
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: Reprinted with permission of Health Effects Institute, Boston, MA; Burbacher et al. (1999a).
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.
To analyze and integrate the PK data of Burbacher et al. (1999a), the one-compartment
model used by Burbacher et al. (1999b) and Burbacher et al. (1999a) was extended by the
addition of a chamber compartment to capture the kinetics of concentration change in the
exposure chamber, as shown in Figure B-16. The data in Figure B-16 [digitized from Figure 5 of
B-36
-------
Burbacher et al. (1999a)] show an exponential rise to and fall from the approximate target
concentration during the exposure period. The use of a single-compartment model for the
chamber allows this dynamic exposure period to be captured, so that the full concentration-time
course is used in simulating the monkey internal concentration rather than an approximate step
function (i.e., rather than assuming an instantaneous rise and fall). The pair of equations
representing the time-course in the chamber and monkey are as follows (bolded parameters are
fit to data):
Chamber: dCCh/dt = [(CcM«S - CCh) • FCh - Rinh]/VCh
Monkey: dCmk/dt = [Rinh - Vmax.Cmk/ (Km + Cmk)]/ (Vmk.BW)
with Rmh = Cch»Rc (1,000-BW)° 74»F and Cnet = Cmk + Cbg.
d: delta, change
Cch.: instantaneous chamber concentration (mg/L)
t: time (hour)
COM: chamber in-flow methanol concentration (mg/L), which was set to the concentrations corresponding to those reported in
Table 2 of Burbacher et al. (1999aX using the "Breeding" column for the NP (87 days pre-exposed; values in Table B-7)
S: exposure switch, set to 1 when exposure is on (first 2 hours) and 0 when off
Fch: chamber air-flow, 25,200 L/hr, as specified by Burbacher et al. (2004a) and Burbacher et al. (2004b)
Rinh: net rate of methanol inhalation by the monkeys (mg/hr)
Vch (1,220 L): chamber volume, initially set to 1 ,380 L ("accessible volume" stated by Burbacher et al. (2004a) and Burbacher et
al. (2004b), but allowed to vary below that value to account for volume taken by equipment, monkey, and to empirically fit
the mixing time to the observed data (Figure B-16).
Cm]j: instantaneous inhalation-induced monkey blood methanol concentration (mg/L); this is added to the measured
background/endogenous concentration before comparison to data
Vmax (32.5 mg/hr): fitted (nonscaled) Michaelis-Menten maximum elimination rate
K,,, (14.4 mg/L): fitted (nonscaled) Michaelis-Menten saturation constant
Vmk (0.623 L/kg): fitted volume of distribution for monkey
BW: monkey body weight (kg); for NP monkeys set to group average values in data of Burbacher et al. (l_999a) and Burbacher et
al. (1999b)
Rc: allometric scaling factor for total monkey respiration (0.12 L/hours/g° 74 = 2 mL/minute/g0 74), as used by Burbacher et al.
(1999a; 1999b) (note that scaling is to BW in g, not kg)
F: fractional absorption of inhaled methanol; set to 0.5 (50%), 2/3 the value fitted for humans using the human PBPK model (see
Appendix B, Section B.2); for the monkey F and Vmk cannot be uniquely identified, given the model structure; since the
monkey model uses total ventilation (defined by Rc) as the driver, while the human model uses alveolar ventilation which is
assumed to be 2/3 of total ventilation, F was set to 2/3 the human value to obtain a realistic estimate of Vmk
Cnet: net blood concentration, equal to sum of the inhalation-induced concentration (C^) and the background blood level (Cbg)
(mg/L)
Cbg: background (endogenous) methanol concentration, set to the pre-exposure group-specific mean from the data of Burbacher
et al. (1999a) and Burbacher et al. (1999b)
B-37
-------
2000 -j
* onn
itiuu -
— 1600 -
E
0-1400 -
O.
^1200 -
/ 1800 ppm
/
/
/ Chamber concentration profile
k
•g 800 -]|
| 600
o 400
200
0
600 ppm
V
\l 200 ppm
1^--*- -*-
\ \
^ * * *" v 1\
. T^HB
0.5
1 1.5
Tirne(hr)
Source: Reprinted with permission of the Health Effects Institute, Boston, MA; Burbacheret al. (|999a).
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.
The model was specifically fit to the 2nd trimester monkey data, assuming that the
parameters were the same for all the exposure groups and concentrations. While the data show
little difference between the NP and two pregnancy groups, the 2nd trimester group was
presumed to be most representative of the average internal dosimetry over the entire pregnancy.
Further, the results of Mooney and Miller (2001) show that developmental effects on the monkey
brain stem following ethanol exposure are essentially identical for monkeys exposed only during
early pregnancy versus full-term, indicating that early pregnancy is a primary window of
vulnerability.
Model simulation results are the lines shown in Figures B-15 and B-16. The model
provides a good fit to the monkey blood and chamber air concentration data. The chamber
volume was treated as a fitted parameter, decreasing the "accessible volume" of 1,380 L,
provided by Burbacher et al. (1999a) to 1,220 L, which calibrated the mixing time in the
chamber to match the chamber concentration data (Figure B-16). An adjustment of the
"accessible volume" also accounts for any volume filled by the monkey and other chamber
equipment. A detailed description of the chamber set-up is found in Burbacher et al. (1999a). The
model does an adequate job of fitting the data for all exposure groups without group-specific
parameters. In particular, the data for all exposure levels can be adequately fit using a single
value for the volume of distribution (Vmk) as well as each of the metabolic parameters. While one
may be able to show statistically distinct parameters for different groups or exposure levels (by
fitting the model separately to each), as was done by Burbacher et al. (1999a), it is unlikely that
such differences are biologically significant, given the fairly large number of data points and the
B-38
-------
large variability evident in the blood concentration data. Thus, the single set of parameters listed
with the parameter descriptions above will be used to estimate internal blood concentrations
(Cmax above background) for the dose-response analysis described in Appendix D. The chamber
concentrations for "pregnancy" exposures recorded by Burbacher et al. (1999a: Table 2) and
average body weights for each exposure group at the 2nd trimester time point were used along
with the model to calculate Cmax values above background (Table B-7).
Table B-7 Monkey group exposure characteristics for Burbacher et al. (1999a).
Exposure concentration (ppm)a Group average BW (kg)b Cmax above background (mg/L)c
0 3.93 0
206 3.46 2.87
610 408 10.38
1,822 3.83 38.51
aReprinted with the permission of the Health Effects Institute, Boston, MA; from Burbacher et al. (1999a) and Burbacher et al.
f(1999b). 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 (Cmk)].
Model simulations were also conducted to predict internal doses for the NEDO (1987)
monkey studies. Specifically, simulations were conducted for 21 h/d exposures to 10, 100, and
1,000 ppm methanol with an average animal BW of 2.2 kg. Exposures were simulated for 7 days
to assure that periodicity had been reached, and internal metrics calculated for the 7th day. Visual
inspection of simulated blood levels indicated that periodicity was in fact attained by the 2nd or
3rd day. Results are provided in Table B-8.
Table B-8 Monkey group exposure characteristics ofr NEDO (1987)."
Exposure concentration (ppm) Caverage above background (mg/L)b Cmax above background (mg/L)
10 0.09 0.11
100 0.97 111
1,000 17.9 21.5
aMetrics calculated for 2.2 kg BW animals exposed for 21 h/d, on the 7th day of simulated exposure.
bNet (total) blood concentration averaged over 24 hours minus the background level of 2 mg/L.
°Peak blood concentration minus the background level of 2 mg/L.
B-39
-------
B.4. Conclusions and Discussion
Rat and human methanol PBPK models have been developed and calibrated to data in the
open literature. EPA developed its own model because none of the existing models satisfactorily
fulfilled all of the criteria specified in Section 3.4.1.2. Further, none of the existing models had
been calibrated or tested against the larger collection of data considered for each species here. As
a result, while each model may fit the subset of the data to which it had been calibrated better
than the final model described here, without adjustment of parameters from those published,
each model either had features which made it incompatible with risk extrapolation (e.g.,
parameters which vary with dose in an unpredictable way) or had an inadequate fit to other data
considered critical for establishing overall model soundness. The EPA model simplifies the
structure used by Ward et al. (1997) in some aspects while adding specific refinements (e.g., a
standard lung compartment and a two-compartment GI tract).
Although the developmental endpoints of concern are effects, which result from in utero
and (to a lesser extent) lactational exposure, it is not necessary for a methanol PBPK model to
specifically describe pregnancy (i.e., specify a fetal/gestational/conceptus compartment) and
lactation in order for it to provide better cross-species extrapolation of risk than default methods.
Representation of the unique physiology of pregnancy and the fetus/conceptus would be
necessary if methanol pharmacokinetics differed significantly during pregnancy or if the
observed partitioning of methanol into the fetus/conceptus versus the mother showed a
concentration ratio significantly greater than or less than 1. Further details on the reasoning for
not including a pregnancy description are given in Section 3.4.1.2.
While lactational exposure is less direct than fetal exposure and blood or target-tissue
levels in the breast-feeding infant or rat pup are likely to differ more from maternal levels, the
health-effects data indicate that most of the effects of concern are due to fetal exposure, with
only a small influence due to postnatal exposures. Separating out the contribution of postnatal
exposure from prenatal exposure to a given endpoint in a way that would allow the risk to be
estimated from estimates of both exposure levels would be extremely difficult, even if one had a
lactation/child PBPK model that allowed for prediction of blood (or target-tissue) levels in the
offspring. Target tissue concentrations in the offspring would still be expected to be closely
related to maternal blood levels (which depend on ambient exposure and determine the amount
delivered through breast milk), with the relationship between maternal levels and those in the
offspring being similar across species.
Therefore, the development of a lactation/child PBPK model appears not to be supported,
given the minimal change that is likely to result in risk extrapolations and use of (NP) maternal
blood levels as a measure of risk in the offspring is still considered preferable over use of default
extrapolation methods. In particular, the existing human data allow for accurate predictions of
B-40
-------
maternal blood levels, which depend strongly on the rate of maternal methanol clearance. Failing
to use the existing data (via PBPK modeling) for human methanol clearance (versus that in other
species) would be to ignore this very important determinant of exposure to breast-fed infants.
And since bottle-fed infants do not receive methanol from their mothers, they are expected to
have lower or, at most, similar overall exposures for a given ambient concentration than the
breast-fed infant, so that use of maternal blood levels for risk estimation should also be
adequately protective for that group.
During model development, several inconsistencies between experimental blood
methanol kinetic data embedded in the Ward et al. (1995) model and the published figures first
reporting these data were discovered. Therefore, data were digitized from the published literature
when a figure was available, and the digitized data was compared to the provided data. When the
digitized data and the data embedded in the computational files (i.e., provided to Battelle under
contract from EPA) were within 3% of each other, the provided data was used; when the
difference was greater than 3%, the digitized data was used. Often, using the published figures as
a data source resulted in substantial improvements of the fit to the data in the cases where the
published figures were different from the embedded data.
The final methanol PBPK model fits inhalation-route blood kinetic data from separate
laboratories in rats and humans fairly well. The low-dose exposures of all routes were considered
the most important for model optimization, since these doses are most relevant to a health
assessment.
Figure B-17 illustrates the changes in blood methanol concentrations predicted by the
human PBPK model for exposures to either the RfC or RfD alone, or a combined exposure to
both the RfC and RfD, with oral exposure assumed divided into six daily boluses as described
previously (Section B.2.7) and inhalation exposure assumed to be continuous. The predictions
are shown for an individual starting with an average background level of 1.5 mg/L blood
methanol, relative to one standard deviation of the background (grey area).
B-41
-------
2.4-
2.1-
1.8-
-.5 1.5
c
(V
o
c
o
o
TJ
O
_O
CO
1.2-
0.9-
0.6-
0.3-
0
RfC only
RfD only
RfC & RfD
mean background
background +/- 1 SD
12 24 36 48
Time (hr)
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 hrto 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.
B-42
-------
APPENDIX C. HUMAN CASE STUDIES
An extensive library of case reports has documented the consequences of acute
accidental/intentional methanol poisoning. Nearly all have involved ingestion, but a few have
involved percutaneous and/or inhalation exposure. As discussed in Section 4.1.1, the CNS
damage seen in acute overdose exposures is most likely from acidosis and not from methanol per
se. As many of the case reports demonstrate, the association of Parkinson-like symptoms with
methanol poisoning is related to the observation that lesions in the putamen are a common
feature both in Parkinson's disease and methanol overexposure. A brief discussion of the terms
cited in case report literature follows.
Basal ganglia, a group of interconnected subcortical nuclei in each cerebral hemisphere,
refers to various structures in the grey matter of the brain that are intimately involved, for
example, in coordinating motor function, maintaining ocular and respiratory function, and
consciousness. The connectivity within the basal ganglia involves both excitatory and inhibitory
neurotransmitters such as dopamine (associated with Parkinson's disease when production is
deficient).
