EPA/635R-00/004
TOXICOLOGICAL REVIEW
OF
VINYL CHLORIDE
(CAS No. 75-01-4)
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
May 2000
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
This document may undergo revisions in the future. The most up-to-date version will be
made electronically via the IRIS Home Page at http://www.epa.gov/iris.
11
-------
CONTENTS—TOXICOLOGICAL REVIEW FOR VINYL CHLORIDE
(CAS No. 75-01-4)
FOREWORD v
AUTHORS, CONTRIBUTORS, AND REVIEWERS vi
1. INTRODUCTION 1
2. CHEMICAL AND PHYSICAL INFORMATION RELEVANT TO ASSESSMENTS 2
3. TOXICOKINETICS/TOXICODYNAMICS RELEVANT TO ASSESSMENTS 2
4. HAZARD IDENTIFICATION 5
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY, CASE REPORTS, CLINICAL
CONTROLS 5
4.1.1. Cancer Effects 5
4.1.2. Noncancer Effects 12
4.2. PRECHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN
ANIMALS—ORAL AND INHALATION 14
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES-
ORAL AND INHALATION 20
4.4. OTHER STUDIES 23
4.4.1. Neurological 23
4.4.2. Genotoxicity 23
4.4.3. Noncancer Mechanism 26
4.5. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS AND
MODE OF ACTION 27
4.6. WEIGHT-OF-EVIDENCE EVALUATION AND CANCER
CHARACTERIZATION 28
4.7. SUSCEPTIBLE POPULATIONS 29
4.7.1. Possible Childhood Susceptibility 29
4.7.2. Possible Gender Differences 35
5. DOSE-RESPONSE ASSESSMENTS 36
5.1. ORAL REFERENCE DOSE (RfD) 36
5.1.1. Choice of Principal Study and Critical Effect 36
5.1.2. Methods of Analysis—Including Models (PBPK, BMD, etc.) 37
5.1.3. RfD Derivation 40
5.2. INHALATION REFERENCE CONCENTRATION (RfC) 41
5.2.1. Choice of Principal Study and Critical Effect 41
5.2.2. Methods of Analysis—Including Models (PBPK, BMC, etc.) 42
5.3. CANCER ASSESSMENT 44
5.3.1. Choice of Study/Data With Rationale and Justification 44
5.3.2. Dose-Response Data 47
iii
-------
CONTENTS (continued)
5.3.3. Dose Conversion 47
5.3.4. Extrapolation Method(s) 50
5.3.5. Oral Slope Factor and Inhalation Unit Risk 50
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND DOSE
RESPONSE 57
6.1. HUMANHAZARD POTENTIAL 57
6.1.1. Hazard Identification for Cancer Effects 57
6.1.2. Hazard Identification for Noncancer Effects 58
6.2. DOSE RESPONSE 58
6.2.1. Dose Response for Cancer Effects 59
6.2.2. Dose Response for Noncancer Effects 61
7. REFERENCES 64
APPENDIX A. Comparison of PBPK Models for Vinyl Chloride A-l
APPENDIX B. The Development and Validation of a PBPK Model for
Vinyl Chloride (VC) and its Application in a Carcinogenic Risk Assessment B-l
APPENDIX C. Vinyl Chloride PBPK Model Code (ACSL Version: VCPBPK.CSL) C-l
APPENDIX D. The Application of a PBPK Model for Vinyl Chloride in a
Noncancer Risk Assessment D-l
APPENDIX E. External Peer Review—Summary of Comments and Disposition E-l
IV
-------
FOREWORD
The purpose of this Toxicological Review is to provide scientific support and rationale
for the hazard and dose-response assessment in IRIS pertaining to chronic exposure to vinyl
chloride. It is not intended to be a comprehensive treatise on the chemical or toxicological
nature of vinyl chloride.
In Section 6, EPA has characterized its overall confidence in the quantitative and
qualitative aspects of hazard and dose response. Matters considered in this characterization
include knowledge gaps, uncertainties, quality of data, and scientific controversies. This
characterization is presented in an effort to make apparent the limitations of the assessment and
to aid and guide the risk assessor in the ensuing steps of the risk assessment process.
For other general information about this assessment or other questions relating to IRIS,
the reader is referred to EPA's Risk Information Hotline at 202-566-1676.
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
Chemical Manager/Author
William E. Pepelko, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Co-Author
Gary L. Foureman, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Additional Contributors
Vincent J. Cogliano, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Reviewers
This document and summary information on IRIS have received peer review both by
EPA scientists and by independent scientists external to EPA. Subsequent to external review
and incorporation of comments, this assessment has undergone an Agency-wide review process
whereby the IRIS Program Manager has achieved a consensus approval among the Office of
Research and Development; Office of Air and Radiation; Office of Prevention, Pesticides, and
Toxic Substances; Office of Solid Waste and Emergency Response; Office of Water; Office of
Policy, Planning, and Evaluation; and the Regional Offices.
Internal EPA Reviewers
Robert Bellies, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Jerry Blancato, Ph.D.
VI
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
National Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Vanessa Vu, Ph.D.
Office of Pollution Prevention and Toxics
U.S. Environmental Protection Agency
Washington, DC
External Peer Reviewers
James J. Beaumont, Ph.D.
University of California
Davis, CA
Gunther Craun, S.M., MPH
Gunther Craun & Associatess
Staunton, VA
Michael L. Dourson, Ph.D.
Toxicology Excellence for Risk Assessment
Cincinnati, OH
Patrick R. Durkin, Ph.D.
Syracuse Environmental Research Associates, Inc.
Fayetteville, NY
Victor J. Feron, Ph.D.
TNO Nutrition and Food Research Institute
Utrechtseweg, The Netherlands
Clay Frederick, Ph.D., DABT
Rohm and Haas Company
Spring House, PA
Jay Gandy, Ph.D.
Center for Toxicology & Environmental Health, L.L.C.
Little Rock, AR
Michael L. Gargas, Ph.D., DABT
Chem Risk
Cleveland, OH
Dawn G. Goodman, V.M.D.
vn
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
Private Consultant
Potomac, MD
Bryan D. Hardin, Ph.D.
National Institute for Occupational Safety and Health
Washington, DC
Gregory L. Kedderis, Ph.D.
Chemical Industry Institute of Toxicology
Research Triangle Park, NC
Nancy K. Kim, Ph.D.
NYS Department of Health
Albany, NY
Norbert P. Page, Ph.D.
Page Associates
Gaithersburg, MD
Colin Park, Ph.D.
The Dow Chemical Company
Midland, MI
Chris Portier, Ph.D.
NIEHS
Res Triangle Park, NC
Richard H. Reitz, Ph.D., DABT
RHR Toxicology Consulting
Midland, MI
Carlo H. Tamburro, M.D., M.P.H.
University of Louisville
Louisville, KY
Elizabeth Weisburger, Ph.D.
Private Consultant
Rockville, MD
Sharon B.Wilbur
Agency for Toxic Substances and Disease Registry
U.S. Department of Health and Human Services
Atlanta, GA
Vlll
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
Summaries of the external peer reviewers' comments and the disposition of their
recommendations are in Appendix E.
IX
-------
1. INTRODUCTION
This document presents background and justification for the hazard and dose-response
assessment summaries in EPA's Integrated Risk Information System (IRIS). IRIS summaries
may include an oral reference dose (RfD), inhalation reference concentration (RfC), and a
carcinogenicity assessment.
The RfD and RfC provide quantitative information for noncancer dose-response
assessments. The RfD is based on the assumption that thresholds exist for certain toxic effects
such as cellular necrosis, but may not exist for other toxic effects, such as some carcinogenic
responses. It is expressed in units of mg/kg-day. In general, the RfD is an estimate (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 noncancer effects during a lifetime. The inhalation RfC is analogous to the oral RfD,
but provides a continuous inhalation exposure estimate. The inhalation RfC considers toxic
effects for both the respiratory system (portal-of-entry) and for effects peripheral to the
respiratory system (extrarespiratory or systemic effects). It is generally expressed in units of
mg/m3.
The carcinogenicity assessment provides information on the carcinogenic hazard
potential of the substance in question and quantitative estimates of risk from oral exposure and
inhalation exposure. The information includes a weight-of-evidence judgment of the likelihood
that the agent is a human carcinogen and the conditions under which the carcinogenic effects
may be expressed. Quantitative risk estimates are presented in three ways. The slope factor is
the result of application of a low-dose extrapolation procedure and is presented as the risk per
mg/kg/day. The unit risk is the quantitative estimate in terms of either risk per |ig/L drinking
water or risk per |ig/m3 air breathed. Another form in which risk is presented is a drinking water
or air concentration providing cancer risks of 1 in 10,000; 1 in 100,000; or 1 in 1,000,000.
Development of these hazard identification and dose-response assessments for vinyl
chloride has followed the general guidelines for risk assessment as set forth by the National
Research Council (1983). EPA guidelines that were used in the development of this assessment
may include the following: the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 1986a),
Guidelines for the Health Risk Assessment of Chemical Mixtures (U.S. EPA, 1986b), Guidelines
for Mutagenicity Risk Assessment (U.S. EPA, 1986c), Guidelines for Developmental Toxicity
Risk Assessment (U.S. EPA, 1991), Proposed Guidelines for Neurotoxicity Risk Assessment
(U.S. EPA, 1995a), Guidelines for Neurotoxicity Risk Assessment (M.S. EPA, 1998b), Proposed
Guidelines for Carcinogen Risk Assessment (1996a), and Reproductive Toxicity Risk Assessment
Guidelines (U.S. EPA, 1996b); Recommendations for and Documentation of Biological Values
for Use in Risk Assessment (U.S. EPA, 1988); (proposed) Interim Policy for Particle Size and
Limit Concentration Issues in Inhalation Toxicity (U.S. EPA, 1994a); Methods for Derivation of
Inhalation Reference Concentrations and Application of Inhalation Dosimetry (U.S. EPA,
1994b); Peer Review and Peer Involvement at the U.S. Environmental Protection Agency (U.S.
EPA, 1994c); Use of the Benchmark Dose Approach in Health Risk Assessment (U.S. EPA,
1995b); Science Policy Council Handbook: Peer Review (U.S. EPA, 1998a); and memorandum
1
-------
from EPA Administrator, Carol Browner, dated March 21, 1995, Subject: Guidance on Risk
Characterization (U.S. EPA, 1995c).
Literature search strategies employed for this compound were based on the CASRN and
at least one common name. At a minimum, the following databases were searched: RTECS,
HSDB, TSCATS, CCRIS, GENETOX, EMIC, EMICBACK, DART, ETICBACK, TOXLINE,
CANCERLINE, MEDLINE, and MEDLINE backfiles. Any pertinent scientific information
submitted by the public to the IRIS Submission Desk was also considered in the development of
this document.
2. CHEMICAL AND PHYSICAL INFORMATION RELEVANT TO ASSESSMENTS
Common synonyms of vinyl chloride (VC) include chloroethene, chloroethylene,
ethylene monochloride, and monochloroethene. Some relevant physical and chemical properties
of VC are listed below (Sax and Lewis, 1989):
CASRN: 75-01-4€
Empirical formula: C2H3C1€
Structural formula: CH2 = CHC1€
Molecular weight: 62.5€
Vapor pressure: 2,660 mm Hg at 25°C€
Water solubility: 2,763 mg/L (U.S. EPA, 1985); 1,100 mg/L (Cowfer and Magistro,€
1983)
LogKow: 1.36(NIOSH, 1986)
Conversion factor: 1 ppm = 2.60 mg/m3, 1.0 mg/m3 = 0.39 ppm
VC is a synthetic chemical used as a chemical intermediate in the polymerization of
polyvinyl chloride. At room temperature and pressure, it is a colorless gas with a mild, sweet
odor. As the data shown above indicate, VC is poorly soluble in water. Structurally, VC is a
haloalkene and is related to vinylidene chloride and trichloroethylene. In the following pages
VC refers to the monomer and PVC to polyvinylchloride, the polymerized form.
3. TOXICOKINETICS/TOXICODYNAMICS RELEVANT TO ASSESSMENTS
Human and animal data indicate that VC is rapidly and efficiently absorbed via the
inhalation and oral routes, is rapidly converted to water-soluble metabolites, and is rapidly
excreted. At low concentrations, VC metabolites are excreted primarily in urine, while at high
exposure concentrations, unchanged VC is also eliminated in exhaled air. Overall, the data
indicate that neither VC nor its metabolites are likely to accumulate in the body.
Absorption of VC in humans after inhalation exposure is rapid. A study conducted in
five young adult male volunteers inhaling VC at concentrations of 7.5 to 60 mg/m3 showed that
42% was retained, maximum retention was reached within 15 minutes, and the percent retention
was independent of inspired VC concentration. Individual variation, however, was high, with
2€
-------
mean retention values after 6 hours exposure to 30 mg VC/m3 ranging from 30% to 71%. After
cessation of exposure, the VC concentration in expired air decreased rapidly within 30 minutes to
4% of the inhaled concentration (Krajewski et al., 1980). Animal inhalation studies also show
that VC is rapidly absorbed. Exposure of male Wistar rats (number/group unspecified) to 1,000,
3,000, or 7,000 ppm VC (99.9% pure) for 5 hours using a head-only apparatus resulted in rapid
uptake into the blood, as measured by gas-liquid chromatography (GLC) (Withey, 1976).
Equilibrium blood levels were achieved within 30 minutes for all exposures. Upon cessation of
exposure, blood levels declined to a barely detectable level after 2 hours. Rat studies show that
the distribution of VC is rapid and widespread, but the storage of VC in the body is limited by its
rapid metabolism and excretion (Bolt et al., 1977).
No human studies of absorption of ingested VC were located. Animal studies show that
VC absorption following oral exposure is rapid and complete. Peak blood levels were reached
within 10 minutes when VC was administered to male rats by gavage in an aqueous solution at
doses up to 92 mg/kg (Withey, 1976). In the same study, more complex and slightly delayed
absorption was observed following VC gavage in oil, although peak blood levels were reached
within 40 minutes (Withey, 1976). At 72 hours after a single gavage dose of 100 mg/kg VC in
oil, unmetabolized VC was detected in exhaled air, indicating that metabolism was saturated
(Watanabe and Gehring, 1976; Watanabe et al., 1976a). Saturation of VC metabolism has also
been observed following inhalation exposure (Watanabe and Gehring, 1976; Watanabe et al.,
1976b). In rats fed VC monomer in a PVC powder, the average amount of VC detected in feces
was 8%, 10%, and 17% for oral intake of 2.3, 7.0, and 21.2 mg/kg-day (Feron et al., 1981).
Because the remaining material was reported as still enclosed in PVC granules, free VC
monomer is considered nearly, if not completely, absorbed in the GI tract. In using this study for
quantitating risk, as is done in this assessment, dose was considered to be the amount ingested
minus that recovered in the feces. Complete absorption is assumed for humans ingesting VC
monomer.
Numerous studies on the pharmacokinetics and metabolism of VC have been conducted,
with the majority of these studies conducted in rats (Withey, 1976; Hefner et al., 1975;
Guengerich and Watanabe, 1979; Bolt et al., 1976, 1977; Watanabe et al., 1976a,b, 1978;
Jedrychowski et al., 1984, 1985; Tarkowski et al., 1980). As discussed in Sections 5.1.2, 5.2.2,
and 5.3.3, both the cancer and noncancer assessments were conducted using a physiologically
based pharmacokinetic (PBPK) model (Clewell et al., 1995a,b) in which VC metabolism was
hypothesized to occur via two saturable pathways. Therefore, VC metabolism is discussed in
some detail here as part of the background for the development of the model. A simplified
diagram of the metabolism of VC is shown in Figure 1. The primary route of metabolism of VC
is by the action of cytochrome P450 or CYP on VC to form chloroethylene oxide (Bolt et al.,
1977; Plugge and Safe, 1977). Chloroethylene oxide (CEO) is a highly reactive, short-lived
epoxide, some of which rapidly rearranges to form chloroacetaldehyde (CAA), a reactive • -
halocarbonyl compound; CEO is also a substrate for epoxide hydrolase (Pessayre et al., 1979).
These two metabolites are detoxified mainly via glutathione (GSH) conjugation
(Jedrychowski et al., 1985; Leibman, 1977; Tarkowski et al., 1980). This hypothesis is
supported by the observation of decreased nonprotein sulfhydryl concentrations at high VC
-------
Vinyl
Chloride
P450
CO2
H2O
^
Chloroethylene
Epoxide
w
DNA
Adducts
GSH
Epoxide
Hydrolase
Glutathione
Conjugates
GSH
Chloroacetaldehyde
w
Tissue
Adducts
Figure 1. Metabolism of vinyl chloride.
exposure concentrations (Jedrychowski et al., 1985; Tarkowski et al., 1980), as well as by the
excretion of GSH-conjugated metabolites in the urine, observed in rats following exposure to VC
(Watanabe et al., 1976c; Hefner et al., 1975). CAA may also combine directly or enzymatically
with GSH via glutathione transferase (GST) to form S-formylmethylglutathione. S-
formylmethylglutathione, through direct interaction with GSH-derived cysteine, can be excreted
as N-acetyl-S-(2-hydroxyethyl)cysteine, another major urinary metabolite of VC (Green and
Hathway, 1975). The GSH conjugates are then subject to hydrolysis, resulting in excretion of
cysteine conjugates in the urine (Hefner et al., 1975). Two of the three major urinary
metabolites of VC in rats have been identified as N-acetyl-S-(2-hydroxyethyl)cysteine and
thiodiglycolic acid (Watanabe et al., 1976b).
The specific isozymes of the P450 system involved in the metabolism of VC have not yet
been unequivocally established. However, it is clear from both in vitro and in vivo studies that
several isozymes can play a role. High-affinity, low-capacity oxidation by CYP2E1 is probably
responsible for essentially all of the metabolism of VC at low concentrations in uninduced
animals and humans (Guengerich et al., 1991). There is also evidence for a significant increase
in metabolism in animals pretreated with phenobarbital (Ivanetich et al., 1977), suggesting that
CYP2B1 also metabolizes VC. At high concentrations in vivo, the metabolism of VC in rats
leads to a destruction of P450 enzyme (Reynolds et al., 1975), which is greatly enhanced in
phenobarbital- or Aroclor-induced animals (Aroclor induces CYP1A2). The loss of P450 has
been suggested to result from the production of reactive intermediates during the metabolism of
VC (Guengerich and Strickland, 1977) and is inhibited by GSH in vitro (Ivanetich et al., 1977).
-------
Induction of P450 by phenobarbital or Aroclor was also necessary to produce acute
hepatotoxicity from VC in rats (Jaeger et al., 1977), indicating that VC toxicity is increased by
increased P450 activity.
The contribution of several P450 isozymes to the metabolism of the related compound
trichloroethylene (TCE) has been studied in the male Wistar rat and male B6C3F1 mouse
(Nakajima et al., 1993). Using monoclonal antibodies specific to each isozyme, the investigators
were able to determine that CYP2E1 contributes more to the metabolism of TCE in mice than in
rats, whereas CYP2C11/6, a constitutive, noninducible isozyme present only in male rodents,
contributes more to the metabolism of TCE in rats than in mice. The investigators also found
that CYP1A1/2 contributes to the uninduced metabolism of TCE in mice but not in rats and that
CYP2B1 does not contribute to the metabolism of TCE in naive animals of either species. Thus,
assuming that the same isozymes are responsible for metabolism of TCE and VC, it appears that
at low concentrations the initial metabolism of VC is primarily due to CYP2E1, but that at
higher concentrations, where CYP2E1 becomes capacity limited, other CYP isozymes may
contribute to its metabolism. The extent of this higher capacity metabolism is likely to vary
across animal species, strain, and sex. To the extent that such higher capacity, lower affinity
metabolism (referred to henceforth as "non-2El" metabolism) may be important in conducting a
risk assessment for VC, it will have to be characterized separately for each species, strain, and
sex of interest. From a pharmacokinetic modeling perspective, non-2El metabolism would be
handled as a second saturable metabolic pathway with a larger concentration for the Michaelis-
Menten constant (KM). For example, it has been demonstrated that the metabolism of another
related compound, vinyl bromide, is best described with two distinct saturable pathways having
different affinities (Gargas and Andersen, 1982). Of major importance for human risk
assessment, some of the low-affinity, high-capacity constitutive (2C11/6) and inducible (2B1/2)
P450 isozymes in the rodent may have no human correspondents (Guengerich, 1987).
Reflecting the dose-dependent, saturable nature of VC metabolism, the route and nature
of VC elimination is also dose related (Green and Hathway, 1975; Bolt, 1978; Hefner et al.,
1975; Gehring et al., 1978). Following exposure via oral or inhalation routes to low doses of
VC, metabolites are excreted primarily in the urine. However, once the saturation point for
metabolism is reached, VC is eliminated via other routes, primarily exhalation of the parent
compound (Watanabe et al., 1976b; Watanabe and Gehring, 1976). The route of elimination of
VC also depends on the route of administration. Urinary excretion is favored more following
oral or intraperitoneal administration, while 99% of the same dose administered intravenously
was exhaled (Bolt, 1978). This may be the result of a high peak concentration with intravenous
administration, combined with a relatively low blood-to-air partition coefficient, resulting in
elimination from the blood via the lungs before a significant amount of urinary clearance can
occur.
4. HAZARD IDENTIFICATION
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY, CASE REPORTS, CLINICAL
CONTROLS
4.1.1. Cancer Effects
-------
Several independent retrospective and prospective cohort studies demonstrate a
statistically significant elevated risk of liver cancer, primarily angiosarcomas, a neoplasm arising
from vascular endothelial cells in the liver, but in many cases hepatocellular carcinoma (a
neoplasm arising from epithelial cells in the liver) as well, from exposure to VC (Byren et al.,
1976; CMA et al., 1998a; Du and Wang, 1998; Fox and Collier, 1977; Jones et al., 1988;
Monson et al., 1975; Pirastu et al., 1990, 1998; Simonato et al., 1991; Tabershaw and Gaffey,
1974; Waxweiler et al., 1976; Weber et al., 1981; Wong et al., 1991; Wu et al., 1989). Although
Duck et al. (1975) failed to find a significant increase in liver cancer, they did report one case of
liver angiosarcoma. The possible association of brain soft tissue and nervous system cancer with
VC exposure was also reported in some studies (Byren et al., 1976; CMA et al., 1998a; Cooper,
1981; Tabershaw and Gaffey, 1974; Waxweiler et al., 1976; Weber et al., 1981; Wong et al.,
1991; Wu et al., 1989), although it should be noted that four of these studies are based upon the
same cohort, which has been updated periodically. Several studies have found an association
between VC exposure and cancer of the hematopoietic and lymphatic systems (Simonato et al.,
1991; Weber et al., 1981). Observed increases in other studies fell below statistical significance
because of the small numbers of these types of cancers (Tabershaw and Gaffey, 1974). VC
exposure has also been associated with lung cancer (Buffler et al., 1979; Monson et al., 1975;
Waxweiler et al., 1976); however, the evidence is weaker than for liver cancer and may be due to
inhalation of PVC. Ott et al. (1975) reported an increase in deaths due to all malignancies,
although none of them were due to angiosarcoma. An excess of melanoma was reported in one
study (Heldaas et al., 1984), but other studies have not substantiated this report.
The first report of an association between exposure to VC and cancer in humans was
published by Creech and Johnson (1974): three cases of liver angiosarcoma were reported in
men employed in a PVC plant. Angiosarcoma of the liver is considered to be a very rare type of
cancer, with only 20-30 cases per year reported in the United States (Gehring et al., 1978;
ATSDR, 1995). As described in the following paragraphs, greater than expected incidences of
angiosarcoma of the liver have since been reported in a number of other cohorts of workers
occupationally exposed to VC.
In a proportionate mortality study analyzing the causes of death of 142 workers exposed
to VC or VC/PVC, Monson et al. (1975) found an excess incidence of liver cancer (8 observed
vs. 0.7 expected). Five of these were angiosarcomas. The study also found an excess of brain
cancer (5 observed vs. 1.2 expected) and lung cancer (13 observed vs. 7.9 expected); all three of
the brain tumors for which the type was identified were glioblastoma multiformae. No statistical
analysis was conducted by tumor target.
Byren et al. (1976) reported a significantly elevated risk of pancreas/liver cancer (4
observed vs. 0.97 expected) in a cohort of 750 Swedish workers exposed to VC. Two of the four
were identified as angiosarcomas of the liver only after reevaluation. The excess risk increases
when latency is considered. The expected number of deaths was 0.68 for a latency period of >10
years, whereas all 4 observed deaths were exposed earlier than 10 years before death. This study
also found a small excess of brain cancer (2 observed vs. 0.33 expected).
Waxweiler et al. (1976) found a significantly elevated risk (7 observed vs. 0.6 expected)
of liver cancer in a cohort of 1,294 workers exposed to VC for a minimum of 5 years and
followed for 10 or more years. In a separate phase of the study, the authors identified 14 cases
-------
of liver and biliary cancer, 11 of which were angiosarcomas. Several of the identified subjects
were not included in the main study because they were still alive, or because they did not meet
the minimum criteria for inclusion in the cohort. Brain cancer incidence was significantly
increased in workers observed for 15 years or more after initial exposure (3 observed vs. 0.6
expected); a nonsignificant increase was observed for a 10-year latency. An additional seven
cases of brain cancer were identified in subjects who did not qualify for inclusion in the cohort
study. Nine of the 10 brain cancers were glioblastoma multiforme; a histological analysis was
not available for the tenth. By contrast, the study authors stated that this distribution of cell type
typically occurs in only 33% of brain cancer deaths. The cohort study also found a slight excess
risk of lymphatic and hematopoietic system cancer (4 observed vs. 2.5 expected). Of the 14
cases of primary lung cancer identified, 5 were large cell undifferentiated, 3 were
adenocarcinomas, and there were no squamous cell or small cell bronchiogenic carcinomas,
suggesting that these cancers were not associated with smoking. In a later study of 4,806
workers at the same plants, for workers exposed to polyvinyl PVC dust and several other
chemicals, but not VC, an elevated risk of lung cancer was found (Waxweiler et al., 1981). The
study authors considered PVC to be the likely etiologic agent inducing lung cancer. While the
association with PVC dust could have been due to VC trapped in the dust, this did not explain
the fact that exposure to VC alone was not associated with lung cancer in their study.
While a large number of occupational studies reported an association between VC and
liver angiosarcoma, quantitative exposure information is available for only a few studies. Fox
and Collier (1977) reported four cases of liver cancer, two of which were angiosarcomas, in a
cohort oil,111 British VC workers. The study authors grouped the subjects by estimated
exposure levels and exposure duration. From these data, average exposure levels have been
estimated as 12.5, 70, and 300 ppm (Clement Associates, 1987) or 11, 71, and 316 ppm (Chen
and Blancato, 1989). Because workers were classified based on the maximum exposure for each
worker, cumulative exposure is overestimated, leading to a probable underestimation of risk
using these data. Both angiosarcoma cases were considered to have had high exposure to VC at
the level of 200 ppm and above time-weighted average. There was no effect on other cancers in
comparison with cancer rates in England and Wales. In a follow-up study, Jones et al. (1988)
analyzed mortality in 5,498 male VC workers. This study found a significant excess of primary
liver tumors, with 11 deaths, 7 of which were angiosarcomas. The median latency for
angiosarcomas was 25 years.
Weber et al. (1981) examined mortality patterns in 7,021 German and Austrian VC/PVC
workers and 4,007 German PVC processing workers. Comparisons were with West German
population death rates. A significantly elevated risk of liver cancer (12 observed vs. 0.79
expected) was observed in the VC/PVC cohort, but a significant increase (4 observed vs. 1
expected) was also observed in an unexposed reference group. However, the risk in the VC
cohort increased with exposure duration. The study authors implied that four cases of
angiosarcoma were identified in the study cohort, although it was not clear if all of the cases
belonged to this cohort. A significant excess risk of brain cancer (Obs = 5, SMR = 535,
p < 0.05) was also observed in the PVC processing workers, but not in VC/PVC workers. Risk
of lymphatic and hematopoietic cancer (Obs =15, SMR = 214) was significantly increased in
VC/PVC production workers, and there was a tendency for increased risk at longer exposure
durations.
-------
In a preliminary mortality follow-up study of 464 workers at an Italian VC production
facility, a significant excess of respiratory cancers was observed (Obs = 5, SMR = 289, p<
0.03). The excess remained after correction for smoking and was associated with longer
exposure durations and higher exposure levels (Belli et al., 1987). A significant excess of lung
cancer was also noted in a cohort of 437 VC/PVC workers.
Smulevich et al. (1988) investigated a cohort of 3,232 workers (2,195 men, 1,037
women) in a Soviet VC/PVC chemical plant. No cases of angiosarcoma or other liver tumors
were reported. Workers who were highly exposed to VC (> 300 mg/m3) had a significantly
elevated risk of lymphomas and leukemias (apparently 7 observed vs. about 1.1 expected for
combined men and women, but there are inconsistencies in the reported numbers). The risk of
brain cancer was elevated in women (Obs = 2, SMR = 500), but the effect was not statistically
significant and the incidence in men was unaffected. This is the only study to date that included
a significant number of females in the cohort. It is of interest that, of the 19 malignancies
reported in this cohort, none were mammary tumors, although mammary cancer increases were
found in some of the animal bioassays of VC.
Simonato et al. (1991) reported on the results of a large multicentric cohort study of
12,706 VC/PVC workers in European plants. A significant increase in liver cancer deaths was
observed (Obs = 24, SMR = 286). Workers were classified based on maximum exposure level
into ranges of < 50 ppm, 50-499 ppm, and • 500 ppm. Estimating an average exposure duration
of 9 years, average exposure levels for these groups can be estimated at 25, 158, and 600 ppm.
Histopathology was available for 17 of the liver cancers; 16 were confirmed as angiosarcoma
and 1 was a primary liver cancer. The excess risk from liver cancer was related to the time since
first exposure, duration of exposure, and estimated total exposure. A nonsignificant increase of
lymphosarcoma was observed (SMR = 170, 95% CI = 69-351). While there appeared to be a
small positive trend with increasing rank of exposure, there was no relationship to duration of
employment. Brain cancer had an elevated risk in certain analyses, but there was no clear
relationship to exposure duration; there was no excess risk of lung cancer.
Lelbach (1996) reported on the course of VC-induced disease in 21 PVC production
workers. Death was due to liver cancer in 19 of these cases. While the predominant tumor type
was angiosarcoma, hepatocellular and colangiocellular carcinoma were also found. Latency
periods ranged from 12 to 34 years, with a mean of 22 years. Younger age at first exposure,
younger than 27 years, seemed to have been accompanied by shorter latency periods.
Lee et al. (1996) described the time course and pathology of 20 patients who died from
angiosarcoma of the liver after occupational exposure to VC in Great Britain. Exposure periods
ranged from 3 to 29 years, with tumors developing after 9 to 35 years from beginning of
exposure.
The annual incidence of angiosarcoma of the liver in Great Britain from all sources was
estimated to be about 1.4 cases per 10 million population (Elliot and Kleinschmidt, 1997). Of 10
cases that were confirmed as angiosarcomas by histological analysis, 9 were VC workers. The
other individual was employed at a VC factory, although not as a VC worker. Since even this
individual could be presumed to have some exposure to VC, it was concluded that there were no
-------
confirmed nonoccupationally exposed cases of angiosarcoma among residents living near a VC
site in Great Britain.
Pirastu et al. (1990) evaluated clinical, pathological, and death certificate data for 63
deaths in three VC/PVC manufacturing or PVC extruding plants in Italy. Fourteen deaths from
primary liver cancer were found, seven of which were identified as angiosarcoma and two of
which were hepatocellular carcinoma. No comparison to a control population was conducted.
However, the authors stated that this study indicated a relationship between VC exposure and
primary liver cancer, as well as with angiosarcoma.
In an update of this cohort, Pirastu et al. (1998) evaluated cause-specific mortality rates
among male workers employed in VC manufacture and polymerization in the three Italian
plants: Ferrara, Rosignano, and Ravenna. The cohorts included all workers hired between start
of operation and 1985, 1978, and 1985, respectively, amounting to 418, 206, and 635 subjects
followed up for mortality until 1996 (Ferrara and Rosignano) and 1997 (Ravenna). The study
detected an increased mortality for primary liver cancer in all three plants; SMR values were 444
in Ferrara (4 Obs. 90% CI = 160-1069), 200 in Rosignano (1 Obs. 90% CI = 10-869), and 375 in
Ravenna (3 Obs. 90% CI = 108-390). In one plant, Ferrara, observed mortality was also above
expected for lung cancer, SMR = 146 (14 Obs. 90% CI = 89-229) and for larynx cancer, SMR =
500 (4 Obs. 90% CI = 174-1167). The possibility that lung cancer induction was caused by
PVC, however, could not be ruled out.
Du and Wang (1998) reported morbitiy odds ratio (MOR) for 2,224 workers with
occupational exposure to VC in Taiwan. A significantly increased risk of hospital admission
among VC workers due to primary liver cancer (MOR 4.5-6.5), cirrhosis of the liver (MOR 1.7-
2.1), and other chronic diseases (MOR 1.5-2.0) was found. There were eight cases of primary
liver cancer, all with heavy previous exposure to VC. Another four cases of liver cancer in PVC
workers were found in the death registry. Ten of 11 cases of liver cancer with detailed medical
information were carriers of hepatitis B virus. Of the 11 cases of liver cancer, four were
confirmed to be hepatocellular carcinoma by histology. Two others had extremely high
concentrations of a-fetoprotein, an indicator of hepatocellular carcinoma. The diagnosis of the
remaining six cases was uncertain.
In a preliminary report with only 85% follow-up completed, Tabershaw and Gaffey
(1974) compared mortality in a cohort of 8,384 men occupationally exposed to VC with death
rates among U.S. males. Each VC plant classified workers as exposed to high, medium, or low
levels of VC, but no quantitative estimate of exposure was provided, and no attempt was made to
establish consistent gradations of exposure between plants or exposure periods. No significant
increases in any general cancer classification were found. However, six cases of angiosarcoma
identified by other investigators occurred in the study population; only two of these were
identified as angiosarcomas on the death certificate. The study authors also noted that 6 of 17
(40%) deaths in the category "other malignancies" were due to brain cancer. The authors stated
that only 22% of the deaths in this category would be expected to be due to this cause, but they
did not provide any supporting documentation. This preliminary report also noted a slight
excess risk of lymphomas (5 observed vs. 2.54 expected) in the group with the higher exposure
index.
-------
Cooper (1981) enlarged the Tabershaw and Gaffey (1974) study to include 10,173 VC
workers; vital status was ascertained for 9,677 men. Cooper noted that, of the nine
angiosarcomas identified in the United States during the study period, eight were included in the
study cohort. Statistical analyses were conducted for broad categories of tumors; a significant
increase (Obs = 12, SMR = 203, p < 0.05) was observed for brain and central nervous system
malignancies.
Wu et al. (1989) investigated a cohort of 2,767 VC workers in a single plant that was a
part of the industry wide cohort studied by Cooper (1981). Most of these workers had been
employed for fewer than 5 years. There was a significant excess risk of liver cancer (14
observed vs. 4.2 expected). The incidence of angiosarcomas was not reported, but 12/18 liver
cancers were angiosarcomas in a larger cohort of 3,620 workers that included workers exposed
to PVC, as well as the VC workers. In a case-control study with the controls taken from a
National Institute for Occupational Safety and Health (NIOSH) database, angiosarcomas were
related to higher cumulative exposure to VC, but other liver cancers were not. Brain and lung
cancer were elevated for the combined cohort, which was exposed to at least 19 other potentially
carcinogenic compounds in the plant, but were not elevated for the subcohort of VC workers.
In an update of the Cooper et al. (1981) cohort Wong et al. (1991) also found an
association between VC exposure and liver angiosarcoma. Fifteen deaths from angiosarcoma
were identified, a clear excess over the incidence in the general population, although no
statistical analysis was conducted for this malignancy. This study also attempted to determine
whether other cancers were associated with VC exposure. Excluding the 15 angiosarcomas
identified from death certificates, a significant increase was observed in liver and biliary tract
cancers alone (Obs = 22, SMR = 386,/? < 0.02). However, the study authors suggested that
these 22 cancers probably included some cases of angiosarcoma that were misdiagnosed. Based
on a comparison of death certificates and pathology records in 14 cases, the authors estimated
that the correct number of primary liver/biliary tract cancers (excluding angiosarcomas) was 14,
which was still significantly increased over background (SMR = 243, p < 0.01). Although this is
an estimate, liver cells were the primary target site in 8 of the 14 pathology records. It can thus
be assumed that VC is capable of inducing both liver angiosarcoma and hepatocellular
carcinoma. This study also found a significantly increased risk of cancer of the brain and central
nervous system (Obs = 23, SMR = 180,/> < 0.05). There was no excess in cancer of the
respiratory system or the lymphatic and hematopoietic systems. Expected deaths were based
upon U.S. mortality rates, standardized for age, race, and calendar time.
CMA (1998a) updated the Wong et al. (1991) study through 1995. This study was also
designed to evaluate possible induction of cancer at sites other than the liver. In this study all
liver and biliary cancers were included in a single category. Mortality rate for these cancers,
based upon 80 deaths, was again significantly increased (SMR = 359; 95% CI = 284-446). The
SMRs increased with duration of exposure from 83 (95% CI = 33-171) to 215 (95% CI= 103-
396) to 679 (95% CI = 483-929) and to 688 (95% CI = 440-1023) for those exposed from 1-4
years, 5-9 years, 10-19 years and 20 years or more, respectively. Mortality from brain and CNS
cancer showed an excess based on 36 deaths (SMR = 142; 95% CI = 100-197). The elevation
was statistically significant for those exposed 5-9 years (SMR =193; 95% CI = 96-346) and for
those exposed 20 years or more (SMR = 290; 95% CI = 132-551). Finally, mortality from
connective and other soft tissue cancers, based upon 12 deaths, was also increased significantly
10
-------
(SMR = 270; 95% CI = 129-472). The increases were significant for those exposed 10-19 years
(SMR = 477; 95% CI = 155-1113) and 20 or more years (SMR = 725; 95% CI = 197-1856). The
latter cause of death category had not been evaluated in the Wong et al. (1991) study. While
SMRs for brain cancer were elevated among men exposed 20 or more years, the SMR was
highest for those hired during the period 1960-1972 and lowest for those hired before 1950.
Also, no trend was seen for higher SMRs based on time since first exposure occurred. The
authors suggested that some of the earlier cohort may have been exposed to another carcinogen
prior to employment.
In conclusion, there exists strong evidence of a causal relationship between exposure to
VC in humans and a significant excess risk of liver angiosarcoma. There is also highly
suggestive evidence of a causal relationship with hepatocellular carcinoma, despite some
uncertainty regarding incidence of hepatocellular tumors, because of some angiosarcomas
possibly being misdiagnosed as hepatocellular carcinoma. Because of the likelihood that both
types of tumors are induced by VC, and because misdiagnosis is likely in some of the studies, it
is reasonable to include both tumor types in any risk analysis. Lung cancer has also been
associated with VC exposure in some studies, but based on the data of Waxweiler et al. (1981),
the increased risk of lung cancer observed in some cohorts may be due to exposure to PVC dust
rather than VC. A relationship among brain cancer and soft-tissue lymphopoietic and
hematopoietic cancers has been noted in some studies, although it is weaker than for liver
cancer. In the review article by Blair and Kazerouni (1997) it is stated that because of the large
size of the cohorts examined, demonstrating a strong exposure-response relationship for
angiosarcoma of the liver and at the same time showing no evidence of an exposure-response
gradient for other nonliver tumors (e.g., leukemia, brain, lung, pancreas, mammary), vinyl
chloride is not likely to be associated strongly with cancers other than liver in humans.
Nevertheless, on the basis of small but statistically significant increases in brain and soft tissue
sarcomas in the large updated cohort reported on by CMA (1998a), the evidence for induction of
cancer at these sites may be considered suggestive.
As discussed in Section 5.3, the dose-response assessment for cancer is based on liver
angiosarcomas, angiomas, hepatomas, and neoplastic nodules because liver tumors lead to the
strongest causal association with VC exposure and because angiosarcomas in particular are rare
in unexposed humans and laboratory animals. Blair and Kazerouni (1997) indicated that the data
in humans suggested VC is not likely to be associated with cancers other than the liver. Further
attempts to estimate cancer risk based upon tumor induction in animal bioassays at other sites,
such as mammary glands, resulted in much greater uncertainty because responses were quite
variable and not always statistically significant, and because the magnitude of the cancer risk
estimated was, with few exceptions, considerably less than the risk of liver tumors. Finally,
although cancer incidence was reported to be significantly increased at two other sites in a recent
epidemiology study (CMA, 1998a), the association is weak and any estimated increase in
mortality from cancer at these sites is likely to be less than for liver cancer. Upon the basis of
the available evidence it was therefore concluded that the liver is the most sensitive site and, as a
result, protection against liver cancer should be protective against other cancers as well.
11
-------
4.1.2. Noncancer Effects
Several epidemiology and case studies have associated chronic occupational exposure to
VC with impaired liver function and/or biochemical or histological evidence of liver damage,
notably subcapsular, portal, and perisinusoidal fibrosis; hyperplasia of hepatocytes and
sinusoidal cells; and portal hypertension (Buchancova et al., 1985; Doss et al., 1984; Gedigk et
al., 1975; Lilis et al., 1975; Marsteller et al., 1975; Popper and Thomas, 1975; Tamburro et al.,
1984). Focal hepatocellular hyperplasia and focal mixed (hepatocytes and sinusoidal cells)
hyperplasia are early histological alterations indicative of VC exposure (Popper and Thomas,
1975) and are the principal anatomic lesions in VC-associated liver disease (Berk et al., 1976).
Doss et al. (1984) reported coproporphyrinuria in 46 males occupationally exposed to VC for 18
months to 21 years. Gedigk et al. (1975) correlated liver damage manifested as parenchymal
damage, fibrosis, and proliferation of the sinusoidal cells with duration of exposure to VC in 51
patients. The severity of degenerative lesions increased with increasing duration of exposure
and appeared to be reversible upon exposure cessation. Another study reported the progressive
nature of the liver changes that resulted in "chronic hepatitis" (Lilis et al., 1975). Thresholds for
hepatotoxicity cannot be identified because data regarding exposure concentrations and duration
were not available. The symptoms and signs of liver disease associated with occupational
exposure to VC include pain or discomfort in the right upper quadrant of the abdomen,
hepatomegaly, splenomegaly, and thrombocytopenia, in addition to fibrosis, cirrhosis, and portal
hypertension; however, these observations are not pathognomonic for VC-induced liver disease
(Lilis et al., 1975; Marsteller et al., 1975; Popper and Thomas, 1975). Fibrosis frequently occurs
in the elderly and in patients with diabetes mellitus (Popper and Thomas, 1975).
Ho et al. (1991) reported liver dysfunction in 12 of 271 workers (4.8%) who were
reportedly exposed to environmental levels of 1 to 20 ppm VC, with a geometric mean of 6 ppm
(15 mg/m3). The affected workers, ranging from 19 to 55 years of age, were identified as a
result of a medical surveillance program of various nonspecific biochemical liver function tests.
In addition to repeated abnormalities in these tests, four workers had hepatomegaly, four had
hepatosplenomegaly, two others had splenomegaly, and the remaining two were normal. An
improvement in liver function testing was claimed to be noted in some (number unclear from the
text) of these affected workers within 6 months to 2 years after removal from exposure; liver
function tests for 2 of these workers who returned to work were reported to have became
abnormal again. Although this study suggests effects in humans at very low levels of VC
monomer exposure, the lack of specificity of liver tests, the small number of workers involved,
the fact that 8 of the 12 affected workers were current or ex-drinkers, and aspects of the exposure
assessment make the results problematic to interpret. For example, although exposures of 1-20
ppm are claimed in the report, all affected workers were reported to experience nausea, and 4 of
the 12 reported dizziness, effects that would be expected to occur at or above the odor threshold,
which is around 3,000 ppm (Amoore and Hautala, 1983). The affected subjects are
acknowledged in the study as having been involved in washing tanks where VC concentrations
as high as several thousand ppm were possible. Also, the significance of nonspecific clinical
chemistry effects and their relationship to hepatic toxicity caused by vinyl chloride have been
considered problematic by Feron et al. (1979), who state that there are few if any suitable
parameters for early diagnosis of VC monomer disease in humans. On the other hand, Du et al.
(1995) found that serum levels of gamma-glutamyl transferase (GOT), but not other indicators of
liver function, were associated with exposure in a group of 224 VC workers with time-weighted
12
-------
average exposure ranging from 0.36 to 74 ppm (0.92 to 189 mg/m3). Hepatomegaly, altered
liver function as shown by biochemical tests, and Raynaud's phenomenon (RP, cold sensitivity
and numbness of fingers) were reported in chemical plant workers exposed to 25 to 250 ppm VC
(64 to 639 mg/m3) (Occidental Chemical Corporation, 1975, levels much higher than those
claimed by Ho and associates. The major obvious problems in the Ho study may thus include
grossly underreported exposure estimation, lack of a plausible association between VC exposure
and minor nonspecific liver dysfunction, confounding from alcohol and other unknown factors
that could affect liver function, and even lack of information or rates on liver dysfunction in the
general population in this part of the world (southern Asia). However, this study does engender
some uncertainty about the possibility that the effects seen could be due to human variability in
response to the effects of VC monomer and should be considered in characterizing the human
response, at least to the noncancer effects, from exposure to VC monomer.
An occupational study attempted to correlate the effects of VC with the liver function of
exposed workers (77 total), as measured by the plasma clearance of the 99mTc-N-(2,4-
dimethylacetanilido)iminodiacetate (HEPIDA) complex (Studniarek et al., 1989). The duration
of exposure varied from 3 to 17 years. Personal air samplers were used to determine the mean
VC concentrations in 1982 at various regions of the plant. Polymerization operators (n = 13) had
the highest mean exposure to VC, 30 mg/m3, with a mean duration of employment of 10 years.
Autoclave cleaners (n = 9) and auxiliary personnel (n = 12) in polymerization rooms were
exposed to mean concentrations of 9 mg/m3 for a mean duration of 8 and 12 years, respectively,
while technical supervisors (n = 6) had the lowest mean VC exposure of 6 mg/m3 for a mean
duration of 13 years. The investigators found a significant correlation between degree of
exposure to VC and the frequency of low clearance values; however, no concentration-response
relationship was detected among the groups with respect to plasma clearance of 99mTc-HEPIDA.
This study is of limited value because personal air sampling was conducted for only 1 year. The
yearly geometric means of VC atmospheric concentrations in various departments of the plant
were provided, but these concentrations fluctuated dramatically between 0.1 and 600 mg/m3
from 1974 to 1982.
There was no evidence of decrements in pulmonary function over the course of a work
shift in a group of 53 chemical, plastics, and rubber workers exposed to higher VC levels (up to
250 ppm, 639 mg/m3) (Occidental Chemical Corporation, 1975). In an analysis of causes of
death in a cohort of 10,173 VC workers for up to 30 years after the onset of exposure, the only
noncancer cause for which the SMR was significantly elevated was emphysema (Dow Chemical
Company, 1986). There was no correlation with exposure duration or latency. There was also
no control for smoking, although there was no excess of lung cancer.
Insufficient data exist to evaluate the teratogenicity of VC in humans. Several
epidemiology studies have investigated the effects of inhalation exposure to VC on the incidence
of fetal loss and birth defects (Hatch et al., 1981; Infante et al., 1976; Waxweiler et al., 1977);
however, no solid association has been found. Studies of communities near VC plants (Edmonds
et al., 1978; Theriault et al., 1983) have found no clear association between parental residence in
a region with a VC plant and the incidence of birth defects in the exposed community.
Fontana et al. (1995) reported a 9% occurrence of clinical symptoms of RP in 128 retired
patients who were exposed occupationally to VC. Although RP secondary to VC exposure can
13
-------
still persist after the end of exposure, capillary lesions did not appear as the main physiological
factor in the persistence of the RP.
4.2. PRECHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN
ANIMALS—ORAL AND INHALATION
Feron et al. (1981) administered diets containing 10% PVC with varying proportions of
VC to Wistar rats. Diets were available to experimental animals for 4 hours per day, and food
consumption and VC concentrations were measured at several times during the feeding period in
order to account for loss of VC from the diet due to volatilization. This information was used to
calculate the ingested dose. Evaporative loss averaged 20% over 4 hours. The ingested dose
was adjusted downward by the amount of VC measured in the feces to arrive at the bioavailable
doses of 0, 1.7, 5.0, or 14.1 mg/kg-day that were fed to Wistar rats (n = 80, 60, 60, and 80,
respectively) for a lifetime. An additional group of 80/sex were administered 300 mg/kg-day by
gavage in oil for 5 days/week for 83 weeks. The rats were 5 weeks old at the start of the study.
They were weighed at 4-week intervals throughout the study. Hematological values were
obtained at 13, 26, 52, 78, and 94 weeks, and blood chemistry was performed at 13, 26, 52, and
106 weeks (n = 10). Urinalysis was performed on 10 animals per group at 13, 25, 52, 78, and 94
weeks. All surviving animals were necropsied at week 135 (males) or week 144 (females).
Interim sacrifices of 10 animals at 26 and 52 weeks included animals from the control and high-
dose groups.
Feron et al. (1981) reported that there was no difference in body weights in the VC-
treated animals, although all groups (including the control) weighed significantly less than the
controls fed ad lib (treated animals had access to food for only 4 hours/day). Significant clinical
signs of toxicity in the 5.0 and 14.1 mg/kg-day groups included lethargy, humpbacked posture,
and emaciation. Significantly increased mortality was seen consistently in males at 14.1 mg/kg-
day and in females at 5.0 and 14.1 mg/kg-day. No treatment-related effects on hematology,
blood chemistry, or urinalysis parameters were observed. Relative liver weight was significantly
increased at 14.1 mg/kg-day but was not reported for the other dose groups.
In the Feron et al. (1981) study, a variety of liver lesions were observed histologically to
be dose related and statistically significant in male and female rats. These included clear cell
foci, basophilic foci, eosinophilic foci, neoplastic nodules, hepatocellular carcinoma,
angiosarcoma, necrosis, cysts, and liver cell polymorphism. Several of these endpoints were
significantly increased in the group exposed to 1.7 mg/kg-day. Furthermore, basophilic foci
were significantly (p < 0.05) increased at doses as low as 0.014 mg/kg-day and liver cell
polymorphisms at doses as low as 0.13 mg/kg-day in a related study conducted at lower doses
(Til et al., 1983, 1991); the Til et al. (1983) study is described in more detail below. The above
lesions, with the exception of the angiosarcoma and bile duct cysts, derive from hepatocytes;
angiosarcoma is derived from sinusoidal cells and cysts from bile duct epithelium. Because the
neoplastic nodules and altered hepatocellular foci are proliferative lesions indicative of changes
in the cells from which hepatocellular carcinomas may be derived, and because these lesions
occur at lower doses and higher incidences than the hepatocellular carcinomas, these lesions are
likely to be preneoplastic. In addition, the fact that they occur at doses one to two orders of
magnitude lower than other liver lesions, such as necrosis, indicates that these lesions probably
14
-------
occur via a genotoxic mechanism, consistent with the known mechanism of VC carcinogenicity.
By contrast, there are no indications that VC causes cancer via a cytotoxic mechanism, so the
necrosis is not considered a preneoplastic effect. The incidence of necrosis was increased in a
dose-related manner in both males (4/55 in controls, 4/58, 8/56, and 23/59 low to high dose) and
females (5/57 in controls, 6/58, 19/59, and 27/57, low to high dose). The incidence was
statistically significant in males receiving 14.1 mg/kg-day and in females receiving 5.0 mg/kg-
day. Liver cell polymorphism, another endpoint not considered preneoplastic (Schoental and
Magee, 1957, 1959), was also significantly increased in males only (4/55 in controls, 16/58,
28/56, and 42/59 low to high dose). Hepatic cysts were increased in females in a dose-related
manner (9/57 in controls, 30/58, 41/59, and 49/57 low to high dose), whereas in males they were
significantly increased only at the highest dose (16/59). Proliferation of sinusoidal cells, the
source of angiosarcomas, showed a dose-related increase in males but did not achieve statistical
significance. Increased tumor incidence was noted in all treated groups. Almost exclusively
angiosarcomas were observed in males and females administered 300 mg/kg-day by gavage,
while a mixture of angiosarcomas and hepatocellular carcinomas was observed at the mid and
high dietary doses. Only hepatocellular carcinomas were reported at the low dose. Several other
rare tumors were identified as possibly being associated with VC exposure. At least some of the
observed pulmonary angiosarcomas (significant alp < 0.05) and extrahepatic abdominal
angiosarcomas appeared to be primary tumors, since they were observed in animals with no liver
angiosarcomas. The incidence of Zymbal gland tumors, a rare tumor type, also increased. These
neoplasms occurred at and above doses of 5 mg/kg-day. Abdominal mesotheliomas were
elevated over controls in all dosed groups, but with no clear dose response. Incidence and
analysis of tumors and noncancerous lesions in this study are presented below.
The lifetime dietary study of Til et al. (1983, 1991) was performed in order to study a
range of oral doses below that delivered in the Feron et al. (1981) study, since tumors were
observed at all doses in the previous study. The oral doses were delivered in the same way
except that the diets contained a final concentration of 1% PVC, rather than 10%. Wistar rats,
beginning at 5 weeks of age (100/sex/dose) were administered doses (corrected for evaporative
loss and the nonabsorbed portion in the feces) of 0, 0.014, 0.13, or 1.3 mg VC/kg/day for 149
weeks. Mortality differences were not remarkable for males but were slightly increased for
females receiving 1.3 mg/kg-day. Relative organ weights were not evaluated. Angiosarcomas
were observed in one high-dose male and two high-dose females. Other significant increases in
tumors were limited to neoplastic nodules in females and hepatocellular carcinomas in males.
No Zymbal gland tumors or abdominal mesotheliomas were observed. Testicular effects were
not evaluated. An increased incidence of basophilic foci in liver cells was observed in both
sexes at 1.3 mg/kg-day and only in females in the two lower dosage groups. Significant
increases in females having "many" hepatic cysts (3/98 in controls, 4/100, 9/96, and 24/29 low
to high dose) as well as liver cell polymorphism in males (incidence of moderate + severe of
5/99 in controls, 5/99, 8/99, and 13/49 low to high dose) and females (incidence of moderate +
severe of 16/98 in controls, 16/100, 12/96, and 24/49 low to high dose) were reported. Since
these latter two endpoints were not considered to be neoplastic or preneoplastic, they were
considered suitable for development of RfDs and RfCs.
As described in Section 5.1.2, the PBPK model of Clewell et al. (1995b) was used to
derive dose metrics that were then used to convert the exposure levels for the endpoints of
interest in the animal studies to equivalent human exposure levels. In addition, because there are
15
-------
no direct effects at the portal of entry, the PBPK model was also used to derive dose metrics that
were then used to convert the oral exposure levels used by Til et al. (1983, 1991) to a continuous
human inhalation exposure concentration that would result in the same internal dose as occurred
in the animal study. The study of Til et al. (1983, 1991) defines a no-observed-adverse-effect
level (NOAEL) of 0.13 mg/kg-day and a lowest-observed-adverse-effect level (LOAEL) of 1.3
mg/kg-day for liver effects that are not considered to be preneoplastic. Using the PBPK model
of Clewell et al. (1995b), aNOAEL(HEC [human equivalent concentration]) and LOAEL(HEC)
of 2.5 and 25 mg/m3, respectively, were calculated. Benchmark dose (BMD) modeling was then
conducted on the internal dose metrics calculated using the PBPK model, and the BMD at a
benchmark response of 10% extra risk (BMD10) was calculated and evaluated. Due to
limitations in the data and variable outputs from the BMD models, the NOAEL was chosen for
use in further quantitative analysis.
Bi et al. (1985) exposed Wistar rats (apparently 75 per group) to 0, 10, 100, or 3,000 ppm
VC (99.99% pure) for 6 hours/day, 6 days/week (duration adjusted to 0, 5.5, 55, 1,643 mg/m3,
respectively) for up to 12 months. Animals were weighed monthly and observed daily for
clinical signs. Interim sacrifices were reported at 3 (n = 8), 6 (n = 30), 9 (n = 6), and 12 (n = 10)
months, with surviving animals examined after 18 months (6 months after the end of exposure).
Organ weights and histopathology were reported to have been assessed on lung, liver, heart,
kidney, testes, spleen, and brain, but only partial organ weight information was presented, and
only testicular histopathology results are discussed in the report. Body weight was significantly
decreased in the mid- and high-exposure groups (320, 310, 280, and 240 g in 0, 10, 100, and
3,000 ppm groups, respectively). Liver-to-body weight ratios were increased in a concentration-
dependent manner after 6 months at all dose levels. At 12 months, increased relative liver
weight was observed only in the 3,000 ppm group, although the power to detect this effect was
limited by the small number of animals examined. No effect on liver weight persisted at 18
months after the start of the exposure. Relative kidney weight in the 3,000 ppm group was
increased at 3 and 12 months but not at 6 or 18 months, and in the 100 ppm group only at 18
months. Relative testes weight was decreased in the 100 and 3,000 ppm groups at 6 months, but
the effect was not concentration related in that the relative testes weight was less at 100 than at
3,000 ppm and no other time points showed significant effects. There were several groups with
significant differences in relative heart or spleen weights, but these were not consistent across
exposure concentrations or durations and thus do not appear to be exposure related. The study
did not report absolute organ weights, relative weights for groups with no significant differences,
standard deviations, or histopathology results (except in the testes), making the organ weight
differences in tissues other than the liver and testes difficult to interpret, although spleen size has
been reported in other animal and human studies. The incidence of damage to the testicular
seminiferous tubules in rats (n = 74) exposed to 0, 10, 100, or 3,000 ppm was 18.9%, 29.7%,
36.5%, and 56%, respectively. The incidence was statistically elevated at 100 and 3,000 ppm
(duration adjusted to 55 and 1,643 mg/m3, respectively) (p < 0.05 andp < 0.001, respectively)
compared with controls and was concentration related. This damage consisted of cellular
alterations, degeneration and necrosis. Thus, 10 ppm (duration adjusted to 5.5 mg/m3) is
considered a LOAEL for liver weight changes and the NOAEL for biologically significant
testicular degeneration.
As described for the Til et al. (1983, 1991) study, this concentration was converted to an
HEC using the PBPK model of Clewell et al. (1995b), and benchmark modeling was then
16
-------
conducted on the dose metric when possible. Thus, the LOAEL(HEC) for increased relative
liver weight is 28 mg/m3, and the NOAEL(HEC) for increased testicular degeneration is 42
mg/m3. The testicular degeneration was the only effect in this study that was suitable for
benchmark modeling because no measure of variability (e.g., standard deviation) was provided
for the liver weight endpoint. The HEC based on the benchmark analysis benchmark
concentration (BMC)(HEC) and the PBPK model is 182 mg/m3. The liver is more sensitive, and
the LOAEL(HEC) is the most appropriate dose-response value in this study.
Du et al. (1979) exposed male Sprague-Dawley rats for 2-8 hours/day over periods of 1-5
weeks to 15,000 ppm VC. The total accumulated exposure period varied from 14 to 137 hours.
Activity of glucose-6-phosphatase in the microsomal fraction decreased 25% with respect to
controls after 70 hours of exposure. Glucose-6-phosphate dehydrogenase activity increased
twofold after more than 100 hours of exposure. Nonprotein sulfhydryl levels (glutathione and/or
cysteine) showed a slight but progressive elevation, whereas glutathione reductase increased
50%-60% during exposure. Ultrastructural alterations including dilatation of rough endoplastic
reticulum and patchy lesions near the plasmalemma were also noted. The pathology and early
enzymatic changes were considered a reflection of mild early injury to liver cells.
In a study by Sokal et al. (1980), male Wistar rats (7-34/sex/group) were exposed to 0,
50, 500, or 20,000 ppm VC for 5 hours/day, 5 days/week (duration adjusted to 0, 19, 190, or
7,607 mg/m3, respectively) for 10 months. Hematological indices, blood chemistry, and
urinalysis were evaluated after 1, 3, 6, and 10 months of exposure (n = 7-10). Histopathology
was conducted on all major organs, including the lungs, with groups sacrificed at 1.5, 3, 6, and
10 months of exposure. The number of animals in each group is not clear from the report.
Ultrastructural examination of the liver was carried out at 3, 6, and 10 months. No statistically
significant differences were observed for urinalysis, hematological, or biochemical indices. No
adverse effects on the lung were reported. There was a statistically significant (p < 0.05)
decrease in body weight at 10 months in all treatment groups relative to the controls that was
biologically significant (i.e., > 10%) in the high-exposure group only. Organ weights were
reported for groups of seven animals exposed for 10 months. Relative spleen, kidney, and heart
weights were significantly elevated in some groups, but there was no change in absolute weight
and no histological changes or effects on kidney function to corroborate an adverse effect in
these organs. Relative liver weight was increased at 500 and 20,000 ppm, and absolute liver and
testes weights were increased at 50,000 ppm. Treatment-related histological changes developed
in the liver and testes. After 10 months, there was a significant increase in polymorphism of
hepatocytes (2/28, 5/21, 18/34, and 10/17 in 0, 50, 500, and 20,000 ppm groups, respectively)
and proliferation of reticuloendothelial cells lining the sinusoids (3/28, 3/21, 13/34, and 8/17 in
0, 50, 500, and 20,000 ppm groups, respectively). These effects were also seen at 6 months in
the 500 and 20,000 ppm groups (incidences not reported). Fatty degeneration was also observed,
and Ultrastructural changes, including proliferation of smooth endoplasmic reticulum and lipid
droplets, were reported, but no data were given. The report indicated that more detailed
description of the histopathology and ultrastructure would be published separately, but no such
record was found. Damage to the spermatogenic epithelium was significantly higher than in
controls following exposure to 500 ppm (3/28, 3/21, 13/34, and 5/17 in the 0, 50, 500, and
20,000 ppm groups, respectively). A NOAEL of 50 ppm was identified for hepatocellular and
testicular histopathology. Using the PBPK model of Clewell et al. (1995b), the NOAEL of 50
ppm corresponds to a duration-adjusted NOAEL(HEC) of 93 mg/m3 for liver effects and a
17
-------
NOAEL(HEC) of 145 mg/m3 for testicular effects. Applying benchmark modeling using the
dosimetry provided by the PBPK model in the same manner as described for Til et al. (1983,
1991), the BMC(HEC) values are 59-168 mg/m3 for liver effects (59 mg/m3 for polymorphism of
hepatocytes, 92 mg/m3 for proliferation of reticuloendothelial cells, 122 mg/m3 for testicular
effects, and 168 mg/m3 for the continuous endpoint of increased relative liver weight).
In a related study (Wisniewska-Knypl et al., 1980), male Wistar rats (7-10/group) were
exposed under conditions to nominal concentrations of 50, 500, or 20,000 ppm VC or to air only
for 5 hours/day, 5 days/week (duration adjusted to 19, 190, or 7,607 mg/m3, respectively) for 10
months with interim sacrifices at 1, 3, and 6 months. This study appears to be a different
experiment from that reported by Sokal et al. (1980) based on different initial animal weights
and chemical purity, although this is not entirely clear. Body weight was significantly affected
only in the 20,000 ppm group exposed for 10 months. Tissue examinations were limited to the
liver. Relative liver weight was increased at all sacrifice times at 500 and 20,000 ppm.
Ultrastructural examination of liver tissue from animals exposed to 50 ppm showed
hepatocellular changes characterized by proliferation of smooth endoplasmic reticulum at 3
months and accumulation of lipid droplets at 10 months. Rats exposed to 500 ppm for 3 months
exhibited hypertrophy of the smooth endoplasmic reticulum, distension of canals of rough-
surfaced membranes, swelling of mitochondria, and an increased number of lipid droplets in
cytoplasm; these changes were more intensive at 20,000 ppm. No quantitative information is
provided on the liver ultrastructural effects. This study identifies a minimal LOAEL of 50 ppm
(duration adjusted to 19 mg/m3) for minor liver histopathology and a NOAEL of 50 ppm for
liver weight effects. Based on the PBPK model of Clewell et al. (1995b), this corresponds to a
duration-adjusted LOAEL(HEC) of 79 mg/m3. Applying benchmark modeling to the liver
weight data in the same manner as described for Feron et al. (1981), the BMC(HEC) values are
168 mg/m3 for increased relative liver weight. The liver ultrastructural data are not amenable to
benchmark analysis because only descriptive information was presented.
In a study by Torkelson et al. (1961), several species of animals were exposed to 0, 50,
100, 200, or 500 ppm VC via inhalation for up to 6 months. Hematologic determinations,
urinalysis, clinical biochemistry, organ weight measurement, and histopathology examination
were conducted. Rats (24/sex/group), guinea pigs (12/sex/group), rabbits (3/sex/group) and dogs
(I/sex/group) exposed to 50 ppm (127.8 mg/m3) for 7 hours/day for 130 days in 189 days did not
exhibit toxicity as judged by appearance, mortality, growth, hematology, liver weight, and
pathology. At an exposure concentration of 100 ppm administered 138-144 times in 204 days, a
statistically significant increase in the relative liver weight of male and female rats was noted.
Exposure to 200 ppm (138-144 times in 204 days) for 6 months resulted in increased relative
liver weight in male and female rats, but there was no biochemical or microscopic evidence of
liver damage. Rabbits exposed under the same conditions exhibited histological changes
(characterized as granular degeneration and necrosis with some vacuolization and cellular
infiltration) in the centrilobular area of the liver. There was no effect at this level in guinea pigs
or dogs. Histopathological lesions of the liver (centrilobular granular degeneration) and
increased organ weight occurred in rats exposed to 500 ppm. Although relative liver weights
were slightly elevated in male rats (n = 5) exposed to 100 or 200 ppm for 2-4 hours/day
(duration adjusted to 15-30 and 30-60 mg/m3, respectively), the increases were not statistically
significant. A NOAEL for liver effects of 50 ppm (duration adjusted to 25.6 mg/m3) is identified
in this study. Based on the PBPK model of Clewell et al. (1995b), this corresponds to a
18
-------
duration-adjusted NOAEL(HEC) of 162 mg/m3. These data were not amenable to benchmark
analysis because standard deviations on the weight measurements were not reported.
Maltoni et al. (1980, 1981, 1984) exposed Sprague-Dawley or Wistar rats to 1 to 30,000
ppm VC for 4 hours/day, 5 days/week for 52 weeks, and mice and hamsters to 50 to 30,000 ppm
VC for 30 weeks, beginning at about 12 weeks of age. Animals were observed throughout their
lifetime (135 weeks). Tumor incidence and shortening of latency for liver angiosarcomas were
concentration dependent. Additional tumor types seen in rats included liver hepatoma,
nephroblastoma, neuroblastoma of the brain, Zymbal gland tumors, and mammary carcinomas.
The study authors particularly noted the rarity of angiosarcoma, hepatoma, nephroblastoma, and
neuroblastoma in their animal colony. The following types of tumors were observed in exposed
mice: mammary, liver (including angiosarcomas), forestomach, lung, and epithelial. Tumor
types in hamsters were liver (including angiosarcomas), forestomach, and epithelial. The
incidence and analysis of the tumors reported in this study are presented in Section 5.3.2.
The incidence of neoplastic or potentially preneoplastic lesions, including "hepatomas,
neoplastic liver nodules, nodular hyperplasia of the liver, and diffuse hyperplasia of the liver,"
was also presented (Maltoni et al., 1980, 1981, 1984). Because morphological descriptions were
not provided, it is not clear why different terms were used. The largest incidences were reported
for diffuse hyperplasia, generally ranging from 1% to 10% for males and females combined, but
occurring at 20%-28% in a single experiment at 100-200 ppm. The incidence of nodular
hyperplasia was about 1% in the combined controls and at • 5 ppm and about 10%-17% at
higher levels. However, although lesions as well as hepatomas and neoplastic nodules were
increased in the exposed groups, there was no clear concentration-response relationship for these
lesions.
Other inhalation experiments support the carcinogenicity of VC. Rats and mice exposed
to 0, 50, 250, or 1,000 ppm for 6 hours/day, 5 days/week for up to 6 months (mice), 10 months
(rats) (Hong et al., 1981), or 12 months (mice and rats) (Lee et al., 1978) had a significantly
increased incidence of angiosarcoma of the liver at • 250 ppm. Animals were sacrificed 12
months after the end of exposure. Mice in this study exposed to • 250 ppm also had an increase
in bronchioloalveolar adenoma of the lung and mammary gland tumors in females
(adenocarcinomas, squamous and anaplastic cell carcinomas). Male rats exposed to
concentrations as low as 100 ppm for 6 hours/day, 6 days/week for 12 months and sacrificed at
18 months (6 months after the end of exposure) had significantly increased incidences of
angiosarcoma of the liver (Bi et al., 1985). Rats exposed to 3% VC (30,000 ppm) for 4
hours/day, 6 days/week for 12 months had significantly increased incidences of epidermoid
carcinoma of the skin, adenocarcinoma of the lungs, and osteochondroma in the bones (Viola et
al., 1971), and rats exposed to 0 to 5,000 ppm for 52 weeks had primary tumors in the brain,
lung, Zymbal gland, and nasal cavity (Feron and Kroes, 1979). Keplinger et al. (1975) provided
a preliminary report of a concentration-dependent increase in tumor formation (alveologenic
adenomas of the lung, angiosarcomas of the liver, and adenosquamous carcinoma of the
mammary gland) in mice exposed to 0, 50, 200, or 2,500 ppm VC.
Suzuki (1978, 1983) investigated the effect of VC on lung tumor formation. In a
preliminary study conducted with a limited number of animals, alveologenic lung tumors
developed in 26 of 27 mice exposed to 2,500 or 6,000 ppm for 5-6 months (Suzuki, 1978). A
19
-------
concentration-related increase in the incidence of alveogenic tumors was observed in a study in
which 30-40 mice/group were exposed to 1-660 ppm VC or filtered air for 4 weeks and then
observed for up to 41 weeks postexposure (Suzuki, 1983). An increase in bronchioloalveolar
adenoma was observed in a lifetime study of mice exposed to 50 ppm VC for 100 1-hour
exposures and 5,000 or 50,000 ppm for a single 1-hour exposure (Hehir et al., 1981). The
statistical significance of these observations was not presented.
Overall, the available evidence from inhalation studies in animals supports the findings in
humans that VC is a carcinogen by this route of exposure. Although human carcinogenicity data
are lacking via the oral route, definitive responses in animal studies by both the oral and
inhalation route, and evidence that VC is well absorbed by the oral route, support a conclusion
that ingested VC is carcinogenic in humans.
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES—ORAL AND INHALATION
Inhalation experiments in animals have associated developmental toxicity only with
concentrations at or above those associated with maternal toxicity. In a two-generation
reproduction study done in accordance with GLP (CMA, 1998b), rats (CD, 30/sex/group) were
exposed by whole-body inhalation for 6 hours /day to concentration levels of 0, 10, 100, and
1,100 ppm. The groups of PI females were exposed 5 days/week beginning at 6 weeks of age
for 10 weeks (premating exposure period), and then daily through mating and gestation.
Treatment was discontinued at gestation day (GD) 20 for delivery of the Fl generation and
resumed on a daily basis on lactation day 4 until sacrifice, which occurred as a group after the
last litter was weaned. At weaning, two Fl pups were selected randomly from each litter and
exposed as the PI parents through this postweaning period until all litters were weaned, about 3
weeks total. Animals were then randomly chosen from this pool of Fl animals, designated the
P2 generation, and the 10-week premating exposure period initiated. At this period, the P2
animals were presumably near to 6 weeks in age, as were the PI generation at the beginning of
their premating exposures. Daily exposures were continued through mating and gestation.
Treatment was discontinued at GD20 for delivery of the F2 generation and resumed on a daily
basis on lactation day 4 until sacrifice, which occurred as a group after the last litters were
weaned. Both generations of males were exposed in a manner parallel to the females.
Evaluation for the parental animals included body weights and food consumption.
Estrous cycling was evaluated during the last 3 weeks of the premating period. Fertility and
reproductive performance (pregnancy rates and male fertility indices) were recorded. Sperm
assessments (motility, caudal epididymal sperm count, and morphology) were performed for 15
PI and P2 males/group. At necropsy, reproductive and other tissues (including brain, lungs,
nasal turbinates [four sections] and mammary glands) were taken from the control and 1,100
ppm groups for gross and microscopic examination. The livers of all parental animals from all
dose groups were examined microscopically. Pups were examined and weighed at birth and
days 4, 7, 14, 21, and 25 (Fl only) during lactation. At weaning one pup/sex/litter was randomly
selected, sacrificed, and given a macroscopic exam with selected tissues (including liver,
ovaries, and testes) weighed and preserved. The remaining pups were examined, sacrificed, and
discarded.
20
-------
No adverse effect of treatment was seen in the parental generations, including mortality,
clinical findings, body weight, food consumption, or effects on fertility or reproductive
performance. No adverse effect of treatment was indicated in the Fl and F2 pups from survival
or growth in either generation. The NOAEL for reproductive effects is > 1,100 ppm.
Liver effects, including hepatocellular foci, centrolobular hypertrophy, and increased
liver weights, were noted in parental (PI and P2) animals. Liver weights were significantly
increased in males only (13% to 20% increase over controls) in the PI animals at 10, 100, and
1,100 ppm and in P2 animals at 100 and 1,100 ppm. Centrilobular hypertrophy (not hepatocyte
polymorphism) was noted in a dose-related manner in PI and P2 males at the two highest
concentrations and in PI and P2 females at all three levels of exposure. Both these effects are
considered as nonadverse adaptive responses to VC exposure (Sipes and Gandolfi, 1991).
Altered hepatocellular foci (basophilic, acidophilic, and clear cell) were observed in the
livers of PI and P2 males and in P2 females. Cellular atypia was generally absent from these
lesions which were also noted as usually occurring one per animal. All foci were noted as being
graded at the minimum level of severity. Among PI males exposed to 1,100 ppm, only a single
basophilic and a single acidophilic foci were noted among the 30 livers. Among P2 males
exposed to 1100 ppm, this incidence was increased with 8 basophilic foci, 5 clear cell foci, and 5
acidophilic foci noted among the 30 livers. In addition, 5 acidophilic foci were observed from
among the 30 P2 livers observed at the next highest concentration of 100 ppm and 1 was
observed from among the 30 P2 male control livers. No foci were observed among the livers of
any PI female, exposed or control. Among livers from P2 females exposed to 1,100 ppm,
however, 11 basophilic and 8 acidophilic foci were noted. A single basophilic foci was observed
among the 30 livers from P2 females exposed at the next highest concentration of 100 ppm. No
foci of any type was observed in either sex of either parental generation at the lowest exposure
level of 10 ppm, the NOAEL for parental effects.
A possible explanation for the increased incidence of altered hepatocellular foci seen in
the P2 versus the PI generation is that the P2 generation was exposed throughout those periods
of the life cycle (in utero and throughout most of the postnatal period) that are generally
accepted as being of increased susceptibility to tissue injury, whereas the PI generation was not.
In establishing the exposure scenario in this reproductive study, however, the P2 generation was
necessarily exposed for a longer period of time, approximately 6 weeks longer if in utero time is
considered, than was the PI generation. Indeed, the increased incidence of liver effects in the P2
generation is more consistent with an increased dose, rather than with a period of susceptibility
to VC toxicity during which other types of effects would have the opportunity to become
manifest. The increase in liver effects seen in the P2 generation relative to the PI generation
could be due to reasons other than in utero or juvenile susceptibility, as the P2 animals were
exposed not only younger than the PI animals, but also longer and on a daily basis during the
postnatal period when body weights and metabolic and respiratory functions are increasing
dramatically. This confounding makes speculative any claim of neonatal or childhood
susceptibility to VC exposure for this study. However, tumor incidence has been documented to
increase at maturity among laboratory animals treated with vinyl chloride during the first 6
months of life when compared to those exposed during the second or third 6-month period of life
(Maltoni et al., 1981; Drew et al., 1983; Section 4.7.1).
21
-------
PBPK analysis of this reproductive study (Appendix D) indicates that the dose to the
liver at 10 ppm (the NOAEL for hepatocellular foci was markedly higher than the corresponding
metric derived from the NOAEL in the chronic study of Til et al. (1983). It is likely that the
metric associated with the no-effect 10 ppm dose level may well have been even higher if
consideration of the physiological and biochemical exposure parameters during the earlier
phases of development and growth were able to be considered by the PBPK model used in this
assessment. This conservative estimate of the dose metric in the liver where no effects were
observed in this reproductive study is near to the metric in the chronic study of Til et al. (1983)
where liver effects were observed. Thus, the chronic study of Til et al. (1983) demonstrates
adverse liver effects at tissue concentrations considerably lower than this reproductive study.
John et al. (1977) examined the effects of inhaled VC on the fetuses of mice, rats, and
rabbits. Pregnant CF1 mice (30-40/group) were exposed to 0, 50, or 500 ppm VC on gestational
days 6-15. Sprague-Dawley rats (20-35/group) and New Zealand white rabbits (15-20/group)
were administered 0, 500, or 2,500 ppm VC for 7 hours/day on gestational days 6-15 for rats and
6-18 for rabbits. Parameters of maternal and developmental toxicity were evaluated; both the
fetuses and litter were evaluated. Mice were more sensitive to the toxic effects of VC than either
rats or rabbits. In mice, concentrations of 500 ppm induced maternal effects that included
increased mortality, reduced body weight, and reduced absolute but not relative liver weight.
Fetotoxicity also occurred in mice at 500 ppm and was manifested as significantly increased
fetal resorption, decreased fetal body weight, reduced litter size, and retarded cranial and
sternebral ossification. However, there was no evidence of a teratogenic effect in mice at either
concentration. In rats exposed to 500 ppm, but not to 2,500 ppm, maternal effects were
restricted to reduced body weight. Maternal effects in rats at 2,500 ppm were death of one rat,
elevated absolute and relative liver weights, and reduced food consumption. A significant
reduction in fetal body weight and an increase in the incidence of lumbar spurs were observed
among rats exposed to 500 ppm but not 2,500 ppm and are not considered signs of VC-induced
fetotoxicity. At 2,500 ppm, an increased incidence of dilated ureters was observed, which may
represent a chemical-induced effect. No signs of maternal or developmental toxicity were
observed in rabbits at either dose. This study identifies a NOAEL of 50 ppm for maternal
toxicity and fetotoxicity in mice and a NOAEL of 2,500 ppm for rabbits.
Ungvary et al. (1978) exposed groups of pregnant CFY rats continuously to 1,500 ppm
(4,000 mg/m3) on gestational days 1-9, 8-14, or 14-21 and demonstrated that VC is not
teratogenic and has no embryotoxic effects when administered during the second or last third of
pregnancy. During the first third of pregnancy, maternal toxicity was manifested by increased
relative liver weight; increased fetal mortality and embryo toxic effects were evident. Slightly
reduced body weight gain was noted in dams exposed on days 14-21.
VC does not appear to produce germinal mutations as manifested by a dominant lethal
effect in male rats. In a dominant lethal study, Short et al. (1977) exposed male CD rats to 0, 50,
250, or 1,000 ppm VC for 6 hours/day, 5 days/week for 11 weeks. At the end of the exposure
period, the exposed males were mated with untreated females, and there was no evidence of
either preimplantation or postimplantation loss in pregnant females. However, reduced fertility
was observed in male rats exposed to 250 and 1,000 ppm VC.
22
-------
4.4. OTHER STUDIES
4.4.1. Neurological
Occupational studies of exposure to VC have reported a variety of central nervous system
effects of VC, including headaches, drowsiness, dizziness, ataxia, and loss of consciousness
(Lilis et al., 1975; Langauer-Lewowicka et al., 1983; Waxweiler et al., 1977). Exposure
information was not available, but the reports of loss of consciousness indicate that at least
periodic high exposures were involved. Central nervous system symptoms associated with VC
(nausea, dizziness) were also reported in volunteers exposed to • 12,000 ppm for 5 minutes
(Lester et al., 1963). Tingling of the extremities (paresthesia), and sometimes finger numbness
and pain, has also been reported. At least some of the symptoms in the extremities appear to be
associated with anoxia due to vascular insufficiency; numbness of fingers and cold sensitivity
are symptoms of Raynaud's phenomenon, which is associated with VC exposure (Lilis et al.,
1975; Occidental Chemical Corporation, 1975). However, VC may also act directly on the
peripheral nerves. Decreased nerve conduction velocities and altered electromyographic
findings were also reported in VC workers, but the decreased velocities did not achieve
statistical significance, and control data were not reported for the electromyographic findings
(Perticoni et al., 1986). Exposure data were not reported for this study.
These occupational reports are supported by animal data. Decreased responses to
external stimuli and disturbed equilibrium were observed in male Wistar rats exposed for 4
hours/day, 5 days/week for 10 months to 30,000 ppm VC (Viola, 1970). Histopathological
examination at 12 months revealed diffuse degeneration of gray and white matter of the brain,
including numerous atrophied nerves and pronounced cerebellar degeneration of the Purkinje
cell layer. Peripheral nerve endings were surrounded and infiltrated with fibrous tissue.
4.4.2. Genotoxicity
Several lines of evidence indicate that VC metabolites are genotoxic, interacting directly
with DNA. In vitro genotoxicity assays indicate that VC is mutagenic in the presence of
exogenous metabolic activation but not in the absence of activation. Similar assays show that
the major VC metabolite, chlorethylene oxide (CEO), is positive in genotoxicity tests. In vivo
genotoxicity tests with VC also provide evidence of genotoxicity. Finally, DNA adducts formed
from VC metabolites have been identified; certain persistent adducts are believed to be
associated with the development of carcinogenicity.
Several occupational studies reported genotoxic effects of VC. Sinues et al. (1991)
examined the incidence of micronuclei and sister chromatid exchanges (SCEs) in a group of 52
nonsmokers exposed to VC and 41 nonsmoking controls. The exposure level was estimated at
1.3-16.7 ppm (high-exposure group) and 0.3-7.3 ppm (low-exposure group), with an average
duration of 17 years. Increases in both SCEs and micronuclei were observed, and the increase
correlated with exposure levels. An increase in chromosome aberrations in peripheral
lymphocytes that correlated with exposure duration was observed in a cohort of 57 VC workers,
compared with 19 on-site controls and 5 off-site controls. Current average exposure was 5 ppm,
but excursions up to 1,000 ppm were reported (Purchase et al., 1978). Hansteen et al. (1978)
23
-------
investigated chromosome aberrations in a group of VC workers exposed to 25 ppm and then
again after the workers had not been exposed for 2-2.5 years. Chromosome aberrations in
lymphocytes were elevated relative to controls at the initial sampling but not after exposure
ceased.
VC-induced mutations were noted in the Salmonella typhimurium reverse mutation
assay, both using vapor exposure (Bartsch et al., 1975) and incorporation into the medium
(Rannug et al., 1974). The mutagenic activity was decreased or eliminated in the absence of
exogenous metabolic activation. By contrast, the VC metabolites CEO and CAA increased the
reversion rate even in the absence of exogenous activation (Bartsch et al., 1975; Rannug et al.,
1976). The highly reactive metabolite CEO was much more potent than the CAA, inducing
mutations at exposures as low as 0.1 mM for 1 hour.
Single-strand breaks (SSBs) have been detected in liver DNA following inhalation
exposure of mice to VC (Walles et al., 1988). (It is generally assumed that SSBs represent an
intermediate stage in the excision repair of DNA adducts.) The occurrence of SSBs reached a
maximum at exposures of 500 ppm, consistent with saturation of metabolism. It was found that
20% of the SSBs remained after 20 hours.
The p53 tumor suppressor gene is often mutated in a wide variety of cancers. VC has
been associated with specific A -> T transversions at codons 179, 249, and 255 of the p53 gene.
The mutations result in transversions of His -> Leu at residue 179, Arg -> Trp at residue 249,
and He -> Phe at residue 255 in highly conserved regions of the DNA-binding core domain of
the P53 protein. The latter two mutants were shown to contain certain common regions that
differ substantially in conformation from the wild-type structure (Chen et al., 1999). By the use
of anti-p53 antibodies, increased incidences of mutations in this gene were detected in workers
occupationally exposed to VC. Even higher incidences were noted in occupationally exposed
workers with angiosarcoma of the liver (Hollstein et al., 1994; Trivers et al., 1995), while similar
mutations have not been identified in liver angiosarcomas not induced by VC (Soini et al.,
1995). More recently Smith et al. (1998) were able to demonstrate a dose-response relationship
between VC exposure and increases in mutant p53 in French workers occupationally exposed to
VC. Adjusted odds ratios for estimated ppm years of exposure equaled 4.16 (95% CI = 1.63-
10.64 for • 500 ppm); 5.76 (95% CI = 2.39-13.85 for 501-2,500 ppm); 10.24 (95% CI = 4.20-
24.95 for 2,501-5,000 ppm); 13.26 (95%CI = 5.52-31.88 for >5,000 ppm). Similar increases
were also reported in Taiwanese VC workers (Luo et al., 1999). Thirty-three of 251 (13.2%) VC
workers tested positive for p53 overexpression (10% with positive mutant p53 protein and 3.6%
with positive anti-p53). The results indicate that this serum biomarker for p53 protein is related
to vinyl chloride exposure and may be an early indicator of carcinogenic risk in exposed
populations.
The genotoxic potential of VC and its metabolites has also been investigated by assaying
the formation of DNA adducts. Although 7-(2-oxoethyl)guanine (OEG) has been identified as
accounting for approximately 98% of all VC adducts formed in vivo (Swenberg et al., 1992), this
adduct is very rapidly repaired and does not appear to lead to miscoding during DNA replication.
Therefore, it is not considered important for carcinogenesis (Laib, 1986; Swenberg et al., 1992).
Instead, VC carcinogenicity is attributed to four etheno-DNA adducts that are formed at much
lower concentrations than OEG but that are more persistent (Swenberg et al., 1992) and can lead
24
-------
to defective transcription (Singer et al., 1987) and presumably also defective replication. For
example, ethenoguanine (EG) produces a base pair mismatch (G* A transition) in bacterial
assays (Cheng et al., 1991). These adducts are: 1,N2-EG; N2,3-EG; l,N6-etheno-2'-
deoxyadenosine (EDA), and 3,N4-etheno-2'-deoxycytidine (EDC) (Laib, 1986; Fedtke et al.,
1990; Dosanjh et al., 1994).
It is still not possible to determine which, if any, of the DNA-adducts identified from VC
exposure may be responsible for the observed carcinogenicity of VC. The likelihood that a
given DNA-adduct will lead to a neoplastic transformation depends on many factors, including
its persistence and the consequences of its repair or failure to be repaired. The persistence of a
given adduct depends on both the rate of formation and the rate of repair (Singer, 1985); in
humans, all of the etheno adducts appear to be repaired by the same DNA glycosylase but not at
the same rate (Dosanjh et al., 1994). In particular, the repair of the ethenoguanines appears to be
much slower than that of the other etheno-adducts in humans (Dosanjh et al., 1994).
This was in contrast to the results of a similar study in rats, where N2,3-EG was repaired
with a half-life of about 30 days, while there was no evidence that EDA and EDC were repaired
at all (Swenberg et al., 1992). Swenberg et al. (1999), however, using a more sensitive method
of analysis, found that the apparent persistence of etheno adducts is actually due to endogenous
production. The amounts of endogenous N2, 3-EG was measured in liver DNA of rats and
humans, with a mean 0.21 ± 0 .07 per 106 unmodified guanine reported for humans, compared
with 0.09 ± 0.04 for rats. The ratio of OEG to N2,3-EG was similar across all tissues measured,
suggesting that DNA repair was not tissue specific. By use of 13C2-VC, endogenous and
exogenous N2,3-EG could be monitored in the same animal. Lack of change for endogenous
N2,3-EG in rats exposed to 1,100 ppm suggests that repair is not saturated at this concentration.
In the same series of studies it was shown that increases in N2,3-EG in rats exposed 4 weeks to
VC are consistent with long-term cancer bioassays in rats, with a steep slope between 0 and 100
ppm and relatively little increase at 1,100 ppm (Morinello et al., 1999). Controls averaged 0.08
± 0.04 and 0.11 ± 0.05 per 106 unmodified dGua in 1 and 4 weeks controls, respectively.
Exposure to 10, 100, and 1,100 ppm VC for 1 wk increased the N2,3-EG adducts to 0.20 ± 0.05,
0.68 ± 0.09, and 1.25 ± 0.20 per 106 dGua, respectively. After 4 weeks exposure, the
corresponding amounts were 0.53 ±0.11, 2.28 ± 0.18, and 3.78 ± 0.55 N2,3-EGper 106 dGuo.
These data provide support for the use of linearized model for low-dose extrapolation of cancer
risk.
Swenberg et al. (1999) also measured the amount of N2,3-EG in both hepatocytes and
sinusoidal cells, the latter being the most common site for liver cancer induction by VC.
Although exposures in this case were to vinyl fluoride (VF), the mechansims of cancer induction
by the two chemicals are considered to be the same. Despite the fact that the sinusoidal cells had
little of the enzyme CYP 2E1, indicating that epoxidation occurs primarily in the hepatocytes,
the amount of N2, 3-EG sinusoidal cells was three times that of the hepatocytes. Moreover, N-
methylpurine-DNA glycosylase mRNA, a DNA repair enzyme capable of removing etheno
DNA adducts (Dosanjh et al., 1994), was expressed in sinusoidal cells at only 20% that of
hepatocytes. Thus even though the sinusoidal cells are exposed to lower concentrations of the
epoxide, because it must diffuse from the hepatocytes, limited repair capability apparently
renders these cells more susceptible to carcinogenic effects of vinyl halides.
25
-------
The overall evidence indicates that VC must be metabolized to cause carcinogenicity. A
reactive, short-lived metabolite that achieves only low steady-state concentrations is thought to
be responsible for the toxic effects of VC (Bolt, 1978). CEO is believed to be the ultimate
carcinogenic metabolite of VC. Both CEO and CAA have been evaluated as possible
carcinogenic metabolites of VC, and the overall evidence indicates that CEO is the reactive
metabolite responsible for VC carcinogenicity. CEO is carcinogenic in skin and acts as an
initiator in the initiation/promotion protocol, while CAA is negative in these assays (Zajdela et
al., 1980). Moreover, CEO has been found to display 400-fold greater mutagenic potency than
CAA in bacterial mutagenicity assays (Perrard, 1985). In a comparison of VC and 2,2'-
dichlorodi ethyl ether, a precursor of CAA but not of CEO (Bolt, 1986), preneoplastic
hepatocellular ATP-deficient foci were reported in rats following exposure to VC but not 2,2'-
dichlorodi ethyl ether (Gwinner et al., 1983). Similarly, DNA adduct formation was observed in
rats dosed with VC but not with 2,2'-dichloroethylether. Finally, inadequate DNA repair is
likely responsible for the sensitivity of liver sinusoidal cells to carcinogenic effects of vinyl
halides.
In summary, recent studies have provided increasing evidence linking etheno-DNA
adducts with the observed carcinogenicity of VC. The recent study by Swenberg et al. (1999)
showed a good correlation between tissue concentrations of a specific adduct and the risk of
cancer in that tissue. Smith et al. (1998) also showed a positive dose-response relationship
between VC exposure in workers and mutant serum p53. However, until carinogenesis can be
quantitatively related to specific DNA adduct(s), or to specific mutations, the amount of
metabolism remains the best dose metric for comparison with tumor incidence. Additionally,
use of DNA adduct data for extrapolation of risk from animals to humans would require
comparative data on DNA repair efficiency in humans.
4.4.3. Noncancer Mechanism
A reactive, short-lived metabolite that achieves only low steady-state concentrations is
thought to be responsible for the toxic effects of VC (Bolt, 1978); the rapid elimination of VC
and its major metabolites is consistent with this hypothesis (Bolt et al., 1977). Both CEO and
CAA can react with tissue nucleophiles, but CAA appears to be the most important source of
protein adducts. The metabolism of VC to produce irreversibly bound adducts to DNA and
protein was examined in vitro with rat liver microsomes (Guengerich et al., 1981). Inhibition
studies were performed with alcohol dehydrogenase, which is the enzyme that catalyzes the
breakdown of CAA to the corresponding alcohol, and epoxide hydrolase, which is the initial
enzyme involved in the breakdown of CEO to oxalic acid. Alcohol dehydrogenase was effective
in inhibiting the binding of VC metabolites to protein, while epoxide hydrolase was effective in
inhibiting the binding of VC metabolites to DNA. These results support the conclusion that the
epoxide is the carcinogenic moiety, but that CAA may also produce toxic manifestation.
26
-------
4.5. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS AND
MODE OF ACTION
That VC is rapidly absorbed and distributed throughout the body via oral and inhalation
routes and leads to similar effects (i.e., liver) via the same modes of action (activation of parent
compound within liver tissues) provides an empirical mechanistic rationale for performing route-
to-route extrapolation.
The liver is clearly the primary target organ for cancer, as evidenced by the rare tumor
type (liver angiosarcomas) occurring in both human and animal toxicity studies. The long-term
repeated dose animal studies of oral (Feron et al., 1981; Til et al., 1983, 1991) and inhalation VC
exposure (Sokal et al., 1980) report a wide spectrum of liver histopathology that is considered to
be neoplastic or preneoplastic in character. There is, however, liver histopathology, such as
cysts and liver cell polymorphisms, reported in these studies that is considered nonneoplastic.
VC-induced liver-cell polymorphisms are very similar to the changes observed in liver
parenchymal cells after administration of several pyrrolizidene alkaloids in which some cells
have diameters at least 4x normal (Schoental and Magee, 1957, 1959). There has been no clear
indication whether these affected cells could develop into hyperplastic nodules or hepatomas
(Afzelius and Schoental, 1968). Other studies have reported increased liver weight in laboratory
animals with repeated dosing (Bi et al., 1985; Sokal et al., 1980; Torkelson et al., 1961;
Wisniewska-Knypl et al., 1980) and in parental animals in inhalation reproductive studies
(CMA, 1998a). Occupational studies have also associated VC exposure with impaired liver
function and/or biochemical or histological evidence of liver damage (Buchancova et al., 1985;
Doss et al., 1984; Gedigk et al., 1975; Lilis et al., 1975; Marsteller et al., 1975; Popper and
Thomas, 1975; Tamburro et al., 1984). Thus, the liver is clearly the primary target of the
noncancer VC effects also.
Both cancer and noncancer liver effects are associated with metabolism of VC. The
putatative epoxide metabolite of VC, CEO, would most likely be reactive enough to manifest
genotoxic damage, whereas the rearrangement product, chloracetaldehyde (CAA), would not.
On the other hand, CEO and CAA both could be involved in the noncancer hepatic effects.
Therefore the mode of action of VC for noncancer hepatic effects is not clear as that for liver
cancer.
Noncancer effects of VC have also been reported in the testes, with lesions observed in
two inhalation studies (Bi et al., 1985; Sokal et al., 1980). Since there is evidence of P450
activity in the testes, it is reasonable to expect that testicular effects result from a locally
generated reactive metabolite. Short et al. (1986) reports male reproductive complications
subsequent to inhalation exposure of VC although a more complete reproductive study showed
liver but no reproductive effects in either sex (CMA, 1998b). Thus, the critical effect (i.e., the
one that occurs first as dose increases) requires resolution, ideally through a comparison among
liver, testicular, and reproductive effects.
A PBPK model as used in this assessment could allow direct comparison of various
effects with common measures of dosimetry associated with those effects. The manner in which
the PBPK model converts external exposures, both inhalation and oral, to common measures of
dosimetry is explained in detail in Section 5.1.2 and in Appendices B and D. A concept central
27
-------
to the use of a common measure of dose (or common dose metric) for VC is that the toxicity of
VC is directly related to metabolism of the parent compound to a more reactive and toxic
species. The PBPK model can be exercised to estimate the amount of metabolism that would
occur in a specific exposure scenario with specific physiological/biochemical parameters.
The PBPK model used in this assessment (Clewell et al., 1995b) was exercised to derive
two different dose metrics associated with various effects: the total amount of metabolism in
the liver/volume of the liver (RISK) and the total amount of metabolism/body weight (AMET).
This conversion of external exposures to common dose metrics is then utilized to elucidate the
adverse endpoint that appears first as the exposure (and dose) increases.
4.6. WEIGHT-OF-EVIDENCE EVALUATION AND CANCER CHARACTERIZATION
Under EPA's Risk Assessment Guidelines of 1986 (U.S. EPA, 1986a, 1987), VC is
classified into cancer weight-of-evidence Category A. Chemicals classified into this category
are considered to be known human carcinogens, based upon sufficient evidence for
carcinogen!city in humans. In the case of VC, sufficient evidence in experimental animal studies
provides additional support for this classification. Under the Proposed Guidelines for
Carcinogen Risk Assessment (U.S. EPA, 1996a, Review Draft, 1999), VC is a known human
carcinogen by the inhalation and oral routes of exposure and highly likely to be carcinogenic by
the dermal route of exposure. This conclusion is based on: (1) consistent epidemiologic
evidence of a causal association between occupational exposure to VC via inhalation and
development of angiosarcoma and hepatocellular carcinoma, the former an extremely rare
tumor; (2) consistent evidence of carcinogenicity in rats, mice, and hamsters via the oral and
inhalation routes, with the critical target site (the liver) being the same in animals and humans;
(3) mutagenicity and DNA adduct formation by VC and its metabolites in numerous in vivo and
in vitro test systems; and (4) efficient absorption via all routes of exposure tested, followed by
rapid distribution throughout the body. The critical target site is the same, and VC is well
absorbed orally. Evidence has also been reported indicating increased sensitivity during early-
life exposure. In light of the very high percentage of angiosarcomas nationwide that are
associated with VC exposure, the evidence for carcinogencity is considered to be strong.
VC carcinogenicity occurs via a genotoxic pathway and is understood in some detail.
VC is metabolized to a reactive metabolite, probably CEO, believed to be the ultimate
carcinogenic metabolite of VC. The reactive metabolite then binds to DNA, forming DNA
adducts that, if not repaired, ultimately lead to mutations and tumor formation. Therefore, a
linear extrapolation was used in the dose-response assessment. An inhalation unit risk of 4.4 x
10"6 per • g/m3 for lifetime exposure during adulthood and 8.8 x 10"6per • g/m3for lifetime
exposure from birth was based on chronic inhalation studies in rats. Because of uncertainty
regarding exposure levels, occupationally exposed cohorts were not utilized to quantitate risk.
28
-------
4.7. SUSCEPTIBLE POPULATIONS
4.7.1. Possible Childhood Susceptibility
In addition to the lifetime cancer studies summarized in Section 4.2, several studies of
partial lifetime exposure suggest that the lifetime cancer risk depends on age at exposure, with
higher lifetime risks attributable to exposures at younger ages. Drew et al. (1983) studied the
effect of age and duration of VC exposure on cancer incidence. Groups of female Fischer-344
rats, Syrian golden hamsters, B6C3F1 mice, and CD-I Swiss mice inhaled VC at 100 ppm for
durations of 6, 12, 18, or 24 months beginning after 0, 6, 12, or 18 months. (Prior to exposure,
animals were 5-6 weeks old when they were received at the testing laboratory; then they were
weighed and observed for 3 weeks.) VC induced angiosarcomas and mammary gland
carcinomas in all four species/strains; in addition, there were hepatocellular carcinomas in rats,
stomach adenomas and skin carcinomas in hamsters, and lung carcinomas in CD-I Swiss mice.
In general, cancer incidence increased with duration of exposure and decreased with age at first
exposure. While early exposure appeared to increase susceptibility, it should be noted that the
animals were near adulthood at the beginning of exposure. Tumor incidences are summarized in
Tables 1 through 4.
Maltoni et al. (1981) investigated the effect of age at exposure as part of a comprehensive
VC study. Groups of male and female Sprague-Dawley rats inhaled 6,000 or 10,000 ppm VC
for 100 hours under different exposure schedules, three groups beginning at 13 weeks of age and
one group beginning at 1 day of age (4 hours/day, 5 days/week for 5 weeks). The angiosarcoma
incidence for rats exposed for 5 weeks as newborns was higher than that for rats exposed for
52 weeks beginning at 13 weeks of age. Moreover, hepatoma incidence, virtually nonexistent in
rats exposed for 52 weeks when mature, approached 50% in rats exposed for 5 weeks as
newborns. While 2-year exposures are likely to induce greater responses, nevertheless it appears
that early-life exposure is at least as effective in liver tumor induction as lifetime exposure
during adulthood. Tumor incidences are summarized in Tables 5 and 6.
29
-------
Table 1. Effects of VC on Fischer-344 rats exposed at different ages
Months exposed
None
0-6
6-12
12-18
18-24
0-12
6-18
12-24
0-18
0-24
Angio-
sarcoma"
incidence
2/112
4/76
2/53
0/53
0/53
12/56
5/55
2/50
15/55
24/55
Mammary
carcinoma
incidence
5/112
6/76
2/53
3/53
2/53
11/56
4/55
0/50
9/55
5/55
Hepato-
carcinomab
incidence
5/112
18/75
16/52
2/51
5/53
24/56
5/54
4/49
15/55
15/55
Mean
induction
time0
NRe
716
613
—
—
671
537
390
643
666
Mean
survival
timed
703
682
703
688
708
634
659
717
575
622
"All sites.
Includes neoplastic nodules.
"Average time required to induce death from angiosarcomas, in days from the day each animal was first exposed.
dAverage lifetime in days from the day the first animals were exposed.
eNot reported.
Source: Qewetal., 1983.
-------
Table 2. Effects of VC on golden Syrian hamsters exposed at different ages
Months
exposed
None
0-6
6-12
12-18
18-24
0-12
6-18
12-24
0-18
0-24
Angio-
sarcomaa
incidence
0/143
13/88
3/53
0/50
0/52
4/52
1/44
0/43
2/103
NRd
Mammary
carinoma
incidence
0/143
28/87
2/52
0/50
1/52
31/52
6/44
0/42
47/102
NRd
Stomach
adenoma
incidence
5/138
23/88
15/53
6/49
0/52
3/50
10/44
3/41
20/101
NRd
Skin
carinoma
incidence
0/133
2/80
0/49
0/46
0/50
2/80
0/38
0/50
3/90
NRd
Mean
induction
timeb
—
NRd
NRd
—
—
NRd
NRd
—
NRd
NRd
Mean
survival
time0
463
390
468
456
499
355
455
424
342
347
aAll sites.
bAverage time required to induce death from angiosarcomas, in days from the day each animal was first exposed.
"Average lifetime in days from the day the first animals were exposed.
dNot reported.
Source: Qewetal., 1983.
-------
Table 3. Effects of VC on B6C3F1 mice exposed at different ages
Months
exposed
None
0-6
6-12
12-18
0-12
6-18
12-24
0-18
Angio-
sarcomaa
incidence
4/69
46/67
27/42
30/51
69/90
30/48
29/48
37/46
Mammary
carcinoma
incidence
3/69
29/67
13/42
4/51
37/90
9/48
4/48
NRb
Mean
induction
time"
NRd
343
344
343
313
319
304
313
Mean
survival
time0
780
316
480
695
301
479
632
304
aAll sites.
bAverage time required to induce death from angiosarcomas, in days from the day each animal was first exposed.
"Average lifetime in days from the day the first animals were exposed.
dNot reported.
Source: Qewetal., 1983.
32
-------
Table 4. Effects of VC on CD-I Swiss mice exposed at different ages
Months
exposed
None
0-6
6-12
12-18
0-12
6-18
12-24
0-18
Angio-
sarcoma"
incidence
1/71
29/67
11/49
5/53
30/47
17/46
3/50
20/45
Mammary
carcinoma
incidence
2/71
33/67
13/49
2/53
22/47
8/45
0/50
9/55
Lung
carcinoma
incidence
5/112
18/75
16/52
2/51
24/56
5/54
4/49
22/45
Mean
induction
timeb
NRd
369
340
226
350
323
124
350
Mean
survival
time0
474
340
472
521
347
443
472
321
aAll sites.
bAverage time required to induce death from angiosarcomas, in days from the day each animal
awfirst exposed.
"Average lifetime in days from the day the first animals were exposed.
dNot reported.
Source: Qewetal., 1983.
33
-------
Table 5. Comparison of newborn and later short-term exposure to VC
Administered
concentration
(ppm) Angiosarcomasa Hepatomas
4 hours/day, 5 days/week for 5 weeks starting at age 13 weeks:
6,000 3/120 0/120
10,000 2/118 1/118
1 hour/day, 4 days/week for 25 weeks starting at age 13 weeks:
6,000 5/118 0/118
10,000 4/119 0/119
4 hours/day, 1 day/week for 25 weeks starting at age 13 weeks:
6,000 4/120 2/120
10,000 4/120 0/120
4 hours/day, 5 days/week for 5 weeks starting at age 1 day:
6,000 20/42 20/42
10,000 18/44 20/44
"All sites, including angiomas.
Source: Maloni et al., 1981 (experiments BT14 and BT1).
Table 6. Comparison of newborn exposure and later chronic exposure to VC
Administered
concentration
(ppm)
10,000
6,000
Angio-
sarcomasa in
newborn ratsb
18/44
20/42
Angio-
sarcomas" in
mature rats0
13/46
22/42
Hepatomas in
newborn ratsb
20/44
20/42
Hepatomas in
mature ratsc
1/24
1/27
aAll sites, including angiomas.
bExposed 4 hours/day, 5 days/week for 5 weeks beginning at 1 day of age.
"Exposed 4 hours/day, 5 days/week for 52 weeks beginning at 13 weeks of age.
Source: Maloni et al., 1981 (experiments BT14 and BT1).
34
-------
Mechanistic studies are consistent with these tumor findings and suggest factors
associated with early-life sensitivity. Laib et al. (1979) found that VC induces preneoplastic foci
in newborn, but not mature, rats. In a subsequent study, Laib et al. (1985) studied the effect of
age on induction by VC of hepatic adenosine-5'-triphosphatase (ATPase) deficient enzyme-
altered foci, a putative precursor of hepatocellular carcinoma. Groups of newborn male and
female Wistar rats inhaled 2,000 ppm VC for different periods of time; their livers were
evaluated at 4 months. The investigators concluded that "the induction of pre-neoplastic
hepatocellular lesions in rats by vinyl chloride is restricted to a well defined period
(approximately day 7 to 21) in the early lifetime of the animals." The lack of response in the first
5 days to the lack of hepatocellular proliferation and the low rate of VC metabolism at this stage
of development.
Laib et al. (1989) found that inhaled radiolabeled VC was incorporated into physiological
purines of 11-day-old Wistar rats at eightfold higher levels than in similarly treated adult rats
(presumably reflecting DNA replication activity), and roughly fivefold higher levels of the DNA
adduct OEG were found in the livers of young animals (reflecting an increased alkylation rate).
Although OEG is not believed to be a precarcinogenic lesion, it is reasonable to expect that its
levels are correlated with levels of other precarcinogenic adducts. In a similar study, Fedtke
et al. (1990) observed roughly fourfold greater concentrations of both OEG and N2,3-EG in
preweanling rats exposed to VC.
An increased incidence of altered hepatocellular foci was noted among mature animals
that were exposed in utero and neonatally as compared to those that were not (CMA, 1998b).
This increased incidence could have been due to exposure during these susceptible periods of the
life cycle but could also have been due merely to longer overall exposure. Also, basophilic foci
were observed in female rat liver in the study of Til et al. (1983, 1991) among animals that were
exposed beginning at 5 weeks of age.
As discussed above and in Section 5.3.5.1, several studies provide evidence for increased
sensitivity to VC-induced carcinogenesis in early-life and prenatal exposures in experimental
animals. Early-life data on humans, however, are lacking because most exposures have been
limited to occupational groups. Nevertheless, many of the factors likely to be responsible for
early-life sensitivity in animals are present in humans. Recommended adjustments to
quantitative risk estimates to account for early-life sensitivity are given in Section 5.3.5.1.
4.7.2. Possible Gender Differences
Human evidence is unavailable regarding possible sex differences in sensitivity to health
effects from exposure to VC. Cohorts evaluated in epidemiology studies have been primarily
male workers. Evidence from case reports is also lacking. Maltoni et al. (1981, 1984) reported
only small differences in liver cancer susceptibility in either rats or mice exposed via inhalation
to VC, although female rats did show the greatest response In feeding studies with rats,
neoplastic nodules and preneoplastic alterations such as basophilic foci were induced at lower
concentrations in females (Til et al., 1983, 1991). There was also some indication of increased
susceptibility to induction of nonneoplastic pathological changes such as liver cysts. In this
study, females had higher incidences of liver tumors than males. While no definite conclusions
35
-------
can be made regarding possible human sex differences in susceptibility to liver tumor induction
by VC, a conservative approach of basing both oral and inhalation risk estimates on the female
rat data was nevertheless considered to be prudent.
There is some evidence for an increase in mammary tumors in female rats, However,
these tumors in rats occurred sporadically, without a positive dose-response relationship, and
appear to occur in strains with a high background rate of mammary tumors. Moreover, there
have been no reports of breast cancer induction in humans associated with VC exposure.
5. DOSE-RESPONSE ASSESSMENTS
5.1. ORAL REFERENCE DOSE (RfD)
5.1.1. Choice of Principal Study and Critical Effect
Two related chronic dietary studies of VC in rats exist (Feron et al., 1981; Til et al., 1983,
1991). Til et al. (1983, 1991) are the unpublished and published versions of the same study,
conducted under the same conditions as the Feron et al. (1981) study, but at lower doses. As
discussed in Section 4.2, altered hepatocellular foci observed in the Til et al. study (1983, 1991)
are likely to be preneoplastic lesions produced via a genotoxic mechanism, consistent with the
known mechanism of VC carcinogen!city. The Agency for Toxic Substances and Disease
Registry (ATSDR, 1995) derived a chronic oral minimal risk level (MRL) based on the
basophilic foci observed in the Til et al. (1983, 1991) study at the lowest administered dose
tested (0.018 mg/kg-day). However, that document does not address the preneoplastic nature of
this lesion, and the authors do not appear to have considered whether a preneoplastic endpoint is
appropriate for the derivation of an MRL.
Based on these considerations of protocol and results, the Til et al. (1983, 1991) study
was used in the derivation of the RfD. This was a well-conducted chronic dietary study with
adequate numbers of rats that found an increased incidence of two nonneoplastic endpoints, liver
cell polymorphism and cysts, at a LOAEL of 1.3 mg/kg-day and a NOAEL at 0.13 mg/kg-day.
Cysts, described as proliferating bile duct epithelium, are not considered to be precursors of
hepatocellular tumors because tumors did not develop from this location. Liver cell
polymorphism was described as affecting both the nucleus and cytoplasm of the liver cells and is
considered to be a toxic rather than a carcinogenic effect (Schoental and Magee, 1957, 1959;
Afzelius and Schoental, 1967). All other significant findings in this study were either neoplastic
or preneoplastic (see Section 4.2).
36
-------
5.1.2. Methods of Analysis—Including Models (PBPK, BMD, etc.)
5.1.2.1. PBPKModd
The oral RfD, inhalation RfC, oral cancer slope factor, and inhalation unit risk were all
derived using a PBPK model to extrapolate animal exposure data to humans. Therefore, general
aspects of the model are described here, and aspects specific to inhalation noncancer toxicity and
to carcinogenesis are described in Sections 5.2.2 and 5.3.2, respectively.
The PBPK model for VC developed by Clewell et al. (1995a) is shown in Figure 2. The
model is basically an adaptation of a previously developed PBPK model for vinylidene chloride
(D'Souza and Andersen, 1988). This model was also used to develop independent cancer risk
estimates for VC (Clewell et al., 1995c). For a poorly soluble, volatile chemical like VC, only
four tissue compartments are required: a richly perfused tissue compartment that includes all of
the organs except the liver, a slowly perfused tissue compartment that includes all of the muscle
and skin tissue, a fat compartment that includes all of the fatty tissues, and a liver compartment.
The model also assumes flow-limited kinetics, or venous equilibration, that is, that the transport
of VC between blood and tissues is fast enough for steady state to be reached within the time it is
transported through the tissues in the blood.
Metabolism of VC was modeled by two saturable pathways, one high affinity, low
capacity (with parameters VMAX1C and KM1) and one low affinity, high capacity (with
parameters VMAX2C and KM2). Subsequent metabolism is based on the metabolic scheme
shown in Figure 1: the reactive metabolites (whether CEO, CAA, or other intermediates) may
then either be metabolized further, leading to CO2, react with GSH, or react with other cellular
materials, including DNA. Because exposure to VC has been shown to deplete circulating levels
of GSH, a simple description of GSH kinetics was also included in the model.
The model is capable of route-to-route extrapolation, as either oral and inhalation
exposures may be entered and common dose metrics calculated either at the liver or in the whole
body. The model is also capable of interspecies extrapolation because it is parameterized for
humans and several different rodent species such that common dose metrics can be calculated
for any of these species. Conversion of various oral, intermittent animal, and intermittent human
exposures to a continuous human exposure concentration (i.e., an HEC) is accomplished by
comparing the common dose metrics to those obtained from running the model with human
parameters under continuous exposure conditions. For example, a specific mg/kg-day dose from
an animal feeding study can be converted by the animal-parameterized model to a dose metric at
the liver in terms of mg metabolites/volume of liver. This dose metric can be compared with
those calculated from the human-parameterized model (also in terms of mg metabolites/volume
of liver) run under conditions of a continuous inhalation exposure to obtain a human dose that
would correspond to the specific dose of an animal feeding study.