The structures comprising the basal ganglia include but are not limited to: the putamen
and the globus pallidus (together termed the lentiform nuclei), the pontine tegmentum, and the
caudate nuclei. Dystonia or involuntary muscle contraction can result from lesions in the
putamina; if there are concomitant lesions in the globus pallidus, Parkinsonism can result (Bhatia
and Marsden, 1994). Bhatia and Marsden (1994) have discussed the various behavioral and
motor consequences of focal lesions of the basal ganglia from 240 case-study reports. Lesions in
the subcortical white matter adjacent to the basal ganglia often occur as well (Airas et al., 2008;
Rubinstein et al., 1995; Bhatia and Marsden, 1994). In the case reports of Patankar et al. (1999),
it was noted that the severity and extent of necrosis in the lentiform nuclei do not necessarily
correlate with clinical outcome.
In one of the earliest reviews of methanol overexposure, Bennett et al. (1953) described a
mass accidental poisoning when 323 persons, ranging in age from 10 to 78 years, in Atlanta,
Georgia, consumed "whisky" adulterated with as much as 35-40% methanol. In all, 41 people
died. Of the 323 individuals, 115 were determined to be acidotic with symptoms (visual
impairment, headache [affecting ~62%], dizziness [affecting ~30%], nausea, abdominal pain and
others) beginning around 24 hours post exposure. Visual impairment was mostly characterized
by blurred or indistinct vision; some who were not acidotic experienced transient visual
disturbances. The cardiovascular parameters were unremarkable. The importance of acidosis to
outcome is shown in Table C-l. Among the key pathological features were cerebral edema, lung
C-1
-------
congestion, gastritis, pancreatic necrosis, fatty liver, epicardial hemorrhages, and congestion of
abdominal viscera.
Table C-l Mortality rate for subjects exposed to methanol-tainted whisky in relation to
their level of acidosis.
Subjects3
All patients
Acidotic (CO2 <20 mEq)
Severely acidotic (CC>2 <10 mEq)
Number
323
115
30
Percent deaths
6.2
19
50
aThese data do not include those who died outside the hospital or who were moribund on arrival.
Source: Reprinted with permission of Lippincott, Williams & Wilkins; Bennett et al. (1953).
Riegel and Wolf (1966), in a case report involving a 60-year-old woman who ingested
methanol, noted that nausea and dizziness occurred within 30 minutes of ingestion. She
subsequently passed out and remained unconscious for 3 days. Upon awakening she had
paralysis of the vocal cords and was clinically blind in one eye after 4 months. Some aspects of
Parkinson-like symptoms were evident. There was a pronounced hypokinesia with a mask-like
face resembling a severe state of Parkinson's disease. The patient had difficulty walking and
could only make right turns with difficulty. There was no memory loss.
Treatment of a 13-year-old girl who ingested an unspecified amount of a windshield-
washer solution containing 60% methanol was described by Guggenheim et al. (1971). She
displayed profound acidosis; her vital signs, once she was treated for acidosis, were normal by
36 hours after hospital admission. During the ensuing 6 months after discharge from the hospital,
visual acuity (20/400, both eyes) worsened, and she experienced muscle tremors, arm pain, and
difficulty in walking. A regimen of levodopa treatment greatly improved her ability to function
normally.
Ley and Gali (1983) also noted symptoms that are Parkinson like following methanol
intoxication. In this case report respiratory support was needed; the woman was in a coma. Once
stabilized, she exhibited symptoms similar to those noted in other case study reports, such as
blurred vision, movement difficulty, and tremors. Computerized Axial Tomography scan findings
highlighted the central nervous system (CNS) as an important site for methanol poisoning.
Rubinstein et al. (1995) presented evidence that a methanol blood level of 360 mg/L is
associated with a suite of CNS and ocular deficits that led to a 36-year-old man (who
subsequently died) becoming comatose. CT scans at 1-2 days following ingestion were normal.
However, MRI scans at day 4 revealed lesions in the putamen and peripheral white matter of the
cerebral and cerebellar hemispheres. Bilateral cerebellar cortical lesions had been reported in an
earlier case of methanol poisoning by Chen et al. (1991).
C-2
-------
Finkelstein and Vardi (2002) reported that long-term inhalation exposure of a woman
scientist to methanol without acute intoxication resulted in a suite of delayed neurotoxic
symptoms (e.g., hand tremor, dystonia, bradykinesia, and other decrements in body movement).
Despite treatment with levodopa, an increase in the frequency and severity of effects occurred.
Exposure to bromine fumes was concomitant with exposure to methanol.
Hantson et al. (1997b) found, in four cases, that MRI and brain CT scans were important
tools in revealing specific brain lesions (e.g., in the putamina and white matter). The first subject
was a 57-year-old woman who complained of blurred vision, diplopia, and weakness 24 hours
after ingesting 250 mL of a methanolic antifreeze solution. Upon hospital admission she was
comatose and in severe metabolic acidosis. An MRI scan at 9 days indicated abnormal
hyperintense foci in the putamina (decreased in size by day 23) and subtle lesions (no change by
day 23) in the white matter. Upon her discharge, bilateral deficits in visual acuity and color
discrimination persisted.
Similar deficits (metabolic acidosis, visual acuity, and color discrimination) were seen in
a man who ingested 300 mL of 75% methanol solution. His blood methanol level
wasl,630 mg/L. An MRI administered 24 hours after hospital admission revealed abnormal
hyperintense foci in the putamina, with less intense lesions in the white matter. Like the first
subject, a subsequent MRI indicated the foci decreased in size over time, but visual impairments
persisted.
The third individual, a male, ingested an unspecified amount of a methanolic solution.
His blood methanol level was 12,900 mg/L, and he was in a coma upon hospital admission. An
MRI revealed lesions in the putamina and occipital subcortical white matter. A follow-up CT
scan was performed after 1 year and showed regression of the putaminal lesions but no change in
the occipital lesions. Upon his discharge, severe visual impairment remained but no
extrapyramidal signs were observed.
The last case was a man who became comatose 12 hours after ingesting 100 mL
methanol. His blood methanol level at that time was 600 mg/L. An MRI revealed lesions in the
putamina; at 3 weeks these lesions were observed to have decreased in size. Upon his discharge,
the neurological signs had improved but optic neuropathy (in visual evoked potential) was
observed.
In a separate publication, Hantson et al. (1997a) reported a case of a 26-year-old woman
who had ingested 250-500 mL methanol during the 38th week of pregnancy. Her initial blood
methanol level was 2,300 mg/L (formate was 336 mg/L), yet only a mild metabolic acidosis was
indicated. No distress to the fetus was observed upon gynecologic examination. Six days after
therapy was initiated (methanol was not present in blood), she gave birth. No further
complications with either the mother or newborn were noted.
C-3
-------
There have been several case reports involving infant or toddler exposures to methanol
(De Brabander et al.. 2005: Wuetal.. 1995: Brent etal.. 1991: Kahn and Blum. 1979). The
report by Wu et al. (1995) involved a 5-week-old infant with moderate metabolic acidosis and a
serum methanol level of 11,480 mg/L, a level that is ordinarily fatal. However, this infant
exhibited no toxic signs and survived without any apparent permanent problems. De Brabander
et al. (2005) reported the case of a 3-year-old boy who ingested an unknown amount of pure
methanol; at 3 hours after ingestion, the blood methanol level was almost 300 mg/L. Ethanol
infusion as a therapeutic measure was not well tolerated; at 8 hours after ingestion, fomepizole
(4-methylpyrazole) was administered to inhibit the metabolism of methanol by ADH1, and blood
methanol levels stabilized below 200 mg/L, a level above which is considered to be toxic by the
American Academy of Clinical Toxicology (Barceloux et al., 2002). Neither metabolic acidosis
nor visual impairment was observed in this individual. Hantson et al. (1997b), in their review,
touted the efficacy of fomepizole over ethanol in the treatment of methanol poisoning
Bilateral putaminal lesions, suggestive of nonhemorrhagic necrosis in the brain of a man
who accidentally ingested methanol, were reported by Arora et al. (2005). Approximately
10 hours after MRI examination, he developed blurred vision and motor dysfunction. After
5 months, visual deficits persisted along with extrapyramidal symptoms. Persistent visual
dysfunction was also reported in another methanol poisoning case (Arora et al., 2007); the vision
problems developed -46 hours subsequent to the incident.
Vara-Castrodeza et al. (2007) applied diffusion-weighted MRI on a methanol-induced
comatose woman. Diffusion-weighted MRI provides an image contrast distinct from standard
imaging in that contrast is dependent on the molecular motion of water (Schaefer et al., 2000).
The neuroradiological findings were suggestive of bilateral putaminal hemorrhagic necrosis,
cerebral and intraventricular hemorrhage, diffuse cerebral edema, and cerebellar necrosis.
Diffusion-weighted MRI allows for differentiation of restricted diffusion which is indicative of
nonviable tissue. In this case, treatment for acidosis (blood methanol levels had risen to
1,000 mg/L) was unsuccessful and the patient died.
Emergency treatment was unable to save the life of a 38-year-old man who presented
with abdominal pain and convulsions after methanol intoxication (Henderson and Brubacher,
2002). A review of a head CT scan performed before the individual went into respiratory arrest
revealed bilateral globus pallidus ischemia.
Discrete lesions of the putamen, cerebral white matter, and corpus callosum were
observed upon MRI (8 days post ingestion) in a man exposed to methanol (blood level
370 mg/L) complaining of vision loss (Keles et al., 2007). Standard treatments corrected the
acidosis (pH 6.8), and at 1-month follow-up, his cognitive function improved but blindness and
bilateral optic atrophy were described as permanent. The follow-up MRI showed persistent
putaminal lesions with cortical involvement.
C-4
-------
Fontenot and Pelak (2002) described a case of a woman who presented with persistent
blurred vision and a worsening mental status 36 hours after ingestion of an unspecified amount
of methanol. The initial CT scan revealed mild cerebral edema. The blood methanol level at this
time was 860 mg/L. A repeat CT scan 48 hours after presentation showed hypodensities in the
putamen and peripheral white matter. One month after discharge, cognitive function improved,
and the patient experienced only a mild lower-extremity tremor.
Putaminal necrosis and edema of the deep white matter (the corpus callosum was not
affected) was found upon MRI examination of a 50-year-old woman who apparently ingested an
unknown amount of what was believed to be pure laboratory methanol (Kuteifan et al., 1998).
Her blood methanol level was 1,272 mg/L upon hospital admission and dropped to 1,020 mg/L at
10 hours and to 710 mg/L at 34 hours. The woman, a chronic alcoholic, was in a vegetative state
when found and did not improve over the course of a year.
MRI and CT scans performed on a 51-year-old man with generalized seizures who had a
blood methanol level of 3,044 mg/L revealed bilateral hemorrhagic necrosis of the putamen and
caudate nuclei (Gaul et al., 1995). In addition, there was extensive subcortical necrosis and
bilateral necrosis of the pontine tegmentum and optic nerve. The patient died several hours after
the scans were performed.
The relation of methanol overexposure to brain hemorrhage was a focus of the report by
Phang et al. (1988), which followed the treatment of 7 individuals, 5 of whom died within
72 hours after hospital admission. In two of the deceased individuals, CT scans and autopsy
revealed putaminal hemorrhagic necrosis. The investigators postulated that the association of
methanol with hemorrhagic necrosis may be complicated by the use of heparin during
hemodialysis treatment for acidosis
Treatment of two men who had drunk a solution containing 58% methanol and presented
with impaired vision, coma, and seizures was discussed in a case report by Bessell-Browne and
Bynevelt (2007). A CT scan, on one individual, revealed bilateral putaminal and cerebral lesions.
Blood methanol levels were 21 mg/L. This individual, despite standard treatments, never
regained consciousness. The second individual, upon MRI, showed scattered hemorrhage at the
grey-white interface of the cerebral hemispheres.
There have been case reports that involved percutaneous and inhalation exposure (Adanir
et al., 2005; Downieetal., 1992). Use of a methanol-containing emollient by a woman with
chronic pain led to vision loss, hyperventilation and finally, coma (Adanir et al., 2005).
Subsequent to standard treatment followed by hospital discharge, some visual impairment and
CNS decrements remained. The methanol blood threshold for ocular damage and acidosis
appeared to be -20 mg/L. Dutkiewicz et al. (1980) have determined the skin absorption rate to be
0.192 mg/cm2/minute. In the case report of Aufderheide et al. (1993), two firefighters were
C-5
-------
transiently exposed to methanol by inhalation and the percutaneous route. Both only complained
of a mild headache and had blood methanol levels of 230 and 160 mg/L, respectively.
Bebarta et al. (2006) conducted a prospective observational study of seven men who had
purposefully inhaled a methanol-containing product. Four had a blood methanol level upon
hospital presentation of >240 mg/L; the mean formic acid level was .71 mg/L. One individual
had a blood methanol level of 860 mg/L and a blood formic acid level of 250 mg/L upon hospital
admission. This latter individual was treated with fomepizole. No patient had an abnormal
ophthalmologic examination. All seven stabilized quickly and acidosis was normalized in 4
hours.
Numerous other case reports documenting putaminal necrosis/hemorrhage and/or
blindness have been reported (Blanco et al., 2006; Feany etal., 2001; Hsu et al., 1997; Pelletier
etal.. 1992: Chen etal.. 1991).