37
-------
*[
kl
Lung
QC
Fat
QF
Rich
CA
QR
CVS
slow
CA
QS
CVL
Liver
VMAX1
KM1
CO2
KCO,
QL
VMAX2
KM2
Reactive
Metabolites
(RISK)
KGSM
KZER
""KA
Glutaihione
Conjugate
(RISKG)
KFEE
I
KGSM
GSH
Tissue/DNA
Adducts
(RISKM)
KSTKO
KB
Figure 2. The PBPK model for vinyl chloride developed by Clewell et al. (1995a).
38
-------
A complete description of the model, including the rationale for parameter choices in
animals and humans, choice of dose metric, and experimental information used to calibrate and
optimize the model, is in Appendix B. It is noted here and elsewhere in this document (Section
6 and Appendices A, B and D) that the inhalation portion of this model is well documented, with
experimental inhalation data sufficient to impart a relatively high degree of confidence in dose
metrics derived from inhalation scenarios. Dose metrics derived from oral scenarios do not have
nearly the amount of data necessary to impart an equivalent level of confidence. To compensate
for this uncertainty, procedures have been instituted in the oral exposure input to ensure that
estimates of oral dose metrics would be "worst case" and conservative of public health.
Based on the analysis in Section 4.5, the liver toxicity endpoints in the Feron et al. (1981)
and Til et al. (1983, 1991) studies were considered appropriate for the derivation of the RfD. As
discussed in Section 4.4, the noncancer effects are believed to be due to reactive metabolites,
possibly CAA. The most appropriate pharmacokinetic dose metric for a reactive metabolite is
the total amount of the metabolite generated divided by the volume of the tissue into which it is
produced (Andersen et al., 1987) and is the designated "RISK" in the output of the PBPK model.
For liver toxicity/carcinogenicity, all metabolism was assumed to occur in the liver, while
testicular toxicity was assumed to be due to metabolism that occurred in the testes. The dose
metric chosen for the testes is the total amount of the metabolite generated (scaled across species
based on body weight) divided by body weight and is designated "AMET" in the output of the
model.
Reitz et al. (1996) developed a similar PBPK model, with a description of parent
chemical kinetics and total metabolism based on the styrene model of Ramsey and Andersen
(1984). Metabolism of VC was modeled with a single saturable pathway, and the kinetic
constants were estimated from fitting of closed chamber gas uptake data with rats. The structure
of the parent chemical portion of the Reitz et al. (1996) and Clewell et al. (1995a) models is
essentially identical; only the descriptions of metabolism in the two models differ substantially.
As discussed above, the model of Clewell et al. (1995a) includes a more complex description of
metabolism, with two saturable oxidative pathways rather than one, and with a description of
GSH conjugation of the oxidative metabolites. Nevertheless, dose metrics calculated on a
common set of data using the two models are in close agreement, as demonstrated in
Appendix A.
For the noncancer oral and inhalation assessments for VC, dose metrics were calculated
for liver cell polymorphisms reported in the chronic rat dietary study of Til et al. (1983, 1991)
(Appendix D). In order to convert these dose metrics to the human equivalent dose, a human
dose metric was generated from a sample continuous human exposure scenario where
stimulation of ingestion of 1 ppm in water (0.0286 mg/kg-day assuming 2 L/day/70 kg person)
yielded a human dose metric of 1.01 mg/L liver. Because VC metabolism is linear in this dose
range, the ratio of the intake and dose metric provides a factor (1.01/.0286 = 35.31) for
converting from male and female rats at the NOAEL ([3.03 + 2.96] H- [2 x 35.31]) to obtain the
NOAEL(HEC) = 0.09 mg/kg-day. The corresponding LOAEL(HEC) is 0.85 mg/kg-day ([30.2 +
29.5]-[2 x 35.31]).
5.1.2.2. BMD Calculation
39
-------
The same dose metric (mg metabolite/L liver) was calculated for the dose groups (males
and females combined) and the benchmark analysis performed on this metric and the incidence
of liver cell polymorphism (males and females combined) reported in the Til et al. (1983, 1991)
study. The (Appendix D, Table D-6) analysis shows data limitations (only one nonzero
datapoint in the dataset and wide dose-spacings) and wide variability in the responses from
various models. As a consequence, the benchmark analysis was not used for quantitation in this
assessment.
5.1.3. RfD Derivation
The NOAEL for liver cell polymorphism in the Til et al. (1983, 1991) study is 0.13
mg/kg-day, and the LOAEL is 1.3 mg/kg-day. Using the PBPK model of Clewell et al.
(1995a,b), the corresponding human NOAEL and LOAEL are 0.09 and 0.9 mg/kg-day,
respectively.
An uncertainty factor of 10 was used for protection of sensitive human subpopulations
and 3 for animal-to-human extrapolation. The uncertainty factor for intraspecies variability
includes the variability in risk estimates that would be predicted by the model for different
individuals because of variability in physiology, level of activity, and metabolic capability. A
factor of 3 was used for interspecies extrapolation because, although PBPK modeling refines the
animal-to-human comparison of delivered dose, it does not address the uncertainty regarding the
toxicodynamic portion of interspecies extrapolation (relating to tissue sensitivity). As the mode
of action for the noncancer hepatic effects is perhaps more unclear than that of cancer (see
Section 4.5), and as there exists some limited and problematic evidence of human susceptibility
to certain hepatic effects from VC (Ho et al., 1991) the toxicodyanamic component of the
interspecies uncertainty is retained. For cancer effects this situation is somewhat reversed such
that there is less uncertainty about the toxicodynamics for carcinogenic effects. Although there
is relative uncertainty in this assessment with regards to the derivation of the dose metrics from
oral settings, it is offset by the conservative manner in which these metrics were derived, and no
extra uncertainty factors are considered necessary.
No modifying factor is proposed for this assessment. Although testicular effects were
reported in a study by Bi et al. (1985), the effects occurred at exposure levels that would result in
a higher value RfD (Appendix D). Developmental and other effects were noted only at high
concentrations (Appendix D, Table D-2). Based on these considerations, the following RfD was
derived:
RfD = 0.09 mg/kg-day - 30 = 0.003 mg/kg-day = 3E-3 mg/kg-day.
40
-------
5.2. INHALATION REFERENCE CONCENTRATION (RfC)
5.2.1. Choice of Principal Study and Critical Effect
The RfC is based on liver cell polymorphism and cysts observed in the chronic dietary rat
study of Til et al. (1983, 1991). Several lines of reasoning justify this choice. The
NOAEL(HEC) from the Til et al. (1983, 1991) study was calculated (see discussion of PBPK
model below) and in Appendix D) at 2.5 mg/m3. This concentration is far lower than the
LOAEL(HEC) of the transient increase in liver weight observed in the study of Bi et al. (1985) at
28 mg/m3, indicating liver cell polymorphism to be the more sensitive endpoint. More detailed
analysis of data in the Bi et al. (1985) study is not possible, because the study authors reported
only the data for those changes that were considered significant, and body weight data were not
reported. In addition, the power to detect an effect at 12 months was limited by the small
number of animals sacrificed (n = 6), compared with the 30 animals sacrificed at 6 months. The
NOAEL(HEC) for liver effects in the 10-month inhalation study of Sokal et al. (1980) was
calculated at 93 mg/m3 and a NOAEL(HEC) for testicular effects considerably higher at 145
mg/m3. A NOAEL(HEC) of 93 mg/m3 based on liver effects was also estimated for the 6-month
inhalation study of Torkelson et al. (1961). An RfC possibly could be derived from among these
inhalation studies which, with the application of sufficient uncertainty factors, could be made
quantitatively comparable to that derived with the Til study. However, the experimental
strengths of the Til study relative to the inhalation studies, including, in addition to the lifetime
exposure, large group sizes and extensive reporting of results, clearly give a qualitative
advantage to the the Til study that would be reflected not only in lower uncertainty but
concomitantly in higher confidence. Although the attributes of the Til study are offset somewhat
by the uncertainty associated with derivation of the oral dose metrics, it is still judged to be the
most valid choice for the principal study.
The numerous occupational studies reporting incidence of liver angiosarcomas (e.g.,
Creech and Johnson, 1974; Waxweiler et al., 1976; Byren et al., 1976) and other liver effects in
humans (Ho et al., 1991) are of limited usefulness for purposes of quantitative assessment owing
principally to deficiencies in exposure information. These studies, however, do provide a clear
link of relevancy to the animal data of Til et al. (1983, 1991) in that liver tumors and liver effects
remain as the basis of the assessment.
ATSDR (1995) based an intermediate-duration inhalation MRL on increased relative
liver, heart, and spleen weights in the Bi et al. (1985) study. Because a pharmacokinetic model
was not used, the oral studies of Feron et al. (1981) and Til et al. (1983, 1991) were not an option
for ATSDR. Interpretation of the organ weight data in the Bi et al. (1985) study is complicated
by the fact that the study did not report absolute organ weights, relative weights for groups with
no significant differences, standard deviations, or histopathology results (except in the testes).
Other endpoints in these and other studies occurred at higher exposure levels and thus
were not considered as appropriate for the critical effect as in the liver. These endpoints
included increased incidence of damage to the testicular seminiferous tubules in rats (Bi et al.,
1985), increased liver weight and liver lesions (Sokal et al., 1980), increased damage of
spermatogenic epithelium (Sokal et al., 1980), increased liver weight (Torkelson et al., 1961;
41
-------
Wisniewska-Knypl et al., 1980), and lipid accumulation (Wisniewska-Knypl et al., 1980). These
efforts are quantitatively compared in Appendix D.
5.2.2. Methods of Analysis—Including Models (PBPK, BMC, etc.)
5.2.2.1. Route-to-Route Extrapolation
Deriving an inhalation RfC from an oral study requires route-to-route extrapolation.
Agency guidelines (U.S. EPA, 1994b) indicate that VC would be a candidate chemical for this
extrapolation as adequate toxicity data exist from one route (oral), and the observed toxicity is
observed in the liver, remote from the portal of entry. The concurrence of liver as the target
organ across routes, the substantial amount of kinetic information available on VC, the lack of
any reported portal-of-entry effects in existing inhalation studies, and the development of several
PBPK models for VC make this chemical an even more compelling candidate for this
extrapolation procedure.
5.2.2.2. PBPK Model
The PBPK model described in Section 5.1.2 was used to extrapolate inhalation
concentrations from the oral data of Til et al. (1983, 1991) by calculating a dose metric (mg VC
metabolite/L liver) that would be common for both the oral and inhalation routes of exposure.
As information for the oral route of exposure was limited, conservative assumptions of 100%
oral absorption over a continuous 24-hr period were made to maximize the formation of the
reactive species. This same metric served as a basis to calculate HECs such that the overall
transformation of data was from mg/kg-day oral intake in animals to an air concentration for
continuous human exposure.
The following procedure was employed in the route-to-route extrapolation with the
chronic oral study of Til et al. (1983, 1991). The dose metric (termed "RISK") for the animal
NOAEL was determined by the PBPK model, i.e., the value of the total metabolites per liver
volume for rats exposed to 0.13 mg/kg-day. This metric was calculated to be 3.00 mg/L liver
(from the average of the male value of 3.03 and the female value of 2.96). The PBPK model was
then exercised to determine the same dose metric for a continuous human inhalation exposure.
The results from a range of exposure concentrations (1 • g/m3 to 10,000 mg/m3) showed that the
relationship with "RISK" was linear up to nearly 100 mg/m3 with the factor in the linear range
being 1.18 mg/L liver per mg/m3 VC (Appendix D, Table D-3). This factor was then used to
convert this metric to a continuous human inhalation exposure. Conversion of the study NOAEL
of 0.13 mg/m3 was then accomplished by dividing the animal dose metric for this concentration
by the conversion factor (3.00/1.18) to arrive at a NOAEL(HEC) of 2.5 mg/m3. For the
LOAEL(HEC), the figures and calculation are (29.9/1.18) or 25.3 mg/m3.
5.2.2.3. BMC Calculation
42
-------
For the noncancer oral and inhalation assessments for VC, dose metrics were calculated
from the PBPK model for liver cell polymorphism and cysts reported in the chronic rat dietary
study of Til et al. (1983, 1991). The same dose metric (mg metabolite/L liver) was calculated
for the dose groups (males and females combined) and benchmark analysis was performed on
this metric and the incidence of liver cell polymorphism (males and females combined) reported
by Til et al. (1983, 1991). The analysis (Appendix D) shows data limitations (only one nonzero
datapoint in the dataset and wide dose-spacings) and wide variability in the responses from the
various models. As a consequence, the benchmark analysis was not used for quantitation in this
assessment.
Even though benchmark dose/concentration was not appropriate for the analysis of the
critical effects in Til et al. (1983, 1991), BMCs calculated for the other inhalation studies were
shown to be considerably higher than for the liver polymorphism endpoint. BMC(HEC) values
corresponding to a benchmark response (BMR) of 10% extra risk were 182 mg/m3 for damage to
the testicular seminiferous tubules in rats (Bi et al., 1985); 59 mg/m3 for polymorphism of
hepatocytes in the 10-month inhalation study of Sokal et al. (1980); 92 mg/m3 for proliferation of
reticuloendothelial cells, although these may be preneoplastic (Sokal et al., 1980); and 122
mg/m3 for damage to the spermatogenic epithelium (Sokal et al., 1980). Proliferation of
reticuloendothelial cells, however, may be preneoplastic. The only continuous endpoint that
could be modeled was increased liver weight in the studies by Sokal et al. (1980) and
Wisniewska-Knypl et al. (1980), which reported the same data. The most sensitive BMC(HEC)
for this endpoint (168 mg/m3) was obtained with the BMR defined as a change in the mean of
sd0/2, using the polynomial model. The increased relative liver weight observed by Torkelson et
al. (1961) could not be modeled, but the NOAEL(HEC) based on the tissue dose was 93 mg/m3.
Similarly, the LOAEL(HEC) for lipid accumulation (Wisniewska-Knypl et al., 1980) was 79
mg/m3.
5.2.2.4. Application of Uncertainty Factors (UF) and Modifying Factors (MF)
The rationale for choice of the critical effect and principal study for the RfC is the same
used for the oral RfD, i.e., the analysis in Section 4.5. The NOAEL(HEC) derived using the
internal dose metric for liver cell polymorphism and cysts from the oral feeding study of Til et
al., 1983, 1991) was 2.5 mg/m3. Section 4.5 and Appendix D, Table D-2, demonstrated that
NOAEL/LOAELs of other noncancer effects in both oral and inhalation studies were higher then
those noted for the incidence of liver cell polymorphisms and hepatic cysts.
As for the RfD, an uncertainty factor of 10 was used for protection of sensitive human
subpopulations and 3 for animal-to-human extrapolation. The uncertainty factor for intraspecies
variability includes the variability in risk estimates that would be predicted by the model for
different individuals because of variability in physiology, level of activity, and metabolic
capability. A factor of 3 was used for interspecies extrapolation because, although PBPK
modeling refines the animal-to-human comparison of toxicokinetics, it does not address the
uncertainty regarding the toxicodynamic portion of interspecies extrapolation (relating to tissue
sensitivity). As the mode of action for the noncancer hepatic effects is perhaps more unclear
than that of cancer (see Section 4.5), and as there exists some limited and problematic evidence
43
-------
of human susceptibility to certain hepatic effects from VC (Ho et al., 1991), the toxicodynamic
component of the interspecies uncertainty factor is retained. For cancer effects, this situation is
reversed such that there is less uncertainty about the interspecies toxicodynamics for
carcinogenic effects. Although there is relative uncertainty in this assessment with regards to the
derivation of the dose metrics from oral settings, it is offset by the conservative manner in which
these metrics were derived, and no extra uncertainty factors are considered necessary.
No modifying factor is proposed for this assessment because the quality of the critical
study was high, and because effects measured at organs other than the liver occurred only at
considerably greater exposure levels. Based on these considerations, the following RfC was
derived:
RfC = 2.5 mg/m3 - 30 = 1E-1 mg/m3.
5.3. CANCER ASSESSMENT
As discussed in Section 4.6, VC is considered to be a known human carcinogen by the
oral and inhalation route, and highly likely to be carcinogenic by the dermal route of exposure.
The weight of evidence is based upon (1) consistent epidemiologic evidence of a causal
association between occupational exposure to VC via inhalation and development of liver
angiosarcoma; (2) consistent evidence of carcinogenicity in rats, mice, and hamsters via the oral
and inhalation routes; (3) mutagenicity and DNA adduct formation by VC and its metabolites in
numerous in vivo and in vitro test systems; and (4) efficient VC absorption via all routes of
exposure tested, followed by rapid distribution throughout the body.
5.3.1. Choice of Study/Data With Rationale and Justification
5.3.1.1. Human Data
As discussed in Section 4.1, numerous human studies have documented the association
between occupational exposure to VC and the development of angiosarcomas and other cancers.
Three of these studies were used to develop dose-response assessments (Fox and Collier, 1977;
Jones et al., 1988; and Simonato et al., 1991). Because exposure was not adequately
characterized in these studies, recommended potency estimates were based on animal bioassay
data. The cancer potency estimates derived from these studies do, however, provide support for
the recommended values. The most detailed exposure information was provided by Fox and
Collier (1977). In this study, since only four deaths from liver cancer (two of which were
angiosarcoma) were recorded, a high degree of uncertainty in relative risk adds to the exposure
uncertainty. In the Jones et al. study, an update of Fox and Collier, adequate exposure data were
available only for autoclave workers, for which seven liver angiosarcoma deaths were recorded.
The Simonato et al. (1991) study has the largest cohort and the most liver deaths (24) but less
accurate exposure information because data were collected from several different workplaces,
and because of possible misclassification of workers. The PBPK model of Clewell et al. (1995a)
was used to calculate a cumulative internal dose metric for these studies. Because VC
metabolism begins to be nonlinear at the high exposure levels in these studies, cumulative
44
-------
exposure (e.g., ppm-years) was not sufficient for a quantitative assessment. Instead, only data
sets providing information on both exposure level and duration (or cumulative exposure, from
which duration could be estimated) were considered appropriate for modeling. Risk estimates
(95% upper bound) derived using these three studies ranged from 2.8 x 10"7 to 2.8 x 10"6 per
• g/m3 VC. An earlier estimate by Chen and Blancato (1989), based on the results of Fox and
Collier (1977) was also within this range. (See Appendix B for details).
5.3.1.2. Animal Data
Three studies were located that provided data on the oral carcinogen!city of VC. The oral
cancer assessment was based on a well-conducted study by Feron et al. (1981) in which rats
were administered VC in the diet for 135 or 144 weeks. VC volatilization and VC in the feces
were measured to ascertain actual intake of VC. A related dietary study was conducted at lower
doses (Til et al., 1983, 1991), but this study did not provide adequate dose-response information,
since the tumors were found only at the highest dose. Maltoni et al. (1981, 1984) conducted a
carcinogen!city study of VC administered by vegetable oil gavage to male and female rats. The
data from this vegetable oil gavage study were not considered appropriate as the basis for a risk
assessment. Chloroform administered in corn oil has been shown to have stronger
hepatotoxicity than the same doses administered in an aqueous suspension (Bull et al., 1986).
Corn oil has also been shown to increase peroxisomal oxidative enzyme activity in rats
(DeAngelo et al., 1989); peroxisomal proliferators have been shown to be hepatic tumor
promoters. Thus, the toxicity and promotional environment created in the liver by continual
dosing with large volumes of vegetable oils could potentiate the effects of genotoxic carcinogens
in the liver. For this reason, the single gavage dose used in the Feron et al. (1981) study was also
not included in the dose-response assessment. Finally, the PK model used in this assessment has
limited, poor data with which to calibrate this administration route (Appendix B).
Evidence for enhanced sensitivity to the carcinogenic effects of VC during early-life
exposure were provided by Maltoni et al. (1981), Drew et al. (1983), and Laib et al. (1985).
Since these studies are inadequate to develop dose-response estimates, recommended estimates
are based upon the long-term oral study by Feron et al. (1983) and inhalation studies reported by
Maltoni et al. (1981, 1984). The former studies, however, did provide sufficient evidence for
recommending a twofold adjustment of risk for to account for early-life exposure. For details
see Section 5.3.6.
Molecular toxicology data suggest that the VC-induced liver angiosarcomas and
hepatocellular carcinomas in rodents develop via different pathways. Knockout of the p53
tumor suppressor gene in mice results in the spontaneous development of angiosarcoma, along
with malignant lymphoma, but not hepatocellular carcinoma (Donehower et al., 1992). In
contrast, accelerated development of hepatocellular carcinoma in rodents is associated with
overexpression of the myc and ras oncogenes (Sandgren et al., 1989), but not with mutational
loss of p53 function (Greenblatt et al., 1994). The data therefore suggest that the hepatocellular
tumors and possibly the neoplastic nodules observed in rodents may occur via a p53-independent
mechanism, more likely related to myc and ras, while the angiosarcomas develop via a p53-
dependent mechanism.
45
-------
Chemically induced human liver carcinogenicity is associated with mutational alteration
of multiple genes, consistent with a mutagenic mode of action. Mutations in the p53 tumor
suppressor gene are the most common gene alteration identified in human cancers and have been
associated with aflatoxin-induced human hepatocellular carcinoma (Greenblatt et al., 1994). Ras
oncogene mutations have also been found in human liver cancers (Bos, 1989), and VC-induced
human angiosarcoma is associated with frequent mutation of ras oncogenes (DeVivo et al.,
1994). In fact, the presence of both mutant ras and p53 tumor suppressor genes has a predictive
value of 0.67 for liver tumors in humans (Marion et al., 1996). On the basis of these studies, it
has been suggested that chemicals that act through a p53-dependent process are more likely to be
trans-species carcinogens than those that act through a p53-independent process such as ras
activation (Tennant et al., 1995; Goldsworthy et al., 1994). As noted above, however, both p53
and ras mechanisms appear to be implicated in human liver cancers.
According to EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 1986a),
when significant increases in tumor induction occur at more than one site, animals with tumors
at all such sites are included in the total, unless mechanistic data are sufficient to discount them.
Animals with either liver angiosarcoma or hepatocellular tumors were included for quantitating
risk because both tumor types were significantly increased in the Feron et al. (1981) oral
exposure study. Neoplastic nodules in the liver were also included because they are considered
equivalent to adenomas and also because it is considered likely that they will progress to
carcinomas if survival duration is sufficient. Although the increase in hepatocellular tumors was
smaller and nonsignificant in the Maltoni et al. (1981, 1984) inhalation studies, they were
counted for quantitation because liver tumors were associated with VC exposure in the Feron et
al. (1981) study. This decision is supported by evidence that, even though the majority of liver
tumors reported in VC-exposed workers were angiosarcomas, some hepatocellular tumors, also a
rare tumor type in humans, were usually noted (Wong et al., 1991; CMA 1998a). Lack of
individual animal data could result in counting some animals twice. However, because of the
small number of hepatocellular tumors any errors are likely to be minimal.
Several animal studies investigated the carcinogenicity of VC via the inhalation route.
Maltoni et al. (1984) conducted the most thorough analysis, in which male and female mice and
rats were exposed to a wide range of VC concentrations for 30 weeks (mice) or 1 year (rats) and
then followed through 135 weeks after the initiation of exposure. Other studies did not
characterize the concentration-response curve as well (Bi et al., 1985; Hong et al., 1981;
Keplinger et al., 1975; Lee et al., 1978) or did not observe angiosarcomas (Feron and Kroes,
1979; Viola et al., 1971). For a review of recent results of human and animal exposures to VC,
mechanistic data, DNA reactivity, and attempts at cross-species extrapolation of cancer risk see
Whysneretal. (1996).
In accordance with the Proposed Guidelines for Carcinogen Risk Assessment (U.S. EPA,
1996a), the potential for using preneoplastic endpoints as a basis for the cancer assessment was
evaluated. Two potential preneoplastic changes were considered: altered hepatocellular foci,
i.e., clear cell foci, basophilic foci, and eosinophilic foci (Feron et al., 1981; Til et al., 1983,
1991) and DNA adducts (Swenberg et al., 1992; Morinello et al., 1999). The altered
hepatocellular foci might be used to extend the tumor dose-response curve to lower doses,
reducing the amount of extrapolation necessary to reach the exposure levels of interest. In order
to conduct such an extrapolation, it would be necessary to determine a correspondence factor
46
-------
between the incidence of foci and the tumor incidence in the portion of the dose-response curve
where both foci and tumors are observed. No attempt was made to conduct such a calculation,
however, because the observed foci are precursors to hepatocellular carcinoma, while
angiosarcomas, the tumor type of greatest relevance to human risk assessment, are derived from
sinusoidal cells. Proliferation of sinusoidal cells was also observed in these studies, but the
incidence did not achieve statistical significance, and any increased response did not extend to
doses below those at which angiosarcomas were observed.
As discussed in Section 4.4, VC exposure results in the formation of DNA adducts, and
four highly persistent ethenoguanine-DNA adducts have been associated with VC
carcinogenicity (Swenberg et al., 1992). More recently Morinello et al. (1999) reported a steep
dose-response for N2,3-ethenoguanine adducts at low VC exposure concentrations, with a
leveling off at higher concentrations, a response consistent with both metabolic activation rates
and tumor induction. Adduct levels normally cannot be used directly to extend tumor dose-
response data to lower doses, since tumor formation from adducts depends on many factors,
including the consequences of adduct repair or failure to be repaired. Thus, although a
quantitative analysis of the relationship between VC metabolism, adduct formation, and tumor
formation is likely to be a fruitful area for additional research, it is premature to attempt to
establish a quantitative link between the tissue concentrations of a specific adduct and the risk of
cancer in that tissue.
5.3.2. Dose-Response Data
Oral cancer risk was calculated based on the incidence of combined liver angiosarcomas,
hepatocellular carcinomas, and neoplastic nodules in female Wistar rats in the dietary study of
Feron et al. (1981). Data on females was utilized because their greater sensitivity. The
administered doses and tumor incidences are shown in Table 7.
Inhalation cancer risk was calculated based on the incidence of liver angiosarcoma,
angioma, hepatoma, or neoplastic nodules in the inhalation study of Maltoni et al. (1981, 1984),
conducted with female Sprague-Dawley rats. The incidence is shown in Table 8.
5.3.3. Dose Conversion
Doses were not converted to human equivalents prior to the calculation of risk. Instead,
the risk modeling (linearized multistage [LMS] or the dose associated with a lifetime cancer risk
of 10% [LED 10]) was conducted based on the animal dose metric to the liver "RISK." Then,
consistent with the statement that". . . tissues experiencing equal average concentrations of the
carcinogenic moiety over a full lifetime should be presumed to have equal lifetime cancer risk"
47
-------
Table 7. Dose and tumor incidence from oral administration of vinyl chloride to
female Wistar rats (from Feron et al., 1981).
Administered dose Human equivalent dose3 Tumor incidence
(mg/kg-day) (mg/kg-day) female rats
0 0 2/57
1.7 1.07 28/58
5.0 3.13 49/59
14.1 8.77 56/57
a Continuous human exposure over a lifetime required to produce an equivalent mg metabolite/L liver.
Table 8. Dose and tumor incidence from inhalation of vinyl chloride by female
Sprague-Dawley rats (from Maltoni et al., 1981,1984).
Exposure concentration
(ppm)a
0
1
5
10
25
50
100
150
200
250
500
2,500
6,000
Human equivalent
concentration
(ppm)b
0
0.20
0.98
1.95
4.60
10.1
19
26
31
35
40
48
51
Tumor incidence'
0/141
0/55
0/47
1/46
5/40
1/29
1/43
5/46
10/44
3/26
11/28
10/24
13/25
aAnimals exposed 4 hours/day 5 days/week for 52 weeks.
bContinuous human exposure concentration over a lifetime required to produce an equivalent mg metabolite/liter of
liver.
cAnimal numbers were adjusted to include those surviving until detection of the first liver tumor.
48
-------
(U.S. EPA, 1992), the calculated risk values based on the dose metric were assumed to
correspond to the same risk for the same human dose metric. In order to convert the human dose
metric to a human dose, the model was run for a sample human continuous oral exposure (1
mg/L in drinking water) to determine the dose of metabolites to the human liver corresponding
to a given ingested dose. Since VC metabolism is linear in the human in the dose range of
interest, this equivalence factor could be used to convert the risk based on the dose metric (now
in humans) into the human oral dose. Similarly, the equivalence factor for inhalation exposure
was calculated by determining the human dose metric for continuous human inhalation exposure
to a range of exposure concentrations (1 • g/m3 to 10,000 mg/m3). This calculation showed that
the model was linear up to nearly 100 mg/m3, and the calculated equivalence factor was used to
convert the risk from the inhalation experiments conducted in animals (in the units of the dose
metric) to human risk values.
An area of uncertainty in any risk assessment utilizing animal data is cross-species
extrapolation of dose. Data collected on chemotherapeutic agents (Freireich et al., 1966)
supports a roughly 10-fold lower minimally toxic dose in humans compared with rodents. These
data served as the principal basis for the use of a body surface area scaling as the default method
in cancer risk assessments. Empirically, the best estimate of scaling is bw3/4 (U.S. EPA, 1992).
These findings reflect general expectations of more rapid detoxification by smaller animals
resulting from faster metabolic rate. This renders them less susceptible to a given dose per unit
body weight. The PBPK model accomodates adjustments for metabolic rate as well as other
species-related dosimetric variables such as blood-to-air partition coefficients, liver perfusion
rates, etc. The model therefore provides a more accurate estimate of steady-state target site
concentration than use of default methods. On the other hand, while the PBPK model is
explicitly designed as a dosimetric adjustment, the presence of a toxicodynamic component is
not explicitly addressed. Barton et al. (1998) suggested use of a default of 10 in the absence of
either pharmacokinetic or pharmacodynamic data. He also suggested a default of 1.0 if either a
PBPK model or bw3/4 is employed for scaling, although based upon data the value might range
between 0.1 and 10.
The animal-to-human extrapolation factor employed in derivation of the noncancer RfD
and RfC is viewed by this Agency and others as comprising a pharmacokinetic (PK dose to
tissue) and a pharmacodynamic (PD; tissue response) component. In adjusting for animal-to-
human differences, PBPK models utilize pharmacokinetic information to adjust for dose; they do
not adjust for pharmacodynamic differences. Hence the application of a partial uncertainty
factor for pharmacodynamics as discussed in the derivation of the RfD/C assessments.
When used in a cancer dose-response assessment involving animal-to-human
extrapolation as with VC, a PBPK model similarly accounts for pharmacokinetic but not
pharmacodynamic differences. A primary issue is whether there is a need to a need to account
for species differences in PD or not. If sufficient information exists to provide a rationale that
there are no differences, i.e., that they are the same or the adjustment factor is 1, then there exists
no need to adjust. This appears to be the situation for VC based on the following reasons.
Cancer risk estimates based on epidemiologic data provided no evidence for greater
carcinogenic sensitivity to VC in humans than rats or mice. Chen and Blancato (1989) derived
lifetime risks for liver cancer based on epidemiology studies of 2.7 x 10"7 to 1.6 x 10"6 per • g/m3.
49
-------
Clewell et al. (Appendix B) developed unit risks of 0.46 x 10"6 to 2.8 x 10"6 per • g/m3 based
upon the Fox and Collier (1977) study, 0.65 x 10"6to 2.4 x 10"6 per • g/m3 based upon the Jones et
al. (1988) study, and 0.27 x lQ-6to 0.53 x 10'6 per • g/m3 based upon the Simonato et al. (1991)
study. Reitz et al. (1996) while not deriving a formal quantitative risk estimate based on the
Simonato et al. (1991) study, nevertheless reported that a unit risk estimate of 5.7 x 10"7 per
• g/m3using the Maltoni et al. (1981, 1984) data overestimated human risk 10- to 35-fold.
Swenberg et al. (1999) reported that N2,3-ethenoguanine (N2,3-EG) has demonstrated miscoding
potential (see Section 4.4.2). Numbers of these adducts correlate very closely with concentration
of chlorethylene oxide the presumed active metabolite of VC. Both humans and rats have
similar amounts of endogenous N2,3-EG and it is reasonable to assume that they would show a
similar exposure response. In summary, the epidemiologic data, while individually weak,
collectively suggest that humans are no more susceptible to VC than are laboratory species and
may be less so. Limited mechanistic data also provide no evidence for greater sensitivity in
humans. A pharmacodynamic adjustment of 1.0 is therefore considered to be adequately
protective. Storm and Rozman (1997) in an extensive review reached similar conclusions.
5.3.4. Extrapolation Method(s)
Two methods were used to extrapolate to low doses. Linear extrapolation is the
appropriate methodology for VC, a chemical known to act via a genotoxic mechanism. The first
extrapolation method used was the linearized multistage model (extra risk), in accordance with
the current risk assessment guidelines (U.S. EPA, 1987). In accordance with the Proposed
Guidelines for Carcinogen Risk Assessment (U.S. EPA, 1996), the LED 10/linear method was
also employed. This method draws a straight line between the point of departure from the
observed data, generally the LED10 (the lower 95% limit on a dose that is estimated to cause a
10% response) and the xy axis.
5.3.5. Oral Slope Factor and Inhalation Unit Risk
The oral slope factor and inhalation unit risk calculated for VC are presented in Table 9
(LMS model) and Table 10 (95% lower bound on the ED 10). The values calculated using these
two methods were very similar. The oral slope factor using the LMS model was determined to be
7.2 x 10'1 per (mg/kg)/day. Inhalation unit risk estimates of 2.6, 2.1, 1.0, and 4.4 x lO'6 per
• g/m3 for male mice, female mice, male rats, and female rats, respectively were derived. The
more conservative estimate of 4.4 x 10"6 per • g/m3 is recommended. The risk estimates are based
upon the assumption of continuous lifetime exposure beginning at adulthood. If exposure begins
early in life, addition of a twofold uncertainty factor is recommended. The basis for this application
is discussed below.
Extrapolation of the oral risk estimate to an inhalation unit risk results in a value of about
1 x 10"4 per • g/m3. The difference in potency estimates appears to be due primarily to the large
number of neoplastic nodules reported in the Feron et al. (1981) study, but not seen in the
Maltoni et al. (1981, 1984) studies. This difference may be due to different strains of rats used,
50
-------
Table 9. Lifetime human cancer risk estimates based on incidence of liver
tumors in animal bioassays with extrapolation using the LMS
Study
Rats, dietary
(Feronetal., 1981)
Mice, inhalation
(Maltoni et al., 1981,
1984, BT4)
Rats, inhalation
(Maltoni et al., 1981,
1984, BT1, BT2, and
Oral slope
Inhalation riska factor15
Sex (per • g/m3) (mg/kg-day)
F 7.2 x 104
M 2.6 x 1Q-6
F 2.1X1Q-6
M l.OxlQ-6
F 4.4 x lO'6
BT15)
aBased on incidence of liver angiosarcomas.
bBased on combined incidence of liver angiosarcomas, hepatocellular carcinomas, and neoplastic nodules.
Table 10. Lifetime human cancer risk estimates based on incidence of liver
tumors in animal bioassays with extrapolation using the EDIO/linear method
Study
Rats, dietary
(Feronetal., 1981)
Mice, inhalation
(Maltoni et al., 1981,
1984, BT4)
Rats, inhalation
(Maltoni et al., 1981,
1984, BT1, BT2, and
Oral slope
Inhalation riska factor15
Sex (per • g/m3) (mg/kg-day)
F 7.5 x 104
M 2.4 x 1Q-6
F 2.7 x 1Q-6
M 0.9 x lO'6
F 4.2 x 1Q-6
BT15)
aBased on incidence of liver angiosarcomas.
bBased on combined incidence of liver angiosarcomas, hepatocellular carcinomas, and neoplastic nodules
51
-------
different ages at the start of the study, or other unknown factors. Since not all neoplastic nodules
are likely to progress to carcinomas, the oral risk estimate is considered to be quite conservative.
Procedures were also instituted in the model to ensure the most conservative estimate in
extrapolating from the oral to the inhalation route, such as assuming 100% absorption over a
24-hour period.
Despite the limitations of the data, the possibility of additional risk due to tumor
induction at nonliver sites deserves consideration. To accommodate the possibility of increased
risk from nonliver tumors, potency estimates based on the induction of either kidney or
mammary tumors were derived, even though the incidence of these tumors were quite sporadic.
Since the PBPK model does not contain a mammary tissue compartment, and since there are no
adequate data on the metabolism of VC in mammary tissue to construct one, a "zero order
approximation" approach was utilized in which metabolism of VC in the liver was used as a
surrogate for in situ metabolism in mammary tissue. Thus, the same liver dose-metrics, "RISK,"
i.e., steady-state concentration of the active metabolite per L liver tissue, were calculated for the
conditions and doses of the bioassays showing increased incidence of mammary tumors. See
Table 8 in Appendix B. The 95% upper confidence limits of mammary tumor risk in female
mice and rats, based upon the dose metric "RISK" for studies in which an increase in mammary
tumors was seen, are listed below. Use of "RISK" results in a conservative estimate of cancer
potency because it is assumed that mammary tissue metabolizes at the same rate as liver tissue,
which is considered unlikely.
Estimated risks from mammary tumors in several studies reported by Maltoni et al.
(1981, 1984) ranged from 5 x 10"7to 5 x 10"6 per • g/m3, with one exception, for which a risk of 2
x 10"4 per • g/m3 was derived for females and 1 x 10"5 per • g/m3 for males (Table 10 in Appendix
B). In the latter case, increases occurred against a very high background for females (• 57%),
raising the possibility that VC was promoting or synergizing with an ongoing process, if indeed
there was any biological increase at all. The high background incidence in all the Sprague-
Dawley groups renders conclusions based upon this strain uncertain. Moreover, in this study the
greatest response occurred at 5 ppm, while in similar studies, reported by Maltoni et al. (1981,
1984) using the same strain, no mammary tumor responses were noted at exposure
concentrations of several hundred or even several thousand ppm.
Wistar rats have a much lower background incidence of mammary tumors. In studies
reported by Maltoni et al. (1981, 1984) using Wistar rats, control incidence was less than 5%;
among groups exposed up to 10,000 ppm, the incidences were even lower and in some cases
zero. In the Feron et al. (1983) study, while slightly elevated but not statistically significant
increases were reported for mammary carcinomas, mammary fibroadenomas showed a
significantly decreased incidence.
Human data regarding the possible induction of breast cancer in females is very limited
because few women are employed in the VC/PVC industry. Smulevitch et al. (1988) did not
report any breast cancer cases in a cohort of 1,037 women employed in a Soviet VC/PVC plant.
No significant increases in breast cancer were reported in other studies of male workers. Breast
cancer can occur in males, although it is uncommon. Because of the high degree of variability
and lack of positive dose responses in the animal studies, as well as little indication of effects in
52
-------
limited human data, the breast is considered unlikely to be a sensitive target site in humans if
indeed it is a target site at all. This conclusion is supported by evidence that the major site of VC
activation is the liver, combined with the likelihood that little of the active metabolite will escape
the liver because of its high degree of reactivity. For these reasons any additional adjustment to
account for possible breast cancer induction is considered to be unnecessary.
Increases in nephroblastoma were noted only in the Maltoni et al. (1981, 1984) studies.
Risk estimates ranged from 1.5 x 10"7 to 2.2 x 10"6 per • g/m3 (Table 9 in Appendix B). No
evidence for nephroblastoma was reported in the Feron et al. (1983) study, even though a high
incidence of liver tumors occurred. Concern regarding risk from induction of nephroblastoma is
also decreased because increases in these tumors were not observed in the epidemiology studies.
No evidence for induction of kidney tumors in humans by VC has been reported. Since
studies of occupational cohorts, with one exception, included only males, information regarding
possible breast cancer induction is very limited. In one study that included female workers no
breast cancer cases were reported (Smulevich et al., 1988). Suggestive evidence for tumors at
other sites in humans has also been noted, but increases were generally small compared to liver
tumors. For example, in a recent update of the "American" cohort by CMA (1998b), significant
increases in brain and and connective tissue tumors were reported. However, increases in
relative risks were quite small compared with those from liver cancer. While the possibility of
cancer induction by VC at nonliver sites remains, the evidence indicates that the liver is the most
sensitive target site. Protection against liver cancer is therefore considered to be protective
against the possibility of tumor induction at other sites.
5.3.5.1. Basis for Recommending Adjustment in Cancer Risk Estimates to Account for Early-
Life Sensitivity
Several studies have compared the carcinogenic effects of VC in newborn and adult
animals. Maltoni et al. (1981) reported liver angiosarcomas in 40.5% and hepatomas in 47.6%
of rats exposed from 1 day of age for 5 weeks to 6,000 ppm VC. At 10,000 ppm angiosarcomas
were noted in 34.1%, and hepatomas in 45.4%. By contrast, angiosarcomas were noted in 33.3%
and 11.7% of rats exposed to VC at 6,000 and 10,000 ppm, respectively, beginning at 3 months
of age, while hepatomas were noted in only 1.7% of either group. Consistent with this
observation, VC was found to induce preneoplastic foci in newborn rats, but not in adult rats
(Laib et al., 1979). Interestingly, in the same study it was found that VC did induce
preneoplastic foci in adult rats after partial hepatectomy, indicating that the appearance of foci,
and presumably of hepatocellular carcinoma, in neonatal animals was a consequence of the
increased rate of cell proliferation at that age. Similarly, Laib et al. (1989) found that inhaled
radiolabeled VC was incorporated into physiological purines of 11-day-old Wistar rats at
eightfold higher levels than in similarly treated adult rats (presumably reflecting the DNA
replication activity), and roughly fivefold higher levels of the DNA adduct 7-N-(2-
oxyethyl)guanine (OEG) were found in the livers of young animals, reflecting an increased
alkylation rate. It should be noted, however, that neoplastic nodules and hepatocellular
carcinoma were induced in rats exposed to VC in the diet. Although OEG is not believed to be a
precarcinogenic lesion, it is reasonable to expect that levels of this adduct would correlate with
53
-------
the levels of the precarcinogenic VC adducts. In a similar study, roughly fourfold greater
concentrations of both OEG and EG were also seen in preweanling rats exposed to VC than
adults (Fedtke et al., 1990).
Drew et al. (1983) studied the effects of age and exposure duration on cancer induction
by VC in rats, mice, and hamsters. Female golden Syrian hamsters, F344 rats, Swiss CD-I mice,
and B6C3F1 mice were exposed for 6 hours/day, 5 days/week to VC (50, 100, or 200 ppm for
mice, rats, and hamsters, respectively) for 6, 12, 18, or 24 months, with the exception of mice,
which were exposed only up to 18 months. All animals were sacrificed at month 24 or 18
(mice), and about 50 animals/species/group were tested. Other groups of rodents were held 6-12
months, and then exposed for 6 or 12 months, and also sacrificed at month 24. Unfortunately,
time-to-tumor data were not reported in this study, making it impossible to deconvolute the
impact of survival on the observation of tumors from later exposure periods. Because both mice
and hamsters showed significant survival effects (life-shortening) from the VC exposures, only
the data on exposures of rats during the first 12 months of life are appropriate for analysis. In
the rats, exposure from 0 to 6 months showed an overall similar potency to exposure from 6 to
12 months of life. In particular, the incidence of hepatocellular carcinoma combined with
neoplastic nodules and hemangiosarcoma was 24% and 5%, respectively, in rats exposed from 0
to 6 months, whereas for exposure from 6 to 12 months, the incidence was 31% and 4%,
respectively. In this study, however, even the 0- to 6-month animals were 8-9 weeks old at the
start of exposure and thus approaching maturity.
Although the reactive nature of the carcinogenic metabolites and the lack of P450 activity
in rodent fetuses would suggest that VC is not a transplacental carcinogen (Bolt et al., 1980),
data from Maltoni et al. (1981) suggest that it may be. Pregnant rats were exposed from
gestation day 12-18 to 6,000 or 10,000 ppm VC for 4 hours/day, and tumors were ascertained at
143 weeks postexposure. Nephroblastomas, forestomach tumors, epithelial tumors, and
mammary gland carcinomas were observed only in the offspring, and the incidence of Zymbal
gland carcinomas was higher in transplacentally exposed animals than in maternal animals.
Since the dams and offspring were followed for the same period, latency is not an issue for this
experiment. However, it is important to note that the offspring were exposed during
organogenesis, a period of rapid cell division, and any genotoxic carcinogen would be expected
to have a higher potency during this period. This apparent increased sensitivity of newborn
animals occurs in spite of a much lower metabolic capability at birth: during the first week of
life, the P450 activity in the liver of rats increases from about 4% to about 80% of adult levels
(Filser and Bolt, 1979). As the fetuses did not possess the capability to metabolize VC, these
data suggest that CEO was produced by the dam and then transported to the fetuses.
The preceding studies provide evidence for increased sensitivity to VC-induced
carcinogenesis in early-life and prenatal exposures in experimental animals. Early-life data on
humans, however, are lacking because most exposures have been limited to occupational groups.
Nevertheless, many of the factors likely to be responsible for early-life sensitivity in animals are
present in humans. Because of more rapid cell division and dosimetric considerations (increased
respiration or liquid intake per unit body weight, more rapid blood flow to liver), an additional
correction to account for early-life exposure is recommended. Guidance has previously been
given to the Regional Offices to double the lifetime risk estimate for VC to account for the
54
-------
additional risk attributable to early-life exposures (Cogliano, 1989, 1990; Cogliano and Parker,
1992).
Several observations can be made about the early-life studies:
1. Exposure periods in the early-life studies (Sprague-Dawley rats exposed 5 weeks,
beginning at 1 day of age, Maltoni et al., 1981, and days 7-21 post conception for
Laib et al., 1985) do not overlap those of the chronic studies (weeks 14-65) from
which chronic slope factors and unit risks are derived.
2. The angiosarcoma incidence after short-term, early-life exposure is approximately
equal to that of long-term exposure starting after maturity (see Table 6), although
hepatoma incidences differ.
3. Because the effects of early-life exposure are qualitatively and quantitatively
different from those of later exposures, it would not be appropriate to prorate early-
life exposures as if they were received at a proportionately lesser rate over a full
lifetime.
The first observation (nonoverlapping exposure periods) suggests that the full lifetime
cancer risk can be approximated by adding risks from the nonoverlapping exposures in early life
and later. The second observation suggests that the angiosarcoma risks from these
nonoverlapping periods are approximately equal. The third observation suggests that the risk
from early-life exposure should not be prorated over a longer duration. The experimental studies
suggest that the risk from short-term exposure immediately after birth may not be reversible
even in the absence of further exposure later in life. This would effectively double the VC slope
factors and unit risks; one portion would apply to any early-life exposure; the other to exposures
later in life.
In applying these results to partial lifetime exposure, the later-life portion can be
apportioned according to a curve that declines with age (Cogliano, 1989, 1990; Cogliano and
Parker, 1992; Cogliano et al., 1996; Hiatt et al., 1994). In contrast, early-life exposures would
not be prorated over a longer duration. (A simpler approach would be to prorate later-life
exposures over the life span, while not prorating early-life exposures.) The following examples
illustrate these adjustments.
Example 1. Full lifetime exposure (birth through death) to 1 • g/m3.
Continuous lifetime exposure during childhood: 8.8 x 10"6 x (1 • g/m3) = 8.8 x 10"6
Total risk: 8.8 x 10'6
Here the total risk is a single unit risk estimate.
Example 2. Exposure to 2 • g/m3 from ages 30 to 60.
Early-life risk: Not applicable.
Later-life risk: (4.4 x 10'6 per • g/m3) x (2 • g/m3) x (30/70) = 3.8 x 10'6.
55
-------
Total risk: 3.8 x 10'6
Here exposure begins at age 30, so there is no early-life component. The later-life component is
prorated as a duration of 30 years over an assumed life span of 70 years.
Example 3. Exposure to 5 • g/m3 from ages 0 to 10.
Early-life risk: (4.4 xlQ'6 per • g/m3) x (5 • g/m3) = 22 x 10'6
Later-life risk: (4.4 xlQ'6 per • g/m3) x (5 • g/m3) x (10/70) = 3.1 x 10'6
Total risk: 25 x lQ-6 = 2.5 x 10'5
In this instance, both "continuous lifetime exposure from birth" and "continuous exposure during
adulthood" components of risk would apply. The first component would be the early-life risk,
which can be apportioned from the "exposure from birth" minus "exposure during adulthood"
components at 8.8 - 4.4 = 4.4 x 10"6. A second component of risk would be another
apportionment from "exposure during adulthood" for later-life risk. Because the exact age
window of susceptibility in humans is not known, but is likely to be much shorter in duration
than 10 years, risk outside this window of susceptibility should be considered, but at the level of
later-life risk, 4.4 x 10"6. Furthermore, this risk would have to be apportioned based on the
fractional life span of the exposure, i.e., 10/70 years. The total risk would be summed from
these two components to be 25 x 10"6 = 2.5 x 10"5. It is recognized that the period of
susceptibility is accounted for in both of these components. It should be noted, however, that the
total risk in this instance is far less than what it would be from continuous lifetime exposure
from birth at (8.8 x IQ'6) x (5 • g/m3) = 44 x 10'6.
In general, the potential for added risk from early-life exposure to VC is accounted for in the
quantitative cancer risk estimates by a twofold uncertainty factor. If exposure occurs only
during adult life, the twofold factor need not be applied.
5.3.5.2. Confidence in the Dose-Response Assessment
Confidence in the dose-response assessment is medium to high for a number of reasons.
VC has been shown to be carcinogenic in a large number of animal bioassays as well as in
epidemiologic studies. The primary target site and major tumor types are also the same in
experimental animals and humans. VC is a well-characterized genotoxic carcinogen. Its
carcinogenic activity is attributed to the formation of DNA adducts by the highly reactive VC
metabolite CEO. There is strong evidence linking etheno-DNA adducts with observed
carcinogen!city. This increases confidence in extrapolating to low doses using either the ED 10
method or the LMS model.
Recommendations from epidemiologic-based estimates were not made because of the
limitations of the studies. The low dose-response estimates from several of these epidemiology
studies (see Appendix B) nevertheless provide support for assuming the animal-based estimates
are sufficiently conservative. The epidemiologic studies also suggest that although tumors may
be induced at other sites, the liver is the most sensitive site. Protection against liver cancer is
56
-------
therefore expected to provide protection against cancer at other sites. It should be noted that
these risk estimates are based primarily upon health male workers.
The use of a PBPK model to determine target site concentration of the active metabolite
allows a more accurate estimate of dose-response than default methods. Uncertainty in the PBPK
model was determined by conducting a Monte Carlo analysis, in which risk is calculated by
sampling the distributions of the parameters used in the model, resulting in a distribution of
calculated risks. This analysis for the VC model found that the 95th percentile of the distribution
of upper confidence limit (UCL) risks was within 50% of the mean UCL risk. Furthermore, in a
sensitivity/uncertainty analysis of the parameters used in the model, none of the parameters
displayed sensitivities markedly greater than 1.0, indicating that there is no amplification of error
from the inputs to the outputs. This is, of course, a desirable trait in a model to be used for risk
assessment. The parameters that did have a significant impact on the calculated dose metric (and
thus the risk) were body weight, alveolar ventilation, cardiac output, liver blood flow and
volume, blood/air partition coefficient, the capacity and affinity for metabolism and, in the case
of oral gavage, the oral uptake rate. All of these parameters could be reasonably well
characterized from experimental data. The sensitivity of the risk predictions to the human values
of these key determinative parameters implies that the risk from exposure to VC could vary
considerably from individual to individual, depending on the various combinations and
permutations of specific physiology, level of activity, and metabolic capability.
Pharmacodynamics were not addressed by the PBPK model. Since the dose metric is the
amount of reactive metabolite (CEO) produced, and the putative reactive metabolite (CEO) is
believed to interact directly with DNA, pharmacodynamics in animals and humans would be
expected to be similar. Moreover, evidence from bioassays and epidemiologic data suggests that
humans are no more sensitive to VC than are laboratory animals and indeed may be less
sensitive.
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND DOSE
RESPONSE
6.1. HUMAN HAZARD POTENTIAL
6.1.1. Hazard Identification for Cancer Effects
The association between occupational exposure to VC and the development of liver
angiosarcomas is one of the best characterized cases of chemical-induced car cinogeni city in
humans. Liver angiosarcomas are an extremely rare tumor, with only 20-30 cases per year
reported in the United States. Since the introduction of VC manufacturing, nearly all of the
reported cases have been associated with VC exposure. VC exposure, including polyvinyl
chloride, has also been associated with increased death due to primary liver cancer, as well as
cancer of the brain, lung, and lymphopoietic system. The association of VC with angiosarcoma
in numerous epidemiologic studies has been supported by findings in rats, mice, and hamsters
administered VC via the oral and inhalation routes. The mode of action is also well understood
57
-------
and documented; VC is metabolized to a reactive metabolite, probably CEO, which interacts
with DNA, forming DNA adducts and ultimately leading to tumor formation.
On the basis of sufficient evidence for carcinogenicity in human epidemiology studies,
VC is therefore considered to best fit the weight-of-evidence Category "A," according to current
EPA Risk Assessment Guidelines (U.S. EPA, 1986). Agents classified into this category are
considered to be known human carcinogens. Under the Proposed Guidelines for Carcinogen
Risk Assessment (U.S. EPA, 1996), it is concluded that VC is a known human carcinogen by the
inhalation route of exposure based upon human evidence, and by the oral route on the basis of
extensive positive data in oral animal studies and the knowledge that VC is well absorbed by the
oral route. VC is also considered highly likely to be carcinogenic by the dermal exposure route
because it is well absorbed by this route and is a systemic carcinogen.
6.1.2. Hazard Identification for Noncancer Effects
The liver is the primary target for the noncancer effects of VC in animals (Bi et al., 1985;
Feron et al., 1981; Sokal et al., 1980; Til et al., 1983, 1991) and humans (Buchancova et al.,
1985; Doss et al., 1984; Gedigk et al., 1975; Lilis et al., 1975; Marsteller et al., 1975; Popper and
Thomas, 1975; Tamburro et al., 1984). Pathological effects such as liver necrosis, liver cell
polymorphism, and cysts as well as alterations in liver function have been reported.
Other effects reported in some occupational studies are associated with exposure levels
much higher than those that cause liver injury. Acroosteolysis, or resorption of the terminal
phalanges of the fingers, was observed in workers occupationally exposed to VC (Lillis et al.,
1975; Marsteller et al., 1975), often preceded by clinical signs of RP (Fontana et al., 1995). This
was most often seen in tank cleaners and is apparently associated with dermal exposure.
Occupational exposures at high concentrations may induce headaches, drowsiness, dizziness,
ataxia, and loss of consciousness (Lilis et al., 1975; Langauer-Lewowicka et al., 1983;
Waxweiler et al., 1977).
Reproductive effects and testes damage occurred in rats exposed to VC (Short et al.,
1977; CMA, 1988a; Bi et al., 1985). These endpoints, however, were generally noted at
concentrations greater than those necessary to cause liver damage.
Although most of the animal and human data result from inhalation studies, these data
are directly applicable to oral exposure, because VC is rapidly absorbed and distributed
throughout the body following oral or inhalation exposure. First-pass metabolism is not a major
issue because the initial function of the liver is activation rather than inactivation. However,
initial liver concentration may be greater via oral dosing because essentially all absorbed VC
passes through the liver before possibly entering the systemic circulation.
6.2. DOSE RESPONSE
58
-------
6.2.1. Dose Response for Cancer Effects
Cancer potency of VC in humans is based on animal experiments because of uncertain
exposure levels in epidemiology studies. Estimated risk from continuous inhalation exposure to
VC during adult life is 4.4 x 10"6 per • g/m3. Estimated lifetime cancer risk from oral exposure to
VC is 7.2 x 104 per mg/kg-day.
Quantitation of risk is based upon tumor incidences in female rats in the feeding study
reported by Feron et al. (1983) and female rat inhalation studies reported by Maltoni et al. (1981,
1984). The Maltoni et al. (1981, 1984) studies included both mice and rats exposed to a wide
range of concentrations. The Feron et al. (1983) studies included three exposure levels and is
supported by a subsequent study by (Til et al., 1991), conducted under nearly identical
conditions that included two lower exposures. The studies were well designed and utilized
adequate numbers of animals.
Risk estimates were based upon estimates of the concentration of the active metabolite of
VC, CEO, in the liver. Concentrations were derived using a PBPK model that accounted for
species differences in factors such as ventilatory exchange rates, blood-air partition coefficients,
metabolic activation rates, organ volumes and flows, etc. Use of this model allows a more
accurate estimation of risk than default models such as body surface area correction, which was
not applied because the PBPK model adjusts for differences in metabolic rate among species.
Pharmacodynamics were not addressed by the PBPK model. However, the dose metric is the
amount of reactive metabolite produced which is believed to interact directly and
indiscriminately with DNA, either animal or human. Given the assumptions about various
aspects of pharmacodynamics such as similar repair (or lack of repair) rates between animals and
humans, the validity of the extrapolation of physiological time between animals and humans, and
the use of a dose metric that is normalized for the size of the liver (i.e., amount of metabolite
produced per liter liver), the pharmacodynamics of the vinyl chloride cancer response in animals
and humans may be reasonably expected to be quite similar.
Several studies have provided evidence for early-life sensitivity to VC-induced tumors.
Maltoni et al. (1981) reported markedly increased cancer incidence in rats exposed via inhalation
beginning at 1 day of age compared with those exposed beginning at 13 weeks of age. Mice,
rats, and hamsters were shown to be more sensitive to cancer induction if exposed at a younger
age (Drew et al., 1983). Vinyl chloride induction of preneoplastic liver foci in rats is restricted
to exposures at approximately 7 to 21 days of age (Laib et al., 1979). None of these studies
were considered to be suitable for deriving recommended unit risk estimates because of short
exposure durations, single exposure levels, or reporting of endpoints other than cancer. The
Maltoni et al. (1991), Drew et al. (1983), and Laib et al. (1979) studies, however, provide a basis
for recommending a twofold adjustment of estimated cancer risk to account for early-life
exposure.
Although results of several epidemiology studies were positive for liver cancer, exposure
concentrations were of sufficient uncertainty to preclude recommendation of risk estimates
derived from these studies. Considerable variation in exposure was likely in the larger studies
that included cohorts from several facilities. Duration of exposure at high concentrations was
often unavailable. Despite these limitations, several epidemiology studies have been used to
59
-------
estimate cancer risk. The study by Fox and Collier (1977) provided the best data set with
respect to providing information regarding duration of employment and exposure level
groupings. Chen and Blancato (1989) used this study to derive a unit risk estimate of about 3 x
10"6 per • g/m3. The weakness of the study is the small cohort with only two cases of liver
angiosarcoma. An estimate of 2 x 10"6 per • g/m3 was derived based upon autoclave workers in
the Jones et al. (1988) study, an update of the Fox and Collier (1977) study. Reitz et al. (1996)
reported that the unit risk estimate of 5.7 x 10"7 per • g/m3 they derived from animal data was as
much as 35-fold greater than the predicted tumor rates in humans derived from Simonato et al.
(1991). While the Simonato study had a larger cohort and more deaths due to liver cancer (24),
exposure uncertainty was also greater because data were collected from many different
workplaces in several countries. It should also be noted that many of the workers were still alive
when these calculations were made, with the likelihood of further deaths from liver cancer.
While epidemiology-based risk estimates are conservative compared with animal-based ones, the
occupational cohorts used lack females, children, and other potentially sensitive members of the
population.
As discussed in Section 5.3.1, use of DNA adduct levels could be considered as the basis
for a VC risk estimate. However, adduct levels cannot be used directly to extend tumor dose-
response data to lower doses, since tumor formation from adducts depends on many factors,
including the nature of the adduct and the consequences of adduct repair or failure to be repaired.
Thus, although a quantitative analysis of the relationship between VC metabolism, adduct
formation, and tumor formation is likely to be a fruitful area for additional research, it is
premature to attempt to establish a quantitative link between the tissue concentrations of a
specific adduct and the risk of cancer in that tissue.
Confidence in the risk assessment is increased by the availability of appropriately
designed and conducted studies, understanding of the mechanisms of VC carcinogenicity,
allowing risk to be based upon concentration of active metabolite, and the fact that risks based
on liver angiosaromas (rare tumors in both animals and humans) are in close agreement. VC is a
well-characterized genotoxic carcinogen. Carcinogenic activity of VC is attributed to the
formation of DNA adducts by the highly reactive VC metabolite CEO. Therefore, there is
considerable justification for extrapolating to low doses using either the LED 10 method or the
LMS model. A Monte Carlo analysis, in which risk is calculated by sampling the distributions
of the parameters used in the model, resulting in a distribution of calculated risks, determined
that the 95th percentile of the distribution of upper confidence limit (UCL) risks was within 50%
of the mean UCL risk. Furthermore, in a sensitivity/uncertainty analysis of the parameters used
in the model, none of the parameters displayed sensitivities markedly greater than 1.0, indicating
that there was no amplification of error from the inputs to the outputs.
Confidence in the risk estimates is decreased somewhat by uncertainty regarding the
possible effect of nonliver tumors on cancer potency. This is especially true for endpoints such
as the mammary gland, for which some animal studies suggest an additional risk, but for which
human data are limited. However, both animal and epidemiologic data suggest that the liver is
the most sensitive target site, and protection against liver cancer should therefore protect against
cancer at nonliver sites.
Overall, confidence in the assessment is medium to high.
60
-------
6.2.2. Dose Response for Noncancer Effects
The quantitative estimates of human risk as a result of low-level chronic exposure to VC
are based on animal experiments because of uncertainties regarding human exposure levels to
VC.
The human oral dose that is likely to be without an appreciable risk of deleterious
noncancer effects during a lifetime (the RfD) is 3E-3 mg/kg-day.
Confidence in the principal study is high. The study of Til et al. (1983, 1991) used
adequate numbers of animals, was well controlled, and reported in detail on the histological
effects on the liver. There are several corroborative inhalation studies that observed effects on
the liver and testes in rodents following inhalation exposure. Medium confidence in the database
results from a lack of a two-generation reproductive study. Other gaps in the oral database can
be filled on the basis of inhalation toxicity data. Two developmental inhalation studies (John et
al., 1977; Ungvary et al., 1978) were located that reported embryotoxic effects only at levels
much higher than those causing maternal toxicity in mice, rats, or rabbits. There is no evidence
for other effects at doses as low as those inducing effects in the Til et al. (1991) study. The two-
generation reproductive study of CMA (1998) demonstrates liver effects at concentrations where
reproductive effects were absent, indicating the sensitivity of the liver relative to any effects on
reproduction. Also, in a dominant lethal study of VC, reduced fertility was observed at a
concentration greater than that inducing liver effects in rats (Short et al., 1977). These data
impart considerable certainty that in the dose-response relationship of VC, liver effects would
occur before reproductive-related effects. Therefore, the confidence in the database is
considered high to medium.
Qualitative differences exist between the dose metrics generated from the PBPK model
used in this assessment. This difference is due principally to the extent of information available
for validating the dose metrics derived from different routes of exposure, i.e., inhalation and oral.
As documented in Appendix B, numerous data sets are available via the inhalation route to both
parameterize and judge the ability of the model to characterize aspects of VC dosimetry,
including the dose metrics used in this assessment, that occur in an inhalation scenario. Data
sets for the oral route, however, are few and problematic (Appendix B), which limits the ability
to either parameterize or to judge performance of the model for this particular route. Thus, a
higher degree of confidence is placed in model outputs (dose metrics) derived from inhalation
scenarios than in those derived from oral scenarios. To attempt to compensate for this
qualitative difference between the oral and inhalation dose metrics, certain procedures were
instituted within the model when calculating oral dose metrics, including assumption of a
maximum rate of VC uptake (i.e., designating it a zero-order process) and spreading the applied
dose over a 24-hr period, which would maximize the likelihood that the parent VC would be
metabolized to reactive species (i.e., the basis of this assessment, mg VC metabolized).
The high degree of confidence in the principal study of Til et al. (1983, 1991), combined
with the high-to-medium assessment of the database and less-than-high confidence in the
qualitative aspects of the PBPK, is considered to result in an overall medium confidence in the
RfD.
61
-------
An uncertainty factor of 10 was used for protection of sensitive human subpopulations
and 3 for animal-to-human extrapolation. The uncertainty factor of 10 for intraspecies
variability includes the variability in risk estimates that would be predicted by the model for
different individuals, due to variability in physiology, level of activity, and metabolic capability.
A factor of 3 was used for interspecies extrapolation because although PBPK modeling refines
the animal-to-human comparison of delivered dose, it does not address the uncertainty regarding
the toxicodynamic portion of interspecies extrapolation (relating to tissue sensitivity). The mode
of action and reactive species are not as clear for noncancer effects as for cancer effects. This
lack of clarity concerning the reactive species (i.e. CEO or CAA) has consequences regarding
the toxicodynamic components of this UF. Little, if any, difference would be anticipated to
occur between animal and human DNA and CEO, whereas considerable differences could exist
between animal and liver proteins/components and CAA. Therefore, the partial UF for TD is
retained for the noncancer but not the cancer assessment. No uncertainty factor was considered
necessary for deficiencies in this relatively complete database.
Daily inhalation exposure to a human population that is likely to be without an
appreciable risk of deleterious effects during a lifetime (the RfC) is 1E-1 mg/m3. The RfC is
based on the same study used to derive the RfD (Til et al., 1983, 1991). As noted above,
confidence in this study is high. An oral study was used to derive the RfC because it was the
best study available, effects were reported at lower doses than in any of the inhalation studies,
and use of the PBPK model allowed route extrapolation of reactive metabolite.
The overall confidence in the RfC is medium. The study of Til et al. (1983, 1991) used
adequate numbers of animals, was well controlled, and reported in detail on the histological
effects on the liver. Since the PBPK model can be used to calculate tissue doses for oral and
inhalation exposure, detailed information on a range of endpoints is available. There are several
corroborative inhalation studies that observed effects on the liver and testes in rodents following
inhalation exposure. Two developmental inhalation studies (John et al., 1977; Ungvary et al.,
1978) were located that reported embryotoxic effects only at levels much higher than those
causing maternal toxicity in mice, rats, or rabbits. Results from both the reproductive study of
CMA (1998) and, to a lesser degree, the dominant lethal study of Short et al. (1977) clearly
indicate that liver effects occur at exposures to VC much less than any reproductive effect or
parameter examined in these studies.
As discussed for the RfD, there exist qualitative differences between dose metrics
generated from oral and inhalation routes by the PBPK model used in this assessment. Data sets
for the oral route are problematic and few (Appendix B), which limits the ability to either
parameterize or to judge performance of the model for this particular route. This RfC is based
on dose metrics derived from the dietary administration study of Til et al. (1983, 1991). Actions
taken to compensate for this qualitative deficiency were those described above for the RfD, the
overall intent being to maximize the likelihood of the administered dose to be transformed to
reactive metabolites in the liver, to obtain the maximum dose metric from any oral dose.
Comparable to the RfD, an uncertainty factor of 10 was used for protection of sensitive
human subpopulations and 3 for animal-to-human extrapolation. The uncertainty factor of 10
for intraspecies variability includes the variability in risk estimates that would be predicted by
the model because of population variability, and a factor of 3 was used for interspecies
62
-------
extrapolation to address uncertainty relating to potential interspecies differences in tissue
sensitivity.
Since VC toxicity results from a reactive metabolite generated by P450 enzymes,
individuals who generate increased amounts of the toxic metabolite through the induction of
these enzymes may comprise a sensitive population. The P450 inducers phenobarbital and
Aroclor 1254 induce VC metabolism and have been shown to increase VC toxicity (Jaeger et al.,
1977; Jedrychowski et al., 1985; Reynolds et al., 1975). Increased sensitivity to the effects of
VC would also be expected in people with preexisting liver disease.
Although VC has often been cited as a chemical for which saturable metabolism should
be considered in the risk assessment, saturation appears to become important only at very high
exposure levels (greater than 250 ppm by inhalation or 25 mg/kg-day orally) compared with
levels associated with the most sensitive noncancer effects or tumorigenic levels, and thus has
little impact on the risk estimates made in the relevant range. The important contribution of
pharmacokinetic modeling is to provide a more biologically plausible estimate of the effective
dose and to compensate for the nonuniform ratio of this biologically effective dose to exposure
concentration or administered dose across routes and species. Therefore, any estimate of
administered dose other than that generated in consideration of pharmacokinetics is less
adequate for performing route-to-route and interspecies extrapolation of risk.
The major area of scientific uncertainty in this assessment is a quantitative
characterization of the variability in the human population and the increased sensitivity of
sensitive populations. This area is compensated for with a default uncertainty factor. As noted
in Section 5.1.1, the LOAEL used by ATSDR (1995) in its calculation of a chronic oral MRL is
considerably lower than the NOAEL identified for the RfD (without consideration of
pharmacokinetics). This discrepancy resulted because ATSDR did not take into consideration
the preneoplastic nature of its critical effect, the proliferative basophilic foci in the Til et al.
(1983, 1991) study. As also noted in Section 5.2.1, ATSDR (1995) considered increased relative
heart and spleen weights (Bi et al., 1985) to be co-critical effects in its calculation of an
intermediate-duration inhalation MRL. These effects were not considered for the derivation of
the RfC because of the absence of a concentration- or duration-related response, and because
they occurred at higher concentrations than liver cell polymorphisms used to derive both the RfC
and RfD.
It should be noted, however, that the most significant effect of VC observed in human
epidemiologic studies is liver cancer. The observation that the cancer effects of VC dominate at
high human exposure concentrations, coupled with the fact that VC is a genotoxic carcinogen for
which linear low-dose extrapolation is appropriate, suggests that the noncancer effects of VC are
not likely to be as important a concern for chronic human exposure.
63
-------
7. REFERENCES
Afzelius, BA; Schoental, R. (1967) The ultrastructure of the enlarged hepatocytes induced in rats with a single oral
dose of retrorsine, a pyrrolizidine (Senecio) alkaloid. J Ultrastruct Res 20:328-345.
Agency for Toxic Substances and Disease Registry (ATSDR), Public Health Service, U.S. Department of Health
and Human Services. (1995) Toxicological profile for vinyl chloride. Draft for public comment.
Amoore, JE; Huatala, E. (1983) Odor as an aid to chemical safety: odor thresholds compared with threshold limit
values and volatiles for 214 industrial chemicals in air and water dilution. J Appl Toxicol 3(6):272-290.
Andersen, M; Clewell, H; Gargas, M; et al. (1987) Physiologically based pharmacokinetics and the risk assessment
process for methylene chloride. Toxicol Appl Pharmacol 87:185-205.
Bartsch, H; Malaveille, C; Montesano, R. (1975) Human, rat and mouse liver-mediated mutagenicity of vinyl
chloride in S. typhimurium strains. Int J Cancer 15:429-437.
Barton, HA; Andersen, ME; Quell, HJ, III. (1998) Harmonization: developing consistent guidelines for applying
mode of action and dosimetry information to cancer and noncancer risk assessment. Hum Ecol Risk Assess 4:75-
115.
Belli, S; Bertazzi, PA; Comba, P; et al. (1987) A cohort study on vinyl chloride manufacturers in Italy: study design
and preliminary results. Cancer Lett 35:253-261.
Berk, PD; Martin, JF; Young, RS; et al. (1976) Vinyl chloride-associated liver disease. Ann Intern Med 84:717-731.
Bi, W; Wang, Y; Huang, M; et al. (1985) Effect of vinyl chloride on testis in rats. Ecotoxicol Environ Saf 10:281-
289.
Blair, A; Kazerouni, N. (1997) Reactive chemicals and cancer. Cancer Causes and Control 8:473-490.
Bolt, HM. (1978) Pharmacokinetics of vinyl chloride. Gen Pharmacol 9:91-95.
Bolt, HM. (1986) Metabolic activation of vinyl chloride, formation of nucleic acid adducts and relevance to
carcinogenesis. IARC Sci Publ 70:261-268.
Bolt, HM; Kappus, H; Buchter, A; et al. (1976) Disposition of [1,2-14C] vinyl chloride in the rat. Arch Toxicol
35:153-162.
Bolt, HM; Laib, PJ; Kappus, H; et al. (1977) Pharmacokinetics of vinyl chloride in the rat. Toxicology 7:179-188.
Bolt, HM; Laib, RJ; Filser, JG; et al. (1980) Vinyl chloride and related compounds: mechanisms of action on the
liver. In: Frontiers in liver disease. Berk, PD; Chalmers, TC, eds. New York: Thieme/Stratton, p. 80.
Bos, JL. (1989) Ras oncogenes in human cancer: a review. Cancer Res 49:4682-4689.
Buchancova, J; Reznak, I; Horak, V; et al. (1985) Scintigraphic pictures of the liver in workers after a long-term
exposure to vinyl chloride. Pracov Lek 37:190-194.
Buffler, PA; Word, S; Eifler, C; et al. (1979) Mortality experience of workers in a vinyl chloride monomer
production plant. J Occup Med 21:195-203.
Bull, RJ; Brown, JM; Meierhenry, EA; et al. (1986) Enhancement of the hepatotoxicity of chloroform in B6C3F1
mice by corn oil: implications for chloroform carcinogenesis. Environ Health Perspect 69:49-58.
64
-------
Byren, D; Engholm, G; Englund, A; et al. (1976) Mortality and cancer morbidity in a group of Swedish VCM and
PVC workers. Environ Health Perspect 17:167-170.
Chemical Manufacturers Association. (CMA) (1998a) Vinyl chloride combined inhalation two-generation
reproduction and developmental toxicity study in CD rats. Submitted by CMA to ATSDR pursuant to Memorandum
of Understanding, June 26, 1996. Final Report. CMA, Arlington VA.
Chemical Manufactuters Association. (CMA) (1998b) Epidemiological study of men employed in the vinyl
chloride industry between 1942 and 1972:1. Re-analysis of mortality through December 31, 1982: and II. Update
of Mortality through December 31, 1995. Final Report. CMA, Arlington VA. pp. 1-178.
Chen, CW; Blancato, JN. (1989) Incorporation of biological information in cancer risk assessment: example—vinyl
chloride. Cell Biol Toxicol 5:417-444.
Chen, JM; Smith, SJ; Marion, MJ; et al. (1999) Common conformational effects in the p53 protein of vinyl
chloride-induced mutations. J Protein Chem 18:467-472.
Cheng, KC; Preston, BD; Cahill, DS; et al. (1991) The vinyl chloride DNA derivative N2,3-ethenoguanine produces
G* A transitions in Escherichia coli. Proc Natl Acad Sci 88:9974-9978.
Clement Associates. (1987) Investigation of cancer risk assessment methods. Final report. Vol. 1: Introduction and
epidemiology. Prepared for the U.S. Environmental Protection Agency, the U.S. Department of Defense, and the
Electric Power Research Institute.
Clewell, HJ; Gentry, PR; Gearhart, JM; et al. (1995a) The development and validation of a physiologically based
pharmacokinetic model for vinyl chloride and its application in a carcinogenic risk assessment for vinyl chloride.
Prepared by ICF Kaiser for the Office of Health and Environmental Assessment, U.S. Environmental Protection
Agency, and the Directorate of Health and Standards Programs, Occupational Safety and Health Administration,
Washington, DC.
Clewell, HJ; Covington, TR; Crump, KS; et al. (1995b) The application of a physiologically based pharmacokinetic
model for vinyl chloride in a noncancer risk assessment. Prepared by ICF Kaiser/Clement Associates for the
National Center for Environmental Assessment, U.S. Environmental Protection Agency, Washington, DC, under
EPA contract number 68-D2-0129.
Clewell, HJ; Gentry, PR; Gearhart, JM; et al. (1995c) Considering pharmacokinetic and mechanistic information in
cancer risk assessments for environmental contaminants: examples with vinyl chloride and trichloroethylene.
Chemosphere 31:2561-2578.
Cogliano, J. (1989) Status of vinyl chloride assessment (memorandum). National Center for Environmental
Assessment, U.S. Environmental Protection Agency, Washington, DC.
Cogliano, J. (1990) Cancer unit risk estimates for vinyl chloride (memorandum). National Center for Environmental
Assessment, U.S. Environmental Protection Agency, Washington, DC.
Cogliano, VJ; Parker, JC. (1992) Some implications of toxicology and pharmacokinetics for exposure assessment.
J Exp Environ Epidemiol Suppl 1:189-207.
Cogliano, VJ; Hiatt, GFS; Den, A. (1996) A quantitative cancer assessment for vinyl chloride: indications of early-
life sensitivity. Toxicology 111:21-28.
Cooper, WC. (1981) Epidemiologic study of vinyl chloride workers: mortality through December 31, 1972. Environ
Health Perspect 41:101-106.
Cowfer, JA; Magistro, AJ. (1983) Vinyl chloride. In: Encyclopedia of chemical technology, vol. 23. Othmer, K, ed.
New York: Wiley Interscience, pp. 865-885.
65
-------
Creech, JL; Johnson, MN. (1974) Angiosarcoma of the liver in the manufacture of polyvinyl chloride. JOccupMed
16:150-151.
DeAngelo, AB; Daniel, FB; McMillan, L; et al. (1989) Species and strain sensitivity to the induction of peroxisome
proliferation by chloroacetic acids. Toxicol Appl Pharmacol 101:285-298.
DeVivo, I; Marion, MJ; Smith, SJ; et al. (1994) Mutant c-Ki-ras protein in chemical carcinogenesis in humans
exposed to vinyl chloride. Cancer Causes Control 5:273-278.
Doll, R. (1988) Effects of exposure to vinyl chloride. Scand J Work Environ Health 14:61-78.
Donehower, LA; Harvey, M; Slagel, BL; et al. (1992) Mice deficient for p53 are developmentally normal but
susceptible to spontaneous tumors. Nature 356:215-221.
Dosanjh, JK; Chenna, A; Kim, E; et al. (1994) All four known cyclic adducts formed in DNA by the vinyl chloride
metabolite chloroacetaldehyde are released by a human DNA glycosylase. ProcNatl Acad Sci 91:1024-1028.
Doss, M; Lange, CE; Veltman, G. (1984) Vinyl chloride-induced hepatic coproporphyrinuria with transition to
chronic hepatic porphyria. Klin Wochenschr 62:175-178.
Dow Chemical Company. (1986) Co-initial submission: an update of an epidemiologic study with vinyl chloride
workers, 1942-1982 (final report with cover letter dated 07/24/92 and attachments). U.S. Environmental Protection
Agency/OTS public files. Document no. 88-920005649. Fiche no. 0544432 (1).
Drew, RT; Boorman, GA; Haseman, JK; et al. (1983) The effect of age and exposure duration on cancer induction
by a known carcinogen in rats, mice, and hamsters. Toxicol Appl Pharmacol 68:120-130.
D'Souza, RW; Andersen, ME. (1988) Physiologically based pharmacokinetic model for vinylidene chloride.
Toxicol Appl Pharmacol 95:230-240.
Du, CL; Wang, J-D. (1998) Increased morbidity odds ratio of primary liver cancer and cirrhosis of the liver among
vinyl chloride monomer workers. Occup Environ Med 55:528-532.
Du, CL; Kuo, ML; Chang, HL; et al. (1995) Changes in lymphocyte single strand breakage and liver function of
workers exposed to vinyl chloride monomer. Toxicol Lett (Ireland) 77:379-385.
Du, JL; Sandoz, JP; Tseng, MT; et al. (1979) Biochemical alterations in livers of rats exposed to vinyl chloride.
J Toxicol Environ Health 5:1119-1132.
Duck, BW; Carter, JT; Coombes, EJ. (1975) Mortality study of workers in a polyvinyl-chloride production plant.
Lancet 13:1197-1199.
Edmonds, LD; Anderson, CE; Flynt, JW, Jr; et al. (1978) Congenital central nervous system malformations and
vinyl chloride monomer exposure: a community study. Teratology 17:137-142.
Elliot, P; Kleinschmidt, I. (1997) Angiosarcoma of the liver in Great Britain in proximity to vinyl chloride sites.
Occup Environ Med 54:14-18.
Fedtke, N; Boucheron, JA; Walker, VE; et al. (1990) Vinyl chloride-induced DNA adducts. II. Formation and
persistence of 7-(2'-oxoethyl)guanine and N2,3-ethenoguanine in rat tissue DNA. Carcinogenesis 11:1287-1292.
Feron, VJ; Kroes, R. (1979) One-year time-sequence inhalation toxicity study of vinyl chloride in rats. II.
Morphological changes in the respiratory tract, ceruminous glands, brain, kidneys, heart, and spleen. Toxicology
13:131-141.
66
-------
Feron, VJ; Kruysse, A; Til, HP. (1979) One-year time-sequence inhalation toxicity study of vinyl chloride in rats. I.
Growth, mortality, haematology, clinical chemistry and organ weights. Toxicology 13:25-28.
Feron, VJ; Hendriksen, CFM; Speek, AJ; et al. (1981) Lifespan oral toxicity study of vinyl chloride in rats. Food
Cosmet Toxicol 19(3):317-333.
Filser, JG; Bolt, HM. (1979) Pharmacokinetics of halogenated ethylenes in rats. Arch Toxicol 42:123-136.
Fontana, L; Baietto, M; Becker, F; et al. (1995) Etude clinique et capillaroscopique de phenomenes de raynaud chez
des retraites anciennement exposes au chlorure de vinyle monomere. J Mai Vasculaires 20:268-273.
Fox, AJ; Collier, PF. (1977) Mortality experience of workers' exposure to vinyl chloride monomer in the
manufacture of polyvinyl chloride in Great Britain. Br J Ind Med 34:1-10.
Freireich, EJ; Gehan, EA; Rail, DP; et al. (1966) Quantitative comparison of toxicity anticancer agents in mouse,
rat, hamster, dog, monkey and man. Cancer ChemotherRep 50:210-244.
Gargas, ML; Andersen, ME. (1982) Metabolism of inhaled brominated hydrocarbons: validation of gas uptake
results by determination of a stable metabolite. Toxicol Appl Pharmacol 66:55-68.
Gedigk, P; Muller, R; Bechtelsheimer, H. (1975) Morphology of liver damage among polyvinyl chloride production
workers. A report on 51 cases. Ann NY Acad Sci 246:279-285.
Gehring, PJ; Watanabe, PG; Park, CN. (1978) Resolution of dose-response toxicity data for chemicals requiring
metabolic activation: example—vinyl chloride. Toxicol Appl Pharmacol 44:581-591.
Goldsworthy, TL; Recio, L; Brown, K; et al. (1994) Symposium overview: transgenic animals in toxicology.
Fundam Appl Toxicol 22:8-19.
Green, T; Hathway, DE. (1975) Biological fate in rats of vinyl chloride in relation to its oncogenicity. Chem-Biol
Interact 11:545-562.
Greenblatt, MS; Bennett, WP; Hollstein, M; et al. (1994) Mutations in the p53 tumor suppressor gene: clues to
cancer etiology and molecular pathogenesis. Cancer Res 54:4855-4878.
Guengerich, FP. (1987) Mammalian cytochromes P-450. Vol. I. Boca Raton, FL: CRC Press, Inc.
Guengerich, FP; Strickland, TW. (1977) Metabolism of vinyl chloride: destruction of the heme of highly purified
liver microsomal cytochrome P-450 by a metabolite. Mol Pharmacol 13:993-1004.
Guengerich, FP; Watanabe, PG. (1979) Metabolism of [14C]-labeled vinyl chloride in vivo and in vitro. Biochem
Pharmacol 28:589-596.
Guengerich, FP; Mason, PS; Stott, WT; et al. (1981) Roles of 2-haloethylene oxides and 2-haloacetaldehydes
derived from vinyl bromide and vinyl chloride in irreversible binding to protein and DNA. Cancer Res 41:43 91 -
4398.
Guengerich, FP; Kim, D; Iwasaki, M. (1991) Role of human cytochrome P-450 IIE1 in the oxidation of many low
molecular weight cancer suspects. Chem Res Toxicol 4:168-179.
Gwinner, IM; Laib, RJ; Filser, JG; et al. (1983) Evidence of chloroethylene oxide being the reactive metabolite of
vinyl chloride towards DNA: comparative studies with2,2'-dichloroethylether. Carcinogenesis 4:1483-1486.
Hansteen, IL; Hillestad, L; Thiis-Evensen, E; et al. (1978) Effects of vinyl chloride in man: a cytogenetic follow-up
study. MutatRes 51:271-278.
67
-------
Hatch, M; Kline, J; Stein, Z. (1981) Power considerations in studies of reproductive effects of vinyl chloride and
some structural analogs. Environ Health Perspect 41:195-201.
Hefner, RE, Jr; Watanabe, PG; Gehring, PJ. (1975) Preliminary studies of the fate of inhaled vinyl chloride
monomer in rats. Ann NY Acad Sci 246:135-148.
Hehir, RM; McNamara, BP; McLaughlin, J, Jr; et al. (1981) Cancer induction following single and multiple
exposures to a constant amount of vinyl chloride monomer. Environ Health Perspect 41:63-72.
Heldaas, SS; Langard, SL; Anderson, A. (1984) Incidence of cancer among vinyl chloride andpolyvinyl chloride
workers: further evidence for an association with malignant melanoma. Br J Ind Med 41:25-30.
Hiatt, GFS; Cogliano, VJ; Becker, RA; et al. (1994) Vinyl chloride action levels: indoor air exposures at a
Superfund site. In: Hazardous waste and public health: International Congress on the Health Effects of Hazardous
Waste. Andrews, JS, Jr; Frumkin, H; Johnson, BL; et al.; eds. Princeton, NJ: Princeton Scientific Publishing, pp.
525-529.
Ho, SF; Phoon, WH; Gan, SL; et al. (1991) Persistent liver dysfunction among workers at a vinyl chloride monomer
polymerization plant. J Soc Occup Med 41:10-16.
Hong, CB; Winston, JM; Thornburg, LP; et al. (1981) Follow-up study on the carcinogenicity of vinyl chloride and
vinylidene chloride in rats and mice: tumor incidence and mortality subsequent to exposure. J Toxicol Environ
Health 7:909-924.
Infante, PF; Wagoner, JK; Waxweiler, PJ. (1976) Carcinogenic, mutagenic and teratogenic risks associated with
vinyl chloride. Mutat Res 41:131-142.
Ivanetich, KM; Aronson, I; Katz, ID. (1977) The interaction of vinyl chloride with rat hepatic microsomal
cytochrome P-450 in vitro. Biochem Biophys Res Commun 74:1411-1418.
Jaeger, RJ; Murphy, SD; Reynolds, ES; et al. (1977) Chemical modification of acute hepatotoxicity of vinyl
chloride monomer in rats. Toxicol Appl Pharmacol 41:597-607.
Jedrychowski, RA; Sokal, JA; Chmielnicka, J. (1984) Influence of exposure mode on vinyl chloride action. Arch
Toxicol 55:195-198.
Jedrychowski, RA; Sokal, JA; Chmielnicka, J. (1985) Comparison of the impact of continuous and intermittent
exposure to vinyl chloride, including phenobarbital effects. J Hyg Epidemiol Microbiol Immunol 28:111-120.
John, JA; Smith, FA; Leong, BKJ; et al. (1977) The effects of maternally inhaled vinyl chloride on embryonal and
fetal development in mice, rats, and rabbits. Toxicol Appl Pharmacol 39:497-513.
Jones, RW; Smith, DM; Thomas, PG. (1988) A mortality study of vinyl chloride monomer workers employed in the
United Kingdom in 1940-1974. Scand J Work Environ Health 14:153-160.
Keplinger, ML; Goode, JW; Gordon, DE; et al. (1975) Interim results of exposure of rats, hamsters, and mice to
vinyl chloride. Ann NY Acad Sci 248:219-224.
Krajewski, J; Dobecki, M; Gromiec, J. (1980) Retention of vinyl chloride in the human lung. Br J Ind Med 37: 373-
374.
Laib, RJ. (1986) The role of cyclic base adducts in vinyl-chloride-induced carcinogenesis: studies on nucleic acid
alkylation in vivo. I ARC Sci Publ 70:101-108.
68
-------
Laib, RJ; Stoeckle, G; Bolt, HM; et al. (1979) Vinyl chloride and trichloroethylene. Comparison of alkylating
effects of metabolites and induction of preneoplastic enzyme deficiencies in rat liver. J Cancer Res Clin Oncol
94:139-147.
Laib, RJ; Klein, KP; Bolt, HM. (1985) The rat liver foci bioassay. I. Age-dependence of induction by vinyl chloride
of ATPase-deficient foci. Carcinogenesis 6:65-68.
Laib, RJ; Bolt, HM; Cartier, R; et al. (1989) Increased alkylation of liver DNA and cell turnover in young versus
old rats exposed to vinyl chloride correlates with cancer susceptibility. Toxicol Lett 45:231-239.
Langauer-Lewowicka, H; Kurzbauer, H; Byczkowska, Z; et al. (1983) Vinyl chloride disease-neurological
disturbances. Int Arch Occup Environ Health 52:151-157.
Lee, CC; Bhandari, JC; Winston, JM; et al. (1978) Carcinogenicity of vinyl chloride and vinylidene chloride.
J Toxicol Environ Health 4:15-30.
Lee, FI; Smith, PM; Bennett, B; et al. (1996) Occupationally related angiosarcoma of the liver in the United
Kingdom 1973-1994. Gut 39:312-318.
Leibman, KC. (1977) Current literature: metabolism of vinyl chloride. Drug Metab Disp 5:93-94.
Lelbach, WK. (1996) A 25-year follow-up study of heavily exposed vinyl chloride workers in Germany. Am J Ind
Med 29:446-458.
Lester, D; Greenberg, LA; Adams, WR. (1963) Effects of single and repeated exposures of humans and rats to vinyl
chloride. Am Ind Hyg Assoc J 24:265-275.
Lilis, R; Anderson, H; Nicholson, WJ; et al. (1975) Prevalence of disease among vinyl chloride and polyvinyl
chloride workers. Ann NY Acad Sci 246:22-41.
Luo, JC; Liu, HT; Cheng, TJ; et al. (1999) Plasma p53 protein and anti-p53 antibody expression in vinyl chloride
monomer workers on Taiwan. J Occup Environ Med 41:521-526.
Maltoni, C; Lefemine, G; Ciliberti, A; et al. (1980) Vinyl chloride carcinogenicity bioassays (BT project) as an
experimental model for risk identification and assessment in environmental and occupational carcinogenesis.
Epidemiol Anim Epidemiol Hum: Cas Chlorure Vinyle Monomer (Reun. Club Carcerog. Chim.) 20th meeting date
1979, Publ. Essent, Paris, France, pp. 11-112.
Maltoni, C; Lefemine, G; Ciliberti, A; et al. (1981) Carcinogenicity bioassay of vinyl chloride monomer: a model of
risk assessment on an experimental basis. Environ Health Perspect 41:3-29.
Maltoni, C; Lefemine, G; Ciliberti, A; et al. (1984) Experimental research on vinyl chloride carcinogenesis. In:
Archives of research on industrial carcinogenesis. Vol. 2. Maltoni, C; Mehlman, MA, eds. Princeton, NJ: Princeton
Scientific Publishers, Inc.
Marion, MJ; DeVivo, I; Smith, S; et al. (1996) The molecular epidemiology of occupational carcinogenesis in vinyl
chloride exposed workers. Int Arch Occup Environ Health 68:394-398.
Marsteller, HJ; Lelbach, WK; Muller, R; et al. (1975) Unusual splenomegalic liver disease as evidenced by
peritoneoscopy and guided liver biopsy among polyvinyl chloride production workers. Ann NY Acad Sci 246:95-
134.
Monson, RR; Peters, JM; Johnson, MN; et al. (1975) Proportional mortality among vinyl chloride workers. Environ
Health Perspect 11:75-77.
69
-------
Morinello, EJ; Ham, A-J; Swenberg, JA. (1999) Molecular dosimetry of N2,3-ethenoguanine in control and vinyl
chloride-exposed rats. Toxicol Sci 48 (Suppl 1):232.
Nakajima, T; Wang, R; Elovaara, E; et al. (1993) Cytochrome P450-related differences between rats and mice in the
metabolism of benzene, toluene and trichloroethylene in liver microsomes. Biochem Pharmacol 45:1079-1085.
National Institute for Occupational Safety and Health (NIOSH), Public Health Service, U.S. Department of Health
and Human Services. (1986) Chemical review: vinyl chloride. Dangerous properties of industrial materials report
6:13-43. NIOSH publication no. 00188891. Washington, DC.
National Research Council (NRC). (1983) Risk assessment in the federal government: managing the process.
Washington, DC: National Academy Press.
Occidental Chemical Corporation. (1975) Initial submission: Pottstown vinyl chloride medical survey results (final
report with attachments and cover letter dated 02/21/92). U.S. Environmental Protection Agency/OTS public files
(8eCP). Document no. 88-920001077. Fiche no. 0535773.
Ott, MG; Langner, RR; Holder, BB. (1975) Vinyl chloride exposure in a controlled industrial environment. Arch
Environ Health 30:333-339.
Perrard, MH. (1985) Mutagenicity and toxicity of chloroethylene oxide and chloroacetaldehyde. Experientia
41:676-677.
Perticoni, GF; Abbritti, G; Cantsani, TA; et al. (1986) Polyneuropathy in workers with long exposure to vinyl
chloride. Electrophysiological study. Electromyogr Clin Neurophysiol 26:41-47.
Pessayre, D; Wandscheer, JC; Descatoire, V; et al. (1979) Formation and inactivation of a chemically reactive
metabolite of vinyl chloride. Toxicol Appl Pharmacol 49:505-515.
Pirastu, R; Comba, P; Reggiani, A; et al. (1990) Mortality from liver disease among Italian vinyl chloride
monomer/polyvinyl chloride manufacturers. Am J Ind Med 17:155-161.
Pirastu, R; Bruno, C; De Santis, M; et al. (1998) An epidemiological study of workers exposed to vinyl chloride in
the plants of Ferrara, Rosignano and Ravenna. Epidemiol Prev 22:226-236.
Plugge, H; Safe, S. (1977) Vinyl chloride metabolism—a review. Chemosphere 6:309-325.
Popper, H; Thomas, LB. (1975) Alterations of liver and spleen among workers exposed to vinyl chloride. Ann NY
Acad Sci 246:172-194.
Purchase, IFH; Richardson, CR; Anderson, D; et al. (1978) Chromosomal analysis in vinyl chloride exposed
workers. MutatRes 57:325-334.
Ramsey, JR; Andersen, ME. (1984) A physiologically based description of the inhalation pharmacokinetics of
styrene in rats and humans. Toxicol Appl Pharmacol 73:159-175.
Rannug, U; Johansson, A; Ramel, C; et al. (1974) The mutagenicity of vinyl chloride after metabolic activation.
Ambio 3:194-197.
Rannug, U; Gothe, R; Wachtmeister, CA. (1976) The mutagenicity of chloroethylene oxide, chloro-acetaldehyde, 2-
chloroethanol and chloroacetic acid, conceivable metabolites of vinyl chloride. Chern Biol Interact 12:251-263.
Reitz, RH; Gargas, ML; Andersen, ME; et al. (1996) Predicting cancer risk from vinyl chloride exposure with a
physiologically based pharmacokinetic model. Toxicol Appl Pharmacol 137:253-267.
70
-------
Reynolds, ES; Moslen, MT; Szabo, S; et al. (1975) Vinyl chloride-induced deactivation of cytochrome P-450 and
other components of the liver mixed function oxidase system: an in vivo study. Res Comm Chem Pathol Pharmacol
12:685-694.
Sandgren, EP; Quaife, CJ; Pinkert, CA; et al. (1989) Oncogene-induced liver neoplasia in transgenic mice.
Oncogene 4:715-724.
Sax, NI; Lewis, RJ, eds. (1989) Dangerous properties of industrial materials. 7th ed. New York: VanNostrand
Reinhold Co., pp. 3473-3474.
Schoental, R; Magee, PN. (1957) Chronic liver changes in rats after a single dose of lasiocarpine, a pyrrolizidine
(Senecio) alkaloid. J Pathol Bact 74:305-319.
Schoental, R; Magee, PN. (1959) Further observations on the subacute and chronic liver changes in rats after a
single dose of various pyrrolizidine (Senecio) alkaloids. J Pathol Bact 78:471-482.
Short, RD; Minor, JL; Winston, JM; et al. (1977) Dominant lethal study in male rats after repeated exposures to
vinyl chloride or vinylidene chloride. J Toxicol Environ Health 3:965-968.
Simonato, L; L'Abbe, KA; Andersen, A; et al. (1991) A collaborative study of cancer incidence and mortality
among vinyl chloride workers. Scand J Work Environ Health 17:159-169.
Singer, B. (1985) In vivo formation and persistence of modified nucleosides resulting from alkylating agents.
Environ Health Perspect 62:41-48.
Singer, B; Spengler, SJ; Chavez, F; et al. (1987) The vinyl chloride-derived nucleoside, N2,3-ethenoguanosine, is a
highly efficient mutagen in transcription. Carcinogenesis 8:745-747.
Sinues, B; Sanz, A; Bernal, ML. (1991) Sister chromatid exchanges, proliferating rate index, and micronuclei in
biomonitoring of internal exposure to vinyl chloride monomer in plastic industry workers. Toxicol Appl Pharmacol
108:37-45.
Sipes, IG; Gandolfi, AJ. (1991) Biotransformation of toxicants. In: Casarett and Doull's Toxicology, 4th ed.
Klasson, CD; Amdur, MO; Doull, J, eds. New York: McGraw Hill, pp. 88-126.
Smith, SJ; Yongliang, L; Whitley, R; et al. (1998) Molecular epidemiology of p53 mutations in workers exposed to
vinyl chloride. Am J Epidemiol 147:302-308.
Smulevich, VB; Fedotova, IV; Filatova, VS. (1988) Increasing evidence of the rise of cancer in workers exposed to
vinyl chloride. Br J Ind Med 45:93-97.
Soini, Y; Welsh, JA; Ishak, KG; et al. (1995) p53 mutations in primary hepatic angiosarcomas not associated with
vinyl chloride exposure. Carcinogenesis 16:2879-2881.
Sokal, JA; Baranski, B; Majka, J; et al. (1980) Experimental studies on the chronic toxic effects of vinyl chloride in
rats. J Hyg Epidemiol Microbiol Immunol 24:285-294.
Storm, JE; Rozman, KK. (1997) Evaluation of alternative methods for establishing safe levels of occupational
exposure to vinyl halides. Regul Tox Pharmacol 24:240-255.
Studniarek, M; Durski, K; Liniecki, J; et al. (1989) Effects of vinyl chloride on liver function of exposed workers,
evaluated by measurements of plasma clearance of the 99mTc-N-2,4-dimethylacetanilido-iminodiacetate complex.
J Appl Toxicol 9:213-218.
Suzuki, Y. (1978) Pulmonary tumors induced in mice by vinyl chloride monomer. Environ Res 16:285-301.
71
-------
Suzuki, Y. (1983) Neoplastic effect of vinyl chloride in mouse lung—lower doses and short-term exposure. Environ
Res 32:91-103.
Swenberg, JA; Fedtke, N; Ciroussel, F; et al. (1992) Etheno adducts formed in DNA of vinyl chloride-exposed rats
are highly persistent in liver. Carcinogenesis 13(4):727-729.
Swenberg, JA; Bogdanffy, MS; Ham, A; et al. (1999) Formation and repair of DNA adducts in vinyl chloride- and
vinyl fluoride-induced carcinogenesis. IARC Sci Publ 150:29-43.
Tabershaw, IR; Gaffey, WR. (1974) Mortality study of workers in the manufacture of vinyl chloride and its
polymers. J Occup Med 16:509-518.
Tamburro, CH; Makk, L; Popper, H. (1984) Early hepatic histologic alterations among chemical (vinyl monomer)
workers. Hepatology 4:413-418.
Tarkowski, S; Wisniewska-Knypl, JM; Klimczak, J; et al. (1980) Urinary excretion of thiodigylcolic acid and
hepatic content of free thiols in rats at different levels of exposure to vinyl chloride. J Hyg Epidemiol Microbiol
Immunol 24:253-261.
Tennant, RW; French, JE; Spalding, JW. (1995) Identifying chemical carcinogens and assessing potential risk in
short-term bioassays using transgenic mouse models. Environ Health Perspect 103:942-950.
Theriault, G; Iturra, H; Gingras, S. (1983) Evaluation of the association between birth-defects and exposure to
ambient vinyl-chloride. Teratology 27:359-370.
Til, HP; Immel, HR; Feron, VJ. (1983) Lifespan oral carcinogenicity study of vinyl chloride in rats. Report no. V
83f.285/291099. Zeist, The Netherlands: Civo Institutes TNO.
Til, HP; Feron, VJ; Immel, HR. (1991) Lifetime (149-week) oral carcinogenicity study of vinyl chloride in rats.
Food Chem Toxicol 29:713-718.
Torkelson, TR; Oyen, F; Rowe, VK. (1961) The toxicity of vinyl chloride as determined by repeated exposure of
laboratory animals. Am Ind Hyg Assoc J 22:354-361.
Trivers, GE; Cawley, HL; DeBenedetti, VMG; et al. (1995) J Natl Cancer Inst 87:1400-1407.
Ungvary, G; Hudak, A; Tatrai, E; et al. (1978) Effects of vinyl chloride exposure alone and in combination with
trypanblue applied systematically during all thirds of pregnancy on the fetuses of CFYrats. Toxicology 11(1):45-
54.
U.S. Environmental Protection Agency. (1985) Drinking water criteria document for vinyl chloride. Washington,
DC. Prepared by the Office of Health and Environmental Assessment, Environmental Criteria and Assessment
Office, Cincinnati, OH, for the Office of Drinking Water, NTIS PB86-118320.
U.S. Environmental Protection Agency. (1986) Guidelines for carcinogen risk assessment. Federal Register
51(185):33992-43003.
U.S. Environmental Protection Agency. (1987) Risk assessment guidelines of 1986. Office of Health and
Environmental Assessment, Washington, DC. EPA/600/8-87/045, dated August 1987.
U.S. Environmental Protection Agency. (1988, Feb) Recommendations for and documentation of biological values
for use in risk assessment. Office of Health and Environmental Assessment, Washington, DC. EPA 600/6-87/800.
NTIS PB88-179874/AS.
U.S. Environmental Protection Agency. (1991, Dec. 5) Guidelines for developmental toxicity risk assessment.
Federal Register 56(234):63798-63826.
72
-------
U.S. Environmental Protection Agency. (1992) Request for comments on draft report of cross-species scaling factor
for cancer risk assessment. Federal Register 57:24152.
U.S. Environmental Protection Agency. (1994a, Oct. 26) Interim policy for particle size and limit concentration
issues in inhalation toxicity: notice of availability. Federal Register 59(206):53799.
U.S. Environmental Protection Agency. (1994b, Oct. 26) Methods for derivation of inhalation reference
concentrations and application of inhalation dosimetry. EPA/600/8-90/066F. Prepared by the Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office, Cincinnati, OH.
U.S. Environmental Protection Agency. (1994c) Peer review and peer involvement at the U.S. Environmental
Protection Agency, signed by U.S. EPA Administrator, Carol M. Browner, dated June 7, 1994.
U.S. Environmental Protection Agency. (1994d) Health effects assessment summary tables FY 1994: supplement
number 2. Prepared by the Office of Health and Environmental Assessment, Environmental Criteria and
Assessment Office, Cincinnati, OH. Washington, DC. EPA/540/R/94-114. NTIS PB94-921102.
U.S. Environmental Protection Agency. (1995a, Oct. 4) Proposed guidelines for neurotoxicity risk assessment.
Federal Register 60(192):52032-52056.
U.S. Environmental Protection Agency. (1995b, Feb.) Use of the benchmark dose approach in health risk
assessment. Prepared by the Office of Health and Environmental Assessment, Environmental Criteria and
Assessment Office, Cincinnati, OH. EPA/630/R-94/007.
U.S. Environmental Protection Agency. (1995c) Guidance on risk characterization. Memorandum of the U.S. EPA
Administrator, Carol Browner, dated March 21, 1995.
U.S. Environmental Protection Agency. (1996a) Proposed guidelines for carcinogen risk assessment. Office of
Research and Development. Federal Register 61(79): 17960-18011.
U.S. Environmental Protection Agency. (1996b) Guidelines for reproductive toxicity risk assessment. Federal
Register 61(212):56274-56322.
Viola, PL. (1970) Pathology of vinyl chloride. Med Lav 61:174-180.
Viola, PL; Bigotti, A; Caputo, A. (1971) Oncogenic response of rats' skin, lungs, and bones to vinyl chloride.
Cancer Res 31:516-522.
Walles, SAS; Holmberg, B; Svensson, K; et al. (1988) Induction of single-strand breaks in liver DNA of mice after
inhalation of vinyl chloride. IARC Sci Publ 89:227-231.
Watanabe, PG; Gehring, PJ. (1976) Dose-dependent fate of vinyl chloride and its possible relationship to
oncogenicity in rats. Environ Health Perspect 17:145-152.
Watanabe, PG; McGowan, GR; Gehring, PJ. (1976a) Fate of [14C]-vinyl chloride after single oral administration in
rats. Toxicol Appl Pharmacol 36:339-352.
Watanabe, PG; McGowan, GR; Madrid, EO; et al. (1976b) Fate of [14C]-vinyl chloride following inhalation
exposure in rats. Toxicol Appl Pharmacol 37:49-59.
Watanabe, PG; Hefner, RE, Jr; Gehring, PJ. (1976c) Vinyl chloride-induced depression of hepatic non-protein
sulfhydryl content and effects onbromosulphalein (BSP) clearance in rats. Toxicology 6:1-8.
Watanabe, PG; Zempel, JA; Peggy, DG; et al. (1978) Hepatic macromolecular binding following exposure to vinyl
chloride. Toxicol Pharmacol 44:571-579.
73
-------
Waxweiler, RJ; Stringer, W; Wagoner, JK; et al. (1976) Neoplastic risk among workers exposed to vinyl chloride.
Ann NY Acad Sci 271:40-48.
Waxweiler, RJ; Falk, H; McMichael, A; et al. (1977) A cross-sectional epidemiologic survey of vinyl chloride
workers. NTIS PB274193. Springfield, VA.
Waxweiler, RJ; Smith, AH; Falk, H; et al. (1981) Excess lung cancer risk in a synthetic chemicals plant. Environ
Health Perspect 41:159-165.
Weber, H; Reinl, W; Greiser, E. (1981) German investigations on morbidity and mortality of workers exposed to
vinyl chloride. Environ Health Perspect 41:95-99.
Whysner, J; Conaway, CC; Verna, L; et al. (1996) Vinyl chloride mechanistic data and risk assessment: DNA
reactivity and cross-species quantitative risk extrapolation. Pharmacol Ther 71:7-28.
Wisniewska-Knypl, J; Klimczak, J; Kolakowski, J. (1980) Monooxygenase activity and ultrastructural changes of
liver in the course of chronic exposure of rats to vinyl chloride. Int Arch Occup Environ Health 46(3):241-249.
Withey, JR. (1976) Pharmacodynamics and uptake of vinyl chloride monomer administered by various routes to
rats. J Toxicol Environ Health 1:381-394.
Wong, O; Whorton, MD; Foliart, DE; et al. (1991) An industry-wide epidemiologic study of vinyl chloride workers,
1942-1982. Am J Ind Med 20:317-334.
Wu, W; Steenland, K; Brown, D; et al. (1989) Cohort and case-control analyses of workers exposed to vinyl
chloride: an update. J Occup Med 31:518-523.
Zajdela, F; Croisy, A; Barbin, A; et al. (1980) Carcinogenicity of chloroethylene oxide, an ultimate reactive
metabolite of vinyl chloride, and Bis(chloromethyl)ether after subcutaneous administration and in initiation-
promotion experiments in mice. Cancer Res 40:352-356.
74
-------
APPENDIX A. COMPARISON OF PBPK MODELS FOR VINYL CHLORIDE
The calculations performed in this risk assessment used the PBPK model of Clewell et al.
(1995a). Another model of vinyl chloride (VC) was recently published (Reitz et al., 1996). The
purpose of this appendix is to provide a comparison of the two models and to demonstrate the
similarity of risk calculations based on either model. For comparison purposes, tumor
incidences were based on liver angiosarcoma only, rather than incidence of all liver tumors used
to develop recommended risk estimates.
A.l. REVIEW OF REPORTED PBPK MODELS FOR VC
Five different PBPK models for VC have been described in the literature. The first
(Chen and Blancato, 1989) was a simple description of parent chemical kinetics and total
metabolism based on the styrene model of Ramsey and Andersen (1984). Metabolism of VC
was modeled with a single saturable pathway, and the kinetic constants were estimated from
measurements of whole-body clearance (e.g., Filser and Bolt, 1989). No attempt was made to
validate the model against data on blood time-courses or total metabolism. The model was used
to calculate total metabolism of VC (representing total production of reactive metabolites) as the
dose metric in a carcinogenic risk assessment for VC. Potency estimates based on the internal
dose (mg VC metabolized per kg/day) were derived from inhalation bioassays of VC performed
by Maltoni et al. (1981, 1984), as well as from human epidemiological data. Using the same
internal dose metric (mg metabolized per kg/day), the inhalation potency estimated from
epidemiological data of Fox and Collier (1977) of 3.8 x 10"3/ppm (1.4 x 10"6 per • g/m3) was
essentially identical to the potency estimated from rat inhalation data of 1.7 - 3.7 x 10"3/ppm (0.7
- 1.4 x 10"6 per • g/m3) using body-weight scaling (that is, without applying a body surface area
correction for cross-species scaling). Although the extrapolations performed by Chen and
Blancato were for carcinogenic risk, the PBPK model would be equally effective for noncancer
endpoints.
The second model published for VC (Gargas et al., 1990) was a generic model of volatile
chemical kinetics in a recirculated closed chamber, which was used to identify global metabolic
parameters in the rat for a number of chemicals, including VC. It differed from the model of
Chen and Blancato chiefly by the incorporation of a second, linear metabolic pathway (presumed
to be glutathione conjugation) in parallel with the saturable (oxidative) pathway. Based on gas
uptake studies, both a saturable and a linear metabolic component were postulated for VC.
The different descriptions of metabolism in the two models discussed above were
examined in a more in-depth study of VC pharmacokinetics performed for the U.S. Air Force by
the K.S. Crump Division of Clement International (Clement, 1990). They refitted the one- and
two-pathway descriptions to gas uptake data and then compared their predictions with
measurements of total metabolism by Gehring et al. (1978) and Watanabe et al. (1976).
Although the two-pathway description provided a significantly better fit to the gas uptake data
A-l
-------
(adding parameters nearly always improves a fit), the resulting parameters tended to overpredict
total metabolism at higher concentrations owing to the presence of the first-order component. In
addition, it was not possible to explain the continued increase in glutathione (GSH) depletion
measured at the highest exposure levels (where the saturable component was above saturation)
because only products of the oxidative metabolism of VC have been shown to react with GSH.
In an attempt to provide a better correspondence to the data on both total metabolism and
glutathione depletion, two possible refinements to the model were investigated. In the first,
direct reaction of VC with GSH was postulated, and in the second, the products of both the
saturable and the linear pathways were assumed to react with GSH. Unfortunately, neither
description was able to provide a satisfactory correspondence to both total metabolism and GSH
depletion data. The authors suggested that a different formulation featuring two saturable
oxidative pathways, both producing reactive metabolites, might provide the required behavior.
This suggestion formed the basis for the subsequent development of the PBPK model of Clewell
etal. (1995a).
More recently, a PBPK model of VC was developed by Reitz et al. (1996) and applied to
compare cancer potency in mice, rats, and humans. The structure of the model was similar to
that of Chen and Blancato (1989), providing a description of parent chemical kinetics and total
metabolism based on the styrene model of Ramsey and Andersen (1984). Metabolism of VC
was modeled with a single saturable pathway, and the kinetic constants were estimated from
fitting of closed chamber gas uptake studies with rats. The model was then validated against
data on total metabolism in the rat (Watanabe et al., 1976), gas uptake data in the mouse, and
inhalation data in the human (Baretta et al., 1969). The model was used to calculate total
metabolism of VC as the dose metric in carcinogenic risk assessments for VC. On the basis of
the rat inhalation bioassay of Maltoni et al. (1981, 1984), and using the linearized multistage
model, they estimated that lifetime continuous human exposure to 1.75 • g VC is associated with
an increased lifetime risk of one in a million. This estimate equates to a lifetime risk of
approximately 0.6 x 10"6/* g/m3, in good agreement with the results of Chen and Blancato (1989).
The potency estimates from rats were then shown to be consistent with tumor incidence data in
mice and humans when the pharmacokinetic dose metric was used.
In a parallel effort, a more elaborate PBPK model of VC was developed for OSHA and
EPA to support a cancer risk assessment for VC (Clewell et al., 1995a). This model and the
modeling results are described in more detail in Appendix B. Following the suggestion of
Clement (1990), the initial metabolism of VC was hypothesized to occur via two saturable
pathways, one representing low-capacity-high-affinity oxidation by CYP2E1 and the other
representing higher capacity-lower affinity oxidation by other isozymes of P450, producing in
both cases chloroethylene oxide (CEO) as an intermediate product. The percentage of CEO
converted to CO2 via reaction with H2O was determined from published reports of radiolabeled
VC whole-body metabolism studies. Previous in vitro and in vivo studies support
chloroacetaldehyde (CAA) as the major metabolite of VC through the breakdown of CEO, and
this metabolite was modeled as the major substrate in GSH conjugation, with a lesser amount of
CEO as the glutathione S-epoxide transferase substrate. Depletion of glutathione by reaction
with CAA was also described. The parameter values for the two metabolic pathways describing
the initial step in VC metabolism were determined by simulation of gas uptake data from mice,
A-2
-------
rats, hamsters, monkeys, and controlled human inhalation exposures, as well as from data on
total metabolism and glutathione depletion in both oral and inhalation exposures of rats. The use
of a low-affinity pathway in parallel with the high-affinity pathway was able to successfully
reproduce the continued increases in total metabolism and GSH depletion observed with VC in
rats. The successful simulation of pharmacokinetic data from a large number of studies over a
wide range of concentrations, using multiple routes of exposure, served as evidence that the
PBPK model was valid over the exposure range of interest.