Hovda et al. (2005) presented a combined prospective and retrospective case series study
of 51 individuals in Norway (39 males and 12 females, many of whom were alcoholics) who
were hospitalized after consuming tainted spirits containing 20% methanol and 80% ethanol. In
general, serum methanol concentrations were highest among those most severely affected. The
poor outcome was closely correlated with the degree of metabolic acidosis. It was noted by the
investigators that the concomitant consumption of ethanol prevented more serious sequelae in
2/5 individuals who presented with detectable ethanol levels and were not acidotic despite 2
having the highest blood methanol levels. However, others with detectable levels of ethanol
along with severe metabolic acidosis (two of whom died) presumably had subtherapeutic levels
of ethanol in their system.
In a later report, Hovda et al. (2007) focused on formate kinetics in a 63-year-old male
who died 6 days after being admitted to the hospital with headache, vomiting, reduced vision,
and dizziness. The investigators speculated that the prolonged metabolic acidosis observed (T1/2
for formic acid was 77 hours before dialysis, compared to a typical normal range of
2.5-12 hours) may have been related to retarded formate elimination.
Hovda and colleagues (Hunderi et al., 2006) found a strong correlation between blood
methanol concentration and the osmolal gap (R2= 0.92) among 17 patients undergoing dialysis
after consuming methanol-contaminated spirits. They concluded that the osmolal gap could be
taken as a priori indication of methanol poisoning and be used to guide initiation and duration of
dialysis. As they indicated, many hours of dialysis could be safely dispensed with. The osmolal
gap pertains to the effect that methanol (and other alcohols) has on the depression of the freezing
point of blood in the presence of normal solutes. Braden et al. (1993) demonstrated in case
studies that the disappearance of the osmolal gap correlates with the correction of acidosis; they
cautioned that methanol and ethanol should not be assumed to be the main factors in causing
C-6
-------
osmolal gap as glycerol and acetone and its metabolites can as well. A more detailed discussion
of the anion and osmolal gap has been provided by Henderson and Brubacher (2002).
Hassanian-Moghaddam et al. (2007) compiled data on the prognostic factor relating to
outcome in methanol-poisoning cases in Iran. They examined 25 patients, 12 of whom died; 3 of
the survivors were rendered blind. There was a significant difference in mean pH of the first
arterial blood gas measurements of those who subsequently died compared with survivors. It was
concluded that poor prognosis was associated with pH <7, coma upon admission, and >24-hours
delay from intake to admission.
The use of blood methanol levels as predictors of outcome is generally not recommended
(Barceloux et al., 2002). These investigators cited differences in sampling time, ingestion of
ethanol, and levels of toxic (e.g., formic acid) metabolites among the complicating factors. As an
illustration, the case report by Prabhakaran et al. (1993) cites two women who ingested a
methanol solution (photocopying diluent) at about the same time, were admitted to the hospital
about the same time (25-26 hours after ingestion) and had identical plasma methanol
concentrations (830 mg/L) upon admission, but different outcomes. Patient #1 was in metabolic
acidosis and had an unstable conscious state even after treatment. Upon discharge at day 6, there
were no apparent sequelae. Patient #2 had severe metabolic acidosis, fixed and dilated pupils,
and no brain stem reflexes. This patient died at day 3 even though therapeutic measures had been
administered.
In a discussion of 3 fatal methanol-overexposure cases, Andresen et al. (2008) found
antemortem blood methanol levels of 5,400 and 7,400 mg/L in two individuals. At autopsy brain
stem blood levels were 7,380 and 10,080 mg/L, respectively. These brain levels were much
higher than blood levels postmortem. Autopsy revealed brain and pulmonary edema in all three
individuals; in the two who had the longer survival times, there was hemorrhagic necrosis of the
putamen and hemorrhages of the tissue surrounding the optic nerve. In their study of 26 chronic
users of methylated spirits, Meyer et al. (2000) found that the best predictor of death or a poor
outcome in chronic abusers was a pH <7.0; there was no correlation between blood methanol
levels and outcome. Mahieu et al. (1989) considered a latency period before treatment exceeding
10 hours and a blood formate level >500 mg/L as predictive of possible permanent sequelae. Liu
et al. (1998) in their examination of medical records of 50 patients treated for methanol
poisoning over a 10-year period found that: (1) deceased patients had a higher mean blood
methanol level than survivors; and (2) initial arterial pH levels <7.0 (i.e., severe metabolic
acidosis). Coma or seizure was also associated with higher mortality upon hospital admission.
Numerous cases of methanol poisoning have been documented in a variety of countries.
In Tunisia, 16 cases of methanol poisoning were discussed by Brahmi et al. (2007). Irreversible
blindness occurred in two individuals, with others reporting CNS symptoms, GI effects, visual
disturbances, and acidosis. Putaminal necrosis was also described in case reports from Iran
C-7
-------
(Sefidbakht et al., 2007). Of 634 forensic autopsies carried out in Turkey during 1992-2003,
18 deaths appeared to be related to methanol poisoning (Azmak, 2006). Brain edema and focal
necrosis of the optic nerve were among various sequelae noted. Dethlefs and colleagues (Naraqi
et al., 1979; Dethlefs and Naraqi, 1978) described permanent ocular damage in 8/24 males who
ingested methanol in Papua New Guinea.
In summary, most cases of accidental/intentional methanol poisoning reveal a common
set of symptoms, many of which are likely to be presented upon hospital admission. See Section
4.1.1 for a list of common symptoms.
C-8
-------
APPENDIX D. RFC DERIVATION OPTIONS
D.1. Benchmark Dose Modeling Summary
This appendix provides technical detail on dose-response evaluation and determination of
points of departure (POD) for relevant lexicological endpoints. The endpoints were modeled
using the U.S. EPA's Benchmark Dose Software (BMDS, version 2.2). Sections D.I.I and D.I.2
describe the common practices used in evaluating the model fit and selecting the appropriate
model for determining the POD, as outlined in the Benchmark Dose Technical Guidance
Document (U.S. EPA. 2012a).
D.1.1. Evaluation of Model Fit
For the nested dichotomous endpoint (cervical rib), BMDS nested dichotomous models
were fitted to the data using the maximum likelihood method. Each model was tested for
goodness-of-fit using a chi-square goodness-of-fit test (% p-va\ue < 0.10 indicates lack of fit).
Other factors were also used to assess model fit, such as scaled residuals, visual fit, and adequacy
of fit in the low-dose region and in the vicinity of the BMR.
For each continuous endpoint (brain weight and VDR), BMDS continuous models1 were
fitted to the data using the maximum likelihood method. Model fit was assessed by a series of
tests as follows. For each model, first the homogeneity of the variances was tested using a
likelihood ratio test (BMDS Test 2). If Test 2 was not rejected (%2/'-value > 0.10), the model was
fitted to the data assuming constant variance. If Test 2 was rejected (% p-va\ue < 0.10), the
variance was modeled as a power function of the mean, and the variance model was tested for
adequacy of fit using a likelihood ratio test (BMDS Test 3). For fitting models using either
constant variance or modeled variance, models for the mean response were tested for adequacy
of fit using a likelihood ratio test (BMDS Test 4, with % p-va\ue < 0.10 indicating inadequate
fit). Other factors were also used to assess the model fit, such as scaled residuals, visual fit, and
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.
D-1
-------
D.1.2. Model Selection
For each endpoint, the BMDL estimate (95% lower confidence limit on the BMD, as
estimated by the profile likelihood method) and AIC value were used to select a best-fit model
from among the models exhibiting adequate fit. If the BMDL estimates were "sufficiently close,"
that is, differed by at most 3-fold, the model selected was the one that yielded the lowest AIC
value. If the BMDL estimates were not sufficiently close, the lowest BMDL was selected as the
POD.
D.2. RfC Derivations Using the NEDO Methanol Report (NEDO, 1987)
In the application of the BMD approach, continuous models in EPA's BMDS, version 2.2
(U.S. EPA, 20lib), were fit to data sets for decreased brain weight in male rats exposed
throughout gestation and the postnatal period to 6 weeks and male rats exposed during gestation
on days 7-17 only (NEDO, 1987). Although there remains uncertainty surrounding the
identification of the proximate teratogen of importance (methanol, formaldehyde, or formate),
the dose metrics chosen for the derivation of RfCs were based on blood methanol levels. This
decision was primarily based on evidence that the toxic moiety is not likely to be the formate
metabolite of methanol (NTP-CERHR, 2004) and evidence that levels of the formaldehyde
metabolite following methanol maternal and/or neonate exposure would be much lower in the
fetus and neonate than in adults. While recent in vitro evidence indicates that formaldehyde is
more embryotoxic than methanol and formate, the high reactivity of formaldehyde would
significantly limit its transport from maternal to fetal blood, and the capacity for the metabolism
of methanol 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
As discussed in Section 5.1.2.1, brain weight is susceptible to both the level and duration
of exposure suggesting that a dose metric that incorporates a time component would be the most
appropriate metric to use. For these reasons and because it is more typically used in internal-
dose-based assessments and better reflects total exposure within a given day, daily AUC
(measured for 22-hour exposure/day) was chosen as the most appropriate dose metric for
modeling the effects of methanol exposure on brain weights in rats exposed throughout gestation
and continuing into the FI generation.
As is discussed in Section 5.1.3.2.2, the additional routes of exposure to the pups in this
study (lactation and inhalation) present uncertainties in that the average blood levels in pups is
D-2
-------
likely to be greater than those of their dams. The assumption made in this assessment is that, if
such differences exist between human mothers and their offspring, they are not significantly
greater than that which has been postulated for rats. Assuming this is true, the PBPK model-
estimated adult blood methanol level is considered to be an appropriate dose metric for the
purpose of this analysis and the estimation of a human equivalent concentration (HEC).
The first step in the current analysis is to convert the inhalation doses, given as ppm
values from the studies, to an internal dose surrogate or dose metric using the EPA PBPK model
(see Appendix B). Predicted AUC values for methanol in the blood of rats are summarized in
Table D-l. The AUC values above background (AUC - control) are then used as the dose metric
for the BMD analysis of response data shown in Table D-l for decreased brain weight at 6 weeks
in male rats following gestational and postnatal exposure.2 Decreases in brain weight at 6 weeks
(gestational and postnatal exposure), rather than those seen at 3 and 8 weeks, were chosen as the
basis for the RfC derivation because they resulted in lower estimated BMDs and BMDLs. The
details of this analysis are reported below. More details concerning the PBPK modeling were
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 weeks'3
1.78 ±0.07
1.74 ±0.09
1.69±0.06C
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 FI generation. Values are means ± SDSD
°p < 0.01,d p < 0.001, as calculated by the authors.
Data from NEDO (1987)
The EPA BMD technical guidance (U.S. EPA, 2012a) suggests that in the absence of
knowledge as to what level of response to consider adverse, a change in the mean equal to
1 control SDSD from the control mean can be used as a BMR for continuous endpoints.
However, it has been suggested that other BMRs, such as 5% change relative to estimated
control mean, are also appropriate when performing BMD analyses on fetal weight change as a
developmental endpoint (Kavlock et al., 1995). Therefore, in this assessment, both a 1 control
mean SD change and a 5% change relative to estimated control mean were considered. All
2A11 BMD assessments in this review were performed using BMDS version 2.2 (U.S. EPA. 2011^)
D-3
-------
models were fit using restrictions and option settings suggested in the EPABMD Technical
Guidance Document (U.S. EPA. 2012a).
D.2.1.1. BMD Approach with a BMR of 1 Control Mean SD - Decreased
Brain Weight in Male Rats Exposed throughout Gestation and into the
Postnatal Period (NEDO, 1987)
A summary of the results most relevant to the development of a POD using the BMD
approach (BMD, BMDL, and model fit statistics) for decreased brain weight at 6 weeks in male
rats exposed to methanol throughout gestation and continuing into the FI generation, with a
BMR of 1 control mean SD (NEDO. 1987). is provided in Table D-2. Model fit and was
determined by statistics (AIC and $ residuals of individual dose groups) and visual inspection,
as recommended by EPA (U.S. EPA, 2012a). There is a 5.1-fold range of BMDL estimates from
adequately fitting models, indicating considerable model dependence. In addition, the fit of the
Hill and more complex Exponential models are better than the other models in the dose region of
interest as indicated by a lower scaled residual at the dose group closest to the BMD (0.18 and
0.16 versus -1.4) and by visual inspection. In accordance with EPABMD Technical Guidance
(U.S. EPA. 2012a). the BMDL from the Hill model (bolded), is selected as the most appropriate
basis for an RfC derivation because it results in the lowest BMDL from among a broad range of
BMDLs and provides a superior fit in the low dose region nearest the BMD. The detailed results
of the Hill model run, including text and a plot (Figure D-l) are shown after Table D-2. The
BMDLiso was determined to be 858 mg-hr/L using the 95% lower confidence limit of the dose-
response curve expressed in terms of the AUC above background for methanol in blood.