As with the PBPK model of Chen and Blancato (1989), the use of a pharmacokinetic
dose metric reflecting lifetime average daily dose to the target tissue resulted in similar potency
estimates for liver angiosarcoma from VC across different species. The human risk estimates
based on studies with mice (1.0 x 10"6 to 2.3 x 10"6 per • g/m3) agreed very well with those based
on inhalation studies with rats (1.6 x 10"6 to 3.7 x 10"6 per • g/m3), demonstrating the ability of
pharmacokinetics to integrate dose-response information across species. Lifetime risk of liver
cancer from VC exposure estimated from three epidemiological studies was 4.7 x 10"7 to 2.8 x
10"6per • g/m3, in good agreement with the estimates based on animal inhalation data. The risk
estimates obtained with this model are also very similar to those obtained with the simpler PBPK
models of Chen and Blancato (1989) and Reitz et al. (1996), as described above. It should be
noted, however, that human exposure estimates have a considerable degree of uncertainly, so
agreement may be at least to some extent due to chance.
The human inhalation risks were somewhat greater when estimated using data from
female rats exposed orally to VC in the Feron et al. (1981) study. These estimates ranged from
2.0 x 10"6 when based on angiosarcomas alone to 2.1 x 10"4 per • g/m3 when based on all liver
tumors including angiosarcomas, hepatocellular carcinomas, and neoplastic nodules. The
estimate based on angiosarcomas alone is in general agreement with those derived from female
rats using the oral bioassays of Maltoni et al. (1981, 1984), 3.0 x 10"5 per • g/m3. Human cancer
potency estimates based on oral exposure are unavailable, because ingestion is not a common
route of human exposure. It is quite possible, however, that potency for induction of liver cancer
is somewhat greater by the oral route of exposure, because essentially all absorbed VC passes
through the liver before entering the systemic circulation, whereas some of the VC taken up
through the lungs may be metabolized by other tissues before reaching the liver.
In summary, the results of pharmacokinetic risk assessments using three different PBPK
models are in remarkable agreement, with lifetime risk estimates for different species exposed
via the inhalation route that range over about an order of magnitude, from 0.5 x 10"6 to 5 x 10"6
per • g/m3. These pharmacokinetic risk estimates for the inhalation route of exposure are lower
than those currently used in environmental decision making by slightly more than an order of
magnitude. The currently used oral risk estimates, however, agree quite well with previous ones.
The simpler PBPK models of Chen and Blancato (1989) or Reitz et al. (1996) would provide an
acceptable framework for conducting a pharmacokinetically based human risk assessment for
VC, and would provide a more accurate estimate of human risk than external measures of VC
exposure. However, the two-saturable-pathway model structure used by Clewell et al. (1995a) is
better validated because, in addition to the data used to validate the other models, it was
validated against experimental data on both total metabolism and GSH depletion in rats as well
A-3
-------
as closed-chamber VC exposure data in humans.
A.2. COMPARISON OF REITZ AND CLEWELL MODELS
A more complete comparison was performed between the model used in this risk
assessment and the recently published model of Reitz et al. (1996). The structures of the two
models are shown in Figures A-l and A-2.1 It can be seen that the structure of the parent
chemical portion of the models is essentially identical. Only the descriptions of metabolism in
the two models differ substantially. The model of Clewell et al. (1995a) includes a more
complex description of metabolism, with two saturable oxidative pathways rather than one, and
with a description of glutathione conjugation of the oxidative metabolites. The purpose of this
additional complexity was (1) to increase confidence in the ability of the model to correctly
simulate VC metabolism by improving the ability of the model to reproduce data on the dose
response for total metabolism and glutathione depletion in rats, and (2) to investigate alternative
dose metrics representing (a) total oxidative metabolites not detoxified by glutathione and (b)
total glutathione conjugates. As reported in Clewell et al. (1995a), the alternative dose metrics
did not provide any improvement over the use of total metabolism and were not used or
presented in the risk assessment. The model components associated with the formation of
glutathione conjugates and the depletion of glutathione do not have any effect on the calculation
of total oxidative metabolism in the model. Therefore, for the calculation of risks based on liver
metabolism dose metrics, the only structural difference between the two models is the use of one
versus two saturable pathways to describe metabolism.
The parameters used in the two models are shown in Tables A-l and A-2.2 Of the
physiological parameters, the only significant differences are in the alveolar ventilation for the
human, the liver volume for the rat, and the body weight and fat volumes for the rat and mouse.
The Clewell et al. (1995a) model used an alveolar ventilation based on EPA's preferred human
ventilation rate (20 m3/day), based on continuous heavy work, whereas the ventilation rates in
the Reitz et al. (1996) model were taken from the International Radiation Consensus Report on
Reference man and were more typical of humans at rest or engaged in light activity. The rat
liver volumes used were recommended in the recent ILSI Risk Science Institute physiological
'For the purpose of this comparison, it was necessary to add oral uptake to the model of Reitz et al. (1996), which
includes only inhalation exposure. This was accomplished by adding a zero-order input term in the equation for the
liver, in the same fashion as in the model of Clewell et al. (1996b).
2The parameter values for the Reitz et al. (1996) model are taken from Table 1 of that publication, with the
exception of the blood/air partition coefficient in the mouse, which was incorrectly reported as 2.26. The value
shown in Table A-l is the value actually used in the risk calculations (R.H. Reitz, personal communication).
A-4
-------
CI
ex
QC
CVF
CVR
CVS
CVL
~n
I
1
r
1
— 1
j
1
•
Hat L
I c-
14 QS
Slow | CA
OL
VMAX1
KM1
CO,
KCO,
CA
VMAX2
KM2
KZER
"KA
Reactive
Metabolites
(RISK)
KGSM
Glutathione
Conjugate
(RISKG)
KFEE
Tissue / DNA
Adducts
(RISKM)
Figure A-l. The PBPK model for vinyl chloride developed by Clewell et al. (1995a).
Abbreviations: QP = alveolar ventilation; CI = inhaled concentration; CX = exhaled concentration; QC =
cardiac output; QF, CVF = blood flow to, and venous concentration leaving, the fat; QR, CVR = blood
flow to, and venous concentration leaving, the richly perfused tissues (most organs); QS, CVS = blood
flow to, and venous concentration leaving, the slowly perfused tissues (e.g., muscle); QL, CVL = blood
flow to, and venous concentration leaving, the liver; VMAX1,KM1 = capacity and affinity for the high-
affinity oxidative pathway enzyme (CYP 2E1); VMAX2,KM2 = capacity and affinity for the lower
affinity oxidative pathway enzymes (e.g., CYP 2C11/6); KZER = zero-order rate constant for uptake of
VC from drinking water; KA = first-order rate constant for uptake of VC from corn oil; KCO2 = first-
order rate constant for metabolism of VC to CO2; KGSM = first-order rate constant for reaction of VC
metabolites with GSH; KFEE = first-order rate constant for reaction of VC metabolites with other cellular
materials, including DNA; KB = first-order rate constant for normal turnover of GSH; KO = zero-order
rate constant for maximum production of GSH; KS = parameter controlling rate of recovery of GSH from
depletion.
A-5
-------
CI QP CX
^ Lung \-
CVF
CVS
CVL r
L
Fat
CVR I
4 1 Rich
Slow
Liver
VMAX
KM
Reactive
Metabolites
(DDOSE)
QC
QF
CA
QR
r
CA
QS
CA
QL
. , CA
KZER
KA
Figure A-2
Figure A-2. Diagram of the PBPK model of Reitz et al. (1996) for VC. Abbreviations are as in
Figure A-l.
A-6
-------
Table A-l. Comparison of model parameters
BW
QPC
QCC
Body weight (kg)
Scaling factor
Alveolar ventilation
(L/hr)
Cardiac output (L/hr)
Mouse
Clewell
0.040-
0.044a
0.75
30.0
18.0
Reitz
0.0285
0.74C
28.0
28.0
Rat
Clewell
0.245-
0.638b
0.75
21.0
18.0
Reitz
0.225
0.74d
18.0
18.0
Human
Clewell
70.0
0.75
24.0
16.5
Reitz
70.0
0.74d
15.0
15.0
Tissue blood flows (fraction of cardiac output):
QRC
QFC
QSC
QLC
Rapidly perfused
tissues
Fat
Slowly perfused
tissues
Liver
0.51
0.09
0.15
0.25
0.52
0.05
0.19
0.24
0.51
0.09
0.15
0.25
0.52
0.05
0.19
0.24
0.5
0.05
0.19
0.26
0.52
0.05
0.19
0.24
Tissue volumes (fraction of body weight):
vsc
we
VRC
VLC
Slow
Fat
Rapid
Liver
0.77
0.12-0.13b
0.035
0.055
0.7614
0.04
0.05
0.0586
0.75
0.11-0.20b
0.05
0.04
0.7647
0.07
0.05
0.0253
0.63
0.19
0.064
0.026
0.6105
0.231
0.0371
0.0314
Partition coefficients:
PB
PF
PS
PR
PL
Blood/air
Fat/blood
Slow/blood
Rapid^lood
Live^lood
2.26
10.62
0.42
0.74
0.74
2.26e
8.85f
0.93a
0.71a
0.71s
2.4
10.0
0.4
0.7
0.7
1.68
11. 9a
1.25a
0.95a
0.95a
1.16
20.7
0.83
1.45
1.45
1.16
17.2a
1.81"
1.38a
1.38a
Metabolic parameters:
VMAX1C
KM1
Maximum velocity
of first saturable
pathway
(mg/hr)
Affinity of first
saturable pathway
(ms/L)
5.0-8.0a
0.1
8.13
0.28
3.0-4.0b
0.1
2.75
0.04
4.0
0.1
3.97
0.04
A-7
-------
Table A-l. Comparison of model parameters (continued)
VMAX2C
KM2
Maximum velocity
of second saturable
pathway
(mg/hr)
Affinity of second
saturable pathway
(mg/L)
Mouse
Clewell
0.1-3.0b
10.0
Reitz
0.0
g
Rat
Clewell
0.1-2.0b
10.0
Reitz
0.0
—
Human
Clewell
0.1
10.0
Reitz
0.0
—
GSH parameters:
KCO2C
KGSMC
KFEEC
GSO
KBC
KS
KOC
First-order
breakdown to CO2
Conjugation rate
constant
Rate constant with
non-GSH
Initial GSH
concentration
First-order rate
constant for GSH
breakdown
Resynthesis constant
Zero-order
production of GSH
1.6
0.13
35.0
5800.0
0.12
2000.0
28.5
_
—
_
_
_
1.6
0.13
35.0
5800.0
0.12
2000.0
28.5
_
—
_
_
_
1.6
0.13
35.0
5800.0
0.12
2000.0
28.5
_
—
_
_
_
Dosing parameters:
KA
Oral uptake rate (/hr)
3.0
—
3.0
—
3.0
—
"See Table A-2.
bFor the purpose of this comparison, it was necessary to add oral uptake to the model of Reitz et al. (1996), which includes only
inhalation exposure. This was accomplished by adding a zero-order input term in the equation for the liver, in the same fashion
as in the model of Clewell et al. (1996b).
cThe scaling factor for maximum velocity of metabolism is 0.70.
dfhe parameter values for the Reitz et al. (1996) model are taken from Table 1 of that publication with the exception of the
blood/air partition coefficient in the mouse, which was incorrectly reported as 2.26. The value shown in Table A-l is the value
actually used in the risk calculations (R.H. Reitz, personal communication).
'Different from reported value of 2.41 (Reitz et al., 1996), but used in risk calculations (D. Reitz, personal communication).
fThe parameters listed here are the tissue/blood partition coefficients. They were derived from the tissue/air partition coefficients
in Table 1 of Reitz et al. (1996) by dividing by the blood/air partition coefficient.
^ot used in model.
A-8
-------
Table A-2. Species/sex/study-dependent parameter values in Clewell model
Swiss albino mice
(inhalation study)
Sprague-Dawley rats
(inhalation study)
Sprague-Dawley rats
(gavage study)
Wistar rats
(drinking water study)
Male
Female
Male - low dose
Male - high dose
Female - low dose
Female - high dose
Male - low dose
Male - high dose
Female - low dose
Female - high dose
Male
Female
BW
0.044
0.040
0.638
0.433
0.485
0.321
0.632
0.405
0.445
0.301
0.436
0.245
VFC
0.13
0.12
0.19
0.13
0.20
0.14
0.19
0.12
0.18
0.13
0.14
0.11
VMAX1C
8.0
5.0
4.0
4.0
3.0
3.0
4.0
4.0
3.0
3.0
4.0
3.0
VMAX2C
O.la
3.0
2.0
2.0
O.lb
O.lb
2.0
2.0
O.lb
O.lb
2.0
O.lb
aZero was used as the variance for this value of VMAX2C in the PBPK-Sim runs.
bFor the purpose of this comparison, it was necessary to add oral uptake to the model of Reitz et al. (1996), which
includes only inhalation exposure. This was accomplished by adding a zero-order input term in the equation for the
liver, in the same fashion as in the model of Clewell et al. (1996b).
parameter document (ILSI, 1994), whereas the Reitz et al. (1996) model used actual necropsy
results. The Clewell et al. model also used the actual animal body weights reported by the
authors of the bioassays, and calculated the fat volume from the observed relationship between
body weight and fat volume in the rodent (ILSI, 1994). The blood/air and tissue/blood partition
coefficients in the two models are for the most part similar, but the slowly perfused tissue/blood
partition coefficients in the Reitz et al. model are as much as threefold higher than those in the
Clewell et al. model. Metabolic parameters also differ somewhat between the two models,
reflecting the different data sets used to estimate metabolism in different species, strains, and
sexes.
The impact of differences in the model parameters can better be evaluated in light of the
results of the parameter sensitivity analysis conducted on the Clewell et al. (1995a) model. Of
the parameters discussed above for which the two models differ, only the body weight, liver
volume, and metabolism parameters have significant impact on dose metric calculations.
Alveolar ventilation and the blood/air partition coefficient have only a minor impact, whereas the
fat volume and tissue/blood partition coefficients have essentially no impact at all. With respect
to the more important differences between the two models in the body weights, liver volumes,
and metabolism parameters, the Clewell et al. model used the actual reported body weights,
A-9
-------
adopted the most recently recommended liver volumes (ILSI, 1994), and employed a much larger
number of studies to estimate and validate the metabolic parameters.
The best way to compare the impact of model selection on risk estimates is simply to
employ the two models in estimating risks from the same studies. The results of this exercise are
shown in Tables A-3 through A-5. Table A-3 shows the dose metrics calculated with the two
models.3 The dose metrics in every case are very similar. The greatest difference, of about 50%
for the Feron et al. (1981) dietary study, is due to the different values used in the models for the
volume of the liver in the rat. As mentioned above, the liver volume used in the Clewell et al.
(1995a) model is the value recommended by ILSI (1994). Table A-4 compares the cancer ED10s
for angiosarcoma calculated with the dose metric from the two models, and Table A-5 provides
the same comparison for noncancer BMD10s for liver necrosis. It should be noted that although
the NOAEL for liver necrosis is tenfold higher than for liver cell polymorphism, the endpoint
used for development of the RfC and RfD in the present assessment, the model comparison is
still valid. The high level of agreement between the ED10s and BMD10s based on the two different
models demonstrates the reliability of PBPK models that have been properly designed and
validated against experimental data.
3The dose metrics for the Reitz et al. (1996) model were obtained with an ACSL version of the model
(VCDOSE2.CSL) kindly provided by Dr. Reitz. The only modification of the model for use in this study was to add
a zero-order oral input term. The model was run with the parameter values shown in Table A-l and the dose metric
calculations were compared with Tables 4 and 5 in Reitz et al. (1996). The ACSL model reproduced the reported
mouse dose metrics within 2% and reproduced the human dose metrics exactly. The minor differences in the mouse
dose metrics are probably due to rounding off of the parameter values as reported in Table 1 of Reitz et al. (1996)
from those originally used to obtain Tables 4 and 5 of that paper (R.H. Reitz, personal communication).
A-10
-------
Table A-3. Comparison of values for lifetime average delivered dose (mg/L liver)
Reference
Occupational
exposure
Maltoni et al.
(1981, 1984)
(BT4)b
Maltoni et al.
(1981, 1984)
(BT1, BT2,
andBT15)c
Feron et al.
(1981)
Route
Inhalation
Drinking water
Inhalation
Inhalation
Food
Species
Human
Swiss
albino mice
Sprague-
Dawley rats
Wistar rats
Duration
Continuous
4 hr/d, 5 d/wk for
30 of 104 wks
4 hr/d, 5 d/wk for
52 of 147 wks
(BT15)
52 of 135 wks
52 of 143 wks
(BT2)
52 of 135 wks
(BT1)
135 weeks
(males)
144 weeks
(females)
Dose
1 ppm
0.028 mg/kg/day
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10,000 ppm
0 ppm
1 ppm
5 ppm
10 ppm
25 ppm
50 ppm
100 ppm
150 ppm
200 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
0 mg/kg/day
1.7 mg/kg/day
5.0 mg/kg/day
14.1 mg/kg/dav
Ansiosarcomas
Male
0/80
1/30
9/30
6/30
6/29
2/30
1/26
0/108
0/48
0/43
0/42
1/41
0/26
0/37
1/36
7/42
1/28
0/22
6/26
3/17
0/55
0/58
6/56
27/59
Female
0/70
0/30
9/30
8/30
10/30
11/30
9/30
0/141
0/55
0/47
1/46
4/40
1/29
1/43
5/46
5/44
2/26
6/28
7/24
10/25
0/57
0/58
2/59
9/57
LADD (ni2/L liver)
Clewell et al.. 1995
Male
Female
1 75 H 03V
0.58(1.01)a
33.36
159.81
256.57
295.63
304.79
310.22
0.61
3.03
6.05
15.05
32.46
59.70
85.90
107.39
130.25
163.41
220.99
250.71
39.54
116.10
325.85
32.33
138.67
182.81
246.77
276.34
289.56
0.59
2.96
5.90
14.61
31.27
55.95
76.67
90.00
103.45
116.94
134.37
143.72
38.61
113.24
316.63
Reitz et al.,
1996
7.05
0.86
38.91
175.36
269.50
337.01
348.82
354.47
0.74
3.69
7.36
18.37
39.76
73.81
107.36
135.09
162.58
188.89
222.82
245.18
63.67
187.03
525.26
aBased on km value of 0.1 as recommended by the expert review panel.
bThe denominator for the incidence data is the total number of mice, as used by Chen and Blancato (1989).
The denominator is the number of rats alive when the first angiosarcoma was observed, as used by Chen and Blancato (1989). However, the male and female
incidence data shown here differ from that reported by Chen and Blancato (1989), after verification with the original study (Maltoni et al., 1984).
-------
Table A-4. Comparison of ED10s in animals based on angiosarcoma incidence
Study
Rats, inhalation
Maltonietal. (1981,
1984) (BT1, BT2, and
BT15)
Mice, inhalation
Maltonietal. (1981,
1984) (BT4)
Sex
M
F
M
F
Average inhalation
Rats, dietary
Feronetal. (1981)a
M
F
Average oral
Clewell et al. (1995a)
ED10
(mg metabolite/kg/day)
95% lower bound
112.24
53.19
112.07
51.94
82.36
94.93
182.02
138.48
MLE
157.14
74.35
153.16
65.52
112.54
132.32
241.33
186.82
Reitz et al. (1996)
ED10
(mg metabolite/kg/day)
95% lower bound
133.53
76.35
125.59
67.88
100.84
152.89
307.14
227.01
MLE
180.05
105.97
171.62
85.63
135.82
213.28
400.11
306.69
aAll risks from the Feron study shown here were calculated using a quanta! model, multistage option. The risks
presented in Appendix B were calculated using a time-to-tumor model, to account for increased deaths in the mid-
and high-dose groups.
Table A-5. Comparison of BMD10 values for rats (in units of dose metric)
based on liver necrosis
Study
Feronetal. (1981)
Sex
M
F
Average
Clewell et al. (1995a)
BMD10a
70.04
40.41
55.22
MLE"
139.25
54.75
97.00
Reitz et al. (1996)
BMD10
112.90
66.93
89.92
MLEC
224.48
90.68
157.58
aBMD10 is the benchmark dose at 10% extra risk based on dichotomous data.
bMLE=maximum likelihood estimate.
The parameter values for the Reitz et al. (1996) model are taken from Table 1 of that publication with the exception
of the blood/air partition coefficient in the mouse, which was incorrectly reported as 2.26. The value shown in
Table A-l is the value actually used in the risk calculations (R.H. Reitz, personal communication).
A-12
-------
A.3. REFERENCES
Baretta, ED; Stewart, RD; Mutchler, JE. (1969) Monitoring exposures to vinyl chloride vapor: breath analysis and
continuous air sampling. Am Ind Hyg Assoc J 30:537-544.
Chen, CW; Blancato, JN. (1989) Incorporation of biological information in cancer risk assessment: example-
vinyl chloride. Cell Biol Toxicol 5:417-444.
Clement International. (1990) Development and validation of methods for applying pharmacokinetic data in risk
assessment. Final report. Volume V: vinyl chloride. AAMRL-TR-90-072. Prepared for the Department of the Air
Force, Armstrong Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio.
Clewell, HJ; Gentry, PR; Gearhart, JM; et al. (1995a) The development and validation of a physiologically based
pharmacokinetic model for vinyl chloride and its application in a carcinogenic risk assessment for vinyl chloride.
ICF Kaiser report prepared for EPA/OHEA and OSHA/DHSP.
Clewell, HJ; Covington, TR; Crump, KS; et al. (1995b) The application of a physiologically based
pharmacokinetic model for vinyl chloride in a noncancer risk assessment. ICF/Clement report prepared for
EPA/NCAA under contract number 68-D2-0129.
Feron, HJ; Hendriksen, CFM; Seek, HJ; et al. (1981) Linesman oral toxicity study of vinyl chloride in rats. Food
Cosmet Toxicol 19:317-333.
Filser, JE; Bolt, HM. (1979) Pharmacokinetics of halogenated ethylene in rats. Arch Toxicol 42:123-136.
Fox, A; Collier, P. (1977) Mortality experience of workers exposed to vinyl chloride monomer in the manufacture
of polyvinyl chloride in Great Britain. Br J Ind Med 344:1-10.
Gargas, ML; Clewell, HJ, III; Andersen, ME. (1990) Gas uptake techniques and the rates of metabolism of
chloromethanes, chloromethanes, and chloroethylenes in the rat. Inhal Toxicol 2:295-319.
Gehring, PJ; Watanabe, PG; Park, CN. (1978) Resolution of dose-response toxicity data for chemicals requiring
metabolic activation: example—vinyl chloride. Toxicol Appl Pharmacol 44:581-591.
International Life Sciences Institute (ILSI) Risk Science Institute. (1994) Physiological parameter values forPBPK
models. Report prepared for the U.S. EPA Office of Health and Environmental Assessment.
Maltoni, C; Lefemine, G; Ciliberti, A; et al. (1981) Carcinogenicity bioassay of vinyl chloride monomer: a model
of risk assessment on an experimental basis. Environ Health Perspect 41:3-29.
Maltoni, C; Lefemine, G; Ciliberti, A; et al. (1984) Experimental research on vinyl chloride carcinogenesis.
Archives of Research on Industrial Carcinogenesis. Vol.2. Maltoni, C; Mehlman, MA, eds. Princeton, NJ:
Princeton Scientific Publishers.
Ramsey, JR; Andersen, ME. (1984) A physiologically based description of the inhalation pharmacokinetics of
styrene in rats and humans. Toxicol Appl Pharmacol 73:159-175.
Reitz, RH; Gargas, ML; Anderson, ME; et al. (1996) Predicting cancer risk from vinyl chloride exposure with a
physiologically based pharmacokinetic model. Toxicol Appl Pharmacol 137:253-267.
Watanabe, PG; McGowan, GR; Madrid, EO; et al. (1976) Fate of 14C-vinyl chloride following inhalation
exposure in rats. Toxicol Appl Pharmacol 37:49-59.
A-13
-------
APPENDIX B. THE DEVELOPMENT AND VALIDATION OF A PBPK MODEL
FOR VINYL CHLORIDE (VC) AND ITS APPLICATION
IN A CARCINOGENIC RISK ASSESSMENT
This appendix documents the development and documentation of the model used to
estimate VC cancer risk, as well as the results of the modeling. The risk estimates presented in
this appendix were calculated using the one-stage version of the LMS model, so the specific risk
estimates are slightly different from those calculated using the LMS or ED10/linear models.
Currently recommended risk values using the LMS and ED10/linear models are presented in the
main document. Other PBPK models developed for VC, and a comparison between the Clewell
model and another recent model (Reitz et al., 1996), are discussed in Appendix A.
The liver tumor data utilized for model development and presented in this section include
only angiosarcomas, in order to better compare results with other assessments using
angiosarcoma data both for rodents and across species. Cancer risk estimates were subsequently
revised to include all liver tumors. These revised estimates are the ones listed in the
Toxicological Review and the cancer summary. Table B-l summarizes the incidence of
angiosarcomas (in some cases only total angiosarcomas) reported in those chronic animal
bioassays in which a statistically significant increase was observed. For completeness, however,
two other tumor types observed at low concentrations in the rodent were also analyzed:
nephroblastoma and mammary gland adenocarcinoma. Table B-2 summarizes the incidence of
these tumors reported in chronic animal bioassays.
B.I. MECHANISM OF CARCINOGENICITY OF VC
As discussed in the main document, experimental evidence indicates that VC
carcinogenicity is due to a reactive metabolite, probably CEO. The reactive metabolite forms
DNA adducts, and a persistent DNA adduct is believed to lead to tumorigenesis.
The majority of the DNA adduct studies conducted with VC have been conducted on or
related to the parenchymal hepatocyte. However, although VC is primarily metabolized in the
hepatocyte (Ottenwalder and Bolt, 1980), the primary target cell for liver carcinogenicity is the
sinusoidal cell, as indicated by the incidence of liver angiosarcoma in both animals and humans.
Sinusoidal cells show a relatively low activity for transforming VC into reactive, alkylating
metabolites, roughly 12% of the activity of hepatocytes (Ottenwalder and Bolt, 1980). Therefore,
it has been suggested that the carcinogenic metabolites of VC may have to migrate from the
hepatocytes to produce tumors in the sinusoidal cells (Laib and Bolt, 1980). This possibility was
suggested by Laib and Bolt (1980) following their observation that alkylating metabolites of VC
were capable of diffusing through an artificial semipermeable membrane in a model in vitro
system. In studies conducted in vitro with rat hepatocytes by Guengerich et al. (1981), more than
90% of the hexane-insoluble metabolites were found to migrate out of the cell, with more than
B-l
-------
Table B-l. Summary of the angiosarcoma incidence data from vinyl chloride chronic animal bioassays
Reference
Leeetal., 1977, 1978
Feron et al., 1979a,b, Feron
andKroes, 1979
Hongetal, 1981
Drewetal., 1983
Route
Inhalation
Inhalation
Inhalation
Inhalation
Strain/species
Albino CD-I mice (M,F)
CD rats (M,F)
Wistar rats (M,F)
Albino CD-I mice (M,F)
CD rats (M,F)
Fischer 344 rats (F)
Golden Syrian hamsters
(F)
B6C3F1 mice (F)
CD-I mice (F)
Concentration/
dose
0,50,250, l,000ppm
0,250, l,000ppm
0, 5,000 ppm
0,50,250, 1,000 ppm
0,50,250, 1,000 ppm
0, 100 ppm
0, 200 ppm
50 ppm
0, 50 ppm
Incidence
Males - 0/26, 3/29, 7/29*, 13/33*
Females - 0/36, 0/34, 16/34*,
18/36*
Males - 0/35, 0/36, 2/36, 6/34*
Females - 0/35, 0/36, 10/34*,
15/36*
Males - 0/62, 6/62*
Females - 0/62, 16/62*
Males - 0/60, 1/40, 8/44*, 6/38*
Females - 1/60, 1/40, 5/40*,
12/38*
Males - 0/36, 0/30, 1/36, 5/36*
Females - 0/36, 0/36, 4/32*,
9/36*' a
1/1 12 (control), 4/76(0-6)*,
11/55(0-12)*, 13/55(0-18)*,
19/55 (0-24), 2/52 (6-12), 0/51
(12-18), 0/53 (18-24), 5/54 (6-
18)*, 2/49(12-24)
0/143 (control), 13/88 (0-6)*,
4/52 (0-12)*, 2/103 (0-18), 3/53
(6-12), 0/50 (12-18), 0/52 (18-
24), 1/44 (6-18), 0/43 (12-24)b
4/69 (control), 46/67 (0-6)*,
69/90 (0-12)*, 27/42 (6-12)*,
30/51 (12-18)*, 30/48(6-18)*,
29/48 (12-24)*c
1/71 (control), 29/67 (0-6)*,
30/47(0-12)*, 20/45(0-18)*,
11/49(6-12)*, 5/53(12-18),
17/46(6-18)*, 3/50 (12-24)c
Exposure duration
6 hours/day, 5 days/week, 12
months
6 hours/day, 5 days/week, 12
months
7 hours/day, 5 days/week, 12
months
6 hours/day, 5 days/week, and
sacrificed at 1, 3, or 6 months
6 hours/day, 5 days/week, and
sacrificed at 1, 3, 6 or 10
months
6 hours/day, 5 days/week, 6,
12, 18, or 24 months, or held
for 6 or 12 months and then
exposed for 6 or 12 months
6 hours/day, 5 days/week, 6,
12, 18, or 24 months, or held
for 6 or 12 months and then
exposed for 6 or 12 months
6 hours/day, 5 days/week, 6,
12, 18, or 24 months, or held
for 6 or 12 months and then
exposed for 6 or 12 months
6 hours/day, 5 days/week, 6,
12, 18, or 24 months, or held
for 6 or 12 months and then
exposed for 6 or 12 months
td
to
-------
Table B-l. Summary of the angiosarcoma incidence data from vinyl chloride chronic animal bioassays (continued)
Keplingeretal., 1975(8
month interim) MCA, 1980
(in U.S. EPA, 1985)
Bietal., 1985
Maltoni et al., 1981,1984
(BT1)
Maltoni et al., 1981,1984
(BT2)
Maltoni et al., 1981,1984
(BT9)
Maltoni et al., 1981,1984
(BT15)
Maltoni et al., 1981,1984
(BT10)
Maltoni et al., 1981,1984
(BT7)
Maltoni et al., 1981,1984
mT^
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
COBS Charles River rats
(M,F)
CDI Swiss Charles River
mice (M,F)
Syrian Golden hamsters
(M,F)
Wistar rats (M)
Sprague-Dawley rats
(M,F)
Sprague-Dawley rats
(M,F)
Sprague-Dawley rats
(M,F)
Sprague-Dawley rats
(M,F)
Sprague-Dawley rats
(M,F)
Wistar rats (M)
Swiss mice (M,F)
0, 50, 200, 2,500 ppm
0, 50, 200, 2,500 ppm
0, 50, 200, 2,500 ppm
0, 10, 100, 3,000 ppm
0,50,250,500,2,500,
6,000, 10,000 ppm
0,100, 150, 200 ppm
0, 50 ppm
0,1,5, 10, 25 ppm
0 (Group VII), 6,000
(Groups II, IV, VI),
10,000 (Groups I, III, V)
ppm
0,50,250,500,2,500,
6,000, 10,000 ppm
0,50,250,500,2,500,
f, nnn i n nnn ppm
0/143,28/139*, 82/141*,
114/147*
0/97,46/121*, 130/134*,
101/101*
0/83, 7/74*, 12/88*, 56/66*
0/19,0/20,7/19*, 17/20*
0/58,1/60,3/59,6/60,13/60,
13/59*, 7/60d
0/185, 1/120,6/119, 12/120*d
0/98, 14/294*
0/120,0/118,0/119,1/119,
5/120*d
1/118 (Group I), 0/120 (Group
II), 1/1 19 (Group III), 3/118*
(Group IV), 1/119 (Group V),
1/120 (Group VI), 0/227 (Group
VII)
0/38, 0/28, 1/27, 3/28, 3/25,
3/26, 8/27*
0/150, 1/60, 18/60*, 14/60*,
16/sq* imn* in/sfi*
7 hours/day, 5 days/week for
12 months
7 hours/day, 5 days/week for
9 months
7 hours/day, 5 days/week for
12 months
6 hours/day, 6 days/week for
1 8 months
4 hours/day, 5 days/week for
52 weeks (135 weeks)
4 hours/day, 5 days/week for
52 weeks (143 weeks)
4 hours/day, 5 days/week for
52 weeks (142 weeks)
4 hours/day, 5 days/week for
52 weeks (147 weeks)
Groups I and II - 4 hours/day,
5 days/week, 5 weeks
Groups III and IV - 1
hour/day, 4 days/week for 25
weeks
Groups V and VI - 4
hours/day, 1 day/week for 25
weeks (154 weeks)
4 hours/day, 5 days/week, for
52 weeks (165 weeks)
4 hours/day, 5 days/week for
"30 wppV« T81 wppV^
td
-------
Table B-l. Summary of the angiosarcoma incidence data from vinyl chloride chronic animal bioassays (continued)
Reference
Maltonietal., 1988
(BT4001,4006)
Grothetal., 1981
Radike et al., 1981
Feronetal, 1981
Maltonietal., 1981, 1984
(Bill)
Route
Inhalation
Inhalation
Inhalation
Oral - diet
Gavage
Strain/species
Sprague-Dawley rats
(Breeders - F; Embryos -
M,F)
Sprague-Dawley rats
(M,F)(ages6, 18, 32, and
52 weeks)
Sprague-Dawley rats (M)
Wistar rats (M,F)
Sprague-Dawley rats
(M,F)
Concentration/
dose
0, 2,500 ppm
0, 940 ppm
0, 600 ppm
0, 1.7,5.0,14.1
mg/kg/day
0,2.38,11.9,35.7
mg/kg/day
Incidence
Breeders - 0/60, 27/54
Embryos (M) - 0/158 (control),
24/60* (Group I), 36/64* (Group
II)
Embryos (F) - 0/149 (control),
28/60* (Group I), 46/63* (Group
II)
e6 weeks - males - 0/110, 1/83,
females -0/1 10, 2/88
18 weeks - males - 0/119, 2/91,
females - 0/120, 7/97*
32 weeks - males - 1/115, 7/94*,
females - 0/120, 27/98*
52 weeks - males - 0/128,
18/102*,females -0/127, 14/104*
0/80, 18/80*
Males - 0/55, 0/58, 6/56*,
27/59*
Females- 0/57, 0/58, 2/59, 9/57*
0/80, 0/80, 10/80*, 17/80*
Exposure duration
Breeders - 4 hours/day, 5
days/week for 7 weeks and
then 7 hours/day, 5 days/week
for 69 weeks
Embryos - 4 hours/day, 5
days/week for 7 weeks and
then 7 hours/day 5 days/week
for 8 (Group I) or 69 weeks
(Group II)
7 hours/day, 5 days/week for
24 weeks
4 hours/day, 5 days/week for
52 weeks
4 hours/day for 135 or 144
weeks
4 to 5 days/week for 52 weeks
(136 weeks)
td
Significantly different from control at/>=0.05.
incidence for both males and females includes only those animals sacrificed at 6 and 10 months. The incidence data for those animals sacrificed at 1 and 3 months
was not reported.
Incidence reported for hemangiosarcomas at all sites only. The authors reported that these tumors occurred primarily in the skin, spleen, and liver.
cHemangiosarcomas for all sites reported.
dThe denominator shown is the total number of animals examined. However, the denominator used for risk calculations was the number alive when the first
angiosarcoma was observed, as shown in Table C-5.
eReported total angiosarcomas.
-------
Table B-2. Summary of incidence data on other low-dose tumors from vinyl chloride chronic animal bioassays
Reference
Lee et al,
1977, 1978
Drewet al.,
1983
Radike et al.,
1981
Maltoni et al.,
1981, 1984
(BT1)
Maltoni et al.,
1981, 1984
(BT2)
Route
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Species
Albino CD-I
mice (F)
Fischer-344
rats (F)
Fischer-344
rats (F)
Golden
Syrian
hamsters (F)
B6C3F1 mice
(F)
CD-I Swiss
mice (F)
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(M,F)
Sprague-
Dawley rats
(M,F)
Endpoint
Mammary gland tumors
Mammary gland:
fibroadenoma and
adenocarcinoma
Hepatocellular carcinoma
Mammary gland
carcinoma
Mammary gland
carcinoma
Mammary gland
carcinoma
Hepatocellular carcinoma
Nephroblastoma
Mammary malignant
tumor
Nephroblastoma
Mammary malignant
tumor
Dose
0,50,250, 1,000
ppm
0, 100 ppm
0, 100 ppm
0, 200 ppm
0, 50 ppm
0, 50 ppm
0, 600 ppm
0, 50, 250, 500,
2,500, 6,000,
10,000 ppm
0, 50, 250, 500,
2,500, 6,000,
10,000 ppm
0, 100, 150,200
ppm
0, 100, 150,200
ppm
Incidence3
0/36,9/34,3/34, 13/36
Fibroadenoma:
24/112,26/55(0-24)
Adenocarcinoma:
5/112,5/55(0-24)
Females: 1/112,9/55(0-24)
Females: 0/143, 47/102 (0-18)
Females: 3/69, 37/90 (0-12)
Females: 2/7 1,22/45 (0-1 8)
Males: 1/80, 35/80
(0-11.5)
M & F: 0/58, 1/60, 5/59, 6/60,
6/60, 5/59, 5/60
M & F: 0/58, 2/60, 2/59, 1/60,
2/60, 0/59, 3/60
M&F:0/185, 10/120, 11/119,
7/120
M&F: 2/128, 4/120, 6/119,
6/120
Exposure duration
6hr/d, 5d/wk, 12 mo
6 hr/d, 5 d/wk, 6, 12,
18 or 24 mo, or held
for 6 or 1 2 mo and
then exposed for 6 or
12 mo
4 hr/d, 5 d/wk, 52
wks
4 hr/d, 5 d/wk for 52
wk(held!35wk)
4 hr/d, 5 d/wk for 52
wk(held!43wk)
td
-------
Table B-2. Summary of incidence data on other low-dose tumors from vinyl chloride chronic animal bioassays
(continued)
Reference
Maltoni et al,
1981, 1984
(BT4)
Maltoni et al.,
1981, 1984
(BT3)
Maltoni et al.,
1981, 1984
(BT9)
Maltoni et al.,
1981, 1984
(BT15)
Maltoni et al.,
1981, 1984
(BT10)
Route
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Species
Swiss mice
(M,F)
Sprague-
Dawley rats
(M,F)
Sprague-
Dawley rats
(M,F)
Sprague-
Dawley rats
(M,F)
Sprague-
Dawley rats
(M,F)
Endpoint
Mammary carcinoma
Nephroblastoma
Mammary malignant
tumor
Nephroblastoma
Mammary malignant
tumor
Nephroblastoma
Mammary malignant
tumor
Nephroblastoma
Mammary malignant
tumor
Dose
0, 50, 250, 500,
2,500, 6,000,
1 0,000 ppm
0, 50, 250, 500,
2,500, 6,000,
10,000 ppm
0, 50, 250, 500,
2,500, 6,000,
10,000 ppm
0, 50 ppm
0, 50 ppm
0, 1,5, 10,25
ppm
0, 1,5, 10,25
ppm
Gp VII: control;
Gps I, III, V:
10,000 ppm;
Gps II, IV, VI:
6,000 ppm
Gp VII: control;
Gps I, III, V:
10,000 ppm;
Gps II, IV, VI:
6,000 ppm
Incidence3
1/150,12/60,12/60,8/60,
8/59,8/60,13/56
M&F: 0/1 90, 3/58,6/59,0/60,
2/60, 1/60, 1/58
M&F: 5/1 90, 1/58, 1/59, 3/60,
4/60, 1/60, 1/58
M&F: 0/98, 1/294
M&F: 10/98,62/294
M&F: 0/1 20, 0/118, 0/119,
0/119, 1/120
M&F: 7/120, 15/118,22/119,
21/119, 17/120
Gp VII: 0/227; Gps I, III, V:
0/118, 0/119, 0/1 19; Gps II,
IV, VI: 1/120,0/118, 1/120
Gp VII: 1 7/227; Gps I, III, V:
13/118, 16/119, 20/1 19; Gps
II, IV, VI: 13/120, 11/118,
12/120
Exposure duration
4hr/d, 5d/wkfor30
wk (held 81 wk)
4hr/d, 5d/wk, 17wk
4hr/d, 5d/wkfor52
wk(held!42wk)
4 hr/d, 5 d/wk for 52
wk(held!47wk)
Gps I & II: 4 hr/d, 5
d/wk, 5 wks; Gps III
& IV: 1 hr/d, 4 d/wk,
25 wks; Gps V & VI:
4 hr/d, 1 d/wk, 25
wks (held 1 54 wks)
td
-------
Table B-2. Summary of incidence data on other low-dose tumors from vinyl chloride chronic animal bioassays
(continued)
Reference
Feronet al,
1981
Til etal., 1983
Route
Oral - diet
Oral - diet
Species
Wistar rats
(M,F)
Wistar rats
(M,F)
Endpoint
Hepatocellular carcinoma
Neoplastic nodules
Combined incidence of
angiosarcomas,
hepatocellular carcinoma,
and neoplastic nodules
Hepatocellular carcinoma
Neoplastic nodules
Combined incidence of
angiosarcomas,
hepatocellular carcinoma,
and neoplastic nodules
Mammary gland tumors
Dose
0,1.7,5.0,14.1
mg/kg body
weight/day
0,1.7,5.0,14.1
mg/kg body
weight/day
0,1.7,5.0,14.1
mg/kg body
weight/day
0,0.017,0.17,
1.7 mg/kg body
weight/day
0,0.017,0.17,
1.7 mg/kg body
weight/day
0,0.017,0.17,
1.7 mg/kg body
weight/day
0,0.017,0.17,
1.7 mg/kg body
weight/day
Incidence3
Males: 0/55, 1/58, 2/56, 8/59
Females: 0/57, 4/58, 19/59,
29/57
Males: 0/55, 1/58, 7/56, 23/59
Females: 2/57, 26/58, 39/59,
44/57
Males: 0/55,2/58,11/56,
41/59
Females: 2/57, 28/58, 49/59,
56/57
Males: 0/99, 0/99, 0/99, 3/49
Females: 1/98,0/100,1/96,
3/49
Males: 0/99, 0/99, 0/99, 3/49
Females: 0/98, 1/100, 1/96,
10/49
Males: 0/99, 0/99, 0/99, 5/49
Females: 1/98, 1/100, 1/96,
11/49
Males: 5/99, 8/99, 3/99, 0/49
Females: 41/98, 21/100, 28/96,
21/48
Exposure duration
135orl44wks
149wks
td
aTumor incidence provided for longest duration of exposure only.
-------
70% of the total irreversibly bound species found outside the cell. These results were interpreted
to indicate that the majority of the reactive metabolites can leave the intact hepatocyte. On the
other hand, sinusoidal cells do possess the ability to produce reactive metabolites from VC,
albeit at a slower rate than the hepatocyte (Ottenwalder and Bolt, 1980). In either case, the
greater susceptibility of the sinusoidal cells to the carcinogenic effects of VC may result from an
inability of the sinusoidal cells to repair one or more of the DNA adducts produced by VC as
efficiently as the hepatocytes. Furthermore, the same dose metric (e.g., total amount of VC
metabolism divided by the volume of the liver) is applicable whether the carcinogenic
metabolites are produced in the hepatocyte or the sinusoidal cell.
B.2. SELECTION OF RISK ASSESSMENT APPROACH
Based on the information above on the pharmacokinetics, metabolism, and mechanism of
carcinogenicity of VC, it is necessary to determine the appropriate approach for conducting a
human risk assessment. Clearly, the evidence is strong that the carcinogenicity of VC is related
to the production of reactive metabolic intermediates. The most appropriate pharmacokinetic
dose metric for a reactive metabolite is the total amount of the metabolite generated divided by
the volume of the tissue into which it is produced (Andersen et al, 1987a). In the case of VC,
reasonable dose metrics for angiosarcoma would include the total amount of metabolism divided
by the volume of the liver (RISK), or the total amount of metabolism not detoxified by reaction
with glutathione, again divided by the volume of the liver (RISKM). A third, less likely
possibility, that the GSH conjugate of VC is subsequently metabolized to a reactive species that
is responsible for the carcinogenicity, can also be considered by using a dose metric based on the
total amount of reaction with GSH divided by the volume of the liver (RISKG). The assumption
underlying the use of these dose metrics is that the concentration of the actual carcinogenic
moiety, or the extent of the crucial event associated with the cellular transformation, is linearly
related to this pseudoconcentration of reactive intermediates, and that the relationship of the
actual carcinogenic moiety or crucial event to the dose metric is constant across concentration
and species. Specifically, the average amount generated in a single day is used, averaged over
the lifetime (i.e., the lifetime average daily dose, or LADD). The use of a dose rate, such as the
LADD, rather than total lifetime dose, has been found empirically to provide a better cross-
species extrapolation of chemical carcinogenic potency (U.S. EPA, 1992).
Subsequent steps in the carcinogenic mechanism related to specific adduct formation,
detection, and repair, as well as to the consequences of DNA mistranscription/misreplication and
the potential impact of increased cell proliferation, have been only sketchily outlined and have
not yet reached the point where they could be incorporated into a risk assessment or model in
any quantitative form. However, there appears to be sufficient evidence to justify the
assumption that VC acts as a classic initiator, producing genetic transformations through direct
reaction of its metabolites with DNA. Therefore, the traditional assumption of low-dose
linearity of risk appears to be warranted, and the linearized multistage (LMS) model would seem
to be the most appropriate for low-dose extrapolation.
B-8
-------
B.3. DESCRIPTION OF PBPK MODEL FOR VC
B.3.1. General - Model Outputs (Dose Metrics) and Conversion to Human Values
The PBPK model for VC developed in this study was shown in Figure A-l. As
mentioned earlier, the model is basically an adaptation of a previously developed PBPK model
for vinylidene chloride (D'Souza and Andersen, 1988). For a poorly soluble, volatile chemical
like VC, only four tissue compartments are required: a richly perfused tissue compartment that
includes all of the organs except the liver, a slowly perfused tissue compartment that includes all
of the muscle and skin tissue, a fat compartment that includes all of the fatty tissues, and a liver
compartment. All metabolism is assumed to occur in the liver, which is a good assumption in
terms of the overall kinetics of VC, but the assumption would have to be revised to include
target-tissue-specific metabolism if a serious attempt were to be made to perform a VC risk
assessment for a tissue other than the liver (Andersen et al, 1987a). The model also assumes
flow-limited kinetics, or venous equilibration; that is, that the transport of VC between blood and
tissues is fast enough for steady state to be reached within the time it is transported through the
tissues in the blood.
Metabolism of VC is modeled by two saturable pathways, one high affinity, low capacity
(with parameters VMAX1C and KM1) and one low affinity, high capacity (with parameters
VMAX2C and KM2). Subsequent metabolism is based on the metabolic scheme shown in
Figure 1 of the main text of the Toxicological Review. The reactive metabolites (whether CEO,
CAA, or other intermediates) may then either be metabolized further, leading to CO2; react with
GSH; or react with other cellular materials, including DNA. Because exposure to VC has been
shown to deplete circulating levels of GSH, a simple description of GSH kinetics was also
included in the model.
The model is designed for input from inhalation (using inhaled concentration), gavage
(using a first-order rate constant for uptake from corn oil), and drinking water/diet (using a zero-
order rate constant for uptake), although the data available to support these routes (shown below)
vary considerably. Various dose rate scenarios can be accommodated for inhalation (e.g.,
number of hours exposed/day and number of days/week) and for water/diet (e.g., mg/kg
absorbed over a set number of hours). Continuous exposure scenarios can also be simulated. As
discussed above, the most logical output from the model upon which to base this assessment is
the total amount of VC metabolized in the liver divided by the volume of the liver, designated as
"RISK" in the model. The other dose metrics mentioned above, RISKM and RISKG, were
considered but were not used in this assessment. The direct output from the model is the daily
average dose for either the RISK dose metric or for the total amount of VC metabolized/body
weight (designated "AMET"). Lifetime average delivered doses (LADDs) were calculated by
factoring the daily average dose (the actual model output) both by the fraction of the week
exposed (e.g., 5/7 days) and by the fraction of the lifespan the exposure period spanned (e.g.,
52/147 weeks).
B-9
-------
As discussed in the main document, the risk modeling was conducted using the animal
target tissue dose, i.e., the dose metric RISK. The calculated risk values based on the animal
dose metric were assumed to correspond to those from the same human dose metric. The human
dose metric was then converted to a human dose as described in the main document. The
following equations were then used to calculate the risk in the units of mg vinyl chloride
ingested/kg body weight/day (oral) or • g vinyl chloride/m3 (inhalation):
Administered dose slope factor (oral, LED 10 method) = 0.1 ^ tissue dose LED 10 (mg
metabolite/kg tissue/day) x 1.01 [(mg metabolite/kg tissue/day)/(mg/L vinyl chloride in
drinking water)] ^ 2 L water ingested/day x 70 kg
where:
Tissue dose LED 10 is the lower bound on the ED 10, in units of (mg metabolite/kg
tissue/day) and is derived from the TOXRISK output;
0.1 represents the 10% response that is divided by the calculated LED 10 to get the slope
at the LED 10;
1.01= Conversion factor for the dose of metabolites to the human liver from a sample
human continuous oral exposure (1 mg/L in drinking water);
70 kg = Human default body weight;
2 L/d = Default for daily drinking water ingestion.
Using the linearized multistage model, the conversion is as follows:
Administered dose slope factor (LMS) = Target tissue slope factor (mg metabolite/kg
tissue/day)"1 x 1.01 [(mg metabolite/kg tissue/day)/(mg/L vinyl chloride in drinking
water)] ^ 2 L water ingested/day x 70 kg
where the constants in the conversion are as described above.
To calculate the inhalation unit risk using the LED 10 method, the conversion is as follows:
Inhalation unit risk (LED 10 method) = 0.1^- tissue dose LED 10 (mg metabolite/kg
tissue/day) x 3.03 [(mg metabolite/kg tissue/day)/(ppm vinyl chloride)] x 0.039
(ppm/mg/m3) x 10"3 (• g/m3)/(mg/m3)
where:
3.03 = Conversion factor for the dose of metabolites to the human liver from a sample
human continuous inhalation exposure (1 ppm in air).
B-10
-------
B.4. PARAMETERIZATION AND VALIDATION
The parameters for the model are listed in Tables B-3 and B-4. The physiological
parameters are the current EPA reference values (U.S. EPA, 1988), except for alveolar
ventilation in the human, which was calculated from the standard EPA value for the ventilation
rate in the human, 20 m3/day, assuming a 33% pulmonary dead space. The partition coefficients
for Fischer-344 (F344) rats were taken from Gargas et al. (1989), and those for Sprague-Dawley
rats were taken from Barton et al. (1995). The Sprague-Dawley values were also used for
modeling of Wistar rats. Blood/air partition coefficients for the other species were obtained
from Gargas et al. (1989), and the corresponding tissue/blood partition coefficients were
estimated by dividing the Sprague-Dawley rat tissue/air partition coefficients by the appropriate
blood/air value.
The affinity for the 2E1 pathway (KM1) in the rat, mouse, and hamster was set to 0.1 on
the basis of studies of the competitive interactions between CYP2E1 substrates in the rat (Barton
et al., 1995; Andersen et al., 1987b). The affinity used for the non-2El pathway (KM2) in the
mouse and rat was set during the iterative fitting of the rat total metabolism, glutathione
depletion, and rate of metabolism data, described below. The capacity parameters for the two
oxidative pathways (VMAX1C and VMAX2C) in the mouse, rat, and hamster were estimated by
fitting the model to data from closed-chamber exposures with each of the species and strains of
interest (Barton et al., 1995; Bolt et al., 1977; Clement, 1990; Gargas et al., 1990). After the
other parameters were scaled from animal weights obtained from individual studies, the model
was exercised for optimization to a single pair of values, VMAX1C and VMAX2C, to be used
for all of the data on a given sex/strain/species.
Initial estimates for the subsequent metabolism of the reactive metabolites and for the
glutathione submodel in the rat were taken from the model for vinylidene chloride (D'Souza and
Andersen, 1988). These parameter estimates, along with the estimates for VMAX2C and KM2,
were then refined for the case of VC in the Sprague-Dawley rat using an iterative fitting process
that included the closed-chamber data for the Sprague-Dawley and Wistar rat (Barton et al.,
1995; Bolt et al., 1977; Clement, 1990) along with data on glutathione depletion (Jedrychowski
et al., 1985; Watanabe et al., 1976d), and total metabolism (Gehring et al., 1978). The
parameters obtained for the rat were used for the other species with appropriate allometric
scaling (e.g., body weight to the -1/4 for the first order rate constants).
Figures B-la through B-ld show the results of this interactive fitting process for mice
and Figures B-2a through B-2g present the results for several strains of rats, with Figure B-2h
demonstrating the fit to hamster data. Figures B-3a through B-3c demonstrate the capability of
the model to simulate depletion of internal GSH (measured as cytoplasmic nonprotein sulfhydryl
concentration) as a function of external air exposure to various concentrations of VC and as a
function of time after inhalation exposure to VC (Jedrychowski et al., 1985). Figure B-4 shows
data and simulation results from modeling total metabolism (the amount of radiolabeled VC
B-ll
-------
Table B-3. Model parameters and their coefficients of variation for the Vinyl
Chloride Model
Unsealed parameters
BW
QPC
QCC
Body weight (kg)
Alveolar ventilation (L/hr, 1 kg animal)
Cardiac output (L/hr, 1 kg animal)
Mouse (CV-%)a
— b an
30.0(58)
18.0(9)
Rat (CV-%)
— an
21.0(58)
18.0(9)
Human (CV-%)
70.0 (30)
24.0 (16)
16.5(9)
Tissue blood flows (fraction of cardiac output):
QRC
QFC
QSC
QLC
Flow to rapidly perfused tissues
Flow to fat
Flow to slowly perfused tissues
Flow to liver
0.51 (50)
0.09 (60)
0.15(40)
0.25 (96)
0.51 (50)
0.09 (60)
0.15 (40)
0.25 (96)
0.5 (20)
0.05 (30)
0.19(15)
0.26 (35)
Tissue volumes (fraction of body weight):
vsc
we
VRC
VLC
Volume of slowly perfused tissues
Volume of fat
Volume of richly perfused tissues
Volume of liver
0.77 (30)
— (30)
0.035 (30)
0.055(6)
0.75 (30)
— (30)
0.05 (30)
0.04 ( 6)
0.63 (30)
0.19(30)
0.064 (10)
0.026 ( 5)
Partition coefficients:
PB
PF
PS
PR
PL
Blood/air
Fat/blood
Slowly perfused tissue/blood
Richly perfused tissue/blood
Liver/blood
2.26(15)
10.62(30)
0.42 (20)
0.74 (20)
0.74 (20)
2.4(15)
10.0 (30)
0.4 (20)
0.7 (20)
0.7 (20)
1.16(10)
20.7 (30)
0.83 (20)
1.45(20)
1.45(20)
Metabolic parameters:
VMAX1C
KM1
VMAX2C
Maximum velocity of first saturable pathway
(mg/hr, 1 kg animal)
Affinity of first saturable pathway (mg/L)
Maximum velocity of second saturable
pathway (mg/hr, 1 kg animal)
-(20)
0.1(30)
-(20)
-(20)
0.1(30)
- 00)
4.0 (30)
0.1 (50)
0.1(0)
B-12
-------
Table B-3. Model parameters and their coefficients of variation for the Vinyl
Chloride Model (continued)
KM2
Affinity of second saturable pathway (mg/L)
10.0(30)
10.0(30)
10.0 (50)
GSH parameters:
KC02C
KGSMC
KFEEC
GSO
KBC
KS
KOC
First order CEO breakdown to CO2
Conjugated rate constant with metabolite
Conjugated rate constant with non-GSH
Initial GSH concentration
First order rate constant for GSH breakdown
Constant controlling resynthesis
Zero order production of GSH
1.6(20)
0.13(20)
35.0 (20)
5,800.0 (20)
0.12(20)
2,000.0 (20)
28.5 (20)
1.6(20)
0.13(20)
35.0 (20)
5,800.0 (20)
0.12(20)
2,000.0 (20)
28.5 (20)
1.6(20)
0.13 (20)
35.0 (20)
5,800.0 (20)
0.12(20)
2,000.0 (20)
28.5 (20)
Dosing parameters:
KA
Oral uptake rate (/hr)
3.0 (50)
3.0 (50)
3.0 (50)
aCV-%: Coefficient of variation = (Standard deviation/mean) x 100
bSee Table B-4.
Table B-4. Strain/study-specific parameter values
Swiss albino mice
(inhalation study)
Sprague-Dawley rats
(inhalation study)
Sprague-Dawley rats
(gavage study)
Wistar rats
(drinking water study)
Male
Female
Male - low dose
Male - high dose
Female - low dose
Female - high dose
Male - low dose
Male - high dose
Female - low dose
Female - high dose
Male
Female
BW
0.044
0.040
0.638
0.433
0.485
0.321
0.632
0.405
0.445
0.301
0.436
0.245
VFC
0.13
0.12
0.19
0.13
0.200
0.14
0.19
0.12
0.18
0.13
0.14
0.11
VMAX1C
8.0
5.0
4.0
4.0
3.0
3.0
4.0
4.0
3.0
3.0
4.0
3.0
VMAX2C
O.la
3.0
2.0
2.0
O.la
O.la
2.0
2.0
O.la
O.la
2.0
O.la
B-13
-------
Table B-4. Strain/study-specific parameter values (continued)
Scaled Parameters
QP = QPC*BW°75
QC = QCC*BW°75
QR = QRC*QC
QF = QFC*QC
QS = QSC*QC
QL = QLC*QC
QC = QL + QF + QS + QR
Note: Since all of the input parameters are subject to modification by the Monte Carlo analysis, it is necessary to
recompute the total blood flow in order to maintain mass balance (where QCC, QLC, QFC, QSC, and QRC are
subject to modification).
VS = VSC*BW
VF = VFC*BW
VR = VRC*BW
VL = VLC*BW
VMAX1 = VMAX1C*BW075
VMAX1M = VMAX1C*BW075*1000.0/MW
VMAX2 = VMAX2C*BW°75
VMAX2M = VMAX2C*BW075*1000.0/MW
KCO2 = KCO2C/BW0 25
KGSM = KGSMC/BW0 25
KFEE = KFEEC/BW0 25
GSO= VLC*BW*GSO
KB = KBC/BW0 25
KO = KOC*BW°75
Principal Dose Surrogate
RISK = (Total amount metabolized)/VL
Other Dose Surrogates
RISK1 = (Total amount metabolized by pathway 1)/VL
RISKG = (Total amount reacted with glutathione)/VL
RISKM = (Total amount binding to cellular materials)/VL
RISKT = Lifetime Average Daily Dose based on RISK
RISKN = Lifetime Average Daily Dose based on RISKM
RISKR = Lifetime Average Daily Dose based on RISKG
RISKT1 = Lifetime Average Daily Dose based on RISK1
aThe value of this parameter was normally set to zero. It was only set to 0.1 for the PBPK_SIM runs. The variance
for this parameter was set to zero in the PBPK_SDVI runs.
B-14
-------
MflLE B6C3F1 MICE — UPAF5
0.0
6.D
(a)
FEMflLE B6C3FI MtCE — UPftFB
Q.o
T 'T '
T - Time (hrs)
4.a
6.0
(b)
MOLE CD-: MICE ~ UPflFB
FEHRLE CC-1 MICE " UPAFB
(C)
1.2
2.4 3.6
T - Time (hrs)
4.8
6.0
0.0
2.4 3.6
T - Time (hrs)
(d)
Figure B-l. Model predictions (lines) and experimental data (symbols) for the chamber
concentration during exposure of mice or hamsters to VC in a closed, recirculated chamber
(Clement, 1990): (a) male B6C3F1 mice; (b) female B6C3F1 mice; (c) male CD-I mice; (d)
B-15
-------
MALE r344 RATS -- GARGAS
MALE F344 RATS —
2.4 3.6
T - Time (hrs)
Q.Q
(a)
(b)
2.4 3.6
T - T ime (hr si
4.a
6.0
FEMALE F344 SATS -- JPAFB
MALE 50 RftTS — BARTON
0.0
2.4 _ 3.6
T ~ Time (hrs)
(C)
(d)
Figure B-2. Model predictions (lines) and experimental data (symbols) for the chamber
concentration during exposure of rats to VC in a closed, recirculated chamber: (a) male
F344 rats (Gargas et al., 1990); (b) male F344 rats (Clement, 1990); (c) female F344 rats
(Clement, 1990); (d) male Sprague-Dawley rats (Barton et al., 1995).
B-16
-------
MflLE UISTAR RflTS — BOLT
LJISTAR RATS — UIPflFE
(e)
1.2
2.4 3.6
T - T{me (hr s)
1.8
6.0
Q.O
(f)
2.4 3.6
T - T i me (hr s)
6.0
FEHflLE UlSTflR RfiTS — WPAFB
0.0
(g)
2.4 3.6
T - Time (hrs)
6.Q
Figure B-2 (continued): (e) male Wistar rats (Bolt et al., 1977); (f) male Wistar rats
(Clement, 1990); (g) female Wistar rats (Clement, 1990).
B-17
-------
HflLE GOLDEN SYRIflN HflHSTERS — UPAFB
(h)
2.4 3.6
T - T i me (hr s)
1.8 S.Q
Figure B-2 (continued): (h) male Golden Syrian hamsters (Clement, 1990).
B-18
-------
g '-ISLE UISTflR RflTS — JEDRYCHOU5KI
10
20 _ 30
T - Time (hps)
_l
"so
(a)
MflLE WiSTflR RftTS — JEDRYCHOU5K[
r
a.
iQ° 10' Ifl2 103
XCONC - Concentration (PPM)
(b)
g HflLE SD RPlTS —
(c)
XCONC
to1 io2 ioJ
Concentration (FFM)
Figure B-3. Model-predicted (lines) and experimentally determined (symbols): (a) GSH
concentrations (% controls) after 4-hr inhalation exposures to VC at concentrations of (top
to bottom) 15, 50, 150, 500, and 15,000 mg/m3 (Jedrychowski et al., 1985); (b) glutathione
concentrations (% control animal levels) immediately following 4-hr inhalation exposures
to VC (Jedrychowski et al., 1985); (c) GSH concentrations (% controls) immediately
following 6-hr inhalation exposures to VC (Watanabe et al., 1976a).
B-19
-------
"flLESD RATS — GEHRING
10° Id1 102 103
XCONC - Concentration 'PPM)
Figure B-4. Model-predicted (lines) and experimentally determined (symbols) total
amount metabolized during 6-hr inhalation exposures to VC (Gehring et al., 1978).
B-20
-------
remaining in rat carcasses after a 6-hr air exposure to VC) (Gehring et al., 1978). No systematic
errors could be surmised from these results, indicating that the kinetic parameters optimized by
the model were valid for the species/strain/sex over a wide range of external air concentrations.
Parameterization of the P450 metabolism pathways in the human was accomplished as
follows: There is no evidence of high-capacity, low-affinity P450 metabolism for chlorinated
ethylenes in the human; therefore, VMAX2C in the human was set to zero. The ratio of
VMAX1C to KM1 could be estimated by fitting the model to data from closed-chamber studies
with human subjects (Buchter et al., 1978) in a manner entirely analogous to the method used for
the animal closed-chamber analysis. The result of this process is shown in Figures B-6a and B-
6b. The precision and sensitivity of the estimate of VMAX1C/KM1 can be evaluated by a
comparison of the several model runs shown in these figures, as each simulation was based on a
separate designation of VMAX1C. It can be seen that the estimate of VMAX1C/KM1 in each
subject can be determined to within about 30% to 50%, but that the ratio varies between the two
subjects, as represented by the two lines on Figure B-6b. This variability of CYP2E1 activity in
the human is not surprising; several studies have demonstrated a variability of human CYP2E1
activity of roughly an order of magnitude (Reitz et al., 1989; Sabadie et al., 1980). This wide
variability is not observed in the inbred strains typically used in animal studies; for example, the
coefficient of variation (standard deviation divided by the mean) for CYP2E1 activity in rats in
one of these same studies was only 14% (Sabadie et al., 1980). This wide variability in human
CYP2E1 activity is an important consideration for estimating the potential difference between
average population risk and individual risk in a human cancer risk assessment for materials like
VC, whose carcinogenicity depends on metabolic activation.
In order to obtain separate estimates of VMAX1C and KM1 in the human, higher
exposure concentration closer to metabolic saturation would be required. Fortunately, cross-
species scaling of CYP2E1 between rodents and humans appears to follow allometric
expectations for metabolism very closely; that is, the metabolic capacity scales approximately
according to body weight raised to the 3/4 power (Andersen et al., 1987a). Support for the
application of this principle to VC can be obtained from data on the metabolism of VC in
nonhuman primates (Buchter et al., 1980). On the basis of data for the dose-dependent
metabolic elimination of VC in the rhesus monkey, the maximum capacity for metabolism can
be estimated to be about 50 • mol/hr/kg. This equates to a VMAX1C (the allometrically scaled
constant used in the model) of approximately 4 mg/hr for a 1 kg animal, which is in the same
range as those estimated for rodents from the closed chamber exposure data. The similarity of
VMAX1C in humans and rats is also supported by an in vitro study that found the activity of
human microsomes to be 84% of the activity of rat microsomes. Based on these comparisons,
the human VMAX1C was set to the primate value and KM1 was calculated using this value of
VMAX1C and the ratio of VMAX1C/KM1 obtained from the closed chamber analysis. The
ability of the resulting human model to reproduce inhalation exposure data (Buchter et al., 1978;
Baretta et al., 1969) is shown in Figures B-6c, B-6d, and B-7. Note that the reproduction of
parent chemical concentrations for a constant concentration inhalation exposure is not a
particularly useful test of the accuracy of the metabolism parameters in a PBPK model of a
volatile compound. The results of Figure B-7, in which three conditions of metabolism were run
for each concentration (none, optimized value and twice the optimized value for VMAX1C),
B-21
-------
indicate that the discrepancies or agreement between the model and the data are due primarily to
details of the physiological description of the individual, such as fat content, ventilation rate,
blood/air partition, etc., rather than rate of metabolism.
In reviewing the selection of parameters by the expert panel, there was agreement that the
values adopted were suitable, except for the KM1 of 1.0 for humans. It was noted that the
derivation of this Km value was based upon a relative insensitive fitting exercise, in which the
highest value is generally chosen. The highest value, however, does not necessarily reflect the
actual rate of substrate oxidation at the enzyme active site. As noted by Kedderis et al. (1993) in
vitro data with human enzymes do not indicate significant species differences in the kinetic
parameters for CYP2E1 substrates. Therefore, the same KM1 value of the rodents, 0.1, was
recommended and adopted for use as the human parameter. This metabolic value would indicate
the same rapid affinity for VC metabolism in humans as in rats.
Figures B-l through B-4, B-6, and B-7 provide a basis to favorably evaluate the
capability of the inhalation portion of the model and its parameters to reproduce and predict
results from experimental inhalation data. There are, however, limited data to judge the
capability and performance of the oral portion of the model. Figures B-8a, b, and c are data and
model simulations of blood levels of VC after gavage administration of VC at the doses
indicated. Modeled simulations provide poor fits to these depuration data, which are themselves
problematic. A similarly poor data fit was observed with expiration of carbon dioxide in rats
following oral dosing with VC (Figure B-5). There are no experimental data from drinking
water or dietary studies to judge the performance of the oral portion of the model, although they
would be expected to provide a better fit. In the case of gavage dosing, rapid uptake of large
doses will result in a significant percentage of the VC being exhaled unmetabolized.
The significance to the overall assessment of having experimental data to judge
capabilities of a PBPK model relates directly to the confidence in model output, i.e., the dose
metrics. Based on the existing experimental data, a much higher confidence would be placed in
dose metrics derived from inhalation studies than for those derived from oral studies. Strategic
programming within the PBPK model can, however, offset this lack of confidence. This would
be done by maximizing the potential of an oral dose for expressing toxicity, i.e., maximizing the
conversion of the parent dose to the reactive species. This has been accomplished in the oral
dose inputs by designating VC uptake from the dietary/drinking water route as zero-order (i.e.,
independent of concentration) and occurring over a 24-hr period. Thus, for oral inputs the model
calculates total VC uptake spread out over a period where the concentrations would not exceed
the capacity of the metabolic processes to work at maximum efficiency (i.e., where Vmax/Km
are linear). These designations would produce the maximum value of the dose metric (mg
metabolite/L liver) and may be viewed as being conservative or "worst case" with respect to
what may actually occur during an oral dose. Coupled with the use of the same hepatic
metabolic processes for both inhalation and oral inputs, this strategy is considered to increase the
confidence in dose metrics derived from oral inputs.
B-22
-------
B.5. COMPARISON OF RISK ASSESSMENTS FOR VC INHALATION
The model just described was used to calculate the pharmacokinetic dose metrics for
angiosarcoma in the most informative of the animal bioassays (Maltoni et al., 1981, 1984; Feron
et al., 1981), as well as for human inhalation exposure. The results of these calculations are
shown in Table B-5. The 95% upper confidence limits (UCLs) on the human risk estimates for
lifetime exposure to 1 ppm VC were then calculated on the basis of each of the sets of bioassay
data, using the 1-hit version of the LMS model, and the resulting risk estimates are shown in
Table B-6. Because saturation of metabolism occurs well above the 1 ppm concentration in the
human, estimates of risk below 1 ppm can be adequately estimated by assuming linearity (e.g.,
the risk estimates for lifetime exposure to 1 • g/m3 of VC would range from approximately 0.3 x
10"6 to 1.0 x 10"5). It should be noted that although the animal studies represent both inhalation
and oral exposure, the risk predictions in each case are for human inhalation exposure.
MflLE 50 RflTS — UflTQNflBE
io'2' nr1 ' i'o° ib1 ib2
XDQSE - Dose (mq/kq)
103
Figure B-5. Model-predicted (lines) and experimentally determined (symbols) total expired
CO2, as a percent of total metabolism (upper line and symbols) and as a percent of dose
(lower line and symbols), following oral dosing with VC in corn oil (Watanabe and
Gehring, 1976b).
B-23
-------
- HUMflN -- BUCHTER SUBJECT ft
-, HUMflN -- 5UCHTE5 SUBJECT B
-O.OQ
(a)
0.15 0.30 0.4S
T ~ Time (hrs)
-- BUCHTER SUBJECT
0.60
0.73
0.15
0.30 0.45
- Time (hrs)
0.60
0.7S
(b)
^ HUHflN -- BUCHTER SUBJECT 3
-0.0
0.1
0.2 0.3
- Time (hrs)
0.4
-
Q.
x -
CJ
0.5
-0.00
0.30 0.45
- Time (hrsJ
0.60
(C)
(d)
Figure B-6. Model predictions (lines) and experimental data (symbols) for the chamber
concentration during exposure of human subjects to VC in a closed, recirculated chamber
(Buchter et al., 1978). (a) The lines show the model predictions for (left to right) VMAX1C
= 2.5, 3.5, and 4.5. The rest of the model parameters are those shown for the human in
Table A-l. (b) The lines show the model predictions for (left to right) VMAX1C = 10 and
3.5 (compare to Subject A in Fig. B-6a). The rest of the model parameters are those shown
for the human in Table A-l. (c) The lines show the model predictions for (top to bottom)
VMAX1C = 2.5, 3.5, and 4.5. The rest of the model parameters are those shown for the
human in Table A-l. (d) The lines show the model predictions for (top to bottom)
VMAX1C = 10 and 3.5. The rest of the model parameters are those shown for the human
in Table A-l.
B-24
-------
HUMfiN -- -BfiRETTA
5.6 11.2. 16.8
T - T ime (hr s)
22.4
23.0
Figure B-7. Model predictions (lines) and experimental data (symbols) for the exhaled air
concentration following inhalation exposure of human subjects for 8 hr (with a 30-min
break for lunch) to a constant concentration of (top to bottom) 492, 261, and 59 ppm VC
(Baretta et al., 1969). At each concentration the three lines show the model predictions for
(top to bottom) VMAX1C = 0, 4, and 8. The rest of the model parameters are those shown
for the human in Table A-l.
B-25
-------
Male ^ats -- 25.8 mq/kq in oil
0.0
0.4
0.8 _ 1.2
T - Time (hrs)
1.6
2.D
(a)
Male Rats -- 28.52 ma/ka in oil
(b)
0.4 o.a 1.2
T - Time (hrs)
Male Rats — 77.47 no/kg in oj 1
ucn"
O
O
CD
0.55 1.10 1.65
T ~ Time thrs)
2.20 2.75
(C)
Figure B-8. Model-predicted (lines) and experimentally determined (symbols) blood
concentrations following oral dosing with VC in corn oil (Withey et al., 1976): (a) 25.8
mg/kg, (b) 28.52 mg/kg, (c) 77.47 mg/kg. The KA (absorption rate constant) used was (a) 2,
(b) 2, (c) 4.
B-26
-------
Table B-5. Dose metric values for angiosarcomas
Reference
Occupational
exposure
Maltoni et al.,
1981, 1984
(BT4)b
Route
Inhalation
Drinking
water
Inhalation
Snecies
Human
Swiss
Albino
mice (M)
Duration
8 hr/d, 5 d/wk,
50wk/yrforlO
of70yr
8 hr/d, 5 d/wk,
50wk/yrfor20
of70yr
Continuous
exposure
4 hr/d, 5 d/wk for
30 of 104 wk
Dose
50 ppm
100 ppm
200 ppm
500 ppm
1,000 ppm
2,000 ppm
50 ppm
100 ppm
200 ppm
500 ppm
1,000 ppm
2,000 ppm
1 ppm in air*
0.028 mg/kg/d
(1 mg/Lin
drinking water)
Oppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10.000 DDm
Incidence
0/80
1/30
9/30
6/30
6/29
2/30
1/26
Daily dose"
metric
nig/L liver
RISK
26.574s
(44.747)
51.685
(88.631)
97.509
(172.566)
202.356
(353.642)
300.965
(431.489)
386.295
(478.173)
26.574
(44.747)
51.685
(88.631)
97.509
(172.566)
202.356
(353.642)
300.965
(431.489)
386.295
(478.173)
1.74
(3.029)
0.581
(1.010)
161.924
775.615
1,245.220
1,434.800
1,479.270
1.505.580
Lifetime'
average daily
delivered dose
mg/L liver
RISK
2.607
(4.390)
5.071
(8.696)
9.567
(16.932)
19.854
(34.698)
29.530
(42.336)
37.902
(46.917)
5.215
(8.781)
10.142
(17.392)
19.134
(33.863)
39.709
(69.396)
59.059
(84.672)
75.804
(93.833)
1.74
(3.029)
0.581
(1.010)
33.363
159.811
256.570
295.632
304.795
310.216
B-27
-------
Table B-5. Dose metric values for angiosarcomas (continued)
Reference
Maltoni et al.,
1981, 1984
(BT4)b
(continued)
Route
Inhalation
Species
Swiss
Albino
mice (F)
Duration
4hr/d, 5d/wkfor
30 of 104 wk
Dose
Oppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10.000 mm
Incidence
0/70
0/30
9/30
8/30
10/30
11/30
9/30
Daily dose"
metric
nig/L liver
RISK
156.907
673.015
887.253
1,197.670
1,341.160
1.405.330
Lifetime'
average daily
delivered dose
mg/L liver
RISK
32.330
138.671
182.813
246.773
276.338
289.560
Reference
Maltoni et al.,
1981, 1984
(BT1, BT2, and
BT15)C
Route
Inhalation
Species
Sprague-
Dawley
rats (M)
Duration
4hrs/d, 5d/wk
for:
52 of 147 wk
(BT15)
52 of 135 wk
(BT1)
52 of 143 wk
(BT2)
52 of 135 wk
(BT1)
Dose
Oppm
1 ppm
5 ppm
10 ppm
25 ppm
50 ppm
100 ppm
150 ppm
200 ppm
250 ppm
500 ppm
2,500 ppm
6.000 DDm
Incidence
0/108
0/48
0/43
0/42
1/41
0/26
0/37
1/36
7/42
2/26
6/28
7/24
10/25
Daily dose
metric
RISK
2.398
11.985
23.933
59.552
117.989
473.425
593.928
803.198
911.248
Lifetime
average
delivered dose
RISK
0.606
3.028
6.047
15.047
32.463
59.70
85.90
107.39
130.254
163.409
220.986
250.714
B-28
-------
Table B-5. Dose metric values for angiosarcomas (continued)
Reference
Maltoni et al.,
1981, 1984
(BT1, BT2,
andBT15)b
(continued)
Feron et al.,
1981
Route
Inhalation
Diet
Species
Sprague-
Dawley
rats (F)
Sprague-
Dawley
rats (F)
Wistar
rats (M)
Wistar
rats (F)
Duration
4 hrs/d, 5 d/wk
for:
52 of 147 wk
(BT15)
52 of 135 wk
(BT1)
52 of 143 wk
(BT2)
135 wk
144wk
Dose
0 ppm
1 ppm
5 ppm
10 ppm
25 ppm
50 ppm
100 ppm
150 ppm
200 ppm
0.714
mg/kg/d
2.38 mg/kg/d
11. 9 mg/kg/d
35.7 mg/kg/d
0 mg/kg/d
0.021
mg/kg/d
0.214
mg/kg/d
0.714
mg/kg/d
2.38 mg/kg/d
11. 9 mg/kg/d
35.7ni2/k£/d
0 mg/kg/d
1.7 mg/kg/d
5.0 mg/kg/d
14.1 mg/kg/d
0 mg/kg/d
1.7 mg/kg/d
Incidence
0/141
0/55
0/47
1/46
4/40
1/29
1/43
5/46
5/44
1/21
0/34
4/39
8/36
0/73
0/18
1/19
2/29
0/37
6/34
9/35
0/55
(0/55)d
0/58 (2/58)
6/56
(11/56)
27/59
(41/59)
0/57 (2/57)
0/58
(28/58)
Daily dose
metric
RISK
2.343
11.698
23.332
57.838
113.653
16.373
50.390
133.231
203.079
0.477
4.835
15.800
45.330
102.763
143.866
39.539
116.103
325.845
38.611
Lifetime
average
delivered dose
RISK
0.592
2.956
5.895
14.614
31.270
55.95
76.67
90.0
4.472
13.762
36.387
55.463
0.130
1.321
4.315
12.380
28.066
39.291
39.539
116.103
325.845
38.611
B-29
-------
Table B-5. Dose metric values for angiosarcomas (continued)
Reference
Route
Species
Duration
Dose
5.0mg/kg/d
14.1 mg/kg/d
Incidence
2/59
(49/59)
9/57
(56/57)
Daily dose
metric
RISK
113.243
316.628
Lifetime
average
delivered dose
RISK
113.243
316.628
The dose metrics reported here differ from those shown in Table A-3, because of use of different breathing rates. The
standard EPA breathing rate was used to calculate the dose metric shown in Table A-3, and that value was used for the risk
calculations.
bThe denominator for the incidence data is the total number of mice, as used by Chen and Blancato (1989).
The denominator is the number of rats alive when the first angiosarcoma was observed, as used by Chen and Blancato
(1989). However, the male and female incidence data shown here differ from that reported by Chen and Blancato (1989),
after verification with the original study (Maltoni et al., 1984).
dNumber in parentheses is the combined incidence of liver angiosarcomas, hepatocellular carcinomas, and neolastic
nodules.
TDaily dose metric for subject species. Animal metric not converted to human metric.
fConverted to continuous exposure over a lifetime.
8Assuming a km value of 1.0 for humans. The numbers in parenthesese represent metrics based upon a km value of 0.1.
B-30
-------
Table B-6 Human risk estimates for inhalation exposure based
on angiosarcoma incidence in oral and inhalation animal assays
and various dose metrics
Maltoni et al. (1981, 1984)
BT4 - Inhalation
Male mice
Female mice
Maltoni et al. (1981, 1984)
BT15/BT1 - Inhalation
Male rats
Female rats
Maltoni et al. (1981, 1984) BT11- Gavage
Male rats
Female rats
Feronetal. (1981) -Diet
Male rats
Female rats
Risk x 10 3/ppmabc
(95% UCL)
2.6
5.7
9.0
3.9
15.1
27.3
5.3
1.9
P
0.005
0.5
0.1
0.2
0.3
0.07
0.005
0.4
Fit
Reject
Good
Poor
OK
OK
Poor
Reject
Good
aRisks were calculated using the 1-hit version of the LMS model.
bTo convert risk estimates to a ug/m3 basis divide by 2600.
°Based on dose metric "RISK."
B-31
-------
There are no consistent differences between risk estimates based on male and female
animals exposed via inhalation, with the female-based risks being higher than the male-based
risks in some studies and lower in others, but generally agreeing within a factor of 2 to 3. The
human risk estimates based on inhalation studies with mice (0.35 x 10"6 to 2.4 x 10"6 per • g/m3)
agree very well with those based on inhalation studies with rats (1.0 x 10"6 to 4.2 x 10"6 per
• g/m3), demonstrating the ability of pharmacokinetics to integrate dose-response information
across species.
The risks estimated from the dietary administration of VC based upon liver
angiosarcomas alone (0.7 x 10"6 to 2 x 10"6 per • g/m3) are similar to those obtained from the
inhalation bioassays. However, the risk estimates based on combined incidence of
angiosarcoma, hepatocellular carcinoma, and neoplastic nodules are considerably greater. Oral
gavage of VC in vegetable oil also resulted in about a sixfold greater risk than the risk based
upon either angiosarcomas alone in the oral exposures, or the inhalation exposure. It has
previously been noted in studies with chloroform that administration of the chemical in corn oil
results in more marked hepatotoxic effects than when the same chemical is provided in an
aqueous suspension (Bull et al., 1986). It has also been demonstrated that administration of corn
oil alone leads to an increase in peroxisomal oxidative enzyme activity in rats (DeAngelo et al.,
1989). The toxicity and oxidative environment created in the liver by continual dosing with
large volumes of vegetable oil could serve to potentiate the effects of genotoxic carcinogens in
the liver. In support of this suggestion, Newberne et al. (1979) found that incorporation of corn
oil into the diet increased the yield of aflatoxin Brinduced tumors in rats. A similar
phenomenon could be responsible for the apparently higher potency of VC when administered
by oil gavage compared to incorporation in the diet.
The/>-values for the goodness of fit of the one-stage LMS model with the
pharmacokinetic dose metric RISK to the bioassay data are generally acceptable, with only two
data sets meeting the criterion for rejection of the model atp=0.05. The ^-values for goodness of
fit with the different metrics (including RISK, RISKM, and RISKG) were in general very
similar; therefore only a single representative p-value is shown for each bioassay data set. The
similarity of the ^-values makes it impossible to select one metric over another on the basis of
agreement with the dose-response of the incidence data. Fortunately, the risks predicted for each
of the studies by the various dose metrics are quite similar. The RISKM metric, which is the
most biologically plausible, predicts slightly lower risks than the other two dose metrics; the
RISKG metric, which is probably the least likely, predicts the highest risks.
B.5.1. Epidemiological Analysis of Vinyl Chloride Carcinogenicity
In order to evaluate the plausibility of the risks predicted on the basis of the animal data,
risk calculations were also performed on the basis of available epidemiological data. A linear
relative risk dose-response model was used for analysis of the human data:
O = E(l + • *d),
B-32
-------
where O is the observed number of liver tumors, E is the expected number of such tumors apart
from any exposure, d is a cumulative dose metric (see discussion below), and • is a potency
parameter that can be estimated by maximum likelihood techniques. Then it follows that the
lifetime probability of liver cancer, P(d), can be estimated by
P(d) = P0(l + -*d),
where P0 is the background probability of liver cancer death. Actually, the lifetime risk should
be estimated by a lifetable method, but the above approximation should be close enough for the
purpose of these comparative potency estimates.
Now suppose that for a particular exposure scenario (e.g., a VC atmospheric
concentration of 50 ppm, 8 hr a day, 5 days per week), the PBPK model predicts an average
daily internal dose metric of X. Then the cumulative exposure that should be used in the dose-
response model is X*Y, where Y is the number of years of such exposure. Note that to compute
this PBPK-based cumulative dose, one must have an estimate of the "typical" workplace
exposure concentration for each subcohort, separate from the number of years of exposure for
the subcohort, rather than just a cumulative dose estimate. Only after the internal dose has been
calculated with the PBPK model can the duration of exposure be applied to get a cumulative
internal dose.
To obtain pharmacokinetic human-based risk estimates, the PBPK model was run for the
exposure scenario appropriate to each of the selected subcohorts from the studies discussed
below. The resulting internal dose metrics (which included RISKM and RISKG for comparison
with RISK) were multiplied by the appropriate durations to obtain the cumulative internal doses,
which were then input into the relative risk model along with the observed and expected liver
cancer deaths for each subcohort to get an estimate of the maximum likelihood estimate and 95%
confidence interval for • . Then, to determine the risk associated with a continuous lifetime
exposure to 1 ppm for comparison with the animal results, the PBPK model was run for a 1 ppm
continuous exposure and the average daily value of the various internal dose metrics was
calculated. Multiplying the dose metrics by 70 years gives the appropriate cumulative dose for
the relative risk model. For a P0 sufficiently small (which it should be for liver cancer in
humans), the extra risk for a lifetime exposure to 1 ppm VC will be approximately:
p *• *H
ro Qi>
where dx = cumulative internal dose for 1 ppm continuous exposure. Using the 95% upper
bound on the estimate for • provides a 95% upper confidence limit on the lifetime risk per ppm
for comparison with the animal-based results obtained with the LMS model.
Three epidemiological studies that associated increased liver cancer with exposure to
VC, and that provide sufficient information to support separate exposure concentration and
duration estimates (as opposed to just cumulative exposure estimates), were selected for this
study: Fox and Collier (1977), Jones et al. (1988), and Simonato et al. (1991). For each study,
B-33
-------
risk was calculated as a linear function of the product of duration and cumulative tissue dose.
B.5.1.1. Fox and Collier (1977)
This study is probably the best with respect to providing information about duration of
employment for different exposure-level groupings (see their Table 2). The average exposure
levels were estimated to be 12.5, 70, and 300 ppm for the low, medium, and high exposure
groups, respectively (Clement, 1987); for comparison Chen and Blancato (1989) estimated
averages of 11, 71, and 316 ppm. For the constant exposure groups, these concentrations were
input into the human PBPK model, assuming 8 hr/day and 5 days/week exposure, to get average
daily internal dose metrics, which were then multiplied by the duration averages (assumed to be
5, 15, and 27 years) to get cumulative doses. For the intermittent exposure groups, exposure for
2 hr/day, 5 days/week was assumed.
Thus, for each exposure level, six values for the cumulative dose were calculated: one
for each of three exposure durations, under both the intermittent and constant exposure
scenarios. Because observed and expected numbers of liver cancers were reported only by
exposure group, not broken down by duration (see their Table 9), an overall average dose was
needed for each exposure level. Therefore, a weighted average of the six values for the
cumulative dose was calculated for each exposure group (high, medium, and low), averaging
across the duration of exposure categories and constant versus intermittent groups. The
weighting was performed using the number of workers in the various subcohorts (their Table 2).
The resulting weighted dose estimates for each internal dose metric were then input into
the relative risk model along with the observed and expected tumors reported by the
investigators:
Cumulative dose Obs. Exp.
Average low dose 1 0.75
Average medium dose 1 0.77
Average high dose 2 0.13
The resulting risk estimates for each pharmacokinetic dose metric are shown in Table
B-7. The range of risk estimates reflects uncertainty in the appropriate value for P0, the
background probability of death from liver cancer. The lower risk estimate was calculated using
the value of P0 derived in the Fox and Collier study, while the higher risk estimate was
calculated using an estimate of the lifetime liver cancer mortality rate in the U.S. population
(Chen and Blancato, 1989). Note that the "range" of risk estimates reflects the results
corresponding to two assumptions about the background rate of liver cancer in humans, rather
than reflecting a true range. An important factor in interpreting these results is that the
classification into exposure groups in this study was based on the maximum exposure level that a
worker experienced. This leads to overestimation of cumulative exposure, particularly for the
workers in the medium and high groups, and therefore a probable underestimation of risk when
using the linear relative risk model.
B-34
-------
B.5.1.2. Jones etal (1988)
This study was an update of the cohort studied by Fox and Collier. Unfortunately, it does
not provide as much information about duration of exposure, so the analysis must be limited to
the autoclave workers. For those workers, four duration-of-employment categories are given
(see their Table 4); in the present analysis estimated average durations of 1.5, 3, 7.5, and 15
years were used. Their Table 1 shows that the autoclave workers had exposures ranging between
150 and 800 ppm at various points in time. A value of 500 ppm was used in the PBPK model (8
hr/day, 5 days/week) to get the average daily internal doses. The average daily internal doses
were then multiplied by the four average durations of exposure to get cumulative doses for the
four groups:
Cumultive dose
Cumulative dose group (units of RISK dose metric) Obs. Exp.
Low 400 trig, x year 0 0.07
Mid 1 802 m^ x year 1 0.08
Mid 2 2,004 mgL x year 2 0.08
High 4,009 mgL x year 4 0.15
Note that the different cumulative dose groups here reflect different exposure durations to the
same average VC concentrations. Insufficient data were presented in this paper to identify the
number of workers exposed to different exposure levels for different durations.
The resulting risk estimates for each pharmacokinetic dose metric are shown in Table B-
7. In each case the lower risk estimate was calculated using the value of P0 derived in the Jones
et al. (1988) study, while the higher risk estimate was calculated using an estimate of the lifetime
liver cancer mortality rate in the U.S. population (Chen and Blancato, 1989). As with the Fox
and Collier (1977) study, it is important to note that workers were classified into job categories
based on the category with the highest exposure, leading to overestimation of cumulative
exposure.
B-35
-------
Table B-7. Risk estimates for angiosarcoma based on
epidemiological studies (to convert to a unit risk basis
[per • g/m3], divide by 2,600).
Study
Fox and Collier (1977)
Jones etal. (1988)
Simonato et al. (1991)
Risk based on
dose metric RISK
(95%UCLRISK/ppm)
1.2-7.3 x lO'3
1.7-6.3 x icr3
0.70-1.4 x 1Q-3
B.5.1.3. Simonato et al (1991)
This study has the largest cohort and the most liver cancer deaths (24). Unfortunately,
the exposure information may not be as accurate as in the other two studies discussed above,
since it was collected from many different workplaces in several different countries, and since
the original reporting of the exposure levels was relatively crude (ranges of <50, 50-499, and >
500 ppm). As in the Fox and Collier study, the classification was based on the "highest level to
which the workers were potentially exposed." Thus, as with the previous studies, the estimates
of risk from this cohort are probably underestimates of the true risk.
Another problem with the reporting of the results in this study is that the durations of
exposure are not cross-classified according to exposure level as was done in the Fox and Collier
report. In fact, there is very little information about duration of exposure that would allow
estimation of an average value for the entire cohort, let alone the exposure groups. (Note that
one cannot use the cumulative exposure groupings, as discussed above, because the exposure
level must be separated from exposure duration.) The information in Simonato et al. (1991)
Table 2 (person-years of observation by duration of employment) was used to estimate an
average duration under the following assumption: if the follow-up time does not depend on the
duration of employment, then the differences in the person-years of follow-up is due to the
numbers of individuals in each duration category. The weighted average (trying different
averages for the > 20 year group) gives an estimate of 9 years of employment. This duration was
used with model-predicted daily dose metrics for average exposure level estimates of 25, 158,
and 600 ppm. The cumulative internal doses were input into the relative risk model with the
following observed and expected liver cancer deaths reported by the study authors:
B-36
-------
Cumulative Dose Obs. Exp.
Low 4 2.52
Medium 7 1.86
High 12 2.12
The resulting risk estimates for each pharmacokinetic dose metric are shown in Table B-
7. Again, the lower risk estimates were calculated using the value of P0 derived in the Simonato
et al. (1991) study, while the higher risk estimates were calculated using an estimate of the
lifetime liver cancer mortality rate in the U.S. population (Chen and Blancato, 1989).
The comparison in Table B-7 of the analyses of the three sets of data gives some
indication of the consistency of the human results, even before the comparison with the animal
predictions. It is encouraging that the lifetime risk of liver cancer per • g/m3 VC exposure
estimated from the three studies only ranges over about one order of magnitude: from 0.2 to 3 x
10"6 per • g/m3. Moreover, these estimates are in remarkable agreement with the estimates based
on animal data shown in Table B-6. However, any confidence produced by this agreement
should be tempered by the likelihood, discussed above, that misclassification of exposure in the
human studies may somewhat underestimate the true risk at lower doses. Nevertheless, the
agreement of the pharmacokinetic animal-based risk estimates with the pharmacokinetic human-
based risk estimates provides strong support for the assumption used in this study: that cross-
species scaling of lifetime cancer risk can be performed on a direct basis of lifetime average
daily dose (without applying a body surface area adjustment) when the risks are based on
biologically appropriate dose metrics calculated with a validated PBPK model.
Based on a closer consideration of the results, a best estimate of the risk based on the
human data can be calculated. The Simonato et al. (1991) study was excluded from this
consideration because of the considerable uncertainty regarding exposure durations. Between
the remaining two studies, the risk values from Jones et al. (1988) were chosen, since this study
is an update of the Fox and Collier (1977) study. Finally, the higher of the two risk values
calculated for the Jones et al. (1988) study was chosen, reflecting the underestimation of risk due
to classification of workers by the job category with the highest exposure. Based on these
factors, a best estimate of risk from the human studies is 6.3 x 10"3 per ppm (2.4 x 10"6 per
• g/m3). This agrees quite closely with the mean of the risk estimates derived from the Maltoni et
al. (1981, 1984) rat and mouse inhalation studies.
B.5.2. Calculation of Approximate Risk Estimates for Other Tumors
Although there is no evidence of human correspondence for the other tumors that occur
at low doses in animals, it is of interest to attempt to estimate the likely level of risk that might
be predicted for those tumors using a pharmacokinetic approach. Of particular interest are the
nephroblastomas, which are a relatively rare tumor in the experimental species in which they
were observed, and the mammary tumors, which are of concern in human females. Since the
PBPK model does not contain kidney or mammary tissue compartments, and since there are not
adequate data on the metabolism of VC in these tissues to construct them, a "zero-order
B-37
-------
approximation" approach was utilized in which the metabolism of VC in the liver was used as a
surrogate for in situ metabolism in the other tissues. Thus RISK was calculated for the
conditions and doses of the bioassays showing increased incidence of nephroblastoma or
mammary tumors (Table B-2). The results of these dose calculations are shown in Table B-8,
and the resulting upper-bound risk estimates, using the one-hit version of the LMS model, are
shown in Tables B-9 and B-10. Note that there is as yet no evidence regarding the mechanism
underlying the production of either of these tumors, so the use of the LMS model (and the
associated assumption of low-dose linearity) may not be justified. Dose metrics for
hepatocellular carcinoma are also listed in Table B-9; these risks are similar to those for
angiosarcoma.
Given these caveats, it is interesting to observe that the range of risk estimates based on
the incidence of nephroblastomas (0.07 x 10"6 to 2.4 x 10"6 per • g/m3) is very similar to that
obtained for angiosarcomas. As with angiosarcoma, there was no evidence from the goodness-
of-fit tests that any of the dose metrics provided a better fit to the data. Risk estimates based on
the mammary tumors are less consistent, ranging from 0.2 x 10"6 to 1.7 x 10"4 per • g/m3. Given
the extremely high variability of the background incidence for mammary tumors in the
experimental animals, as well as the highly nonlinear dose-response (for most of the studies the
dose-response in the exposed groups is either flat or decreasing) it does not seem reasonable to
perform a quantitative risk estimate based on this tumor outcome. Nevertheless, it is important
to note that human females also demonstrate a background incidence of mammary tumors, and
that the epidemiological cohorts, with one exception, did not include females. Therefore, it
seems reasonable that the evidence of increased mammary tumor incidence from VC should be
considered at least qualitatively during risk management decisions regarding potential human
VC exposure.
B.6. PHARMACOKINETIC SENSITIVITY/UNCERTAINTY ANALYSIS
Table B-l 1 shows the normalized analytical sensitivities for the PBPK model described
above. The normalized analytical sensitivity coefficient represents the fractional change in
output associated with a fractional change in the input parameter. For example, if a 1% change
in the input parameter results in a 2% change in the output, the sensitivity coefficient would be
2.0. In Table B-l 1, the outputs are the dose metrics used in the analysis of angiosarcoma risk.
The parameters in the table are defined in Tables B-3 and B-4. Sensitivity coefficients of less
than 0.01 in absolute value were omitted from the table for clarity, and coefficients greater than
0.2 in absolute value are outlined for emphasis. None of the parameters display sensitivities
markedly greater than 1.0, indicating that there is no amplification of error from the inputs to the
outputs. This is, of course, a desirable trait in a model to be used for risk assessment.
It can be seen that of the 24 parameters in the VC model, 10 have essentially no impact
on risk predictions based on any of the dose metrics, and only 8 have a significant impact on
B-38
-------
Table B-8. Dose metric values for other tumors
Reference
Lee et al.,
1977, 1978
Drew et al.,
1983
Radike et al.,
1981
Maltoni et al.,
1981, 1984
(BT1)
Maltoni et al.,
1981, 1984
(BT1)
(continued)
Route
Inhalation
Inhalation
Inhalation
Inhalation
Inhalation
Species
Albino
CD-I mice
(F)
Fischer-344
rats (F)
Golden
Syrian
hamsters (F)
B6C3F1
mice (F)
CD-I Swiss
mice (F)
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(F)
Duration
6 hr/d, 5
d/wk for 52
wk
6 hr/d, 5
d/wk for 104
wk
6 hr/d, 5
d/wk for 78
wk
6 hr/d, 5
d/wk for 52
wk
6 hr/d, 5
d/wk for 78
wk
4 hr/d, 5
d/wk for 52
wk
4 hr/d, 5
d/wk for 52
of 135 wk
4 hr/d, 5
d/wk for 52
of 135 wk
Dose
0 ppm
50 ppm
250 ppm
1,000 ppm
0 ppm
100 ppm
0 ppm
200 ppm
0 ppm
50 ppm
0 ppm
50 ppm
0 ppm
600 ppm
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10,000 ppm
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10.000 DDm
Incidence
Mamm."
0/36
9/34
3/34
13/36
29/112
31/55
0/143
47/102
3/69
37/90
2/71
22/45
2/29
1/30
0/29
0/30
0/30
0/29
1/30
12/29
11/30
7/30
5/30
5/30
6/30
7/30
Neph."
0/29
0/30
1/29
2/30
5/30
4/29
3/30
0/29
1/30
4/30
4/30
1/30
1/30
2/30
Daily dose
metrics
RISK
235.368
1,008.690
1,524.920
274.462
753.523
242.897
235.368
617.249
117.990
473.425
593.931
803.194
911.248
966.074
113.653
375.989
425.029
488.374
522.359
542.339
Lifetime
average
delivered
dose
RISK
168.12
720.49
1,089.23
196.04
538.23
173.50
168.12
440.89
32.46
130.25
163.41
220.98
250.71
265.80
31.27
103.45
116.94
134.37
143.72
149.21
B-39
-------
Table B-8. Dose metric values for other tumors (continued)
Reference
Maltoni et al.,
1981, 1984
(BT2)
Maltoni et al.,
1981, 1984
(BT4)
Maltoni et al.,
1981, 1984
(BT4)
(continued)
Route
Inhalation
Inhalation
Inhalation
Species
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(F)
Swiss mice
(M)
Swiss mice
(F)
Duration
4 hr/d, 5 d/wk
for 52 of 143
wk
4 hr/d, 5 d/wk
for 30 of 81
wk
4 hr/d, 5 d/wk
for 30 of 81
wk
Dose
0 ppm
100 ppm
150 ppm
200 ppm
0 ppm
100 ppm
150 ppm
200 ppm
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10,000 ppm
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10.000 DDm
Incidence
Mamm/
1/85
0/60
1/59
3/60
20/100
20/60
12/60
20/60
0/80
0/30
0/30
1/30
0/29
0/30
0/26
1/70
12/30
13/30
10/30
9/30
9/30
14/30
Neph."
0/85
8/60
8/59
5/60
0/100
2/60
3/60
2/60
Daily dose
metrics
RISK
229.851
330.722
413.443
215.406
295.167
346.510
161.924
775.615
1,245.220
1,434.800
1,479.270
1,505.580
156.683
672.996
887.322
1,198.110
1,341.100
1.405.300
Lifetime
average
delivered
dose
RISK
59.70
85.90
107.39
55.95
76.67
90.00
42.84
205.19
329.42
379.58
391.34
398.30
41.45
178.04
234.74
316.96
354.79
371.77
B-40
-------
Table B-8. Dose metric values for other tumors (continued)
Reference
Maltoni et al.,
1981, 1984
(BT3)
Maltoni et al.,
1981, 1984
(BT9)
Maltoni et al.,
1981, 1984
(BT9)
(continued)
Maltoni et al.,
1981, 1984
(BT15)
Route
Inhalation
Inhalation
Inhalation
Inhalation
Species
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(F)
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(F)
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(F)
Duration
4 hr/d, 5
d/wk for
17 wk
4 hr/d, 5
d/wk for
52 of 142
wks
4 hr/d, 5
d/wk for
52 of 142
wk
4 hr/d, 5
d/wk for
52 of 147
wk
Dose
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10,000 ppm
0 ppm
50 ppm
250 ppm
500 ppm
2,500 ppm
6,000 ppm
10,000 ppm
0 ppm
50 ppm
0 ppm
50 ppm
0 ppm
1 ppm
5 ppm
10 ppm
25 ppm
0 ppm
1 ppm
5 ppm
10 ppm
25 DDm
Incidence
Mamm."
1/108
0/28
0/30
0/30
3/30
0/30
1/28
14/82
11/30
5/29
12/30
12/30
4/30
6/30
2/48
14/144
27/50
117/150
8/60
8/58
10/59
6/59
11/60
34/60
46/60
57/60
52/60
53/60
Neph."
0/108
1/28
3/30
0/30
2/30
0/30
0/28
0/82
2/30
3/29
0/30
0/30
1/30
1/30
0/48
0/144
0/50
1/150
0/60
0/58
0/59
0/59
1/60
0/60
0/60
0/60
0/60
0/60
Daily dose
metrics
RISK
117.990
473.425
593.931
803.194
911.248
966.074
113.653
375.989
425.029
488.374
522.359
542.339
117.990
113.653
2.398
11.985
23.933
59.552
2.343
11.698
23.332
57.838
Lifetime
average
delivered
dose
RISK
84.28
338.16
424.24
573.71
650.89
690.05
81.18
268.56
303.59
348.84
373.11
387.39
30.86
29.73
0.61
3.03
6.05
15.05
0.59
2.96
5.90
14.61
B-41
-------
Table B-8. Dose metric values for other tumors (continued)
Reference
Maltoni et al.,
1981, 1984
(BT10)
Maltoni et al.,
1981, 1984
(BT10)
(continued)
Feron et al.,
1981
Route
Inhalation
Inhalation
Oral-diet
Species
Sprague-
Dawley rats
(M)
Sprague-
Dawley rats
(F)
Wistar rats
(M)
Wistar rats
(F)
Duration
4 hr/d, 5
d/wk for
5 of 154
wk
1 hr/d, 4
d/wk for
25 of 154
wk
4 hr/d, 1
d/wk for
25 of 154
wks
4 hr/d, 5
d/wk for
5 of 154
wk
1 hr/d, 4
d/wk for
25 of 154
wk
4 hr/d, 1
d/wk for
25 of 154
wk
135
weeks
144
weeks
Dose
Group VII:
Control
Group I: 10,000
ppm
Group II: 6,000
ppm
Group III:
10,000 ppm
Group IV:
6,000 ppm
Group V:
10,000 ppm
Group VI:
6,000 ppm
Group VII:
Control
Group I: 10,000
ppm
Group II: 6,000
ppm
Group III:
10,000 ppm
Group IV:
6,000 ppm
Group V:
10,000 ppm
Group VI:
6,000 ppm
0 mg/kg/day
1.7mg/kg/day
5 mg/kg/day
14.1 mg/kg/day
0 mg/kg/day
1.7 mg/kg/day
5 mg/kg/day
14.1 ms/ka/dav
Incidence
Mamm."
11/107
3/59
3/60
2/59
4/59
9/60
6/60
76/120
36/59
37/60
42/60
40/60
45/59
46/60
Neph."
0/107
0/59
1/60
0/59
0/59
0/60
1/60
0/120
0/59
1/60
0/60
0/59
1/59
0/60
Daily dose
metrics
RISK
966.074
911.248
356.811
319.490
966.074
911.248
542.339
522.359
222.071
202.515
542.339
522.359
37.561
85.345
143.370
34.928
71.008
109.035
Lifetime
average
delivered
dose
RISK
22.40
21.13
33.10
29.64
22.40
21.13
12.58
12.11
20.60
18.79
12.58
12.11
B-42
-------
Table B-8. Dose metric values for other tumors (continued)
Reference
Tiletal, 1983
Route
Oral-diet
Species
Wistar rats
(M)
Wistar rats
(F)
Duration
149
weeks
Dose
0 mg/kg/day
0.014 mg/kg/day
0.13 mg/kg/day
1.3 mg/kg/day
0 mg/kg/day
0.014 mg/kg/day
0.13 mg/kg/day
1.3 mg/kg/day
Incidence
Mamm."
5/100
8/99
3/99
0/49
41/98
21/100
28/96
21/48
Neph."
Daily dose
metrics
RISK
0.326
3.026
30.2
0.318
2.96
29.5
Lifetime
average
delivered
dose
RISK
0.326
3.026
30.2
0.318
2.96
29.5
aMammary gland carcinoma.
bNephroblastoma.
B-43
-------
Table B-9. Human inhalation risk estimates based on the incidence of
hepatocellular carcinoma or nephroblastoma in oral and inhalation animal assays
and various dose metrics
95% UCL of risk x 10 3/ppma "
P
Fit
Hepatocellular carcinoma:
Feronetal. ( 1981) -Diet
Male rats
Female rats
Til etal. (1983) -Diet
Male rats
Female rats
1.2
6.9
3.2
3.8
0.45
0.1
0.7
0.6
Good
Poor
Good
Good
Nephroblastoma:
Maltoni etal. (1981, 1984)
BTl-Inhalation
Male mouse
Female mouse
Maltoni etal. (1981, 1984)
BT2-Inhalation
Male rats
Female rats
Maltoni etal. (1981, 1984)
BT3 -Inhalation (17wks)
Male rats
Female rats
2.1
2.8
5.8
2.6
0.38
0.64
0.7
0.2
0.1
0.8
0.01
0.02
Good
OK
OK
Good
Reject
Reject
aRisk estimates were calculated using the 1-hit version of the LMS model and based on the dose-metric "RISK.'
bTo convert to a unit risk estimate (• g/m3) divide by 2,600.
B-44
-------
Table B-10 Human inhalation risk estimates based on total mammary tumor
incidence in oral and inhalation animal assays and various dose metrics
Lee etal. (1977, 1978)
Female mice
Maltoni etal. (1981, 1984)
BT2-Inhalation
Male rats
Female rats
Maltoni etal. (1981, 1984)
BT4-Inhalation
Female mice
Maltoni etal. (1981, 1984)
BT3 -Inhalation (17 weeks)
Female rats
Maltoni etal. (1981, 1984)
BT15-Inhalation
Male rats
Female rats
Til etal. (1983) -Diet
Female rats
95% UCL of risk x 10 3/ppma "
RISK
1.4 x 10'3
1.4 x lO'3
8.2 x 10'3
5.2 x lO'3
1.6 x 10'3
3.0 x lO'2
4.4 x 10'1
1.3 x 10'2
P
0.0003
0.3
0.1
0.002
0.01
0.7
io-n
0.005
Fit
Reject
OK
Poor
Reject
Reject
Good
Reject
Reject
aRisks were calculated using the 1-hit version of the LMS model and based upon the dose metric "RISK.".
bTo convert to a unit risk estimate (• g/m3) divide by 2,600.
B-45
-------
Table B-ll. Normalized parameter sensitivity in the vinyl chloride PBPK model
Rat inhalation
(50 ppm - 4 hr)
Dose metric RISK AMET
Parameter
BW -0.25 -0.25
QPC 0.30 0.30
QCC 0.58 0.58
QFC — —
QLC 0.58 0.58
VFC — —
VLC -0.99 —
PB 0.67 0.67
PF — —
PS — —
PR — —
PL — —
VMAX1C 0.09 0.09
KM1 -0.09 0.09
VMAX2C — —
KM2 — —
KA — —
KCO2C — —
KGSMC — —
KFEEC — —
GSO — —
KBC — —
KS — —
KOC — —
Human inhalation Human drinking water
(1 ppm - continuous) (1 ppm)
RISK AMET RISK
-0.25 -0.25 — a
0.20 0.20 —
0.74 0.74 -0.06
— — —
0.74 0.74 -0.06
— — —
-0.99 — -0.99
0.79 0.79 —
— — —
— — —
— — —
— — —
0.07 0.07 0.07
-0.07 -0.07 -0.07
— — —
— — —
— — —
— — —
— — —
— — —
— — —
— — —
— — —
— — —
aSensitivity coefficient < 0.01 in absolute value.
predictions based on RISK: the body weight (BW), alveolar ventilation (QPC), cardiac output
B-46
-------
(QCC), liver blood flow (QLC) and volume (VLC), blood/air partition coefficient (PB), the
capacity (VMAX1C) and affinity (KM1) for metabolism by CYP2E1, and in the case of oral
gavage, the oral uptake rate (KA). As discussed in the description of the PBPK model, all of
these parameters could be reasonably well characterized from experimental data. However, the
sensitivity of the risk predictions to the human values of these parameters implies that the risk
from exposure to VC could vary considerably from individual to individual, depending on
specific physiology, level of activity, and metabolic capability.
The other dose metrics, RISKM and RISKG (data not shown), are also sensitive to a
number of the parameters in the model for the subsequent metabolism of the reactive
metabolites, as well as for the GSH submodel. Since these parameters could only be identified
from data in rats, their values in other species are uncertain. Given the sensitivity of RISKM and
RISKG to these less certain parameters, and the general similarity of risks based on these two
metrics to those based on the RISK metric, the RISK metric would seem to be preferable for
quantitative risk assessment. Risk estimates reported in the main body of this document were
calculated using the RISK metric.
B.6.1. Monte Carlo Uncertainty/Varlability Analysis
The sensitivity analysis described above does not consider the potential interactions
between parameters; the parameters are tested individually. Also, sensitivity analysis does not
adequately reflect the uncertainty associated with each parameter. The fact that the output is
highly sensitive to a particular parameter is not important if the parameter is known exactly. To
estimate the combined impact of the uncertainty around the values of all the parameters, a Monte
Carlo analysis can be performed. In a Monte Carlo analysis, the distributions of possible values
for each of the input parameters are estimated. The Monte Carlo algorithm then randomly
selects a value for each parameter from its distribution and runs the model. The random
selection of parameter values and running of the model is repeated a large number of times
(typically hundreds to thousands) until the distribution of the output has been characterized.
To assess the impact of parameter uncertainty on risk predictions, a dose-response model
must be selected. In this case the one-hit version of the linearized multistage model was used,
for the reasons discussed earlier. The actual analysis was performed with the software package
PBPK_SEVI (KS Crump Group, ICF Kaiser International, Ruston, LA), which was developed for
the Air Force specifically to perform such a Monte Carlo analysis on PBPK models. The
PBPK_SEVI program randomly selects a set of parameter values from the distributions for the
bioassay animal and runs the PBPK model to obtain dose metric values for each of the bioassay
dose groups. It then selects a set of parameter values from the distributions for the human and
runs the PBPK model to obtain a dose metric value for a specified human exposure scenario.
Finally, it runs the linearized multistage model (or other specified risk model) with the animal
and human dose metric values to obtain the human risk estimate. This entire process is repeated
a specified number of times until the desired distribution of risks has been obtained.
B-47
-------
Tables B-3 and B-4 list the means (preferred values) and coefficients of variation (CV)
used in a Monte Carlo uncertainty analysis of the TCE/TCA model. Truncated normal
distributions were used for all parameters except the kinetic parameters, which were assumed to
be lognormally distributed. The CVs for the physiological parameters were estimated from data
on the variability of published values (U.S. EPA, 1988; Stan Lindstedt, 1992, personal
communication), while the CVs for the partition coefficients were based on repeated
determinations for two other chemicals, perchloroethylene (Gearhart et al., 1993) and
chloropentafluorobenzene (Clewell and Jarnot, 1994). The CVs for the metabolic and kinetic
constants were estimated from a comparison of reported values in the literature and by
exercising the model against the various data sets to determine the identifiability of the
parameters which were estimated from pharmacokinetic data. The KM1 value for humans of 1.0
(vice 0.1) was used in this analysis.