D-4
-------
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
5,469.53
5,469.53
5,469.53
5,469.53
1.730.35
5,159.24
5,159.24
1,802.01
1,802.01
BMDLiso (AUC,
mg-hr/L)a
4,410.68
4,410.68
4,410.68
4,410.68
858.04
4,118.16
4,118.16
997.71
997.71
p-value
0.1385
0.1385
0.1385
0.1385
0.5920
0.1573
0.1573
0.5513
0.5513
AICb
-201.13
-201.13
-201.13
-201.13
-202.79
-201.38
-201.38
-202.72
-202.72
Scaled residual0
-1.39
-1.39
-1.39
-1.39
0.179
-1.336
-1.336
0.163
0.163
"The BMDL is the 95% lower confidence limit on the AUC estimated to decrease brain weight by 1 control mean SD using
BMDS 2.2 (U.S. EPA. 2011b) and model options and restrictions suggested by EPA BMD technical guidance (U.S. 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).
°chi-squared (X2) 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 SD 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.
Data from NEDO (1987).
D-5
-------
Hill Model. (Version: 2.16; Date: 04/06/2011)
Input Data File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
ConstantVariance-BMRlStd-Restrict.(d)
Gnuplot Plotting File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
ConstantVariance-BMRlStd-Restrict.pit
Tue Mar 27 08:42:04 2012
BMDS Model Run
The form of the response function is:
Y[dose] = intercept + v*dose/xn/(k^n + dose^n)
Dependent variable = Mean
Independent variable = Dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.00539333
rho = 0 Specified
intercept = 1.78
v = -0.26
n = 0.698151
k = 5889.18
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -rho -n have been estimated at a boundary
point, or have been specified by the user, and do not appear in the
correlation matrix )
alpha intercept v k
alpha 1 1.7e-008 2.5e-008 -4e-008
intercept 1.7e-008 1 0.24 -0.62
v 2.5e-008 0.24 1 -0.85
k -4e-008 -0.62 -0.85 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
alpha 0.00498218 0.00100655 0.00300938 0.00695499
intercept 1.77449 0.0177456 1.73971 1.80927
D-6
-------
v -0.3555 0.0666435 -0.486119 -0.224881
n 1 NA
k 6984.58 4505.13 -1845.31 15814.5
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Table of Data and Estimated Values of Interest
Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Res.
0 12 1.78 1.77 0.07 0.0706 0.27
547 12 1.74 1.75 0.09 0.0706 -0.425
2308 11 1.69 1.69 0.06 0.0706 0.179
1.753e+004 14 1.52 1.52 0.07 0.0706 -0.0151
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma/x2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)^2
Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma/x2
Model A3 uses any fixed variance parameters that
were specified by the user
Model R: Yi = Mu + e(i)
Var{e(i)} = Sigma/x2
Likelihoods of Interest
Model Log(likelihood) # Param's AIC
Al 105.539862 5 -201.079724
A2 106.570724 8 -197.141449
A3 105.539862 5 -201.079724
fitted 105.396232 4 -202.792465
R 77.428662 2 -150.857324
Explanation of Tests
Test 1: Do responses and/or variances differ among Dose levels?
(A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2)
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
(Note: When rho=0 the results of Test 3 and Test 2 will be the same.)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 58.2841 6 <.0001
Test 2 2.06173 3 0.5597
D-7
-------
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
Hill Model with 0.95 Confidence Level
1.85
1.75
1.7
1.65
1.6
1.55
1.5
1.45
Hill
\
BMDL
BMD
0
2000 4000 6000 8000 10000 12000 14000 16000 18000
dose
10:01 11/10 2011
Data points obtained from NEDO (1987).
Figure D-l Hill model, BMR of 1 Control Mean SD - Decreased Brain weight in male rats
at 6 weeks age versus AUC above background, Fl Generation inhalational
study.
D-8
-------
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).
A summary of the results most relevant to the development of a POD using the BMD
approach (BMD, BMDL, and model fit statistics) for decreased brain weight at 6 weeks in male
rats exposed to methanol throughout gestation and continuing into the FI generation, with a
BMR of 0.05 change relative to estimated control mean, is provided in Table D-3. Model fit was
determined by statistics (AIC and $ residuals of individual dose groups) and visual inspection,
as recommended by the EPA BMD Technical Guidance (U.S. EPA. 2012a). There is a 4.7-fold
range of BMDL estimates from adequately fitting models, indicating considerable model
dependence. In addition, the fit of the Hill and more complex Exponential models are better than
the other models in the dose region of interest as indicated by a lower scaled residual at the dose
group closest to the BMD (0.18 and 0.16 versus -1.4) and visual inspection. In accordance with
EPA BMD Technical Guidance (U.S. EPA. 2012a). the BMDL from the Hill model (bolded), is
selected as the most appropriate basis for an RfC derivation because it results in the lowest
BMDL from among a broad range of BMDLs and provides a superior fit in the low dose region
nearest the BMD. Output from the Hill model, including text and plot (Figure D-2), is shown
after Table D-3. The BMDLos was determined to be 1,183 mg-hr/L, using the 95% lower
confidence limit of the dose-response curve expressed in terms of the AUC above background for
methanol in blood.
D-9
-------
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
Linear13
2nd degree Polynomial
3rd degree Polynomial
Power
Hill
Exponential 2
Exponential 3
Exponential 4
Exponential 5
BMD05 (AUC,
mg-hr/L)a
6,
6,
6,
6,
2,
6
6
2,
2,
537.04
537.04
537.04
537.04
322.94
,212.5
,212.5
367.26
367.26
BMDUs (AUC,
mg-hr/L)a
5,
5,
5,
5,
1
5,
5,
1,
1,
,614,
,614,
,614,
,614,
,182
,270,
,270,
334,
,334,
.56
.56
.56
.56
.99
.18
.18
.02
.02
p-value
0,
0,
0,
0,
0,
0,
0,
0,
0,
.1385
.1385
.1385
.1385
.5920
.1573
.1573
.5513
.5513
"The BMDL is the 95% lower confidence limit on the AUC estimated to decrease brain weight
201 1 b) and model options and restrictions suggested by EPA BMD Technical Guidance (U.S.
AICb Scaled Residual0
-201.
-201.
-201.
-201.
13
13
13
13
-202.79
-201.
-201.
-202.
-202.
38
38
,72
,72
by 5% using BMDS
EPA. 201 2a).
-1
-1
-1
-1
0.
-1
-1
0.
0.
.39
.39
.39
.39
179
.34
.34
163
163
2.2 (U.S. EPA,
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).
°X2d 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 SD 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.
Data from NEDO (1987)
D-10
-------
Hill Model. (Version: 2.16; Date: 04/06/2011)
Input Data File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
ConstantVariance-BMROS-Restrict.(d)
Gnuplot Plotting File: C:/USEPA/BMDS220/Data/Methanol/hil_NEDOrat-6wk-male_Hil-
ConstantVariance-BMROS-Restrict.pit
Tue Mar 27 10:57:37 2012
BMDS Model Run
The form of the response function is:
Y[dose] = intercept + v*dose/xn/(k^n + dose^n)
Dependent variable = Mean
Independent variable = Dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.00539333
rho = 0 Specified
intercept = 1.78
v = -0.26
n = 0.698151
k = 5889.18
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -rho -n have been estimated at a boundary
point, or have been specified by the user, and do not appear in the
correlation matrix )
alpha intercept v k
alpha 1 1.7e-008 2.5e-008 -4e-008
intercept 1.7e-008 1 0.24 -0.62
v 2.5e-008 0.24 1 -0.85
k -4e-008 -0.62 -0.85 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
alpha 0.00498218 0.00100655 0.00300938 0.00695499
intercept 1.77449 0.0177456 1.73971 1.80927
D-11
-------
v -0.3555 0.0666435 -0.486119 -0.224881
n 1 NA
k 6984.58 4505.13 -1845.31 15814.5
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Table of Data and Estimated Values of Interest
Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Res.
0 12 1.78 1.77 0.07 0.0706 0.27
547 12 1.74 1.75 0.09 0.0706 -0.425
2308 11 1.69 1.69 0.06 0.0706 0.179
1.753e+004 14 1.52 1.52 0.07 0.0706 -0.0151
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma/x2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)^2
Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma/x2
Model A3 uses any fixed variance parameters that
were specified by the user
Model R: Yi = Mu + e(i)
Var{e(i)} = Sigma/x2
Likelihoods of Interest
Model Log(likelihood) # Param's AIC
Al 105.539862 5 -201.079724
A2 106.570724 8 -197.141449
A3 105.539862 5 -201.079724
fitted 105.396232 4 -202.792465
R 77.428662 2 -150.857324
Explanation of Tests
Test 1: Do responses and/or variances differ among Dose levels?
(A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2)
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
(Note: When rho=0 the results of Test 3 and Test 2 will be the same.)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 58.2841 6 <.0001
Test 2 2.06173 3 0.5597
D-12
-------
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
Hill Model with 0.95 Confidence Level
1.85
1.75
1.7
1.65
1.6
1.55
1.5
1.45
Hill
\
BMDL
BMD
0
2000 4000 6000 8000 10000 12000 14000 16000 18000
dose
10:03 11/10 2011
Data points obtained from 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.
D-13
-------
D.2.2. Decreased Brain Weight in Male Rats Exposed During
Gestation Only (GD7-GD17)
As discussed in Section 5.1.2.1, Cmax, as calculated by EPA's PBPK model, was selected
as the dose metric for this exposure scenario. Exposures occurred only during the major period of
organogenesis, during which the level of exposure is believed to be more important than the
duration of exposure.
The first step in the current analysis is to convert the inhalation doses, given as ppm
values from the studies, to an internal dose surrogate or dose metric using the EPA PBPK model
(see Appendix B). Predicted Cmax values for methanol in the blood of rats, with and without
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 weeks'5
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).
""Exposed throughout gestation. Values are means ± SD
°p < 0.01, as calculated by the authors.
Data from NEDO (1987).
The BMD technical guidance (U.S. EPA, 2012a) suggests that in the absence of
knowledge as to what level of response to consider adverse, a change in the mean equal to
1 control SD from the control mean can be used as a BMR for continuous endpoints. However, it
has been suggested that other BMRs, such as 5% change relative to estimated control mean, are
also appropriate when performing BMD analyses on fetal weight change as a developmental
endpoint (Kavlock et al., 1995). Therefore, in this assessment, both a 1 control mean SD change
and a 5% change relative to estimated control mean were considered. All models were fit using
restrictions and option settings suggested in the EPA's BMD Technical Guidance Document
(U.S. EPA. 2012a).
D.2.2.1. BMD Approach with a BMR of 1 Control Mean SD (GD7-GD17)
A summary of the results most relevant to the development of a POD using the BMD
approach (BMD, BMDL, and model fit statistics) (NEDO, 1987) for decreased brain weight at 8
D-14
-------
weeks in male rats exposed to methanol during gestation from days 7-17, with a BMR of 1
control mean S.D, is provided in Table D-5. Male brain weight responses were chosen because
they resulted in lower BMD and BMDL estimates than female responses (data not shown).
Model fit was determined by statistics (AIC and $ residuals of individual dose groups) and
visual inspection, as recommended by EPA (U.S. EPA, 2012a). The Polynomial and Power
models reduced to Linear model and returned identical modeling results. There is a greater than
5-fold range of BMDL estimates from adequately fitting models, indicating considerable model
dependence. In addition, the fit of the Hill and Exponential 4 and 5 models are better than the
other models in the dose region of interest as indicated by a lower scaled residual at the dose
group closest to the BMD (-0.09 versus —0.3) and visual inspection. In accordance with EPA
BMD Technical Guidance (U.S. EPA, 2012a), the BMDL from the Exponential 4 and 5 models
(bolded), is selected as the most appropriate basis for an RfC derivation because it results in the
lowest BMDL from among a broad range of BMDLs and provides a superior fit in the low dose
region nearest the BMD. Output from the Exponential 4 model, including text and plot
(Figure D-3), is shown after Table D-5. The BMDLiso was determined to be 115 mg/L, using the
95% lower confidence limit of the dose-response curve expressed in terms of the Cmax above
background for methanol in blood.
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
Linear
2nd degree Polynomial
3rd degree Polynomial
Power
Hillb
Exponential 2
Exponential 3
Exponential 4
Exponential 5
BMDiso
(Cmax, mg/L)a
960.78
960.78
960.78
960.78
449.28
925.82
925.92
433.46
433.46
BMDLiso
(Cmax, mg/L)a
626.64
626.64
626.64
626.64
115.97
589.97
589.97
114.86
114.86
"The BMDL is the 95% lower confidence limit on the Cmax estimated to
2.1 .1 (U.S. EPA, 2009) and model options and restrictions suaaested
p-value AICb
0.8837 -173.347015
0.8837 -173.347015
0.8837 -173.347015
0.8837 -173.347015
0.9272 -171.586011
0.8910 -173.3635
0.8910 -173.3635
0.9266 -171.5859
0.9266 -171.5859
Scaled
residual0
-0.28
-0.28
-0.28
-0.28
0.0944
-0.2674
-0.2674
0.09421
0.09421
decrease brain weight by 1 control mean SD using BMDS
bv EPA BMD technical auidance (U.S. EPA, 201 2a).
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).
°chi-squared (X2) 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 SD 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.