The results of the Monte Carlo analysis are shown in Table B-12, which lists the
estimated risks associated with lifetime exposure to 1 ppm VC in air or 1 mg/L VC in drinking
water. In all cases, the risk estimates represent the 95% UCL for risk, based on the 1-hit version
of the LMS model. However, in order to characterize the impact of uncertainty in the
pharmacokinetic parameters on the risk estimates, both the mean and the upper 95th percentile of
the distribution of UCL risk estimates are shown. Thus, the mean value represents the best
estimate of the pharmacokinetically based upper-bound risk for VC exposure, and the 95th
percentile provides a reasonable value for the "highest probable" pharmacokinetic risk estimate,
considering both pharmacokinetic uncertainty and uncertainty regarding the low-dose
extrapolation. The small differences between the best estimates from the Monte Carlo analysis
listed in Table B-12 and those listed in columns 3 and 7 result from the way in which they were
calculated. While the values in columns 3 and 7 are the risk estimates using the mean values for
the parameters, the other values are the mean risk estimates based on the distribution of risk
estimates estimates calculated in the Monte Carlo analysis. As can be seen, even the "highest
probable" pharmacokinetic risk estimates were only modestly greater than those using mean
values, giving added confidence to the assessments. As discussed in the Toxicological Review,
these values have been derived using only liver angiosarcomas in order to compare with results
of other modeling approaches, and do not account for hepatocelluar carcinoma or neoplastic
nodule incidence.
B.7. DISCUSSION
Although VC has often been cited as a chemical for which saturable metabolism should
be considered in the risk assessment, saturation is relevant only at very high exposure levels
(greater than 250 ppm by inhalation or 25 mg/kg/day orally) compared to the lowest tumorigenic
levels, and thus has little impact on the quantitative risk estimates. The important contribution
of pharmacokinetic modeling is to provide a more biologically plausible estimate of the effective
B-48
-------
Table B-12. Comparison of human inhalation and oral risk estimates for liver angiosarcoma resulting from a Monte
Carlo analysis, based on a pharmacokinetic dose metric" and using a human km value of 1.0
Animal
route
Inhalation
Inhalation
Oil gavage
Diet
Sex/species
Male mouse
Female
mouse
Male rat
Female rat
Male rat
Female rat
Male rat
Female rat
Ippm
Inhalation
UCLb
1.52
3.27
5.17
2.24
8.68
15,70
3.05
1.10
1 ppm inhalation
Mont Carlo analysis
Mean/
UCL
1.89
3.89
6.80
1.90
9.45
16.35
3.26
1.15
P
0.002
0.25
0.20
0.44
0.57
0.11
0.05
0.43
95th/U
CL
3.38
6.95
14.31
3.81
17.22
29.73
5.26
1.87
1 mg/L
drinking
water
UCLb
0.51
1.10
1.72
0.75
2.90
5.23
1.02
0.37
1 mg/L drinking water
Monte Carlo analysis
Mean/
UCL
0.67
1.39
2.45
0.67
3.36
5.83
1.14
0.41
P
0.002
0.25
0.20
0.44
0.57
0.11
0.05
0.43
95th/
UCL
1.18
2.33
5.60
1.37
5.72
10.54
1.64
0.60
td
T)ose metric = lifetime-average total amount metabolized per day, divided by the volume of liver.
bBased on the incidence of angiosarcoma in the corresponding oral and inhalation animal bioassays.
-------
dose: total production of reactive metabolites at the target tissue. The ratio of this biologically
effective dose to the administered dose is not uniform across routes and species. Therefore, any
estimate of administered dose is less adequate for performing route-to-route and interspecies
extrapolation of risk. The inhalation risk estimates obtained for VC using the pharmacokinetic
dose metric are considerably lower than those obtained with conventional
external dose calculations, and appear to be more consistent with human epidemiological data.
In the pharmacokinetic risk calculations presented in this report, no body weight scaling
adjustment factor was applied to obtain the human risks. Although this may appear to represent
a departure from previous EPA practice in a risk assessment for VC, this marks the first time a
pharmacokinetic dose metric has been used. The dose metric was selected to be consistent with
the position stated in the interagency pharmacokinetics group consensus report on cross-species
extrapolation of cancer (U.S. EPA, 1992) that "...tissues experiencing equal average
concentrations of the carcinogenic moiety over a full lifetime should be presumed to have equal
lifetime cancer risk." As discussed above, this adjustment does not address any pharmaco-
dynamic differences that may exist between rodents and humans. For VC, sufficient information
exists to support the position that rats are at least as sensitive, if not more so, than are humans to
the carcinogenic effects of UCL.
The risk assessment performed in this study has focused on cancer risk from a continuous
lifetime exposure, or at least an exposure over a large fraction of lifetime. Although there are
certainly many uncertainties and unresolved issues regarding cross-species extrapolation of
lifetime risks, there are even greater uncertainties regarding the extrapolation of partial-lifetime
exposures. In particular, studies performed with VC make it evident that extrapolation of partial
lifetime exposure is not straightforward with this chemical. For example, in the comparative
studies of partial lifetime exposure of rats to VC discussed earlier (Drew et al., 1983), whereas
exposure from 0 to 6 months resulted in a similar tumor incidence to exposure from 6 to 12
months of life, exposure from 0 to 12 months produced a significantly different incidence than
would be expected from the sum of the incidences for the two subintervals. For angiosarcomas,
on the one hand, exposure to VC from 0 to 6 months and from 6 to 12 months resulted in
incidences of 5.3% and 3.8%, respectively, while exposure from 0 to 12 months resulted in a
much higher incidence of 21.4%. For hepatocellular carcinomas, on the other hand, exposure to
VC from 0 to 6 months and from 6 to 12 months resulted in incidences of 4.0% and 11.5%,
respectively, while exposure from 0 to 12 months resulted in an incidence of only 7.1%. Thus
this comparative bioassay does not provide support for a simple relationship of the observed
incidence to the fraction of lifetime of the exposure. As discussed earlier, it seems reasonable to
assume that newborns, with their higher rate of cell proliferation, should be at greater risk from
genotoxic carcinogens, and some studies with VC support this assumption (Maltoni et al., 1981;
Laib et al., 1989; Fedtke et al., 1990), although other well-conducted studies with VC do not
(Drew et al., 1983). The issue of sensitive populations has never been seriously dealt with in
quantitative carcinogenic risk assessment, but it would seem to be an appropriate consideration
during risk management for specific potential exposures.
B-50
-------
B.8. REFERENCES
Andersen, M; Clewell, H; Gargas, M; et al. (1987a) Physiologically based pharmacokinetics and the risk
assessment process for methylene chloride. Toxicol Appl Pharmacol 87:185-205.
Andersen, ME; Gargas, ML; Clewell, HJ; et al. (1987b) Quantitative evaluation of the metabolic interactions
between trichloroethylene and 1,1-dichloroethylene by gas uptake methods. Toxicol Appl Pharmacol 89:149-157.
Baretta, ED; Stewart, RD; Mutchler, JE. (1969) Monitoring exposures to vinyl chloride vapor: breath analysis and
continuous air sampling. Am Ind Hyg Assoc J 30:537-544.
Barton, HA; Creech, JA; Godin, CS; et al. (1995) Chloroethylene mixtures: pharmacokinetic modeling and in vitro
metabolism of vinyl chloride, trichloroethylene, and trans-l,2-dichloroethylene in the rat. Toxicol Appl Pharmacol
130(2):237-247.
Bolt, HM; Laib, RJ; Kappus, H; et al. (1977) Pharmacokinetics of vinyl chloride in the rat. Toxicology 7:179-188.
Buchter, A; Bolt, HM; Filser, JG; et al. (1978) Pharmakokinetic und karzinogenese von vinylchlorid.
Arbeitsmedizinische Risikobeurteilung. Verhandlungen der Deutchen Gesellschaft fuer Arbeitsmedizin, Vol. 18,
Centner Verlag, Stuttgart, pp. 111-124.
Buchter, A; Filser, JG; Peter, H; et al. (1980) Pharmacokinetics of vinyl chloride in the Rhesus monkey. Toxicol
Lett 6:33-36.
Bull, RJ; Brown, JM; Meierhenry, EA; et al. (1986) Enhancement of the hepatotoxicity of chloroform in B6C3F1
mice by corn oil: implications for chloroform carcinogenesis. Environ Health Perspect 69:49-58.
Chen, CW; Blancato, JN. (1989) Incorporation of biological information in cancer risk assessment: Example -
vinyl chloride. Cell Biol Toxicol 5:417-444.
Clement Associates. (1987) Investigation of cancer risk assessment methods. Final report. Vol. 1: Introduction
and Epidemiology. Prepared for the U.S. Environmental Protection Agency, the Department of Defense, and the
Electric Power Research Institute.
Clement International. (1990) Development and validation of methods for applying pharmacokinetic data in risk
assessment. Final Report. Volume V: Vinyl Chloride. AAMRL-TR-90-072. Prepared for the Department of the
Air Force, Armstrong Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio.
Clewell, HJ; Jarnot, BM. (1994) Incorporation of pharmacokinetics in non-carcinogenic risk assessment: example
with chloropentafluorobenzene. Risk Anal 14:265-276.
Clewell, HJ; Gentry, PR; Gearhart, JM; et al. (1995a) The development and validation of a physiologically based
pharmacokinetic model for vinyl chloride and its application in a carcinogenic risk assessment for vinyl chloride.
ICF Kaiser report prepared for EPA/OHEA and OSHA/DHSP.
DeAngelo, AB; Daniel, FB; McMillan, L; et al. (1989) Species and strain sensitivity to the induction of peroxisome
proliferation by chloroacetic acids. Toxicol Appl Pharmacol 101:285-298.
Drew, RT; Boorman, GA; Haseman, JK; et al. (1983) The effect of age and exposure duration on cancer induction
by a known carcinogen in rats, mice, and hamsters. Toxicol Appl Pharmacol 68:120-130.
D'Souza, RW; Andersen, ME. (1988) Physiologically based pharmacokinetic model for vinylidene chloride.
Toxicol Appl Pharmacol 95:230-240.
B-51
-------
Fedtke, N; Boucheron, JA; Walker, VE; et al. (1990) Vinyl chloride-induced DNA adducts. II: Formation and
persistence of 7-(2'-oxoethyl)guanine and N2,3-ethenoguanine in rat tissue DNA. Carcinogenesis 11:1287-1292.
Feron VJ; Kruysse A; Til HP. (1979a) One-year time sequence inhalation toxicity study of vinyl chloride in rats. I.
Growth, mortality, haematology, clinical chemistry, and organ weights. Toxicology 13(l):25-28.
Feron VJ; Spit BJ; Immel HR; et al. (1979b) One-year time-sequence inhalation toxicity study of vinyl chloride in
rats. III. Morphological changes in the liver. Toxicology 13(2): 143-154.
Feron, VJ; Hendriksen, CFM; Speek, AJ; et al. (1981) Lifespan oral toxicity study of vinyl chloride in rats. Food
Cosmetol Toxicol 19:317-333.
Fox, AJ; Collier, PF. (1977) Mortality experience of workers exposure to vinyl chloride monomer in the
manufacture of polyvinyl chloride in Great Britain. Br J Ind Med 34:1-10.
Gargas, ML; Burgess, RJ; Voisard, DE; et al. (1989) Partition coefficients oflow-molecular-weight volatile
chemicals in various liquids and tissues. Toxicol Appl Pharmacol 98:87-99.
Gargas, ML; Clewell, HJ, III; Andersen, ME. (1990) Gas uptake techniques and the rates of metabolism of
chloromethanes, chloroethanes, and chloroethylenes in the rat. Inhal Toxicol 2:295-319.
Gearhart, JM; Mahle, DA; Greene, RJ; et al. (1993) Variability of physiologically based pharmacokinetic (PBPK)
model parameters and their effect on PBPK model predictions in a risk assessment for perchloroethylene (PCE).
Toxicol Lett 68:131-144.
Gehring, PJ; Watanabe, PG; Park, CN. (1978) Resolution of dose-response toxicity data for chemicals requiring
metabolic activation: example - vinyl chloride. Toxicol Appl Pharmacol 44:581-591.
Groth, DH; Coate, WB; Thornburg, LP. (1981) Effects of aging on the induction of angiosarcoma. Environ Health
Perspect 41:53-57.
Guengerich, FP; Mason, PS; Stott, WT; et al. (1981) Roles of 2-haloethylene oxides and 2-haloacetaldehydes
derived from vinyl bromide and vinyl chloride in irreversible binding to protein and DNA. Cancer Res 41:4391-
4398.
Hong CB; Winston JM; Thornburg LP et al. (1981) Follow-up study on the carcinogenicity of vinyl chloride and
vinyllidene chloride in rats and mice: tumor incidence and mortality subsequent to exposure. J Toxicol Environ
Health 7(6):909-924.
Jedrychowski, RA; Sokal, JA; Chmielnicka, J. (1985) Comparison of the impact of continuous and intermittent
exposure to vinyl chloride, including phenobarbital effects. J Hyg Epidemiol Microbiol Immunol 28:111-120.
Jones, RW; Smith, DM; Thomas, PG. (1988) A mortality study of vinyl chloride monomer workers employed in the
United Kingdom in 1940-1974. Scand J Work Environ Health 14:153-160.
Kedderis, GL; Carfagna, MA; Held, SD; et al. (1993) Kinetic analysis of furan biotransformation by F-344 rats in
vivo and in vitro. Toxicol Appl Pharmacol 123:274-282.
Laib, RJ; Bolt, HM. (1980) Trans-membrane alkylation: a new method for studying irreversible binding of reactive
metabolites to nucleic acids. Biochem Pharmacol 29:449-452.
Laib, RJ; Bolt, HM; Carrier, R; et al. (1989) Increased alkylation of liver DNA and cell turnover in young versus
old rats exposed to vinyl chloride correlates with cancer susceptibility. Toxicol Lett 45:231-239.
B-52
-------
Lee, CC; Bhandari JC; Winston JM. (1977) Inhalation toxicity of vinyl chloride and vinyl chloride. Environ Health
Perspect 21:25-32.
Lindstedt, S. (1992) Personal communication ~ Draft report to the ILSI RSI Physiological Parameters Working
Group.
Maltoni, C; Lefemine, G; Ciliberti, A; et al. (1981) Carcinogenicity bioassay of vinyl chloride monomer: a model of
risk assessment on an experimental basis. Environ Health Perspect 41:3-29.
Newberne, PM; Weigert, J; Kula, N. (1979) Effects of dietary fat on hepatic mixed function oxidases and
hepatocellular carcinoma induced by aflatoxin B{ in rats. Cancer Res 39:3986-3991.
Ottenwalder, H; Bolt, HM. (1980) Metabolic activation of vinyl chloride and vinyl bromide by isolated hepatocytes
and hepatic sinusoidal cells. J Environ Pathol Toxicol 4:411-417.
Radike MJ; Stemmer KL; Bingham E. (1981) Effect of ethanol on vinyl chloride carcinogenesis. Environ Health
Perspect 41:59-62.
Reitz, RH; Mendrala, AL; Guengerich, FP. (1989) In vitro metabolism of methylene chloride in human and animal
tissues: use in physiologically-based pharmacokinetic models. Toxicol Appl Pharmacol 97:230-246.
Reitz, RH; Gargas, ML; Anderson, ME; et al. (1996) Predicting cancer risk from vinyl chloride exposure with a
physiologically based pharmacokinetic model. Toxicol Appl Pharmacol 136:1-16.
Sabadie, N; Malaveille, C; Carmus, A-M; et al. (1980) Comparison of the hydroxylation of benzo(a)pyrene with the
metabolism of vinyl chloride, TV-nitrosomorpholine, and 7V-nitroso-W-methyrpiperazine to mutagens by human and
rat liver microsomal fractions. Cancer Res 40(1): 119-126.
Simonato, L; L'Abbe, KA; Andersen, A; et al. (1991) A collaborative study of cancer incidence and mortality
among vinyl chloride workers. Scand J Work Environ Health 17:159-169.
U.S. EPA. (1987) Update to the health assessment document and addendum for dichloromethane (methylene
chloride): pharmacokinetics, mechanism of action, and epidemiology. External Review Draft. EPA/600/8-87/030A.
U.S. EPA. (1988) Reference physiological parameters in pharmacokinetic modeling. EPA/600/6-88/004. Office of
Health and Environmental Assessment, Washington, DC.
U.S. EPA. (1992) Request for comments on draft report of cross-species scaling factor for cancer risk assessment.
Fed. Reg. 57:24152.
Watanabe, PG; Hefner, RE, Jr.; Gehring, PJ. (1976a) Vinyl chloride-induced depression of hepatic non-protein
sulfhydryl content and effects on bromosulphthalein (BSP) clearance in rats. Toxicology 6:1-8.
Watanabe, PG; Gehring, PJ. (1976b) Dose-dependent fate of vinyl chloride and its possible relationship to
oncogenicity in rats. Environ Health Perspect 17:145-152.
B-53
-------
APPENDIX C. VINYL CHLORIDE PBPK MODEL CODE
(ACSL VERSION: VCPBPK.CSL)
PROGRAM VCPBPK.CSL - Vinyl Chloride Risk Assessment Model
INITIAL
CAT - BODY WEIGHT
CONSTANT BW - 70 Body Weight (kg)
ENDCAT
CAT - SPECIAL FLOW RATES
CONSTANT QPC-24 Unsealed Alveolar Vent
CONSTANT QCC-16.5 Unsealed Cardiac Output
CAT - FRACTIONAL BLOOD FLOWS TO TISSUES
CONSTANT QLC - 0.24 Flow to Liver as % Cardiac Output
CONSTANT QFC - 0.05 Flow to Fat as % Cardiac Output
CONSTANT QSC - 0.19 Flow to Slow as % Cardiac Output
CONSTANT QRC -0.52 Flow to Rapid as % Cardiac Output
ENDCAT
CAT - FRACTIONAL VOLUMES OF TISSUES
CONSTANT VLC - 0.04 Volume Liver as % Body Weight
CONSTANT VFC -0.19 Volume Fat as % Body Weight
CONSTANT VRC - 0.05 Volume Rapid Perfused as % Body Weight
CONSTANT VSC-0.63 Volume Slow Perfused as % Body Weight
ENDCAT
CAT - PARTITION COEFFICIENTS - GARGAS ET AL. (1989)
CONSTANT PL - 0.95 Liver/Blood Partition Coefficient
CONSTANT PF - 11.9 Fat/Blood Partition Coefficient
CONSTANT PS-1.25 Slow/Blood Partition Coefficient
CONSTANT PR - 0.95 Rapid/Blood Partition Coefficient
CONSTANT PB-1.68 Blood/Air Partition Coefficient
ENDCAT
CAT - KINETIC CONSTANTS
CONSTANT MW - 62.5 Molecular weight (g/mol)
CONSTANT KA - 3.0 Oral uptake rate (/hi)
CONSTANT VMAX1C - 4.0 Scaled Vmax for 1st Saturated Pathway
CONSTANT KM1 -0.1 Km for 1st Saturated Pathway
CONSTANT VMAX2C - 0.0 Scaled Vmax for 2nd Saturated Pathway
CONSTANT KM2 - 10.0 Km for 2nd Saturated Pathway
ENDCAT
C-l
-------
CAT - DOSING INFORMATION
CONSTANT
CONSTANT
CONSTANT
CONSTANT
ENDCAT
PDOSE-0.1
DRINK - 0.0
CONC - 100.0
TCHNG - 6.0
Oral dose (mg/kg)
Dose (mg/kg/day) in H2O
Inhaled concentration (ppm)
CAT - GSH PARAMETERS GROUP 1
CONSTANT KGSMC-0.13
CONSTANT KFEEC -35.0
CONSTANT KCO2C - 1.6
ENDCAT
CAT - GSH PARAMETERS GROUP 2
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
ENDCAT
KOC-28.5
KBC-0.12
KS - 2000.0
GSO - 5800.0
H2O - 55.0
Conjugated rate constant with metabolite
Conjugated rate constant with non-GSH
First-order CEO breakdown to CO9
Zero-order production of GSH
First-order rate constant for GSH breakdown
Constant controlling resynthesis
Initial GSH concentration
Moles of H9O
CAT - SIMULATION LENGTH CONTROL
CONSTANT TSTOP - 24.0
CONSTANT POINTS - 1.0
CONSTANT H-10000.0
ENDCAT
Set initial values
IF (PDOSE.EQ.0.0) KA - 0.0 Parenteral dosing
Scaled parameters
CINT - TSTOP / POINTS
NSTP - CINT*H + 1
QC - QCC*BW**0.75
QP - QPC*BW**0.75
QL - QLC*QC
QF - QFC*QC
QS-QSC*QC
QR - QRC*QC
QC - QL + QF + QS + QR
VL - VLC*BW
VF - VFC*BW
VS - VSC*BW
Cardiac output
Alveolar ventilation
Liver blood flow
Fat blood flow
Slowly perfused tissue blood flow
Richly perfused tissue blood flow
Liver volume
Fat tissue volume
Slowly perfused tissue volume
C-2
-------
VR-VRC*BW Richly perfused tissue volume
GSO - VLC*BW*GSO Initial amount of GSH
KGSM - KGSMC/BW**0.25 Reaction with GSH
KFEE - KFEEC/BW**0.25 Reaction with other tissues
KO - KOC*BW**0.75 Zero-order GSH production
KB - KBC/BW**0.25 Normal GSH turnover
KCO2 - KCO2C/BW**0.25 Production of CO2
VMAX1 - VMAX1C*BW**0.75 Maximum rate of metabolism
VMAX2 - VMAX2C*BW**0.75 Maximum rate of metabolism
VMAX1M - VMAX1C*BW**0.75*1000.0/MW
VMAX2M - VMAX2C*BW**0.75*1000.0/MW
DOSE - PDOSE*BW
KZER - DRINK/24.0 * BW
CIX - CONC*MW/24450.0
WADDF - 5.0/7.0
IF(BW.LT.0.1)THEN
LADDF - WADDF* (30.0/104.0) Mice
ELSE IF (BW.GT.1.0) THEN
LADDF-1.0 Humans
ELSE IF (DRINK.GT.O.O)THEN
LADDF -1.0 Drinking Water
ELSE IF (CONC.GT.30.0) THEN
LADDF - WADDF*(52.0/147.0) Hi
ELSE IF (CONC.GT.0.0) THEN
LADDF - WADDF*(52.0/135.0) Low
ELSE
LADDF - WADDF*(52.0/136.0) Gavage
ENDIF
END END OF INITIAL
DYNAMIC
ALGORITHM IALG - 2
DERIVATIVE
Concentration in Arterial Blood (mg/L)
(Algebraic Solution for CA after gas exchange)
CI - CIX* (1.0 - STEP/TCHNG)
CA - (QC*CV + QP*CI)/(QC + QP/PB)
AUCB - INTEG (CA.0.0)
Amount Exhaled (mg)
CX - CA/PB
CALPPM - CX*24450.0/MW
C-3
-------
CXPPM - (0.7*CX + 0.3*CI)*24450.0/MW
RAX - QP*CX
AX - INTEG (RAX.0.0)
Amount in Liver Compartment (mg)
RAL - QL*(CA-CVL) - RAM + RAO + KZER
AL- INTEG (RAL.0.0)
CVL - AL/(VL*PL)
CL - AL/VL
AUCL - INTEG (CL, 0.0)
Amounts Metabolized in Liver
RAM - VMAX1*CVL/(KM1 + CVL) + VMAX2*CVL/(KM2 + CVL)
AM - INTEG (RAM,0.0)
RISK - AM/VL
RISKT - LADDF*RISK
AMP-AM*1000./MW
RAMP - RAM*1000.0/MW
Amount in Slowly Perfused Tissues (mg)
RAS - QS* (CA - CVS)
AS-INTEC(..AS,0.0)
CVS-AS/( *PS)
CS - AS/VS
Amount in Rapidly Perfused Tissues (mg)
RAR - QR*(CA - CVR)
AR - INTEG (RAR,0.0)
CVR - AR/(VR*PR)
CR-AR/VR
Mixed Venous Blood Concentration (mg/L)
CV - (QF*CVF + QL*CVL + QS*CVS + QR*CVR)/QC
Amount in Fat Compartment (mg)
RAF - QF* (CA - CVF)
AF - INTEG (RAF,0.0)
CVF - AF/(VF*PF)
CF - AF/VF
Total Mass Input from Stomach (mg)
RAO -KA*MR
AO - DOSE-MR
Amount Remaining in Stomach (mg)
C-4
-------
RMR - KA*MR
MR - DOSE*EXP(-KA*T)
Amount of Oxidative Metabolite (uMoles)
RAMM - (VMAX1M*CVL)/(KM1+CVL) + (VMAX2M*CVL)/(KM2 + CVL)
RACMG - RACMEE - RACO2
AMM - INTEG (RAMM,0.0)
CMM-AMM/VL
Glutathione (uMoles)
RAMGSH - KO*(KS+GSO)/(KS+GSH)-KB*GSH*VL-RACMG
AMGSH - INTEG (RAMGSH,GSO)
GSH - AMGSH/VL
GSHP - (AMGSH/GSO)*100
Amount Metabolite Conjugated With GSH (uMoles)
RACMG - KGSM*GSH*CMM*VL
ACMG - INTEG (RACMG,0.0)
RISKG - ACMG/VL
RISKR - LADDF*RISKG
Amount Metabolite Conjugated With Other Things (uMoles)
RACMEE - KFEE*CMM*VL
ACMEE - INTEG (RACMEE,0.0)
RISKM - ACMEE/VL
RISKN - LADDF*RISKM
Amount of CO2 (uMoles)
RAC02 - KCO2*CMM*H2O*VL
AC02 - INTEG (RACO2,0.)
Total Intake of Vinyl Chloride (mg)
AMET-AM/BW
TERMT (T.GE.TSTOP)
END END OF DERIVATIVE
END END OF DYNAMIC
END END OF PROGRAM
C-5
-------
APPENDIX D. THE APPLICATION OF A PBPK MODEL FOR VINYL CHLORIDE
IN A NONCANCER RISK ASSESSMENT
This appendix discusses the application of the PBPK model described in Appendix A for
noncancer risk assessment. A comparison for calculation of human equivalent concentrations
(HECs) is made between the agency default strategy (U.S. EPA, 1994) and the dose metrics
"RISK" and "AMET" calculated from the PBPK model. Applications of the PBPK model using
both the NOAEL/LOAEL approach and the benchmark concentration/dose (BMC/D) modeling
approach also are described, and the results of benchmark modeling for several endpoints
considered.
D.I. SELECTION OF A NONCANCER RISK ASSESSMENT APPROACH
As discussed in Section 4.4 of the main document, evidence is strong that the
carcinogen!city and liver toxicity of VC are related to the production of reactive metabolic
intermediates. The most appropriate pharmacokinetic dose metric for a reactive metabolite is the
total amount of the metabolite generated divided by the volume of the tissue in which it is
produced (Andersen et al., 1987). It has been demonstrated in the case of VC that binding to
liver macromolecules following inhalation exposure of rats correlates well with total metabolism
rather than exposure concentration (Watanabe et al., 1978). Therefore, the most reasonable dose
metric for liver toxicity would be the total amount of metabolism divided by the volume of the
liver. This dose metric, referred to in the PBPK model as "RISK," will be used for evaluating
the dose response for increased liver/body weight ratio and liver nonneoplastic effects.
In the case of toxicity to the testes, as observed by Bi et al. (1985) and Sokal et al.
(1980), the most appropriate dose metric is less certain. However, toxicity from locally
generated reactive metabolites is a reasonable mechanism for an organ for which there is
evidence of P450 activity, such as the testes. The most appropriate dose metric in this case,
analogous to the case of the liver, would be the total amount of metabolism in the testes divided
by the volume of the testes. Unfortunately, there is not adequate information on P450 activity
for VC in the testes to support this approach; therefore a surrogate must be used. If it is assumed
(1) that P450 metabolism in the testes and P450 metabolism in the liver scale across species in
the same way (that is, the proportion of metabolism between the testes and liver is constant) and
(2) that the relative proportion of body weight associated with the testes is the same across
species, then the total amount of metabolism divided by body weight can be used as the
surrogate dose metric for testicular toxicity and is designated as "AMET" in this assessment.
D-l
-------
D.2. COMPARISON OF NONCANCER RISK ASSESSMENTS FOR VC
The Clewell model was used to calculate the pharmacokinetic dose metrics for the most
informative of the animal studies (Bi et al., 1985; Sokal et al., 1980; Wisniewska-Knypl et al.,
1980; Torkelson et al., 1961; Feron et al., 1981; Til et al., 1983, 1991). Because the model
calculates delivered dose at the target tissue, an oral study (Til et al., 1983, 1991) could be
modeled and converted to exposure concentrations without the need for additional route-to-route
extrapolation. Because a high-quality chronic oral study, but no chronic inhalation study, was
available, no attempt was made to conduct a route-to-route extrapolation for oral exposure. The
studies and calculated dose metrics for the endpoints of interest are shown in Table D-l.
D.2.1. Calculation of NOAELHEC by RfC Default Procedures; a First Approximation of a
Human Equivalent Concentration
In derivation of the lifetime estimate of a safe concentration, the RfC, the Agency
employs various default strategies for interspecies extrapolation, i.e., to convert experimental
inhaled exposures in animals to corresponding exposures in humans, termed human equivalent
concentrations (HEC). The particular strategy employed depends both on the character of a gas
and on its effect in biological systems (U.S. EPA, 1994). VC is a water-insoluble lipophilic gas
whose distribution to the body is limited by the amount of blood flowing past VC-laden lungs
into which the gas may partition (i.e., perfusion limited). Effects produced by inhaled or
ingested VC are in the liver, i.e., VC is a systemic toxicant. The default strategy for predicting
an HEC from a gas affecting systemic endpoints is dual-stepped. First, the experimental
concentrations in the animal are adjusted to what they would have been if that same exposure
had been administered in a continuous manner. For example, an exposure of 100 ppm
administered for 6 hours/day would be adjusted to 25 ppm (100 ppm x 6/24 hrs). The number of
days exposed in a week is also taken into account in a parallel manner. Second, a primary
determinant of systemic concentration is considered, in this case the ratio of the blood/air
partition coefficients (• ) in the human and the animal. If the partition coefficient is larger in
animals than in humans (which is the case for VC as shown in Table A-l; PB = 2.26 for mice,
2.4 for rat, and only 1.16 for humans) then a factor of 1 is applied to the time-adjusted
concentration described above to obtain the human equivalent concentration (HEC). These
values are listed in Table D-l under "RfC Default." This default procedure is a reasonable first
approximation to the expected interspecies relationship of exposures to a volatile, lipophilic
chemical such as VC, because in an inhalation exposure, the average blood concentration of such
a chemical during the exposure will be proportional to the air concentration multiplied by the
blood/air partition coefficient.
The above strategy for HEC calculation is for inhaled exposures. The current default
approach for predicting human equivalent exposures from oral exposure to obtain the oral
equivalent of the RfC, the RfD, is to assume equivalent doses for animals and humans on a
mg/kg-day basis.
D-2
-------
Table D-l. Modeled dose metric values RISK (mg reactive metabolites/L liver) and AMET (mg reactive
metabolite/BW)
Reference
Modeled
human
exposure
Bi et al.
(1984)
Sokal et al.
(1980)
Torkelson et
al. (1961)
Route
Inhalation
Drinking
water
Inhalation
Inhalation
Inhalation
Species
Human
Wistar rats (M)
Wistar rats (M)
Wistar rats (M)
Exposure
duration
Continuous
6 hr/d, 6 d/wkc
5 hr/d, 5 d/wk
for 10 months
0.5 hrs/d, 5 d/wk
for 6 months
1 hr/day, 5 d/wk
for 6 months
Experimental
concentration
or dose
1 ppm (2.6
mg/m3
0.0286 mg/kg-
day (Ippm)
0 mg/m3
25.6 mg/m3
256 mg/m3
7,670 mg/m3
0 mg/m3
128 mg/m3
1,280 mg/m3
51, 140 mg/m3
256 mg/m3
511 mg/m3
128 mg/m3
256 mg/m3
511 mg/m3
Human
equivalent
concentration
RfC default0
0 mg/m3
5.5 mg/m3
55 mg/m3
1,654 mg/m3
0 mg/m3
19 mg/m3
190 mg/m3
7,610 mg/m3
3.8 mg/m3
7.6 mg/m3
3.8 mg/m3
7.6 mg/m3
15.2 mg/m3
Modeled daily dose
metrics3
RISK
38
364
1,260
156
779
1,300
25
43.3
26.2
48.8
83
AMET
1.52
14.6
50.4
6.24
31.2
51.9
0.99
1.73
1.05
1.95
Average daily dose
metric1"
RISK
3.029
1.010
32.5
312
1,080
111
556
927
17.8
30.9
18.7
34.8
59.3
AMET
0.0787
0.02625
1.3
12.5
43.2
4.46
22.3
37.1
0.71
1.24
0.75
1.39
2.37
o
oo
-------
Table D-l. Modeled dose metric values RISK (mg reactive metabolites/L liver) and AMET (mg reactive
metabolite/BW) (continued)
Reference
Torkelson et
al. (1961)
continued
Route
Inhalation
Inhalation
Species
Rats (M)
Rats (M)
Exposure
duration
2 hr/day, 5 d/wk
for 6 months
7 hrs/d, 5 d/wk
for 6 months for
4.5 months
7 hrs/d, 5 d/wk
for 6 months for
4.5 months
Experimental
concentration
or dose
128 mg/m3
256 mg/m3
511 mg/m3
128 mg/m3
256 mg/m3
511 mg/m3
1,280 mg/m3
128 mg/m3
256 mg/m3
511 mg/m3
1,280 mg/m3
Human
equivalent
concentration
RfC default0
7.6 mg/m3
15.2 mg/m3
30.4 mg/m3
26.7 mg/m3
53.3 mg/m3
106.5 mg/m3
267 mg/m3
26. 7 mg/m3
53.3 mg/m3
106.5 mg/m3
267 mg/m3
Modeled daily dose
metrics3
RISK
53.4
101
163
183
343
550
691
1,241
327
449
524
AMET
2.14
4.03
6.51
7.31
13.7
22
27.6
6.15
13.1
17.9
21
Average daily dose
metric1"
RISK
38.1
72
116
131
245
393
493
110
234
320
375
AMET
1.53
2.88
4.65
5.22
9.79
15.7
19.7
4.39
9.35
12.8
15
-------
Table D-l. Modeled dose metric values RISK (mg reactive metabolites/L liver) and AMET (mg reactive
metabolite/BW) (continued)
Reference
Wisniewska-
Knypl
Route
Inhalation
Species
Wistar rats (M)
Exposure
duration
5 hr/day, 5 d/wk
for 1 month
5 hr/d, 5 d/wk
for 3 months
5 hr/d, 5 d/wk
for 6 months
Experimental
concentration
or dose
0 mg/m3
128 mg/m3
1,280 mg/m3
51, 100 mg/m3
0 mg/m3
128 mg/m3
1,280 mg/m3
51, 100 mg/m3
0 mg/m3
128 mg/m3
1,280 mg/m3
51,100 mg/m3
Human
equivalent
concentration
RfC default0
0 mg/m3
19 mg/m3
190 mg/m3
7,604 mg/m3
0 mg/m3
19 mg/m3
190 mg/m3
7,604 mg/m3
0 mg/m3
19 mg/m3
190 mg/m3
7,604 mg/m3
Modeled daily dose
metrics3
RISK
144
730
1,230
140
690
1,170
135
704
1,150
AMET
5.75
29.2
49
5.6
27.6
46.8
5.4
28.2
46
Average daily dose
metric1"
RISK
103
521
876
100
493
835
96.4
503
821
AMET
4.11
20.9
35
4
19.7
33.4
3.86
20.1
32.9
-------
Table D-l. Modeled dose metric values RISK (mg reactive metabolites/L liver) and AMET (mg reactive
metabolite/BW) (continued)
Reference
Feron et al.
(1981)
Route
Diet
Species
Wistar rats (M)
Wistar rats (F)
Exposure
duration
5 hr/day, 5 d/wk
for 10 months
135 wks
149 wks
Experimental
concentration
or dose
0 mg/m3
128 mg/m3
1,280 mg/m3
51, 100 mg/m3
0 mg/kg-d
1.7 mg/kg-d
5.0 mg/kg-d
14.1 mg/kg-d
0 mg/kg-d
.014 mg/kg-d
0.1 3 mg/kg-d
1.3 mg/kg-d
Human
equivalent
concentration
RfC default0
0 mg/m3
19 mg/m3
190 mg/m3
7,604 mg/m3
Modeled daily dose
metrics3
RISK
130
649
1,130
39.5
116.1
326
0.318
2.958
29.6
AMET
5.21
26
45.3
1.58
4.64
13
0.013
0.118
1.19
Average daily dose
metric
RISK
93
463
809
39.5
116.1
326
0.318
2.958
29.6
AMET
3.72
18.5
32.4
1.58
4.64
13
0.013
0.118
1.19
a The RfC default (or first approximation) of a human equivalent concentration (HEC) was calculated by first adjustin the experimental concentration to a
continuous, 24 hr exposure and then applying the ratio of animal to human blood/air partition coefficient (•) in accordance with U.S. EPA, 1994, which is 1 in
this instance. For example, the 6 hr exposures of Bi; 25.6 mg/m3 x (6 hr/24 hr) x 6/7 days/week x 1/1 • = 5.5 mg/m3.
b For a given exposure scenario the model calculates a daily dose metric (a concentration) to the liver in the units of mg metabolite/L liver.
0 The daily does metric is converted to an average daily dose metric by expressing the daily dose metric in terms of 7 days/week. For the first daily dose metric
from Bi et al. (1984), the daily dose metric RISK of 38 is factored by 6/7 days = average daily dose of 32.5 mg metabolite /L liver.
-------
Using the default RfD and RfC strategies described above, the candidate NOAELs and LOAELs,
and the corresponding NOAEL^q and LOAEL^q values were determined and are listed
below. Note that the same nominal NOAEL or LOAEL concentration can correspond to
different duration-adjusted values for different inhalation studies, because of different exposure
protocols.
Oral Studies
Tiletal. (1983, 1991)
Increased incidence of hepatic cysts and of liver cell polymorphisms graded moderate
and severe:
NOAEL = 0.13 mg/kg-day; LOAEL = 1.3 mg/kg-day
Feronetal. (1981)
Increased extensive liver necrosis:
In females, NOAEL = 1.7 mg/kg-day; LOAEL = 5.0 mg/kg-day
In males, NOAEL = 5.0 mg/kg-day; LOAEL = 14.1 mg/kg-day
Inhalation Studies
Bietal. (1985)
Increased relative liver weight:
NOAEL, none
LOAEL = 25.6 mg/m3; LOAEL[HEC] = 5.5 mg/m3
Increased testicular degeneration:
NOAEL = 25.6 mg/m3; NOAEL[HEC] = 5.5 mg/m3
LOAEL = 256 mg/m3; LOAELmc] = 55 mg/m3
Sokaletal. (1980)
Increased relative liver weight and liver lesions (nuclear polymorphism of hepatocytes,
proliferation of reticuloendothelial cells):
NOAEL = 128 mg/m3; NOAEL^q = 19 mg/m3
LOAEL = 1,278 mg/m3; LOAEL[HEC] = 190 mg/m3
Increased damage of spermatogenic epithelium:
NOAEL = 128 mg/m3; NOAEL^q = 19 mg/m3
LOAEL = 1,278 mg/m3; LOAEL[HEC] = 190 mg/m3
• Torkelson et al. (1961) (6-month point)
Increased relative liver weight:
NOAEL = 128 mg/m3; NOAELpuq = 26.6 mg/m3
LOAEL = 256 mg/m3; LOAEE^^ = 53.3 mg/m3
• Wisniewska-Knypl et al. (1980) (3-month point)
Lipid accumulation:
LOAEL = 128 mg/m3; LOAEL^^ = 19 mg/m3
D-7
-------
The principal study for the RfD is the chronic study of Til et al. (1983, 1991) in which
the oral animal dose of 0.13 mg/kg-day was a NOAEL with liver cysts and liver cell
polymorphism occurring at the highest dose of the study, 1.3 mg/kg-day. The results of this
study are clearly more sensitive than those reported by Feron et al. (1981).
Without the PBPK modeling making route-to-route extrapolation feasible, the principal
study for the RfC would probably be that of Bi et al. (1985), in which no NOAEL was identified
and the lowest exposure level (10 ppm, or 25.6 mg/m3) was identified as a LOAEL for increased
liver/body weight ratio after a 6-month exposure. Adjusting from exposure for 6 hr/day, 6
days/wk to continuous exposure yields an adjusted LOAEL of 5.5 mg/m3. As shown in Table A-
1 of Appendix A, the measured blood/air partition coefficients (shown as PB in the table, but
referred to in the RfC process as •) in the rat and the human are 2.4 and 1.16, respectively.
Because the ratio of the animal-to-human partition coefficients (2.4/1.14 = 2.1) is greater than 1,
a default value of 1 is used in accordance with EPA guidance (U.S. EPA, 1994). Therefore, the
resulting LOAEL^q is the same as the duration-adjusted value, 5.5 mg/m3.
D.2.2. Calculation of the NOAELHEC/D Using a PBPK (Physiologically Based
Pharmacological Kinetic) Modeling Approach
The default approach can be readily adapted to the use of target tissue dose using the
PBPK model. The principal study and NOAEL/LOAEL are selected in the same way, except
that the exposure is characterized by the target tissue dose from the PBPK model. As discussed
in the earlier section on the selection of the noncancer risk assessment approach, liver toxicity is
assumed to result from reactive species generated during metabolism and is modeled using the
"RISK" dose metric, which is based on the total amount of metabolism divided by the liver
volume (mg metabolites/L liver). For testicular effects, toxicity is assumed to result from locally
generated reactive metabolites, and the dose metric "AMET" with units of mg/kg is used, which
is based on the total amount of metabolism divided by body weight. To obtain an average daily
dose metric that is equivalent to the adjusted concentration in the traditional approach, the PBPK
model is run for 24 hours and the resulting daily dose metric is then adjusted by the number of
days per week the exposure took place. An alternative approach for chemicals that are not as
rapidly cleared as VC is to run the model for several weeks (simulating both exposure days and
nonexposure days), until steady state is reached, and then divide the weekly increase in the dose
metric by 7 to obtain the average daily value. Unlike the default approach, however, no
adjustment is necessary for the number of hours of exposure per day, because the model
incorporates this information into the prediction of the daily dose metric. Consideration of the
average daily dose metric values associated with statistically significant responses indicates
several candidate NOAELs and LOAELs as follows:
Oral Studies
Tiletal. (1983, 1991)
Increased incidence of hepatic cysts and of liver cell polymorphisms graded moderate
and severe:
NOAEL at RISK = 2.96 mg/L; LOAEL at RISK = 29.6 mg/L in females
NOAEL at RISK = 3.03 mg/L; LOAEL at RISK = 30.2 mg/L in males
D-8
-------
Feronetal. (1981)
Increased extensive liver necrosis:
NOAEL at RISK = 38.6 mg/L; LOAEL at RISK = 113 mg/L in females
NOAEL at RISK =116 mg/L; LOAEL at RISK = 326 mg/L in males
Inhalation Studies
Bietal. (1985)
Increased relative liver weight: no NOAEL; LOAEL at RISK = 32.5 mg/L
Increased testicular degeneration: NOAEL at AMET = 1.30 mg/kg; LOAEL at AMET =
12.5 mg/kg
Sokaletal. (1980)
Increased relative liver weight and liver lesions (nuclear polymorphism of hepatocytes,
proliferation of reticuloendothelial cells):
NOAEL at RISK =111 mg/L; LOAEL at RISK = 556 mg/L
Increased damage of spermatogenic epithelium:
NOAEL at AMET = 4.46 mg/kg; LOAEL at AMET = 22.3 mg/kg
Torkelson etal. (1961)
Increased relative liver weight (at 6-month point):
NOAEL at RISK =110 mg/L; LOAEL at RISK = 234 mg/L in females
NOAEL at RISK =131 mg/L; LOAEL at RISK = 245 mg/L in males
• Wisniewska-Knypl et al. (1980) (at 3-month point)
Lipid accumulation: no NOAEL; LOAEL at RISK = 93 mg/L.
This analysis is codified in Table D-2, in which are included the dose metrics calculated
for the reproductive studies of Short et al. (1986) and CMA (1998b). Consideration of either
AMET or RISK shows clearly the sensitivity of the liver endpoint in the Til et al. (1983, 1991)
study when compared with other studies, either inhalation or oral, or other endpoints, either
testicular or reproductive effects.
To convert these dose metrics into an HEC, the PBPK model must be run to determine
the continuous human exposure associated with each dose metric value of RISK and/or AMET.
Table D-3 shows the results of this exercise, where the dose metrics associated with human
continuous exposures range from 1 • g/m3 through 10,000 mg/m3. These results show that in the
case of VC the model is linear to nearly 100 mg/m3. This simplifies the calculation of HECs,
because the appropriate equivalence factor can thus be used: 1.18 (mg/L)/(mg/m3) for RISK, or
0.0308 (mg/kg)/(mg/m3) for AMET. These are reported as human equivalent concentrations, or
HEC. Similarly, the equivalence factor for oral dosing was calculated by determining the human
D-9
-------
Table D-2. Dose metrics (AMET and RISK) derived for oral and inhalation
exposure scenarios using a PBPK model (Clewell et al., 1995b) compared with
effects observed in various studies
Reference
Til etal. (1991)
Til etal. (1991)
Til etal. (1991)
CMA(1998b)
Bi etal. (1985)
Feronetal. (1981)
Wisniewska-Knypl (1980)
Sokal etal. (1980)
Feronetal. (1981)
Short etal. (1977)
Torkelson(1961)
CMA(1998a)
Feronetal. (1981)
Torkelson(1961)
Bi etal. (1985)
Short etal. (1986)
Sokal etal. (1980)
Short etal. (1986)
Sokal etal. (1980)
Bi etal. (1985)
Dose metrics
AMET"
0.013
0.12
1.2
1.3
1.3
1.6
3.7
4.5
4.6
6
9.3
12
13
14
15
21
22
32
37
43
RISKC
0.3
3
30
32
33
39
93
111
116
156
234
298
326
343
364
534
556
800
927
1,080
Effects3
Liver
—
—
+
—
—
+
+
—
+
NE
—
+
+
—
+
+
NE
NE
+
+
Testicular
—
—
—
—
—
—
NE
—
—
NE
—
—
—
—
+
+
+
NE
+
+
Reproductive
NE
NE
NE
—
NE
NE
NE
NE
NE
NE
—
NE
NE
+
+
NE
+
NE
NE
a— = NOAE (No Observed Adverse Effect), + = OAE (Observed Adverse Effect), NE = not examined for.
bAMET: Total amount of VC metabolism divided by body weight (average daily mg metabolite/kg-day).
Male values given.
°RISK: Total amount of VC metabolized by liver divided by volume of liver (average daily mg metabolite/L liver).
Male values given.
D-10
-------
Table D-3. Daily dose metrics (RISK and AMET) obtained by running the
PBPK model (Clewell et al., 1995a) under conditions of continuous human
exposure
Concentration (mg/m3)
0.001
0.01
0.1
1
10
100
1,000
10,000
Dose metric"
RISK
1.18 x ID'3
1.18 x ID'2
0.118
1.18
11.85
117.6
954
1,264
AMET
3.08 x 1Q-5
3.08 x 1Q-4
3.08 x 1Q-3
3.08 x 1Q-2
0.308
3.06
24.8
32.9
a These values reflect dose metrics calculated using KM1=0.1, rather than a value of 1.0 used in the Clewell et al. (1995a)
model. At the suggestion of the external peer review, the human dose metrics based on a KM1 of 0.1 were used for
calculations in the main text.
dose metric corresponding to a sample near-continuous exposure scenario (1 ppm in water,
corresponding to 0.0286 mg/kg-day, assuming ingestion of 2 L/day by a 70 kg person) yielded a
dose metric of 1.010 as shown in Table D-l. These are reported as the human equivalent dose, or
HED. The corresponding values for "AMET" are 0.0308 mg/kg for 1 mg/m3 for inhaled VC
and 0.92 mg/kg for 1 mg/kg-day. As with inhalation exposure, VC metabolism is linear in this
dose range, as per Table D-3 where a RISK value of 100 would correspond to a continuous oral
intake in humans of about 3 mg/kg-day (RISK - 35.31) so the equivalence factor for RISK is
(1.01/0.0286) = 35.31 (mg/L)/(mg/kg-day). Liver toxicity was the only endpoint of concern for
oral exposure, so equivalence factors for the other dose metrics were not calculated. To obtain
the HED or HEC for each animal NOAEL or LOAEL, the animal dose metric was divided by the
human dose metric equivalence factor for RISK:
RISK (mg/L liver) -35.31= oral HED (mg/kg-day)
RISK (mg/L liver) - 1.18 = inhalation HEC (mg/m3)
AMET (mg/kg) - 0.92 = oral HED (mg/kg-day)
AMET (mg/kg) + 0.0308 = inhalation HEC (mg/m3)
Projections based on "RISK" values greater than 100 or "AMET" values greater than 3
are in the range of nonlinearity and are therefore minor overestimates of dose or concentration
(Table D-3). They are provided for comparative purposes only.
D-ll
-------
Oral Studies
Tiletal. (1983, 1991)
Increased incidence of hepatic cysts and of liver cell polymorphisms graded moderate
and severe:
Females
NOAEL[HED/C] at 2.96 mg/L liver =0.08 mg/kg-day or 2.5 mg/m3
LOAELpjED/C] at 29.5 mg/L liver = 0.8 mg/kg-day or 25 mg/m3
Males
NOAEL[HED/C] at 3.03 mg/L liver =0.09 mg/kg-day or 2.6 mg/m3
LOAELpjED/C] at 30.2 mg/L liver = 0.9 mg/kg-day or 26 mg/m3
Feronetal. (1981)
Increased extensive liver necrosis:
Females
NOAEL[HED/C] at 38.6 mg/mL =1.1 mg/kg-day or 33 mg/m3
LOAELpjED/C] at 113 mg/mL = 3.2 mg/kg-day or 97 mg/m3
Males
NOAEL[HED/C] at 116 mg/mL = 3.3 mg/kg-day or 98 mg/m3
LOAELpjED/C] at 326 mg/mL = 9.2 mg/kg-day or 276 mg/m3
Inhalation Studies
Bietal. (1985)
Increased relative liver weight: LOAELpjED/C] at 32.5 mg/L = 0.9 mg/kg-day or 28 mg/m3
Increased testicular degen:
NOAEL[HED/C] at 1.30 mg/kg =1.4 mg/kg-day or 42 mg/m3
LOAELpjED/C] at 12.5 mg/kg =13 mg/kg-day or 400 mg/m3
Sokaletal. (1980)
Increased relative liver weight and liver lesions:
NOAEL[HED/C] at 111 mg/L = 3.1 mg/kg-day or 93 mg/m3
LOAEL at 556 mg/L = 16 mg/kg-day or 470 mg/m3
Increased damage of spermatogenic epithelium:
NOAEL[HED/C] at 4.46 mg/kg = 4.8 mg/kg-day or 145 mg/m3
LOAELpjED/C] at 22.3 mg/kg = 24 mg/kg-day or 700 mg/m3
Torkelson etal. (1961)
Increased relative liver weight:
Females
NOAEL[HED/C] at 110 mg/L = 3.1 mg/kg-day or 93 mg/m3
D-12
-------
at 234 mg/L = 6.6 mg/kg-day or 200 mg/m3
Males
NOAEL[HED/C] at 13 1 mg/L = 3.7 mg/kg-day or 1 10 mg/m3
at 245 mg/L = 7 mg/kg-day or 210 mg/m3
Wisniewska-Knypl et al. (1980)
Hepatic lipid proliferation: LOAELpjED/C] at 93 mg/L = 2.6 mg/kg-day or 80 mg/m3
Summarizing the above results, the lowest LOAEL was for increased relative liver
weight in a subchronic study, just as it was for the traditional approach, but the resulting
LOAELpjEq is 47.8 mg/m3, about ninefold higher than the value of 5.5 mg/m3 arrived at without
considering pharmacokinetics. The reason for the difference is that the default approach is based
on parent chemical exposure, whereas the PBPK approach used a dose metric (RISK)
representing exposure to reactive metabolites. Use of the PBPK model also allows for
extrapolation from the oral route, in which aNOAEL(HEC) of 2.5 mg/m3 (average of male and
female values) was identified. For the oral assessment, a NOAEL of 0.09 mg/kg-day (average of
male and female values) was identified, a dose fairly close to the animal dose of 0.13 mg/kg-day
that would be used as the NOAEL in the absence of the model.
The overall approach just described is actually an approximate method that is acceptable
for VC. The correct approach in general is to apply the desired uncertainty factor to the animal
dose metric to obtain the lower target tissue dose desired in the human; the human PBPK model
is then run iteratively to estimate the concentration associated with the desired human target
tissue dose (Clewell and Jarnot, 1994). However, because of the linearity of the human dose
metric for VC over the region of interest, the two formulations are equivalent in the case of VC.
D.2.3. Benchmark Dose Modeling
When possible, dose-response analysis of the results of the VC studies was also
performed using the benchmark dose (BMD) methodology (Crump, 1984, 1995). When used
with exposure concentrations, this approach is sometimes referred to as the benchmark
concentration (BMC) methodology. The BMD (BMC) is the dose (concentration) predicted to
result in a specified amount of increased risk (called the "benchmark risk"). The BMD or BMC
is calculated using a statistical dose-response model applied to either experimental toxicological
or epidemiological data. It has been proposed that a statistical lower bound on the BMD or
BMC (referred to as the BMDL or BMCL, respectively) may be used in the setting of acceptable
exposure limits as a replacement for the traditional NOAEL, which must be selected from one of
the actual experimental dosing levels (U.S. EPA, 1994; Gaylor and Slikker, 1990).
In the traditional approach for estimating a NOAEL from animal data, the response at
each of the experimental doses is compared statistically with that in the controls, and the
NOAEL is defined as the lowest dose showing no statistical difference. The benchmark
approach has several advantages over the traditional NOAEL approach: (1) the benchmark
D-13
-------
approach makes better use of the dose-response information inherent in the data; (2) the
benchmark approach appropriately reflects the sample size of a study (smaller studies tend to
result in smaller BMDs or BMCs, whereas the opposite is true for traditionally derived
NOAELs); (3) the benchmark approach does not require arbitrary categorization of the data in
epidemiological studies; (4) the benchmark approach does not involve difficult and
argumentative "all or nothing" decisions, such as determining whether or not a NOAEL was
observed in a particular experimental dose or exposure category; and (5) a benchmark estimate
of the NOAEL can be determined even when effects are observed in the lowest experimental
dose group or exposure category. In its report, "Interim Methods for Development of Inhalation
Reference Concentrations" (U. S. EPA, 1994), the EPA stated: "This novel method utilizes more
of the available data than the current methodology .... It also addresses to some degree several
of the criticisms of the current approach, such as the use of dose-response slopes and the number
of animals tested in defining NOELs."
D.2.4. Quantal Benchmark Results
Calculations of BMDs and BMCs for quantal (incidence) data in the present study were
performed with the standard quantal benchmark programs, THRESH and THRESHW (KS
Crump Group, ICF Kaiser International, Ruston, LA), which employ the polynomial and
Weibull models, respectively:
Polynomial model:
P(d) = PO + (1- PO) * (1 • exp{- [• t(d • d0) + • 2(d • d0)2 + ... + • k(d • d0)k]})
Weibull model:
Po + (l-po)*(l- exp{-[-(d- d0)k]})
where p0 is the proportion of responses in the control group, and d0 is a threshold below which
no increase in response is expected to occur.
A key step in the use of BMD modeling for the calculation of RfDs and RfCs is in the
choice of the benchmark response level (BMR). This issue is an area of ongoing research, and
the appropriate choices are better defined for quantal endpoints than for continuous endpoints.