Data from NEDO (1987)
D-15
-------
Exponential Model. (Version: 1.7; Date: 12/10/2009)
Input Data File: C:/USEPA/BMDS220/Data/Methanol/exp_NEDOrat-Gest-Cmax-Std_Exp-
ModelVariance-BMRlStd-Down.(d)
Gnuplot Plotting File:
Tue Mar 27 12:45:12 2012
BMDS Model Run
The form of the response function by Model:
Model 2: Y[dose] = a * expfsign * b * dose}
Model 3: Y[dose] = a * expfsign * (b * dose) ^d}
Model 4: Y[dose] = a * [c-(c-l) * exp{-b * dose}]
Model 5: Y[dose] = a * [c-(c-l) * exp{-(b
Note: Y[dose] is the median response for exposure = dose;
sign = +1 for increasing trend in data;
sign = -1 for decreasing trend.
Model 2 is nested within Models 3 and 4.
Model 3 is nested within Model 5 .
Model 4 is nested within Model 5 .
Dependent variable = Mean
Independent variable = Dose
Data are assumed to be distributed: normally
Variance Model: exp(lnalpha +rho *ln(Y[dose]))
The variance is to be modeled as Var(i) = exp(lalpha + log(mean(i)) * rho)
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
MLE solution provided: Exact
Initial Parameter Values
Variable Model 4
Inalpha 7.32457
rho -18.5236
a 2.1105
b 0.000507001
c 0.816778
d 1
Parameter Estimates
Variable Model 4
Inalpha 6.99305
rho -18.0776
a 2.00632
b 0.000758964
c 0.891583
d 1
D-16
-------
Table of Stats From Input Data
Dose N Obs Mean Obs Std Dev
0 11 2 0.047
7.41 11 2.01 0.075
114.6 12 1.99 0.072
2986 10 1.81 0.161
Estimated Values of Interest
Dose Est Mean Est Std Scaled Residual
0 2.006 0.06098 -0.3437
7.41 2.005 0.06132 0.2651
114.6 1.988 0.06619 0.09421
2986 1.811 0.1536 -0.02792
Other models for which likelihoods are calculated:
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma/x2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)^2
Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = exp(lalpha + log(mean(i)) * rho)
Model R: Yij = Mu + e(i)
Var{e(ij)} = Sigma/x2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 83.20596 5 -156.4119
A2 92.06049 8 -168.121
A3 90.61606 6 -169.2321
R 70.76186 2 -137.5237
4 90.79294 5 -171.5859
Additive constant for all log-likelihoods = -40.43. This constant added to the
above values gives the log-likelihood including the term that does not
depend on the model parameters.
Explanation of Tests
Test 1: Does response and/or variances differ among Dose levels? (A2 vs. R)
Test 2: Are Variances Homogeneous? (A2 vs. Al)
Test 3: Are variances adeguately modeled? (A2 vs. A3)
Test 6a: Does Model 4 fit the data? (A3 vs 4)
Tests of Interest
Test -2*log(Likelihood Ratio) D. F. p-value
Test 1 42.6 6 < 0.0001
Test 2 17.71 3 0.000505
Test 3 2.889 2 0.2359
Test 6a -0.3538 1 N/A
D-17
-------
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 6a is less than .1. Model 4 may not adeguately
describe the data; you may want to consider another model.
Benchmark Dose Computations:
Specified Effect = 1.000000
Risk Type = Estimated standard deviations from control
Confidence Level = 0.950000
BMD = 433.456
BMDL = 114.856
Exponential Model 4 with 0.95 Confidence Level
I
2.05
2
1.95
1.9
1.85
1.8
1.75
1.7
Exponential —
;BMDL
BMD
0
500
1000
1500
dose
2000
2500
3000
11:45 03/27 201 2
Data points obtained from NEDO (1987).
Figure D-3 Exponential model 4, BMR of 1 control mean SD - Decreased brain weight in
male rats at 8 weeks of age versus Cmax above background, gestation only
inhalational study.
D-18
-------
D.2.3. C.1.2.2. BMD Approach with a BMR of 0.05 Change Relative to
Control Mean (GD7-GD17)
A summary of the results most relevant to the development of a POD using the BMD
approach (BMD, BMDL, and model fit statistics) for decreased brain weight at 8 weeks in male
rats exposed to methanol during gestation from days 7 to 17, with a BMR of 0.05 change relative
to estimated control mean, is provided in Table D-6. Model fit was determined by statistics (AIC
and %2 residuals of individual dose groups) and visual inspection, as recommended by EPA
(2012a). Modeling considerations and uncertainties for this data set were discussed in C. 1.2.1
and, as was done for the BMR of 1 SD, the lowest BMDL was chosen for use in the RfC
derivation (NEDO, 1987), which in this case was the BMDL05 of 119.51 mg methanol/L in blood
estimated by the Exponential 5 model. Results from the Exponential 5 model, including text and
plot (see Figure D-4), are shown after Table D-6.
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
Linear13
2nd degreePolynomial
3rd degree Polynomial
Power
Hillb
Exponential 2
Exponential 3
Exponential 4
Exponential 5
BMDos (Cmax,
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
BMDLos (Cmax,
mg/L)a
1,061.91
1,061.91
1,061.91
1,061.91
Not Reported
1,009.52
1,009.52
233.33
119.51
aThe BMDL is the 95% lower confidence limit on the Cmax estimated to
p-value
0.8837
0.8837
0.8837
0.8837
0.9272
0.8910
0.8910
0.9266
0.9266
decrease brain
AICC
-173.347015
-173.347015
-173.347015
-173.347015
-171.586011
-173.3635
-173.3635
-171.5859
-171.5859
weight by 5% using
Scaled residual01
-0.28
-0.28
-0.28
-0.28
0.0944
-0.2674
-0.2674
0.09421
0.09421
BMDS 2.2 (U.S. EPA,
2011 b) and model options and restrictions suggested by EPA BMD Technical Guidance (2012a).
°AIC = Akaike Information Criterion = -2L + 2P, where L is the log-likelihood at the maximum likelihood estimates for the parameters,
and Pis the number of modeled degrees of freedom (usually the number of parameters estimated).
dchi-squared (X2) 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 SD 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.
Data from NEDO (1987).
D-19
-------
Exponential Model. (Version: 1.7; Date: 12/10/2009)
Input Data File: C:/USEPA/BMDS220/Data/Methanol/exp_NEDOrat-Gest-Cmax-Std_Exp-
ModelVariance-BMR05-Down.(d)
Gnuplot Plotting File:
Tue Mar 27 15:30:45 2012
BMDS Model Run
The form of the response function by Model:
Model 2: Y[dose] = a * expfsign * b * dose}
Model 3: Y[dose] = a * expfsign * (b * dose) ^d}
Model 4: Y[dose] = a * [c-(c-l) * exp{-b * dose}]
Model 5: Y[dose] = a * [c-(c-l) * exp{-(b
Note: Y[dose] is the median response for exposure = dose;
sign = +1 for increasing trend in data;
sign = -1 for decreasing trend.
Model 2 is nested within Models 3 and 4.
Model 3 is nested within Model 5 .
Model 4 is nested within Model 5 .
Dependent variable = Mean
Independent variable = Dose
Data are assumed to be distributed: normally
Variance Model: exp(lnalpha +rho *ln(Y[dose]))
The variance is to be modeled as Var(i) = exp(lalpha + log(mean(i)) * rho)
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
MLE solution provided: Exact
Initial Parameter Values
Variable Model 5
Inalpha 7.32457
rho -18.5236
a 2.1105
b 0.000507001
c 0.816778
d 1
Parameter Estimates
Variable Model 5
Inalpha 6.99305
rho -18.0776
a 2.00632
b 0.000758964
c 0.891583
d 1
D-20
-------
Table of Stats From Input Data
Dose N Obs Mean Obs Std Dev
0 11 2 0.047
7.41 11 2.01 0.075
114.6 12 1.99 0.072
2986 10 1.81 0.161
Estimated Values of Interest
Dose Est Mean Est Std Scaled Residual
0 2.006 0.06098 -0.3437
7.41 2.005 0.06132 0.2651
114.6 1.988 0.06619 0.09421
2986 1.811 0.1536 -0.02792
Other models for which likelihoods are calculated:
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma/x2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)^2
Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = exp(lalpha + log(mean(i)) * rho)
Model R: Yij = Mu + e(i)
Var{e(ij)} = Sigma/x2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 83.20596 5 -156.4119
A2 92.06049 8 -168.121
A3 90.61606 6 -169.2321
R 70.76186 2 -137.5237
5 90.79294 5 -171.5859
Additive constant for all log-likelihoods = -40.43. This constant added to the
above values gives the log-likelihood including the term that does not
depend on the model parameters.
Explanation of Tests
Test 1: Does response and/or variances differ among Dose levels? (A2 vs. R)
Test 2: Are Variances Homogeneous? (A2 vs. Al)
Test 3: Are variances adeguately modeled? (A2 vs. A3)
Test 7a: Does Model 5 fit the data? (A3 vs 5)
Tests of Interest
Test -2*log(Likelihood Ratio) D. F. p-value
Test 1 42.6 6 < 0.0001
Test 2 17.71 3 0.000505
Test 3 2.889 2 0.2359
Test 7a -0.3538 1 N/A
D-21
-------
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 7a is less than .1. Model 5 may not adeguately
describe the data; you may want to consider another model.
Benchmark Dose Computations:
Specified Effect = 0.050000
Risk Type = Relative deviation
Confidence Level = 0.950000
BMD = 814.763
BMDL = 119.505
Exponential Model 5 with 0.95 Confidence Level
I
1.75
1.7
14:3003/272012
3000
Data points obtained from 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-22
-------
D.3. RfC Derivations Using Rogers et al. (1993b)
For the purposes of deriving an RfC for methanol from developmental endpoints using
the BMD method and mouse data, cervical rib incidence data were evaluated from Rogers et al.
(1993b). In this paper, Rogers et al. (1993b) also utilized a BMD methodology, examining the
dosimetric threshold for cervical ribs and other developmental impacts by applying a log-logistic
maximum likelihood model to the dose-response data. Using air exposure concentrations (ppm)
as their dose metric, a value for the lower 95% confidence limit on the benchmark dose for 5%
additional risk in mice was 305 ppm (400 mg/m3), using the log-logistic model. Although the
teratology portion of the NEDO study (1987) also reported increases in cervical rib incidence in
Sprague-Dawley rats, the Rogers et al. (1993b) study was chosen for dose-response modeling
because effects were seen at lower doses, it was peer-reviewed and published in the open
literature, and data on individual animals were available for a more statistically robust analysis
utilizing nested models available in HMDS 2.2 (U.S. EPA, 201 Ib).
As described in Section 5.1.2.1, because exposure was during gestation only and due to
the small critical gestational window for cervical rib abnormalities, Cmax of methanol in blood
(mg/L) is chosen as the appropriate internal dose metric. Because the critical window for
methanol induction of cervical rib malformations in CD-I mice is between GD6 and GD7
(Rogers and Mole, 1997; Rogers et al., 1993a), the measured Cmax plasma methanol levels for
gestation day 6 from the Rogers study are used with background levels (1.6 g/L) subtracted. Cmax
values for methanol in the blood of mice are summarized in Table D-7. These Cmax values are
then used as the dose metric for the BMD analysis of the litter-specific cervical rib response. The
overall cervical rib/litter (%) reported by Rogers et al. (1993b) is shown in Table D-7, but litter-
specific response data from this study (170 litters) obtained from John Rogers (personal
communication) was used for the nested BMD analysis. Due to high mortality, the high
(15,000 ppm) dose group (5 litters) was excluded from this analysis. The individual animal
response data for the four dose groups are displayed below in the text output files for the
NLogistic model.
D-23
-------
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
"Reported Cmax background levels of 1.6 mg/L were subtracted from reported Cmax values.
Source: Rogers et al. (1_993b)
A 10% BMR level is the value typically calculated for comparisons across chemicals and
endpoints for dichotomous responses because this level is near the low end of the observable
range for many types of toxicity studies. However, from a statistical standpoint most
reproductive and developmental studies involve a large enough sample size to support a 5%
BMR for determination of a POD (U.S. EPA. 2012a: Allen etal.. 1994a). Rogers et al. (1993b)
utilized a 5% added risk for the BMR in the original study. This assessment utilizes both a 10%
and 5% extra risk level as a BMR for the determination of a POD.3 The nested suite of models
available in BMDS 2.2 (U.S. EPA, 20lib) was used to model the cervical rib data. In general,
data from developmental toxicity studies are best modeled using nested models, as these models
account for any intralitter correlation (i.e., the tendency of littermates to respond similarly to one
another, relative to other litters in a dose group). All models were fit using restrictions and option
settings suggested in the EPAs BMD Technical Guidance Document (U.S. EPA, 2012a).
D.3.1. BMD Approach with a BMR of 0.10 Extra Risk
A summary of the results most relevant to the development of a POD using the BMD
approach (BMD, BMDL, and model fit statistics) for increased incidence of cervical rib in mice
exposed to methanol during gestation from days 6 to 15, with a BMR of 0.10 extra risk, is
provided in Table D-8. Model fit was determined by statistics (AIC and ^ residuals of individual
dose groups) and visual inspection, as recommended by U.S. EPA (U.S. EPA, 2012a). The best
model fit to these data (from visual inspection and comparison of AIC values) was obtained
using the Nested Logistic (NLogistic) model. The textual and graphic (see Figure D-5) output
from this model follows Table D-8. The BMDLio was determined to be 90.9972 mg/L using the
3 Starr and Festa (2003) have argued that the Rogers, et al. Q993b) 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. 20l2a) 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.
D-24
-------
95% lower confidence limit of the dose-response curve expressed in terms of the Cn
methanol in blood (Rogers et al., 1993b).
-for
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
NCTR
Rai and Van Ryzin
BMD10
(Cmax, mg/L)a
140.75
223.55
233.61
BMDLio
(Cmax, mg/L)a
91.00
111.78
116.81
p-value
0.3359
0.2705
0.2625
AICC
1,047.37
1,050.32
1,052.14
Scaled residual01
0.5395
0.5640
0.6043
aCmax values are the blood levels of the damson 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. 2012a).
bModel choice based on adequate p 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 Pis the number of modeled degrees of freedom (usually the number of parameters estimated).
dX2d 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 SD 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.
Data from Rogers et al. (1993b).
D-25
-------
NLogistic Model. (Version: 2.15; Date: 10/28/2009)
Input Data File: C:/Documents and
Settings/llowe/Desktop/ROGERS_CMAX_BMD/ROGERS_CMAX_BMD10/NLog_CR_10. (d)
Fri Dec 16 10:48:13 2011
BMDS Model Run
The probability function is:
Prob. = alpha + thetal*Rij + [1 - alpha - thetal*Rij]/
[1+exp (-beta-theta2*Rij-rho*log (Dose))],
where Rij is the litter specific covariate.
Restrict Power rho >= 1.
Total number of observations =166
Total number of records with missing values = 0
Total number of parameters in model = 9
Total number of specified parameters = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.302379
beta = -7.2579
thetal = 0
theta2 = 0
rho = 1
phil = 0.214334
phi2 = 0.304943
phi3 = 0.220179
phi4 = 0.370088
Parameter Estimates
Variable Estimate Std. Err.
alpha 0.127131 *
beta -4.62736 *
thetal 0.0297845 *
theta2 -0.467856 *
rho 1 *
phil 0.203691 *
phi2 0.305429 *
phi3 0.212663 *
phi4 0.363199 *
Indicates that this value is not calculated.
Log-likelihood: -515.686AIC: 1047.37
D-26
-------
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
-0
-0
2.
-0
-0
1.
0.
-0
0.
-0
-0
0.
-0
0.
-1
-0
-0
1.
-1
-1
-1
-0
-0
-0
-1
1.
0.
0.
-1
0.
1.
0.
-0
0.
0.
0.
-0
2.
-1
-1
-0
1.
-0
-0
-0
-0
1.
-0
0.
0.
-1
-1
-0
0.
-0
2.
0.
.4314
.4314
.6176
6904
.6176
.6176
0364
4142
.7674
4142
.7674
.9007
0136
.9007
0136
.0250
.2824
.2824
2028
.0250
.0250
.0250
.2824
.2824
.2824
.1444
9738
7265
7265
.1444
7265
9738
7265
.5208
7265
1029
1029
.5208
5975
.1444
.1444
.7245
4233
.7245
.1876
.7245
.1876
4233
.1876
3494
3494
.2615
.2615
.1876
5076
.4352
3932
0362
D-27
-------
0.0000 8.0000 0.365 8 2.923 1 -0.9066
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
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.
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
.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
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
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
.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
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
0 -0.7951
0 -0.7951
1.7177
0 -0.8919
1 0.0472
3 2.0021
0.0472
1.0246
3 1.3743
0 -0.9685
0 -1.0189
0 -1.0189
1 -0.3733
1.5633
0 -1.0189
0.5566
0.0105
-0.2131
-0.2131
0.2577
0 -1.1545
-0.2131
-0.2131
-0.4101
8 2.0644
0 -1.2350
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
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
0.716
0.638
0.564
0.564
0.503
0.503
0.503
0.503
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
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
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
.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
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
3.057 1
0.8091
1.0920
-0.9912
2 -0.2032
1.6327
-0.9972
0.3178
0.3178
-0.4350
1.2756
2 -0.4350
0.1352
-0.4350
0 -1.5754
0.7054
0 -1.5754
1.2756
-0.4350
0.7054
0.1352
1.8458
0.1352
1.2756
0.1352
0.4762
0.9813
0 -1.5443
0.9813
-1.0392
0.4762
-0.0289
0.4762
-1.0392
-0.0289
-0.0289
-1.0392
D-28
-------
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.
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.
,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
.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
Combine litters with adjacent levels of the litter-specific covariate
within dose groups until the expected count exceeds 3.0, to help improve
the fit of the X^2 statistic to chi-square.
Grouped Data
Mean Scaled
Dose Lit.-Spec. Cov. Expected Observed Residual
0.0000 1.0000 0.314 0 -0.6101
0.0000 2.0000 1.867 3 0.8381
0.0000 3.0000 2.598 2 -0.3532
0.0000 4.0000 3.940 2 -0.8870
0.0000 5.0000 4.141 2 -0.9178
0.0000 5.0000 4.141 3 -0.4891
D-29
-------
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
5
5
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
8
8
8
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
4.
1.
3.
3.
3.
3.
3.
3.
3.
3.
4.
4.
4.
4.
4.
4.
2.
5.
5.
2.
,141
,380
, 670
, 670
,670
,670
, 670
, 670
,670
,670
, 699
, 699
,699
,699
, 699
, 699
,349
,847
,847
,923
2
1
5
6
3
8
4
4
7
0
6
3
3
7
6
0
2
6
11
1
-
-
0
1
-
1
0
0
1
-
0
0
0
0
1
-0
-
0
0
0
-1
-
0
0
1
-0
.9178
.2824
5865
0275
.2955
9094
1455
1455
4685
.6184
4941
.6450
.6450
8738
4941
.7840
.1876
0512
.7178
.9066
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
1.
2.
3.
4.
5.
5.
5.
6.
7.
7.
7.
8.
8.
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0
1
3
2
3
3
1
3
4
4
4
5
2
.775
.367
.807
.480
.157
.157
.578
.962
.905
.905
.905
.989
.994
0 -1.1245
2 0.5840
7 1.5606
3 0.2870
0 -1.4409
5 0.8414
0 -1.0189
5 0.4010
4 -0.3013
3 -0.6342
4 -0.3013
10 1.1697
0 -1.2350
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
2.
3.
4.
5.
5.
6.
6.
6.
6.
6.
6.
6.
6.
7.
7.
7.
7.
7.
7.
7.
7.
7.
7.
7.
7.
8.
8.
8.
9.
9.
,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
4
5
5
5
5
5
5
5
5
5
5
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
1.432 2
1.915 3
516 3
033 6
033 6
526 7
526 5
526 2
526 4
526 7
526 7
526 9
526 8
3.057 4
3.057 5
057 0
057 5
3.057 1
3.057 4
057 3
057 4
3.057 1
3.057 3
3.057 3
057 1
436 7
3.436 5
436 0
915 0
3.915 6
0.8091
1.0920
-0.8446
0.4494
0.4494
0.5944
-0.2120
-1.4216
-0.6152
0.5944
0.5944
1.4008
0.9976
0.4762
0.9813
-1.5443
0.9813
-1.0392
0.4762
-0.0289
0.4762
-1.0392
-0.0289
-0.0289
-1.0392
1.6134
0.7079
-1.5558
-1.6016
0.8530
D-30
-------
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
D-31
-------
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
Data points obtained from Rogers et al. (1_993b).
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
A summary of the results most relevant to the development of a POD using the BMD
approach (BMD, BMDL, and model fit statistics) for increased incidence of cervical rib in mice
exposed to methanol during gestation from days 6 to 15, with a BMR of 0.05 extra risk, is
provided in Table D-9. Model fit was determined by statistics (AIC and ^ residuals of individual
dose groups) and visual inspection, as recommended by U.S. EPA (20J_2a). The best model fit to
these data (from visual inspection and comparison of AIC values) was obtained using the
NLogistic model. The text and graphic (see Figure D-6) output from this model follow
Table D-6. The BMDLos was determined to be 43.1039 mg/L using the 95% lower confidence
limit of the dose-response curve expressed in terms of the Cmaxfor methanol in blood (Rogers et
al.. 1993b).
D-32
-------
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
BMDos
(Cmax, mg/L)a
66.67
108.83
113.73
BMDUs
(Cmax, mg/L)a
43.10
54.42
56.87
p-value
0.3359
0.2705
0.2625
AICC
1047.37
1050.32
1052.14
Scaled residual01
0.5395
0.5640
0.6043
aCmaxare 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 p 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 Pis the number of modeled degrees of freedom (usually the number of parameters estimated).
dchi-squared (X2) 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 SD 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.
Data from Rogers et al. (1993b).
D-33
-------
NLogistic Model. (Version: 2.15; Date: 10/28/2009)
Input Data File: C:/Documents and
Settings/llowe/Desktop/ROGERS_CMAX_BMD/ROGERS_CMAX_BMD05/NLog_CR_5. (d)
Fri Dec 16 10:56:05 2011
BMDS Model Run
The probability function is:
Prob. = alpha + thetal*Rij + [1 - alpha - thetal*Rij]/
[1+exp (-beta-theta2*Rij-rho*log (Dose))],
where Rij is the litter specific covariate.
Restrict Power rho >= 1.
Total number of observations =166
Total number of records with missing values = 0
Total number of parameters in model = 9
Total number of specified parameters = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.302379
beta = -7.2579
thetal = 0
theta2 = 0
rho = 1
phil = 0.214334
phi2 = 0.304943
phi3 = 0.220179
phi4 = 0.370088
Parameter Estimates
Variable Estimate Std. Err.
alpha 0.127131 *
beta -4.62736 *
thetal 0.0297845 *
theta2 -0.467856 *
rho 1 *
phil 0.203691 *
phi2 0.305429 *
phi3 0.212663 *
phi4 0.363199 *
Indicates that this value is not calculated.
Log-likelihood: -515.686AIC: 1047.37
D-34
-------
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
-0
-0
2.
-0
-0
1.
0.
-0
0.
-0
-0
0.
-0
0.
-1
-0
-0
1.
-1
-1
-1
-0
-0
-0
-1
1.
0.
0.
-1
0.
1.
0.
-0
0.
0.
0.
-0
2.
-1
-1
-0
1.
-0
-0
-0
-0
1.
-0
0.
0.
-1
-1
-0
0.
-0
2.
0.
.4314
.4314
.6176
6904
.6176
.6176
0364
4142
.7674
4142
.7674
.9007
0136
.9007
0136
.0250
.2824
.2824
2028
.0250
.0250
.0250
.2824
.2824
.2824
.1444
9738
7265
7265
.1444
7265
9738
7265
.5208
7265
1029
1029
.5208
5975
.1444
.1444
.7245
4233
.7245
.1876
.7245
.1876
4233
.1876
3494
3494
.2615
.2615
.1876
5076
.4352
3932
0362
D-35
-------
0.0000 8.0000 0.365 8 2.923 1 -0.9066
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
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.
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
.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
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
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
.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
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
0 -0.7951
0 -0.7951
1.7177
0 -0.8919
1 0.0472
3 2.0021
0.0472
1.0246
3 1.3743
0 -0.9685
0 -1.0189
0 -1.0189
1 -0.3733
1.5633
0 -1.0189
0.5566
0.0105
-0.2131
-0.2131
0.2577
0 -1.1545
-0.2131
-0.2131
-0.4101
8 2.0644
0 -1.2350
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
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
0.716
0.638
0.564
0.564
0.503
0.503
0.503
0.503
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.460
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
0.437 7
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
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
.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
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
3.057 1
0.8091
1.0920
-0.9912
2 -0.2032
1.6327
-0.9972
0.3178
0.3178
-0.4350
1.2756
2 -0.4350
0.1352
-0.4350
0 -1.5754
0.7054
0 -1.5754
1.2756
-0.4350
0.7054
0.1352
1.8458
0.1352
1.2756
0.1352
0.4762
0.9813
0 -1.5443
0.9813
-1.0392
0.4762
-0.0289
0.4762
-1.0392
-0.0289
-0.0289
-1.0392
D-36
-------
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.
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.
,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
.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
Combine litters with adjacent levels of the litter-specific covariate
within dose groups until the expected count exceeds 3.0, to help improve
the fit of the X^2 statistic to chi-square.
Grouped Data
Mean Scaled
Dose Lit.-Spec. Cov. Expected Observed Residual
0.0000 1.0000 0.314 0 -0.6101
0.0000 2.0000 1.867 3 0.8381
0.0000 3.0000 2.598 2 -0.3532
0.0000 4.0000 3.940 2 -0.8870
0.0000 5.0000 4.141 2 -0.9178
0.0000 5.0000 4.141 3 -0.4891
D-37
-------
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
5
5
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
8
8
8
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
4.
1.
3.
3.
3.
3.
3.
3.
3.
3.
4.
4.
4.
4.
4.
4.
2.
5.
5.
2.
,141
,380
, 670
, 670
,670
,670
, 670
, 670
,670
,670
, 699
, 699
,699
,699
, 699
, 699
,349
,847
,847
,923
2
1
5
6
3
8
4
4
7
0
6
3
3
7
6
0
2
6
11
1
-
-
0
1
-
1
0
0
1
-
0
0
0
0
1
-0
-
0
0
0
-1
-
0
0
1
-0
.9178
.2824
5865
0275
.2955
9094
1455
1455
4685
.6184
4941
.6450
.6450
8738
4941
.7840
.1876
0512
.7178
.9066
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
61.4000
1.
2.
3.
4.
5.
5.
5.
6.
7.
7.
7.
8.
8.
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0
1
3
2
3
3
1
3
4
4
4
5
2
.775
.367
.807
.480
.157
.157
.578
.962
.905
.905
.905
.989
.994
0 -1.1245
2 0.5840
7 1.5606
3 0.2870
0 -1.4409
5 0.8414
0 -1.0189
5 0.4010
4 -0.3013
3 -0.6342
4 -0.3013
10 1.1697
0 -1.2350
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
485.4000
2.
3.
4.
5.
5.
6.
6.
6.
6.
6.
6.
6.
6.
7.
7.
7.
7.
7.
7.
7.
7.
7.
7.
7.
7.
8.
8.
8.
9.
9.
,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
4
5
5
5
5
5
5
5
5
5
5
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
1.432 2
1.915 3
516 3
033 6
033 6
526 7
526 5
526 2
526 4
526 7
526 7
526 9
526 8
3.057 4
3.057 5
057 0
057 5
3.057 1
3.057 4
057 3
057 4
3.057 1
3.057 3
3.057 3
057 1
436 7
3.436 5
436 0
915 0
3.915 6
0.8091
1.0920
-0.8446
0.4494
0.4494
0.5944
-0.2120
-1.4216
-0.6152
0.5944
0.5944
1.4008
0.9976
0.4762
0.9813
-1.5443
0.9813
-1.0392
0.4762
-0.0289
0.4762
-1.0392
-0.0289
-0.0289
-1.0392
1.6134
0.7079
-1.5558
-1.6016
0.8530
D-38
-------
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 =0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 66.6706
BMDL = 43.1039
D-39
-------
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
Data points obtained from Rogers et al.(1993b).
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)
The BMD approach was utilized in the derivation of potential chronic inhalation
reference values from effects seen in monkeys due to prenatal methanol exposure. Deficits in
VDR were evaluated from Burbacher et al. (1999a: 1999b). In the application of the BMD
approach, continuous models in EPA's BMDS 2.2 were fit to the data set for increased latency in
VDR in neonatal monkeys. The maximum blood methanol values (Cmax) above background
estimated using the PK model described in Appendix B were used as the dose metric.
The VDR test, which assesses time (from birth) it takes for an infant to grasp for a
brightly colored object containing an applesauce-covered nipple, is a measure of sensorimotor
development. Beginning at 2 weeks after birth, infants were tested 5 times/day, 4 days/week.
Performance on that test, measured as age from birth at achievement of test criterion (successful
object retrieval on 8/10 consecutive trials over 2 testing sessions), was reduced in all treated male
infants. The times (days after birth) to achieve the criteria for the VDR test in means ± SD were
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
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
females in the control to 1,800 ppm groups, respectively. As discussed in Section 4.4.2, this type
D-40
-------
of response data is sometimes adjusted to account for premature births by subtracting time (days)
premature from the time (days from birth) needed to meet the test criteria (Wilson and Cradock,
2004). When this type of adjustment is applied, the times (days after birth or, if shorter, days
after control mean gestation length) to achieve the criteria for VDR test in means ± SD were
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
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 =
7) days for females in the control to 1,800 ppm groups, respectively. When these data were
modeled within BMDS 2.1.1 (U.S. EPA, 2009), there was no significant difference between
unadjusted responses and/or variances among the dose levels (indicating lack of a dose-response)
for males and females combined (p = 0.244), for males only (p = 0.321) and for males only with
the high-dose group excluded (p = 0.182), or for adjusted responses of males and females
combined (p = 0.12), males only (p = 0.448) and males only with the high-dose group excluded
(p = 0.586).4 The only data that offered a significant dose-response trend was that for unadjusted
(p = 0.0265) and adjusted (p = 0.009) female responses, largely because of the much larger
overall sample size across dose groups for females versus males (21 females versus 13 males).
However, the model fits for the adjusted female response data were unacceptable. Only the
unadjusted female VDR response data (Table D-10) offered both a dose-response trend and
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 C^x above
background (mg/L)b
0
2.87
10.4
38.4
Days After Birth to Achieve VDR Test
Criteria"
34.2 ±4.0
33.0 ±5.8
27.6 ±6.0
40.0 ±10.6
N
5
4
5
7
"Reprinted with permission of the Health Effects Institute, Boston, MA; from Burbacher et al. (1999a) and Burbacher et al. f(1999b).
Table 2].
bEstimated from the two-compartment PK monkey model described in Appendix B.
°Data reported in means ± standard deviation.
The BMD technical guidance (U.S. EPA, 2012a) suggests that in the absence of
knowledge as to what level of response to consider adverse, a change in the mean equal to
1 control SD from the control mean can be used as a BMR for continuous endpoints. A summary
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.
D-41
-------
of the results most relevant to the development of a POD using the BMD approach (BMD,
BMDL, and model fit statistics) for increased latency of VDR in female neonatal monkeys
exposed to methanol with a BMR of 1 control mean SD is provided in Table D-l 1. Model fit was
determined by statistics (AIC and $ residuals of individual dose groups) and visual inspection,
as recommended by EPA (2012a). The Power model returned a lower AIC than the other
models.5 The text and graphic (see Figure D-7) output from this model follows Table D-10. The
BMDLiso was determined to be 19.59 mg/L, using the 95% lower confidence limit of the dose-
response curve expressed in terms of the ppm of external methanol concentration.
Table D-ll Comparison of BMD modeling results for VDR in female monkeys using Cr
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
BMDiso
(Cmax, mg/L)a
38.92
32.27
33.53
37.50
36.90
36.54
37.32
38.92
37.19
BMDLiso
(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
AICC
110.51
109.49
109.09
108.93
110.93
110.46
110.93
112.51
112.93
Scaled residual01
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 Cmax of a decrease of 1 control mean SD estimated by the model using the likelihood profile
method (U.S. EPA. 2012a).
bModel choice based on adequate p 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).
dX2d 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 SD 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.
Data from Burbacher et al. (1999a).
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.
D-42
-------
Power Model. (Version: 2.16; Date: 10/28/2009)
Input Data File: C:/USEPA/BMDS220/Data/pow_monkey_Pow-ModelVariance-BMRlStd-
Restrict. (d)
Gnuplot Plotting File: C:/USEPA/BMDS220/Data/pow_monkey_Pow-ModelVariance-BMRlStd-
Restrict.plt
Wed Nov 16 11:02:04 2011
BMDS Model Run
The form of the response function is:
Y[dose] = control + slope * dose^power
Dependent variable = Mean
Independent variable = Dose
The power is restricted to be greater than or egual to 1
The variance is to be modeled as Var (i) = exp (lalpha + log (mean (i)) * rho)
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
lalpha = 4.05748
rho = 0
control = 27.6
slope = 3.85161
power = 0.320501
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter (s) -power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
lalpha rho control slope
lalpha 1 -1 -0.29 0.6
rho -1 1 0.27 -0.6
control -0.29 0.27 1 -0.37
slope 0.6 -0.6 -0.37 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
lalpha -13.0645 12.1112 -36.8021 10.673
rho 4.77979 3.44065 -1.96376 11.5233
control 31.5 1.4819 28.5955 34.4045
slope 2.57903e-028 1.21193e-028 2.03691e-029 4.95437e-028
power 18 NA
D-43
-------
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Table of Data and Estimated Values of Interest
Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Res.
0 5 34.2 31.5 4 5.54 1.09
2.87 4 33 31.5 5.8 5.54 0.541
10.4 5 27.6 31.5 6 5.54 -1.57
38.4 7 40 40 10.6 9.81 7.35e-008
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu (i) + e (ij)
Var {e (ij)} = SigmaA2
Model A2: Yij = Mu (i) + e (ij)
Var {e (ij)} = Sigma (i)A2
Model A3: Yij = Mu (i) + e (ij)
Var {e (ij)} = exp (lalpha + rho*ln (Mu (i)))
Model A3 uses any fixed variance parameters that
were specified by the user
Model R: Yi = Mu + e (i)
Var {e (i) } = SigmaA2
Likelihoods of Interest
Model Log (likelihood) # Param's AIC
Al -50.884765 5 111.769529
A2 -47.717070 8 111.434139
A3 -49.215263 6 110.430526
fitted -50.466380 4 108.932759
R -54.905426 2 113.810852
Explanation of Tests
Test 1: Do responses and/or variances differ among Dose levels?
(A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2)
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
(Note: When rho=0 the results of Test 3 and Test 2 will be the same.)
Tests of Interest
Test -2*log (Likelihood Ratio) Test df p-value
Test 1 14.3767 6 0.0257
Test 2 6.33539 3 0.09639
Test 3 2.99639 2 0.2235
Test 4 2.50223 2 0.2862
D-44
-------
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
Power Model with 0.95 Confidence Level
g.
(/}
o>
on
50
45
40
35
30
25
20
Power
/
T
BMDL
BMD
10
15
20
dose
25
30
35
40
11:02 11/16 2011
Data points obtained from Burbacher et al. (1999a: 1999b)
Figure D-7 Third (3rd) degree Polynomial model, BMR of 1 control mean SD - VDR in
female monkeys using Cmax above background of blood methanol as the dose
metric.
D-45
-------
APPENDIX E. DOCUMENTATION OF
IMPLEMENTATION OF THE 2011 NATIONAL
RESEARCH COUNCIL RECOMMENDATIONS
Background: On December 23, 2011, The Consolidated Appropriations Act, 2012, was
signed into law (U.S. Congress, 2011). The report language included direction to EPA for the
IRIS Program related to recommendations provided by the National Research Council (NRC) in
their review of EPA's draft IRIS assessment of formaldehyde (NRC, 2011). The report language
included the following:
"The Agency shall incorporate, as appropriate, based on chemical-specific datasets and
biological effects, the recommendations of Chapter 7 of the National Research Council's Review
of the Environmental Protection Agency's Draft IRIS Assessment of Formaldehyde into the IRIS
process.. .For draft assessments released in fiscal year 2012, the Agency shall include
documentation describing how the Chapter 7 recommendations of the National Academy of
Sciences (NAS) have been implemented or addressed, including an explanation for why certain
recommendations were not incorporated."
The NRC's recommendations, provided in Chapter 7 of their review report, offered
suggestions to EPA for improving the development of IRIS assessments. Consistent with the
direction provided by Congress, documentation of how the recommendations from Chapter 7 of
the NRC report have been implemented in this assessment is provided in the table below. Where
necessary, the documentation includes an explanation for why certain recommendations were not
incorporated.
The IRIS Program's implementation of the NRC recommendations is following a phased
approach that is consistent with the NRC's "Roadmap for Revision" as described in Chapter 7 of
the formaldehyde review report. The NRC stated that "the committee recognizes that the changes
suggested would involve a multi-year process and extensive effort by the staff at the National
Center for Environmental Assessment and input and review by the EPA Science Advisory Board
and others."
The IRIS methanol (noncancer) assessment is in Phase 1 of implementation, which
focuses on a subset of the short-term recommendations, such as editing and streamlining
documents, increasing transparency and clarity, and using more tables, figures, and appendices to
present information and data in assessments. Phase 1 also focuses on assessments near the end of
the development process and close to final posting. Chemical assessments in Phase 2 of
implementation will address all of the short-term recommendations from Table E-l. The IRIS
Program is implementing all of these recommendations but recognizes that achieving full and
robust implementation of certain recommendations will be an evolving process with input and
E-1
-------
feedback from the public, stakeholders, and external peer review committees. Chemical
assessments in Phase 3 of implementation will incorporate the longer-term recommendations
made by the NRC as outlined below in Table E-2, including the development of a standardized
approach to describe the strength of evidence for noncancer effects . On May 16, 2012, EPA
announced (U.S. EPA, 2012c)6 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.
This effort is included in Phase 3 of EPA's implementation plan.
Table E-l. National Research Council recommendation that EPA is implementing in the
short-term
NRC RECOMMENDATIONS THAT
EPA IS IMPLEMENTING IN THE
SHORT-TERM
IMPLEMENTATION IN THE METHANOL
(NONCANCER) ASSESSMENT
General recommendations for completing the IRIS formaldehyde assessment that EPA will adopt for all IRIS
assessments (see 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.
6EPA Announces NAS' Review of IRIS Assessment Development Process (www.epa. gov/iris)
E-2
-------
NRC RECOMMENDATIONS THAT
EPA IS IMPLEMENTING IN THE
SHORT-TERM
IMPLEMENTATION IN THE METHANOL
(NONCANCER) ASSESSMENT
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.
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. Standardized evidence tables are being
developed as a part of Phase 2 of the implementation
process.
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.
Partially 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.
Standardized approaches for evaluating studies are under
development as a part of Phase 2 and 3.
E-3
-------
NRC RECOMMENDATIONS THAT
EPA IS IMPLEMENTING IN THE
SHORT-TERM
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.
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 IN THE METHANOL
(NONCANCER) ASSESSMENT
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 RfCand 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 5-7.
Partially implemented. Weight-of-evidence considerations
were added or revised based on peer review comments
(see Appendix A). Table 5-7 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.
General Guidance for the Overall Process (p. 164 of the NRC Report)
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. Due to timing, and because methanol is a post-
peer review, Phase 1 chemical, the methanol team was
not able to make use of the "overall, documented, and
quality-controlled process" that is now being developed in
response to the NRC recommendations.
E-4
-------
NRC RECOMMENDATIONS THAT
EPA IS IMPLEMENTING IN THE
SHORT-TERM
IMPLEMENTATION IN THE METHANOL
(NONCANCER) ASSESSMENT
Evidence Identification: Literature Collection and Collation Phase (p. 164 of the NRC Report)
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, 1994). 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.
(2004b; 1999b) 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).
Standard protocols for evidence identification and
templates for describing the search approach are being
implemented as a part of Phase 2.
Evidence Evaluation: Hazard Identification and Dose-Response Modeling (p. 165 of the NRC Report)
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.
15. Develop templates for evidence tables,
forest plots, or other displays.
Partially Implemented. 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 methanol. The exposure-
response arrays inform the identification of doses
associated with specific effects and the weight-of-
evidence for those effects. The use of standardized tables
and graphics will be included in assessments that are part
of Phase 2 of the implementation process.
Not Implemented. Evidence table templates and
templates for other graphics are being implemented as a
part of Phase 2.
E-5
-------
NRC RECOMMENDATIONS THAT
EPA IS IMPLEMENTING IN THE
SHORT-TERM
16. Establish protocols for review of major
types of studies, such as epidemiologic and
bioassay.
IMPLEMENTATION IN THE METHANOL
(NONCANCER) ASSESSMENT
Partially Implemented. Formalized protocols for review of
studies will be developed as a part of Phase 2 and 3. 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 of the NRC Report)
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-
7. 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 of the NRC Report)
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.
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.
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.
Not applicable. A cancer unit risk estimate was not
derived in this assessment.
E-6
-------
NRC RECOMMENDATIONS THAT
EPA IS IMPLEMENTING IN THE
SHORT-TERM
IMPLEMENTATION IN THE METHANOL
(NONCANCER) 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.
E-7
-------
Table E-2. National Research Council recommendations that the EPA is generally
implementing in the long-term
NRC RECOMMENDATIONS THAT
THE EPA IS GENERALLY
IMPLEMENTING IN THE
LONG-TERM
IMPLEMENTATION IN THE METHANOL
(NONCANCER) ASSESSMENT
Weight-of-Evidence Evaluation: Synthesis of
Evidence for Hazard Identification (p. 165 of
the NRC Report)
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 announced (U.S.
EPA, 2012c) 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 held a
workshop on August 26, 2013, on issues related to weight-
of-evidence to inform future assessments.
Calculation of Reference Values and Unit Risks
(pp. 165-166 of the NRC Report)
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 RfCs and RfDs from 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.
E-8
-------
REFERENCES
Adanir. J: Ozkalkanti. MY: Aksun. M. (2005). Percutaneous methanol intoxication: Case report [Abstract]. Eur J
Anaesthesiol 22: 560-561.
Aims. 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/i.forsciint.2008.05.014.
Arora. V: Nijjar. IS: Thukral H: Roopa. (2005). Bilateral putaminal necrosis caused by methanol intoxication- a
case report. Neuroradiology 15: 341-342.
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.
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. (1997). Time-resolved cutaneous absorption and permeation rates of methanol in
human volunteers. Int Arch Occup Environ Health 70: 341-351.
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.
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.
R-1
-------
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.
http://dx.doi.org/10.1093/toxsci/64.2.169.
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: Hedhilj 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.
Burbacher. TM: Shen. D: Grant. K: Sheppard. L: Damian. D: Ellis. S: Liberate. N. (1999a). 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: Grant K: Shen. D: Damian. D: Ellis. S: Liberate. N. (1999b). 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: Shen. DP: Lalovic. B: Grant KS: Sheppard. L: Damian. D: Ellis. S: Liberate. N. (2004a).
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.
Burbacher. TM: Grant KS: Shen. DP: Sheppard. L: Damian. D: Ellis. S: Liberate. N. (2004b). 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.
Butchko. HH: Stargel. WW. (2001). Aspartame: scientific evaluation in the postmarketing period. Regul Toxicol
Pharmacol 34: 221-233.
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: 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.
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.
Clary. JJ. (2003). Methanol, is it a developmental risk to humans? [Review]. Regul Toxicol Pharmacol 37: 83-91.
http://dx.doi.org/10.1016/80273-2300(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.
Cook. MR: Bergman. FJ: Cohen. HP: Gerkovich. MM: Graham. C: Harris. RK: Siemann. LG. (1991). Effects of
methanol vapor on human neurobehavioral measures. (ISSN 1041-5505 Research Report Number 42).
Boston, MA: Health Effects Institute.
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.
COT (Committee on Toxicity). (2011). COT Statement on the effects of chronic dietary exposure to methanol.
http://cot.food. gov.uk/cotstatements/cotstatementsvrs/cotstatements201 l/cot201102.
R-2
-------
Davoli E: Cappellini. L: Airoldi L: Fanelli. R. (1986). Seram methanol concentrations in rats and in men after a
single dose of aspartame. Food ChemToxicol 24: 187-189.
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
J Pediatr 164: 158-161. http://dx.doi.org/10.1007/s00431-004-1588-5.
Dethlefs. R: Naraqi S. (1978). Ocular manifestations and complications of acute methyl alcohol intoxication.
MedJAust2:483-485.
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.
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.
Eng. MY: Luczak. SE: Wall TL. (2007). ALDH2, ADH1B, and ADH1C genotypes in Asians: a literature review
[Review]. Alcohol Res Health 30: 22-27.
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
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.
Gaul HP: Wallace. CJ: Auer. RN: Chen Fong. T. (1995). MRI findings in methanol intoxication. AJNR Am J
Neuroradiol 16: 1783-1786.
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. http://dx.doi.org/10.1006/rtph.2002.1540.
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.
R-3
-------
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: Lambermont. JY: Mahieu. P. (1997a). Methanol poisoning during late pregnancy. Clin Toxicol 35:
187-191.
Hantson. P: Duprez. T: Mahieu. P. (1997b). 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.
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: Tafiord. 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.
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. J
Pediatr 94: 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: Tajahmadv. 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.
Lee. EW: Terzo. TS: D'Arcv. JB: Gross. KB: Schreck. RM. (1992). Lack of blood formate accumulation in
humans following exposure to methanol vapor at the current permissible exposure limit of 200 ppm. Am Ind
Hyg Assoc J 53: 99-104. http://dx.doi.org/10.1080/15298669291359357.
Lev. CO: Gali. FG. (1983). Parkinsonian syndrome after methanol intoxication. EurNeurol 22: 405-409.
Liu. JJ: Dava. MR: Carrasquillo. O: Kales. SN. (1998). Prognostic factors in patients with methanol poisoning.
Clin Toxicol 36: 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.org/10.1016/i.Yrtph.201Q. 12.004.
R-4
-------
Magnuson. BA: Burdock. GA: Doull J: Kroes. RM: Marsh. GM: Pariza. MW: Spencer. PS: Waddell WJ:
Walker. R: Williams. GM. (2007). Aspartame: A safety evaluation based on current use levels, regulations,
and lexicological and epidemiological studies [Review]. Crit Rev Toxicol 37: 629-727'.
http://dx.doi.org/10.1080/10408440701516184.
Mahieu. P: Hassoun. A: Lauwerys. R. (1989). Predictors of methanol intoxication with unfavourable outcome.
Hum Toxicol 8: 135-137.
Makar. AB: Tephly. TR: Mannering. GJ. (1968). Methanol metabolism in the monkey. Mol Pharmacol 4: 471-
483.
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. 5 A-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.
NTP-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.
NTP-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.0rg/10.1016/j.reprotox.2003.10.013.
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/j.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. (1995). Comparative toxicokinetics of inhaled methanol in the female
CD-I mouse and Sprague-Dawley rat. Toxicol Sci 28: 245-254.
R-5
-------
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.
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 (Vol. NTIS). (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: 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.
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.
Rubinstein. D: Escort. E: Kelly. JP (1995). Methanol intoxication with putaminal and white matter necrosis: MR
and CT findings. AJNR Am J Neuroradiol 16: 1492-1494.
Sarkola. T: Eriksson. CJP. (2001). Effect of 4-methylpyrazole on endogenous plasma ethanol and methanol
levels in humans. Alcohol Clin Exp Res 25: 513-516.
Sasso. AF: Schlosser. PM: Kedderis. GL: Center. MB: Snawder. JE: Li. Z: Rieth. S: Lipscomb. JC. (2013).
Application of an updated physiologically based pharmacokinetic model for chloroform to evaluate
CYP2E1-mediated renal toxicity in rats and mice. Toxicol Sci 131: 360-374.
http://dx.doi.org/10.1093/toxsci/kfs320
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-1.
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.0rg/10.1016/i.reprotox.2012.ll.006.
R-6
-------
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/j.reprotox.2010.09.014.
Tran. 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.
Turner. C: Spanel P: Smith. D. (2006). A longitudinal study of methanol in the exhaled breath of 30 healthy
volunteers using selected ion flow tube mass spectrometry, SIFT-MS. Physiol Meas 27: 637-648.
http://dx.doi.Org/10.1088/0967-3334/27/7/007.
U.S. Congress. (2011). Consolidated Appropriations Act, 2012. (Pub. L. No. 112-74; 125 STAT. 786). 112th
U.S. Congress. http://www.gpo.gov/fdsvs/pkg/PLAW-l 12publ74/pdf/PLAW-l 12publ74.pdf
USDA (U.S. Department of Agriculture). (2010). Dietary guidelines for Americans, 2010 (7th ed.). Washington,
DC: U.S. Department of Health and Human Services. U.S. Government Printing Office.
http://www.health.gov/dietarvguidelines/2010.asp.
U.S. EPA (U.S. Environmental Protection Agency). (1994). 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 Retrieved July 13,
2010, from http://www.epa.gov/ncea/iris/subst/1001.htm
U.S. EPA (U.S. Environmental Protection Agency). (2000b). Science policy council handbook: Risk
characterization [EPA Report]. (EPA 100-B-00-002). Washington, D.C.: Office of Science Policy, Office of
Research and Development, http://www.epa.gov/osa/spc/pdfs/rchandbk.pdf.
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: Risk Assessment Forum,
U.S. Environmental Protection Agency, 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. http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=157668.
U.S. EPA (U.S. Environmental Protection Agency). (2006b). Peer review handbook (3rd edition) [EPA Report].
(EPA/100/B-06/002). Washington, DC. http://www.epa.gov/peerreview/.
U.S. EPA (U.S. Environmental Protection Agency). (2009). Benchmark dose software (BMDS) Retrieved
December 19, 2009, from http://www.epa.gov/NCEA/bmds
U.S. EPA (U.S. Environmental Protection Agency). (201 la). Addendum to the March 2011 external peer review
draft IRIS lexicological review of methanol (non-cancer) [EPA Report]. (EPA/635/R-11/001). Washington,
D.C. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=233771
R-7
-------
U.S. EPA (U.S. Environmental Protection Agency). (201 Ib). Benchmark Dose Software (BMDS) (Version 2.2
R65 [Build: 04/13/2011]) [Computer Program]. 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). (2011c). Toxicological Review of Methanol (Non-Cancer)
(External Review Draft) [EPA Report]. (EPA/635/R-11/001 A). Washington, D.C.
http ://cfpub .epa. gov/ncea/cfm/recordisplav. cfm?deid=23 3 771
U.S. EPA (U.S. Environmental Protection Agency). (2012b). Methanol PBPK Model [PBPK].
U.S. EPA (U.S. Environmental Protection Agency). (2012c). EPA announces NAS' review of IRIS Assessment
development process. Available online at
http://vosemite.epa.gOv/opa/admpress.nsf/0/lce2a7875daf093485257a000054df547OpenDocument
Vara-Castrodeza. A: Perez-Castrillon. JL: Duenas-Laita. A. (2007). Magnetic resonance imaging in methanol
poisoning. ClinToxicol 45: 429-430. http://dx.doi.org/10.1080/15563650701285313.
Ward. KW: Perkins. RA: Kawagoe. JL: Pollack. GM. (1995). Comparative toxicokinetics of methanol in the
female mouse and rat. Toxicol Sci 26: 258-264.
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.
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/i.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/ipepsv/ish067.
Woo. YS: Yoon. SJ: Lee. HK: Lee. CU: Chae. JH: Lee. CT: Kim. DJ. (2005). Concentration changes of methanol
in blood samples during an experimentally induced alcohol hangover state. Addict Biol 10: 351-355.
http://dx.doi.org/10.1080/13556210500352543.
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.
R-8
------- |