However, the following may be considered. For developmental toxicity, a set of studies
sponsored by EPA (Faustman et al., 1994; Allen et al., 1994a, 1994b; Kavlock et al., 1995) has
indicated choices for the response levels that yield BMDs that are, on average, similar to
corresponding NOAELs. No large-scale studies comparable to those conducted for
developmental toxicity have been completed for other types of toxicity. Thus, it is not as clear
for such endpoints how to define the BMDs. However, Allen et al. (1994a) investigated a
quantal treatment of developmental toxicity endpoints (counting the number of litters per group
with one or more affected fetuses). Such a treatment of developmental toxicity endpoints should
not be much different from any other quantal endpoint. Allen et al. (1994a) determined that a
BMD corresponding to a 10% increase in risk (BMD 10) tended to match the associated
NOAELs better than other choices (5% and 1% increases). About 76% of the BMDs for 10%
D-14
-------
additional risk were less than the corresponding NOAELs, but the median of the relative
differences was a factor of 2. BMDLs corresponding to an additional risk of 10% also have the
advantage that they are likely to depend less on the dose-response model than BMDLs
corresponding to additional risk of 1% or 5% (Crump, 1984). These analyses suggest that use of
a lower bound for 10% additional risk would increase the conservatism in the determination of
RfCs and RfDs by a factor of about 2 to 3 (i.e., would decrease RfCs and RfDs by a factor of 2
to 3 on average) compared to the traditional NOAEL approach. The BMD10 values are
highlighted in the presentation of the benchmark modeling results to indicate the values that
should be compared for different endpoints. However, it should be noted that this study was
conducted using additional risk, whereas EPA is using extra risk as a conservative default. The
concentration corresponding to a given extra risk will always be the same or lower than the
concentration corresponding to the same percentage of additional risk. Additional risk is defined
as P(d) • P(0), while extra risk is defined as [P(d) • P(0)]/[l • P(0)].
The following data sets were suitable for analysis: (1) incidence of testicular
degeneration in rats exposed to VC by inhalation for 3-18 months (Bi et al., 1985), (2) incidence
of extensive necrosis in the liver of rats chronically exposed to VC in the diet (Feron et al.,
1981), (3) incidence of nuclear polymorphism of hepatocytes in rats exposed to VC by inhalation
for 10 months (Sokal et al., 1980), (4) incidence of proliferation of hepatic reticulo-endothelial
cells in rats exposed to VC by inhalation for 10 months (Sokal et al. 1980), and (5) incidence of
damage to spermatogenic epithelium in rats exposed to VC by inhalation for 10 months (Sokal et
al. 1980). No incidence data were reported for lipid proliferation in the study by Wisniewska-
Knypl et al., (1980), so these data could not be modeled. Although modeling of the data was
based on the PBPK dose metrics, selected endpoints also were modeled on the basis of the
exposure levels or administered doses, for comparison.
Benchmark (BM) Analysis on Dose/Exposure Concentration:
The results of the quantal benchmark modeling of the administered dose (in mg/kg-day)
or exposure concentration (in mg/m3) vice the PBPK derived metrics are shown in Table D-4.
An acceptable fit was obtained with all endpoints, with one exception. Poor fit was obtained in
the liver necrosis data of Feron et al. (1981) when the diet and gavage data were combined, but a
good fit was obtained when the diet data alone were modeled. This is the expected result,
because the response and dose metric of the gavage dose were lower than the highest dietary
D-15
-------
Table D-4. Benchmark dose modeling results using exposure concentration or
administered dose
Model
BMR
type
BMR
MLE
BMC
(mg/m3)
Log-likelihood
G-O-F
p-value
Chi-square
Bi et al. (1985): Testicular degeneration
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Extra
Extra
Extra
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
3.25e+01
1.66e+02
3.41e+02
3.25e+01
1.66e+02
3.41e+02
2.18e+01
l.lle+02
2.29e+02
2.18e+01
l.lle+02
2.29e+02
-1.84e+02
-1.84e+02
-1.84e+02
-1.84e+02
-1.84e+02
-1.84e+02
8.02e-02
8.02e-02
8.02e-02
8.02e-02
8.02e-02
8.02e-02
5.05e+00
5.05e+00
5.05e+00
5.05e+00
5.05e+00
5.05e+00
Feron et al. (1981): Liver necrosis females
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Extra
Extra
Extra
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
2.32e-01
1.18e+00
2.43e+00
2.32e-01
1.18e+00
2.43e+00
1.71e-01
8.74e-01
1.79e+00
1.71e-01
8.74e-01
1.79e+00
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
3.26e-01
3.26e-01
3.26e-01
3.26e-01
3.26e-01
3.26e-01
2.24e+00
2.24e+00
2.24e+00
2.24e+00
2.24e+00
2.24e+00
Feron et al. (1981): Liver necrosis females, including gavage dose
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Extra
Extra
Extra
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
8.41e+00
4.29e+01
8.81e+01
8.41e+00
4.29e+01
8.81e+01
4.93e+00
2.51e+01
5.16e+01
4.93e+00
2.51e+01
5.16e+01
-1.65e+02
-1.65e+02
-1.65e+02
-1.65e+02
-1.65e+02
-1.65e+02
le-06
le-06
le-06
le-06
le-06
le-06
3.03e+01
3.03e+01
3.03e+01
3.03e+01
3.03e+01
3.03e+01
Feron et al. (1981): Liver necrosis males
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
\A/piHn11 rpmntnl
Extra
Extra
Extra
Extra
pYtrn
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
s nnp.n?
1.21e+00
3.94e+00
6.23e+00
1.44e+00
^ S7p+nn
2.87e-01
1.46e+00
3.00e+00
2.88e-01
i 47p+nn
-9.14e+01
-9.14e+01
-9.14e+01
-9.14e+01
-Q I4p+m
6.61e-01
6.61e-01
6.61e-01
7.17e-01
7 i7p.ni
1.92e-01
1.92e-01
1.92e-01
1.31e-01
i ^ip.ni
D-16
-------
Table D-4. Benchmark dose modeling results using exposure concentration or
administered dose (continued)
Model
Weibull quantal
BMR
type
Extra
BMR
l.OOe-01
MLE
6.01e+00
BMD
(mg/kg/
day)
3.02e+00
Log-likelihood
-9.14e+01
G-O-F
p-value
7.17e-01
Chi-square
1.31e-01
Feron et al. (1981): Liver necrosis males, including gavage dose
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Extra
Extra
Extra
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
8.73e+00
4.46e+01
9.16e+01
8.73e+00
4.46e+01
9.16e+01
5.31e+00
2.71e+01
5.56e+01
5.31e+00
2.71e+01
5.56e+01
-1.40e+02
-1.40e+02
-1.40e+02
-1.40e+02
-1.40e+02
-1.40e+02
le-05
le-05
le-05
le-05
le-05
le-05
2.64e+01
2.64e+01
2.64e+01
2.64e+01
2.64e+01
2.64e+01
dose, even though the administered gavage dose was higher (300 mg/kg-day versus 14.1 mg/kg-
day). The BMD/C10 (benchmark dose/concentration for 10% extra risk) estimated for the
sample data sets are as follows:
Oral Studies
Til et al. (1983, 1991) (see below)
Feron etal. (1981)
Increased extensive liver necrosis in females: BMD 10 at 1.8 mg/kg-day
Increased extensive liver necrosis in males: BMD 10 at 3.0 mg/kg-day
Inhalation Studies
Bi etal. (1985)
Increased testicular degeneration: BMC 10 at 229 mg/m3
Thus, the BMC10 estimated for the testicular effect of Bi et al. (1985) is higher than the
default LOAELpgq estimated using the NOAEL approach at 55 mg/m3. Analysis of the
modeling output indicates that this overestimate probably results from poor fit in the low-
concentration region. In order to fit the moderate increase in response corresponding to the 3Ox
increase in exposure level between the middle and high concentrations, the model
underestimated the response in the low concentration region. The goodness-of-fit/? value (0.08)
is low but acceptable, indicating the importance of evaluating the fit in the low concentration
region in addition to the overall p value. The BMD for the liver necrosis in the oral study was
comparable to the NOAEL in females, and between the NOAEL and LOAEL in males. The
calculated BMDs were approximately an order of magnitude higher when the gavage dose was
included in the modeling, reflecting the poor model fit with this data set.
D-17
-------
BM Analysis on PBPK Metrics (the delivered dose) :
The results of the quantal benchmark analysis using doses from the PBPK model are
shown in Table D-5 (in units of the appropriate dose metric, and converted to the exposure
concentration in human equivalent doses/concentrations; HED/Cs). The following data sets
were analyzed: (1) incidence of testicular degeneration in rats exposed to VC by inhalation for 3
to 18 months (Bi et al., 1985), (2) incidence of extensive necrosis in the liver of male and female
rats chronically exposed to VC in the diet (Feron et al., 1981), (3) incidence of nuclear
polymorphism of hepatocytes in rats exposed to VC by inhalation for 10 months (Sokal et al.,
1980), (4) incidence of proliferation of hepatic reticulo-endothelial cells in rats exposed to VC
by inhalation for 10 months (Sokal et al., 1980), and (5) incidence of damage to spermatogenic
epithelium in rats exposed to VC by inhalation for 10 months (Sokal et al., 1980). Note that,
although the same dose metric is used for different effects in the same organ, separate BMDs are
calculated because the response data differ. Acceptable fits were obtained for all endpoints, and
the BMCs obtained with the two BMD models are identical or very similar in all cases. The
BMC 10 (for 10% extra risk) estimated for each animal data set was converted to the
corresponding HEC (which will be referred to as the BMC^q) using the appropriate
equivalence factor, 1.18 (mg/L)/(mg/m3) for RISK, or 0.0308 (mg/kg)/(mg/m3) for AMET, in the
same manner as in the PBPK NOAEL/LOAEL approach. Similarly, the BMD 10 was converted
to the equivalent human oral dose using the equivalence factors of 35.31 (mg/L)/(mg/kg-day) for
RISK or 0.92 (mg/kg)/(mg/kg-day) for AMET.
Oral Studies (bolded in Table D-5)
Til et al. (1983, 1991) (see Table D-6 below)
Feron etal. (1981)
Increased liver necrosis in females: BMD at 40.4 mg/L =1.1 mg/kg-day or 34 mg/m3
Increased liver necrosis in males: BMD at 70 mg/L = 2 mg/kg-day or 59 mg/m3
For comparison, the following data were obtained when the gavage dose of Feron et al.
(1981) was included in the benchmark modeling, and the results were converted to a human
inhalation HEC:
Increased liver necrosis in females: BMC^q at 37.4 mg/L = 32 mg/m3
Increased liver necrosis in males: BMC[HEC] at 65.6 mg/L = 56 mg/nf
D-18
-------
Table D-5. Benchmark dose modeling results using pbpk-derived dose metric
Model
BMR type
BMR
Dose metric
MLE
BMD
Calculated
HEC
(mg/m3)
MLEHFr | BMCHFr
Log-likelihood
G-O-F
p-value
Chi-square
Bi et al. (1985): Testicular degeneration
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
Additional
Additional
Additional
Extra
Extra
Extra
Additional
Additional
Additional
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe+00
5.15e+00
1.07e+01
7.63e-01
3.89e+00
8.00e+00
l.OOe+00
5.15e+00
1.07e+01
7.63e-01
3.89e+00
8.00e+00
6.76e-01
3.47e+00
7.18e+00
5.35e-01
2.73e+00
5.60e+00
6.76e-01
3.47e+00
7.18e+00
5.35e-01
2.73e+00
5.60e+00
3.25e+01
1.67e+02
3.48e+02
2.48e+01
1.26e+02
2.60e+02
3.25e+01
1.67e+02
3.48e+02
2.48e+01
1.26e+02
2.60e+02
2.20e+01
1.13e+02
2.33e+02
1.74e+01
0.89e+02
1.82e+02
2.20e+01
1.13e+02
2.33e+02
1.74e+01
0.89e+02
1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
-1.82e+02
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
3.90e-01
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
1.88e+00
Feron et al. (1981): Liver necrosis females
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Additional
Additional
Additional
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
5.68e+00
2.90e+01
5.98e+01
5.22e+00
2.67e+01
4.09e+00
2.09e+01
4.31e+01
3.85e+00
1.97e+01
4.80e+00
2.45e+01
5.05e+01
4.41e+00
2.26e+01
3.45e+00
1.76e+01
3.64e+01
3.25e+00
1.67e+01
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
3.27e-01
3.27e-01
3.27e-01
3.27e-01
3.27e-01
2.23e+00
2.23e+00
2.23e+00
2.23e+00
2.23e+00
-------
Table D-5. Benchmark dose modeling results using PBPK-derived dose metric (continued)
Model
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
BMR type
Extra
Additional
Additional
Additional
Extra
Extra
Extra
BMR
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
Dose metric
MLE
5.47e+01
5.68e+00
2.90e+01
5.98e+01
5.22e+00
2.67e+01
5.47e+01
BMD
4.04e+01
4.09e+00
2.09e+01
4.31e+01
3.85e+00
1.97e+01
4.04e+01
Calculated
HEC
(mg/m3)
MLEHEC
4.62e+01
4.80e+00
2.45e+01
5.05e+01
4.41e+00
2.26e+01
4.62e+01
BMCHFr
3.41e+01
or
1.15e+00
mg/kg/day
3.46e+00
1.76e+01
3.64e+01
3.25e+00
1.67e+01
3.41e+01
Log-likelihood
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
-1.14e+02
G-O-F
p-value
3.27e-01
3.27e-01
3.27e-01
3.27e-01
3.27e-01
3.27e-01
3.27e-01
Chi-square
2.23e+00
2.23e+00
2.23e+00
2.23e+00
2.23e+00
2.23e+00
2.23e+00
Feron et al. (1981): Liver necrosis in females, including gavage dose
Weibull quantal
Weibull quantal
Weibull quantal
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
4.56e+00
2.33e+01
4.78e+01
3.53e+00
1. 80e+01
3.71e+01
3.86e+00
1.97e+01
4.04e+01
2.98e+00
1.52e+01
3.14e+01
-1.52e+02
-1.52e+02
-1.52e+02
2.10e-01
2.10e-01
2.10e-01
4.53e+00
4.53e+00
4.53e+00
Feron et al. (1981): Liver necrosis males
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Additional
Additional
Additional
Extra
Extra
Extra
Additional
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
2.99e+01
9.59e+01
1.51e+02
2.83e+01
9.16e+01
1.44e+02
3.49e+01
6.94e+00
3.55e+01
7.29e+01
6.64e+00
3.39e+01
6.96e+01
6.98e+00
2.53e+01
0.81e+02
1.28e+02
2.39e+01
0.78e+02
1.22e+02
2.95e+01
0.59e+01
S.OOe+Ol
0.62e+02
5.61e+00
2.87e+01
0.59e+02
0.59e+01
-9.146+01
-9.14e+01
-9.146+01
-9.14e+01
-9.146+01
-9.14e+01
-9.146+01
6.62e-01
6.62e-01
6.62e-01
6.62e-01
6.62e-01
6.62e-01
7.18e-01
1.91e-01
1.91e-01
1.91e-01
1.91e-01
1.91e-01
1.91e-01
l.SOe-Ol
-------
Table D-5. Benchmark dose modeling results using PBPK-derived dose metric (continued)
Model
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
Weibull quantal
BMR type
Additional
Additional
Extra
Extra
Extra
BMR
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
Dose metric
MLE
9.39e+01
1.46e+02
3.34e+01
8.99e+01
1.39e+02
BMD
3.57e+01
7.33e+01
6.68e+00
3.41e+01
7.00e+01
Calculated
HEC
(mg/m3)
MLEHEC
0.79e+02
1.24e+02
2.82e+01
0.76e+02
1.17e+02
BMCHFr
3.02e+01
0.62e+02
5.64e+00
2.886+01
0.59e+02
Log-likelihood
-9.14e+01
-9.146+01
-9.14e+01
-9.146+01
-9.14e+01
G-O-F
p-value
7.18e-01
7.18e-01
7.18e-01
7.18e-01
7.18e-01
Chi-square
1.30e-01
l.SOe-Ol
1.30e-01
l.SOe-Ol
1.30e-01
Sokal et al. (1980): Nuclear proliferation of hepatocytes
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Additional
Additional
Additional
Additional
Additional
Additional
Extra
Extra
Extra
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-01
5.00e-02
l.OOe-02
l.OOe-01
5.00e-02
l.OOe-02
1.03e+01
5.26e+01
1.08e+02
1.03e+01
5.26e+01
1.08e+02
9.83e+01
4.78e+01
9.37e+00
9.83e+01
4.78e+01
9.37e+00
6.98e+00
3.57e+01
7.34e+01
6.98e+00
3.57e+01
7.34e+01
6.91e+01
3.36e+01
6.59e+00
6.91e+01
3.36e+01
6.59e+00
0.87e+01
4.45e+01
0.91e+02
0.87e+01
4.45e+01
0.91e+02
0.83e+02
4.04e+01
0.79e+01
0.83e+02
4.04e+01
0.79e+01
0.59e+01
3.02e+01
0.62e+02
0.59e+01
3.02e+01
0.62e+02
0.59e+02
2.84e+01
5.57e+00
0.59e+02
2.84e+01
5.57e+00
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
-5.42e+01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
6.18e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
9.62e-01
-------
Table D-5. Benchmark dose modeling results using PBPK-derived dose metric (continued)
Model
BMRtype
BMR
Dose metric
MLE
BMD
Calculated
HEC
(mg/m3)
MLEHFC | BMCHFC
Log-likelihood
G-O-F
p-value
Chi-square
Sokal et al. (1980): Proliferation, hepatic reticuloendothelial cells
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Additional
Additional
Additional
Additional
Additional
Additional
Extra
Extra
Extra
Extra
Extra
Extra
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-01
5.00e-02
l.OOe-02
l.OOe-01
5.00e-02
l.OOe-02
1. 81e+01
9.27e+01
1.91e+02
1.92e+01
9.57e+01
1.95e+02
1.71e+02
8.30e+01
1.63e+01
1.74e+02
8.57e+01
1.72e+01
1.12e+01
5.73e+01
1.18e+02
1.126+01
5.73e+01
1. 18e+02
1.09e+02
5.33e+01
1.04e+01
1.09e+02
5.33e+01
1.04e+01
1.536+01
0.78e+02
1.62e+02
1.62e+01
0. Sle+02
1.65e+02
1.44e+02
0.70e+02
1.38e+01
1.47e+02
0.72e+02
1.45e+01
0.95e+01
4.84e+01
l.OOe+02
0.95e+01
4.84e+01
l.OOe+02
0.92e+02
4.51e+01
0.88e+01
0.92e+02
4.51e+01
0.88e+01
-5.26e+01
-5.266+01
-5.26e+01
-5.266+01
-5.26e+01
-5.266+01
-5.26e+01
-5.266+01
-5.26e+01
-5.266+01
-5.26e+01
-5.266+01
9.27e-01
9.27e-01
9.27e-01
6.96e-01
6.96e-01
6.96e-01
9.27e-01
9.27e-01
9.27e-01
6.96e-01
6.96e-01
6.96e-01
1.53e-01
1.53e-01
1.53e-01
1.52e-01
1.52e-01
1.52e-01
1.53e-01
1.53e-01
1.53e-01
1.52e-01
1.52e-01
1.52e-01
Sokal et al. (1980): Spermatogenic epithelium damage
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Additional
Additional
Additional
Additional
Additional
Additional
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
1.03e+00
5.29e+00
1.09e+01
1.03e+00
5.29e+00
1.09e+01
5.75e-01
2.94e+00
6.05e+00
5.75e-01
2.94e+00
6.05e+00
3.34e+01
1.72e+02
3.54e+02
3.34e+01
1.72e+02
3.54e+02
1.87e+01
0.95e+02
1.97e+02
1.88e+01
0.95e+02
1.97e+02
-5.20e+01
-5.206+01
-5.20e+01
-5.206+01
-5.20e+01
-5.20e+01
3.84e-01
3.84e-01
3.84e-01
3.84e-01
3.84e-01
3.84e-01
1.91e+00
1.91e+00
1.91e+00
1.91e+00
1.91e+00
1.91e+00
Sokal et al. (1980): Spermatogenic epithelium damage (high dropped)
-------
Table D-5. Benchmark dose modeling results using PBPK-derived dose metric (continued)
Model
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
Polynomial quantal
Polynomial quantal
Polynomial quantal
Weibull quantal
Weibull quantal
Weibull quantal
BMRtype
Additional
Additional
Additional
Additional
Additional
Additional
Extra
Extra
Extra
Extra
Extra
Extra
BMR
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-02
5.00e-02
l.OOe-01
l.OOe-01
5.00e-02
l.OOe-02
l.OOe-01
5.00e-02
l.OOe-02
Dose metric
MLE
1.43e+00
5.91e+00
1.03e+01
1.74e+00
5.74e+00
9.74e+00
9.41e+00
5.38e+00
1.28e+00
8.92e+00
5.27e+00
1.60e+00
BMD
3.85e-01
1.97e+00
4.05e+00
3.85e-01
1.97e+00
4.05e+00
3.75e+00
1.83e+00
3.58e-01
3.75e+00
1.83e+00
3.58e-01
Calculated
HEC
(me/m3)
MLEHEC
4.64e+01
1.92e+02
3.34e+02
5.65e+01
1.86e+02
3.16e+02
3.06e+02
1.75e+02
4.16e+01
2.90e+02
1.71e+02
5.20e+01
BMCmr
1.25e+01
0.64e+02
1.32e+02
1.25e+01
0.64e+02
1.32e+02
1.22e+02
0.59e+02
1.16e+01
1.22e+02
0.59e+02
1.16e+01
Log-likelihood
-4.08e+01
-4.08e+01
-4.08e+01
-4.08e+01
-4.08e+01
-4.08e+01
-4.08e+01
-4.086+01
-4.08e+01
-4.086+01
-4.08e+01
-4.086+01
G-O-F
p-value
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
l.OOe+00
Chi-square
2.25e-21
2.25e-21
2.25e-21
7.31e-21
7.31e-21
7.31e-21
2.25e-21
2.25e-21
2.25e-21
7.31e-21
7.31e-21
7.31e-21
o
to
-------
Table D-5. Benchmark dose modeling results using PBPK-derived dose metric (continued)
Model
BMRtype
BMR
Dose metric
MLE
BMD
Calculated
HEC
(mg/m3)
MLEHFC | BMCHFC
Log-likelihood
G-O-F
p-value
Chi-square
Continuous endpoint
Sokal et al. (1980): Liver to body weight ratios
Polynomial continuous
Polynomial continuous
Polynomial continuous
Polynomial continuous
Weibull continuous
Weibull continuous
Weibull dontinuous
Weibull continuous
Absolute
Relative
Relative
Relative
Absolute
Relative
Relative
Relative
SD0/2
l.OOe-02
5.00e-02
l.OOe-01
SD0/2
l.OOe-02
5.00e-02
l.OOe-01
2.44e+02
8.73e+01
3.34e+02
5.05e+02
3.23e+02
1.95e+02
3.91e+02
5.28e+02
1.35e+02
4.15e+01
2.00e+02
3.70e+02
2.12e+02
1.03e+02
2.73e+02
4.13e+02
2.06e+02
0.74e+02
2.82e+02
4.27e+02
2.73e+02
1.65e+02
3.30e+02
4.46e+02
1.14e+02
3.51e+01
1.69e+02
3.13e+02
1.79e+02
0.63e+02
2.30e+02
3.49e+02
5.21e+01
5.21e+01
5.21e+01
5.21e+01
5.18e+01
5.18e+01
5.18e+01
5.18e+01
1.24e-01
1.24e-01
1.24e-01
1.24e-01
8.19e-02
8.19e-02
8.19e-02
8.19e-02
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
o
to
-------
Inhalation Studies
Bietal. (1985)
Increased testicular degeneration: BMC[HEC] at 5.60 mg/kg = 6.1 mg/kg-day or 180 mg/m3
Sokaletal. (1980)
Nuclear proliferation of hepatocytes: 69.1 mg/L = 2 mg/kg-day or 59 mg/m3
Proliferation of hepatic reticuloendothelial cells: 109 mg/L = 3.1 mg/kg-day or 92 mg/m3
Damage of spermatogenic epithelium: 3.75 mg/kg = 4.1 mg/kg-day or 122 mg/m3
As shown in Table D-5, acceptable fits were obtained for all of the data sets. On the
basis of the "AMET" dose metric, the BMC[HEC] at 180 mg/m3 for testicular degeneration of Bi et
al. (1985) is about a factor of two below the LOAEL[HEC] at 406 mg/m3 for the same study,
whereas the BMC[HEC] at 34-59 mg/m3 for liver necrosis of Feron et al. (1981) is inclusive of the
NOAEL[HEC] of 32.6 mg/m3 based on the "RISK" dose metric. It is also noteworthy that the
BMC[HEC] of 180 mg/m3 for testicular degeneration is about a factor of four above the
NOAEL[HEC] of 42.2 mg/m3 based on the "AMET" dose metric, perhaps demonstrating the value
of the BMD approach when the dose levels in a study are widely spaced. Although similar
BMCs were calculated using extra and additional risk for most endpoints, the BMC for testicular
degeneration using additional risk is about 30% higher than the value calculated using extra risk,
probably because of the high background response for this endpoint.
The response for the testicular endpoint observed by Sokal et al. (1980) at the high
concentration was lower than that at the middle concentration. This decrease does appear to be
biologically meaningful. High VC concentrations destroy P450, decreasing the amount of
reactive metabolites formed (Reynolds et al, 1975; Guengerich and Strickland, 1977). Thus, the
highest exposure level may have knocked out the P450 system and prevented the testicular
toxicity. Although the model fit was acceptable in both cases, dropping the high dose resulted in
much better fit, as expected. The decrease in response was not observed for the liver lesions,
apparently because of the higher P450 levels in the liver. Based on this analysis, the response
was modeled with and without the high dose (additional risk) or with the high dose dropped
(extra risk). As expected, dropping the high dose improved the model fit. The benchmark
values calculated without the high dose are about two-thirds those calculated including the high
dose.
D.2.4.1. Quanta! Benchmark Concentration Analyses of Liver Cell Polymorphism in
Til et al (1983,1991)
(1) Computational Models — Discontinuous Data
The models used are listed in Table D-6. For data inputs, the multistage polynomial was
set to the number of dose groups minus one, the risk type was extra [P(d) • P(0)] / [1 • P(0)],
and no threshold was estimated. For the Weibull, the lower limit of • was set at 1. The
gammahit model was run to convergence (approximately 1,000 iterations).
D-25
-------
Table D-6. BMD10 and maximum likelihood estimates (MLE) values
generated from various model fits to liver cell polymorphism incidence
data from exposure to vinyl chloride monomer (Til et al., 1991)
Model
Weibull (power • 1)
Gammahit
Quantal quadratic
Logistic
Multistage
Probit
Quantal linear
NOAEL
LOAEL
BMD10
(MLE), mg/L liver3
24.0 (26.6)
21.4(23.2)
13.8(16.2)
12.9(13.4)
11.8(16.2)
11.6(12.7)
6.5(9.1)
3.00
(0.13mg/kg-day)
29.9
(1.3 mg/kg-day)
/7-value
0.88
0.88
0.96
0.47
0.79
0.44
0.46
aBMD10 is the lower 95% confidence bound on the MLE of a 10% change in numbers exhibiting polymorphism
evaluated as either moderate or severe. Results shown are generated with dose metrics (mg VC metabolites/L liver;
RISK) and were generated from the PBPK model of Clewell et al. (this document). The NOAEL and LOAEL are
also shown for comparative purposes.
(2) Data
Incidence data from Table 4 in Til et al. (1983, 1991) for both sexes of moderate and
severe grades of liver cell polymorphism were combined and summed to produce one control
group and three exposed groups (incidence of moderate + severe)/total exposed; (21)/197 for
controls, (21)/199 for 0.014 mg/kg, (20)/196 for 0.13 mg/kg, and (37)798 for 1.3 mg/kg.
The doses were further transformed by use of the PBPK model to an average daily
delivered dose in mg/L liver to the following (averaged metric values of male and female): for
0.014 mg/kg-day, 0.321 mg/L; for 0.13 mg/kg-day, 2.98 mg/L; for 1.3 mg/kg-day, 29.8 mg/L.
This metric (mg/L liver) was used in the BMD modeling. BMD results were transformed to the
human equivalent oral concentration by dividing this metric by 35.31 and by 1.18 to obtain the
continuous human equivalent inhalation concentration.
(3) Model Fit
D-26
-------
Model fit was judged by the/>-values given from the analysis of deviance by inspection
of the observed versus predicted output from each model (not shown) and from visual
inspections of the graphical outputs (not shown).
(4) Results
The dose-response character of liver cell polymorphism was limited, appearing as a high-
dose phenomenon only. Nevertheless, all models tested fit these data acceptably, i.e.,/? > 0.05.
Comparison of the observed versus expected values generated from the various models also
showed that all models, save for the quantal linear, gave reasonable approximations of the data.
There was, however, a clear dichotomy between models giving ^-values in the range of 0.8 and
those giving values around 0.4. Visual inspection of the graphicals (not shown) indicate that
model fitting to the 0.13 mg/kg (3.00 mg/L liver) point may have accounted for this dichotomy,
in that those models generating MLEs within the variability of this point gave elevateds-values,
whereas those models with MLEs missing this point had/>-values smaller by one-half.
The range of the outputs for the various models for the BMD10 was over threefold, from
6.5 to 24.01 mg/L liver, with BMDIOs of 11-14 mg/L liver being intermediate. All modeled
BMDIOs were larger than the NOAEL of the study, 3.00 mg/L liver (0.13 mg/kg), with even the
lowest modeled BMD10 (quantal linear) being over twice this value.
(5) Discussion
The only nonzero response in this data set is the highest dose employed, 29.9 mg/L liver,
where the response rate was about 38%. The dose-spacing of the study is such that the nearest
experimental point to the nonzero point is 10-fold different (3.00 versus 29.9 mg/L liver). Some
models (gammahit and Weibull) generated MLE responses that remained close to the control
rate of 10% until near the nonzero point, where they rose steeply. Others (e.g., logistic, probit,
and multistage) generated MLE responses that sloped upward more gradually at doses
considerably less than the nonzero point. The implication for model choice is appreciable. The
range between the NOAEL and the highest BMD10 (Weibull) is eightfold. The range between
the NOAEL and the lowest BMD10 (quantal linear) is slightly more than twofold.
No biological reasoning governs the choice of any of the model outputs from Table D-4.
The character of the dose response is highly uncertain because of the large spacings between
applied doses. Because of this uncertainty and the lack of biological motivation for model
choice, the NOAEL, 3.00 mg/L liver, is chosen for use as the basis for further quantitative
analysis for both oral and inhalation assessments. The HECs are derived from the output of the
PBPK model as shown in this appendix. The oral NOAEL(HEC) is 0.09 mg/kg-day (3.00 mg/L
liver 735.31). The inhalation NOAEL(HEC) is 2.5 mg/m3 (3.00 mg/L liver/ 1.18).
D.4. REFERENCES
Allen, BC; Kavlock, RJ; Kimmel, CA; et al. (1994a) Dose response assessments for developmental toxicity: II.
Comparison of generic benchmark dose estimates with NOAELs. Fundam Appl Toxicol 23:487-495.
D-27
-------
Allen, BC; Kavlock, RJ; Kimmel, CA; et al. (1994b) Dose response assessments for developmental toxicity: III.
Statistical models. Fundam Appl Toxicol 23:496-509.
Andersen, M; Clewell, H; Gargas, M; et al. (1987) Physiologically based pharmacokinetics and the risk assessment
process for methylene chloride. Toxicol Appl Pharmacol 87:185-205.
Bi, W; Wang, Y; Huang, M; et al. (1985) Effect of vinyl chloride on testis in rats. Ecotoxicol Environ Saf 10:281-
289.
Clewell, HJ; Jarnot, BM. (1994) Incorporation of pharmacokinetics in non-carcinogenic risk assessment: example
with chloropentafluorobenzene. Risk Anal 14:265-276.
Crump, K. (1984) A new method for determining allowable daily intakes. Fundam Appl Toxicol 4:854-871.
Crump, K. (1995) Calculation of benchmark doses from continuous data. Pdsk Anal 15:79-89.
Faustman, EM; Allen, BC; Kavlock, RJ; et al. (1994) Dose response assessment for developmental toxicity: I.
Characterization of data base and determination of NOAELs. Fundam Appl Toxicol 23:478-486.
Feron, VJ; Hendriksen, CFM; Speek, AJ; et al. (1981) Lifespan oral toxicity study of vinyl chloride in rats. Food
Cosmet Toxicol 19:317-333.
Gaylor,DW; Slikker, W. (1990) Risk assessment for neurotoxic effects. Neurol Toxicol 11:211-218.
Guengerich, FP; Strickland, TW. (1977) Metabolism of vinyl chloride: destruction of the heme of highly purified
liver microsomal cytochrome P-450 by a metabolite. Mol Pharmacol 13:993-1004.
Kavlock, RJ; Allen, BC; Faustman, EM; et al. (1995) Dose response assessments for developmental toxicity: IV.
Benchmark doses for fetal weight changes. Fundam Appl Toxicol 26(2):211-222.
Reynolds, ES; Moslen, MT; Szabo, S; et al. (1975) Vinyl chloride-induced deactivation of cytochrome P-450 and
other components of the liver mixed function oxidase system: an in vivo study. Res Commun Chem Pathol
Pharmacol 12:685-694.
Sokal, JA; Baranski, B; Majka, J; et al. (1980) Experimental studies on the chronic toxic effects of vinyl chloride in
rats. J Hyg Epidemiol Microbiol Immunol 24:285-294.
Torkelson, TR; Oyen, F; Rowe, VK. (1961) The toxicity of vinyl chloride as determined by repeated exposure of
laboratory animals. Am Ind Hyg Assoc J 22:354-361.
U.S. Environmental Protection Agency (U.S. EPA). (1994) Methods for derivation of inhalation reference
concentrations and application of inhalation dosimetry. EPA/600/8-90/066F. Office of Health and Environmental
Assessment, Washington, DC.
Watanabe, PG; Zempel, JA; Pegg, DG; et al. (1978) Hepatic macromolecular binding following exposure to vinyl
chloride. Toxicol Appl Pharmacol 44:571-579.
Wisniewska-Knypl, JM; Klimczak, J; Kolakowski, J. (1980) Monooxygenase activity and ultrastructural changes
of liver in the course of chronic exposure of rats to vinyl chloride. Int Arch Occup Environ Health 46(3):241-249.
D-28
-------
APPENDIX E. EXTERNAL PEER RE VIEW—SUMMARY
OF COMMENTS AND DISPOSITION
The Toxicological Review and IRIS summary for vinyl chloride have undergone both
internal peer review performed by scientists within EPA and two formal external panel peer
reviews. Comments made by the internal reviewers were addressed prior to submitting the
documents for external peer review and are not part of the appendix. 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.
A summary of comments made by the external reviewers at the first panel meeting based upon
questions posed to them and EPA's response to these comments are presented below. A
summary of the second panel meeting is given in Appendix E-l.
(1) General Question
A. Are there any other data/studies that are relevant (i.e., useful for the hazard
identification or dose-response assessment) for the assessment of the adverse health effects,
both cancer and noncancer, of this chemical?
The 11 external peer reviewers offered editorial comments and many minor but valuable
suggestions, all of which have been considered for incorporation into the text to the extent
feasible. The reviewers identified a number of publications for inclusion in the support
document.
Response: Copies of the publications identified by the reviewers have been acquired and the
appropriate ones were incorporated into the summaries and the support document.
(2) Chemical-Specific Questions Posed to Panelists
A. Is the pharmacokinetic (PBPK) model developed for this assessment adequate for
quantifying cancer and noncancer risk of VC exposure?
Several questions regarding the adequacy of the model were discussed. An aspect of
considerable discussion concerned whether the use of a two-pathway model is an improvement
over previous models using a single metabolic activation pathway. Several individuals felt that
it was not an improvement over the single-pathway models because the lower dose ranges,
where the low-affinity pathway produces few metabolites, are of greatest human interest. It was
also pointed out that there is little evidence for a second metabolic activation pathway in
humans. The general feeling was that a two-pathway model was an unnecessary increase in
complexity but nevertheless acceptable.
E-l
-------
One reviewer questioned the use of an undocumented model when models published in the
peer-reviewed literature were available. However, it was pointed out that both the present model
and one published by Reitz et al. in 1996 predicted similar internal dose measures.
The possibility that the active metabolite chlorethylene oxide (CEO) formed in the liver
could migrate to other tissues was discussed. It was concluded that because of its reactivity this
was unlikely to occur. It was also concluded that other tissues (brain, kidney, etc.) may have a
limited metabolic capability, but it was likely to be small in relation to the liver. Because of
these conclusions it was generally agreed that modeling concentration of CEO in the liver was an
acceptable approach to assess risk from VC exposure.
In general, the reviewers felt that although use of the model presented some uncertainties
that require discussion in the Toxicological Review, it is a fairly standard model and its use for
quantitating risk will lead to acceptable potency estimates.
Response: Use of the two-pathway model was retained. While assumption of a second pathway
in rats and mice (a second pathway is not assumed for humans in the model) may be
unnecessary, its inclusion is not considered likely to introduce errors in estimating liver
concentration of the active metabolite.
The PBPK model used for VC is an adaptation of a model published earlier for vinylidene
chloride, and its use for estimating cancer risk from VC exposure has been described in the
literature. Thus, it is not considered to be undocumented.
B. Should human data be used to quantitate risk, and if not, were the animal studies
selected the proper ones?
Two reviewers believed that human data were adequate. Three studies were cited as being
useful for this purpose and the possibility of obtaining unpublished data was alluded to. This
data, however, has not been provided to EPA to date. Several others believed that an attempt
should be made to evaluate human data further to determine its possible use for either
quantitating cancer risk or confirming the animal-based risks. On the other hand, the only
epidemiologist among the reviewers believed that human exposure levels are too uncertain to
base human risk upon epidemiology data. There was general agreement that the correct animal
studies were selected.
Response: Published epidemiology studies have a considerable uncertainty. In the largest and
best documented one, the mean age of the cohort was only 54 years. The development of liver
tumors in an unknown additional number of subjects can therefore be anticipated. The subjects
worked in a number of different facilities in different countries, so exposures varied and were
very uncertain. Uncertainty is also increased because of the small numbers of subjects with liver
tumors in many of the cohorts.
On the other hand, a large number of animal studies have been carried out utilizing a wide
range of doses. Both oral and inhalation studies of suitable quality are available. The primary
E-2
-------
tumor site, the liver, is relevant to human response. Two species, mice and rats, have been
shown to have similar susceptibilities, increasing confidence that responses may be similar
across species.
While the animal studies used to quantitate risk did not utilize neonates, the animals were
exposed prior to adulthood. Other studies using neonates provided evidence regarding early life
sensitivity not available in the human epidemiology studies. Epidemiology studies also provided
no information regarding possible sex differences in sensitivity.
Although the decision was made to base cancer risk upon animal data, potency estimates
based on epidemiology data are used as support for the recommended values.
C. Are neoplastic nodules reported in the Feron et al. feeding study and/or hepatomas
reported in the Maltoni et al. inhalation studies appropriate for inclusion in the data sets
used for cancer risk assessment?
Four reviewers, including the only animal pathologist on the panel, believed that neoplastic
nodules reported in the Feron and Til studies should be included in cancer quantitation. Two
reviewers disagreed and four did not comment. The dissenters felt that liver angiosarcoma was
the primary endpoint in humans and should be used for quantitation, and that inclusion of all
liver tumors would result in an overestimation of cancer risk. Those favoring inclusion of
nodules believed that the nodules have the potential to progress to malignancy and should
therefore be counted. There was general agreement that hepatomas should be included only if
increases were statistically significant.
Response: Both angiosarcomas and hepatocellular carcinomas are induced by VC in the animal
as well as epidemiology studies. Because neoplastic nodules, according to pathologists, are
tumors (adenomas) and are capable of progressing to carcinoma, it was deemed appropriate to
include them. Hepatocellular tumors were also considered appropriate in assessing risk on the
basis of the inhalation studies. Although their numbers were not statistically significantly
increased in the inhalation studies, they are considered to be associated with VC exposure
because they were significantly increased in the oral studies. As there are few of them, however,
their effect upon cancer potency is minimal.
D. If all liver tumors were included, is risk likely to be overestimated, or would the tumors
counterbalance a possible underestimate of total risk due to possible induction of tumors at
other sites?
Three of the reviewers believed that including all liver tumors would result in an overestimate of
cancer risk. Two felt that it might address the possibility of tumor induction at other sites. The
other reviewers were either uncertain or had no opinion.
Response: There is evidence from both the animal feeding studies and epidemiology studies
that hepatocellular tumors as well as angiosarcomas are induced by VC. On this basis, the
E-3
-------
inclusion of all liver tumors, even from studies in which hepatocellular tumors were not
significantly increased, is considered to be appropriate.
E. Do the confidence statements and weight-of-evidence statements present a clear picture
and accurately reflect the utility of the studies chosen, the relevance of cancer as well as
noncancer data to humans, and the comprehensiveness of the database?
The answers were generally yes, although one reviewer suggested that we should make an
effort to collect all human data, including unpublished data.
Response: Attempts have been made to collect additional human data, however such data are
found to be unpublished or of questionable use for quantitating risk.
F. Because the model accounts for metabolic rate differences among species and provides
an estimate of the steady-state concentration of the active metabolite (chlorethylene oxide)
in the liver, do we still need a scaling factor, i.e., a metabolic rate adjustment?
Six reviewers believed that no surface area correction is required in quantitating cancer risk
if the PBPK model is used. Two agreed that there should be one, and the others made no
comment. The reviewers that believed a surface correction should be included were concerned
that the model accounts for dosimetric considerations, but does not account for possible
toxicodynamic differences among species.
Response: In other assessments, a body surface correction, or metabolic scaling factor, has
been applied to account for the fact that laboratory animals, which are smaller than humans, have
a correspondingly higher metabolic rate and thereby are predicted to detoxify a chemical faster,
resulting in a smaller steady-state concentration of the chemical at the target site. The PBPK
model for VC was developed to predict the steady-state concentration of the active metabolite at
the target site. The model included metabolic rate factors. Also, epidemiological studies support
the conclusion that humans are less sensitive than rodents; therefore applying a scaling factor
would render the results overly conservative. An additional surface area correction is thus
considered to be inappropriate.
G. For the RfD and RfC, has the most appropriate critical effect been chosen (i.e., effect
occurring at the lowest concentration)?
The pathologist present on the panel recommended the use of liver cell polymorphism and
bile duct cysts as the most appropriate endpoints for quantitating noncancer risk. Other panel
members, with one exception, agreed. These endpoints were recommended rather than liver
necrosis, which was used in the draft document, because they occur at lower exposure levels
than liver necrosis, they are not considered to be preneoplastic, and they are indicative of liver
toxicity. One reviewer recommended the use of basophilic foci because they are noted at an
even lower concentration than are polymorphisms.
E-4
-------
Response: The change to utilization of liver cell polymorphism and bile duct cysts was agreed
to and appropriate changes have been made in the noncancer risk estimates. Use of basophilic
foci was rejected because it was pointed out that this endpoint may be a precursor to cancer and
is thus not appropriate for use in development of RfC or RfDs.
H. Were the appropriate studies used for development of the RfD and RfC?
The Til et al. (1991) study in which liver cell polymorphism was identified as the critical
endpoint was also recommended for development of the RfC rather than available inhalation
studies. This recommendation was made because the critical endpoint was detected at a much
lower concentration than any endpoints reported in inhalation studies, and pharmacokinetic data
indicate that accurate route extrapolation is feasible. The Til et al. (1991) study is of higher
quality than available inhalation studies, and it is of chronic duration, unlike the inhalation
studies. The reviewers generally felt comfortable using a route extrapolation because VC is
well absorbed by both routes, the liver is the primary target organ by both exposure routes, and
use of the oral study to derive an RfC results in at least as conservative an assessment of risk as
use of a lower quality inhalation study. One reviewer believed that the Bi et al. (1985) inhalation
study is adequate for derivation of an RfC despite the fact that it is a subchronic study and the
data were unsuitable for derivation of a benchmark dose.
Response: The recommendation regarding the use of the Til et al. study has been agreed to and
appropriate changes have been made to the documents. Route extrapolation results in a
conservative estimate of risk because essentially all vinyl chloride taken up via the oral route
will pass through the liver first. It is thus very unlikely that an equivalent inhaled dose will
result in a greater liver concentration of VC. Use of the Bi et al. (1985) study was not agreed to.
The study is of only subchronic duration, it is of lower quality than the Til et al. (1991) study,
and it identified a considerably higher NOAEL. It was, however, used to support the
recommended RfC.
(3) Additional Comments by Panelists
A. One reviewer recommended combining the Til and Feron oral studies for estimating
cancer potency.
Response: Combining the studies was not considered to be appropriate because the Til et al.
study, while similar in design to the Feron study, was not concurrent and used very low doses
with only a very marginal and not statistically significant response.
B. Some reviewers recommended that the high dose in the Maltoni et al. rat and mouse
studies be eliminated from the data sets used to estimate cancer potency because the
concentrations were well above those required for saturation of activation pathways.
Response: These dose groups were retained because the PBPK model was designed to account
for metabolic saturation, and a larger data set will decrease the uncertainty in estimating the unit
risks.
E-5
-------
C. Liver metabolism may change with time.
One reviewer was concerned that the PK model did not account for possible changes in
rate of liver metabolism through the animal's lifetime. These changes could occur either through
aging or liver injury. Other panel members, however, did not believe that either aging or liver
toxicity were likely to result in sufficient changes in liver metabolism to result in significant
dosimetry errors.
Response: Although this is of concern, data were available indicating that liver metabolism
normally did not show large changes with aging, and few of the doses were expected to induce a
level of toxicity that would alter metabolism greatly. Nevertheless, exposure to the highest
concentrations may approach maximum tolerated doses and result in altered liver metabolism.
D. A range of cancer potency estimates may be more appropriate than the
recommendation of a point estimate of risk.
One of the reviewers felt that the degree of uncertainty precluded recommendation of a
point estimate of cancer risk.
Response: The uncertainty in estimating cancer potency of VC is less than that for many
chemicals evaluated by EPA, for several reasons. Considerable pharmacokinetic data are
available in both animals and humans, allowing accurate determination of active metabolite
concentrations at the target site. There is site concordance for tumors in laboratory animals and
humans. Cancer potency estimates are supported by epidemiologic data.
E. Use of animal data results in an overly conservative estimate of risk.
Some of the reviewers believed that development of cancer risk estimates based upon
animal data overpredict risk because estimates based upon human data result in considerably
lower estimates.
Response: It is true that the animal data predict greater risk. However, the cohorts used in the
epidemiology studies were generally healthy adult males. Risk may be greater during early-life
exposure and among sensitive populations because of genetic factors, health status, etc.
Moreover, as noted previously, exposures in the epidemiology studies were very uncertain.
E-6
-------
APPENDIX E-l. SECOND EXTERNAL PEER REVIEW-
SUMMARY OF COMMENTS AND DISPOSITION
Because of certain issues raised by the EPA consensus reviewers, a second expert peer
review panel was convened June 2, 1999. The consensus reviewers were concerned about the
complexity of the PBPK model and the fact that use of a PBPK model was a relatively new
approach being attempted by EPA. As a result, some felt that the model needed additional
review. Other concerns were related to the possible need for an adjustment to account for
species differences in toxicodyanamics and whether the unit risk estimates adequately accounted
for possible tumor induction at nonliver sites. The panel was also asked to review two major
new studies: a CMA reproductive/developmental study (1998a, Vinyl chloride combined
inhalation two-generation reproduction and developmental toxicity study in CD rats. Final
Report) and a CMA epidemiological study (1998b, Epidemiological study of men employed in
the vinyl chloride industry between 1942 and 1972:1. Re-analysis of mortality through
December 31, 1982, and II. Update of mortality through December 31, 1995. Final Report).
(1) General Questions Posed to the Panelists
A. Are there any other data/studies that are relevant (i.e., useful for the hazard
identification or dose-response estimation) for the assessment of the adverse health
effects, both cancer and noncancer, of this chemical?
The panel concluded that the EPA document was an excellent survey of the vinyl chloride
literature. However, some additional studies, mostly very recent ones, were identified by the
reviewers that merit review by EPA and inclusion in the documents. Some of the important ones
are listed below.
An important paper by Swenberg et al. (1999) corrects a previously reported error that
etheno adducts are highly persistent (Swenberg et al., 1992). As methods improved, etheno
adducts were found to be endogenous (present in unexposed control animals and humans) and
that what had appeared to be the presence of highly persistent etheno adducts was actually a
return to background levels following cessation of exposure. This finding is significant because
it provides evidence that DNA adducts formed by reactive metabolites of VC are additive to the
endogenous DNA adduct levels, supporting a low-dose linear model for liver DNA adduct
formation. Furthermore, a reasonable correlation, consistent with the known metabolism and
genotoxicity of VC, is found between rat liver DNA adduct formation, the reactive metabolite
concentration predicted by the PBPK model in the document, and incidence of rat liver tumors
from the Maltoni et al. (1981, 1984) inhalation study.
An epidemiology study from Japan (Du and Wang, 1998), wherein 2,224 workers had
occupational exposure to vinyl chloride monomer (VCM), provides additional support for liver
toxicity and cancer, but also lacks exposure data and notes a correlation of liver toxicity with
hepatitis B infection. Some of the key points in the paper of Storm and Rozman (1997), i.e., that
humans are less sensitive to the carcinogenic effects of vinyl chloride than are rats and that
E-7
-------
reduced occupational exposures have successfully eliminated VC-induced liver angiosarcoma,
should be addressed.
Response: Copies of the publications identified by the reviewers have been acquired and
incorporated into the summaries and/or the support document.
B. Appropriateness of the noncancer critical effect(s).
The panel agreed that hepatic toxicity is the most appropriate endpoint for RfC/RfD
development. These effects occur at the lowest dose, have been shown in many studies, and
demonstrate site concordance between animal studies and occupationally exposed humans. In
addition, these effects are consistent with known mechanistic, pharmacokinetic, and metabolic
information. However, there was some disagreement regarding the appropriate endpoint for
identifying liver toxicity. While the majority agreed that liver cysts and liver cell
polymorphisms, which are not considered to be precursors to cancer, are the most suitable
endpoint, one reviewer felt these arbitrary morphological endpoints were overly conservative
because it is unclear whether the cells progress to become necrotic or whether the effects are
reversible. Another reviewer stated that preneoplastic liver findings should be included as an
adverse effect in the RfC/RfD derivation (i.e., NOAELs should not be limited to noncancerous
endpoints). Under those circumstances, basophilic foci would be utilized as the critical
endpoint.
Reponse: Because liver cell polymorphism is the noncancer toxic effect noted at the lowest
concentration, this endpoint was retained for the determination of RfC/RfD. According to
present EPA methodology, endpoints that may progress to cancer, such as basophilic foci, are
not considered suitable for noncancer quantitation. Although there is no absolute proof that all
liver cell polymorphisms progress to become necrotic, they are nevertheless considered to be
abnormal, and it is preferable to err in the direction of conservatism.
C. Appropriateness of the endpoints selected for quantitating cancer risk.
The panel agreed that liver tumors and angiosarcomas were the most appropriate endpoints
for the cancer risk assessment. The inclusion of liver neoplastic nodules in the tumor count was
considered to be overly conservative by one reviewer because these lesions regress with
cessation of exposure and not all progress to carcinomas, even in the face of continued exposure.
A question was raised as to whether there was an overestimation of risk by adding together
tumor types of different embryological origin (i.e., liver cells are of endothelial origin whereas
angiosarcomas are derived from cells originating in the mesodermal cell layer). EPA was asked
to address this point in the document.
Response: While not all neoplastic nodules may progress to cancer, it is considered preferable
to err on the side of conservatism. It is true that hepatocellular carcinoma and liver
angiosarcoma have different origins. However, according to standard EPA methodology, when
E-8
-------
tumors are significantly increased at more than one site (or more than one tissue type), animals
with either of these tumors are included in the quantitation.
D. Have the most appropriate studies been utilized as the basis for the noncancer and
cancer assessments?
The panel agreed that the most appropriate studies have been chosen.
Response: No changes were made in the studies selected for quantitating risk.
E. Are the supporting studies used sufficient and appropriate?
Even though supporting/additional studies were considered sufficient and appropriate,
several inconsistencies were noted between the IRIS summaries and the Toxicological Review.
Response: The inconsistencies have been corrected.
F. Consideration of other data for the uncertainty factors or the modifying factors for
the noncancer assessments.
The panel was not aware of other data.
Response: Consideration of other data is considered unnecessary.
G. Accuracy and clarity of confidence statements and weight-of-evidence statements.
These statements were considered to be appropriate. It was suggested that confidence in
the RfD be increased to medium-to-high. One panel member took exception to the statement
that VC is shown to be a human carcinogen by the oral route; the data only show VC to be a
human carcinogen by the inhalation route. One reviewer suggested that EPA could better
support the risk estimates based on animal data by incorporating some of the information from
the risk estimates based on epidemiological data from Appendix B in the discussion of
confidence in the oral and inhalation carcinogenicity.
Response: The medium confidence in the RfD is retained because our confidence in the database
is medium to high, and our confidence in the qualitative aspects of the PBPK model is less than
high. The discussion regarding carcinogenicity has been altered. VC is now considered to be a
human carcinogen by the oral route because of positive animal bioassay data by both the oral
and inhalation route, positive human studies by the inhalation route, site concordance for tumors,
high degree of absorption by both routes, etc. Some of the information regarding estimates from
epidemiology data has been incorporated in the confidence statement.
(2) Chemical-Specific Questions Posed to the Panelists
A. Is the PBPK model used suitable for quantitating both cancer and noncancer risk?
E-9
-------
The panel agreed that the PBPK model is the appropriate approach for assessing potential
human risks from exposure to vinyl chloride. The two-pathway model adequately describes the
underlying biology, although the low-affinity pathway (other CYP, non-2El enzymes) is more of
an academic interest and does not significantly contribute to the low-dose extrapolation because
only the high-affinity pathway (CYP2E1) would be operable at the low doses. The human
model appropriately uses only one CYP pathway. The panel agreed that the average daily dose
of the metabolite (mg metabolite/L liver/day or "RISK") in the liver is a reasonable dosimeter.
However, the experts in PBPK modeling made one important recommendation. It was suggested
that the Km value for the high-affinity pathway estimated in the human model (value of 16 jiM in
the document) should be reevaluated based on in vitro human data or estimated from the
available animal data. It was recommended that the Km be the same as that used in the animal
model for VC. A larger value is not supported because large species differences in rates and
substrate specificity of CYP2E1 have not been described in the literature. Metabolism of most
2E1 substrates is limited by hepatic blood flow (which varies approximately by twofold)
delivering the substrate to the liver, and thus, large variations (e.g., tenfold) between individuals
in metabolic rate would not be likely. The other input parameters appear to be appropriate.
Response: The Km for human metabolism was changed to that used in the animal models. The
quantitative estimates were adjusted accordingly.
B. Should a new reproduction study published since the last review be used to
quantitate noncancer risk?
An overview of the developmental and two-generation reproductive toxicity tests, which
were reported subsequent to the first IRIS peer review panel for vinyl chloride in 1997, was
presented (CMA, 1998a,b). The studies were well conducted and adequately summarized in the
Toxicological Review. Although an error in the Toxicological Review on the incidence of
acidophilic foci was noted, the panel concurred with the parental (systemic effects) NOAEL (10
ppm) and developmental and reproductive NOAELs (1,100 ppm). The basis for the increased
incidence of altered hepatocellular foci in the P2 generation as increased susceptibility of the
fetal or neonatal animals or increased exposure duration of the P2 generation could not be
determined, but was adequately discussed in the document.
The panel agreed that the RfD/RfC should be based on the lifetime Til et al. study (1983;
1991) and that the results of the two-generation study do not call for a change in the RfD. The
addition of the reproduction study strengthens the database, but the high quality of the Til et al.
study and the toxicity observed at the lower doses in this study justify its selection as the
principal study for the RfD/RfC. The liver centrilobular hypertrophy and increased liver weight
in the reproductive/developmental study may be an adaptive response, but several panel
members stated that they considered it minimally adverse.
Response: No changes were made in the studies used to develop the RfC/RfD.
C. Is the use of a route-of-exposure extrapolation suitable to derive an RfC using the
feeding study reported by Til et al. (1983,1991)?
E-10
-------
The panel agreed that it was reasonable to use the PBPK model to extrapolate from the oral
to the inhalation route using the metabolite formed in the liver as the dose surrogate ("RISK").
The critical effect is a systemic effect, and the observation of hepatic toxicity at lower doses in
the oral studies is consistent with the known distribution of blood flow and physiology in
animals. Blood flow and other physiological differences between routes are accommodated by
the PBPK model. The sensitivity and Monte Carlo analyses account for the uncertainty in oral
absorption rate (KA), which may have a significant impact on "RISK." One reviewer noted that
EPA needs to assure that lung toxicity is adequately described to rule out portal-of-entry effects
fortheRfC.
Response: Lung toxicity will be described.
D. Should a body surface area correction be employed to account for species differences
(e.g., toxicodynamic differences) not accounted for by the model?
It is appropriate to use a surface area adjustment to account for differences in physiology
(toxicokinetics), but not for differences in susceptibility (toxicodynamics). For VC, the panel
stated that a body surface area correction is not justified because it would "double count"
physiological factors that are already accommodated by the PBPK model. Furthermore, an
assumption of equal sensitivity across species (not greater sensitivity of humans) seems
reasonable for vinyl chloride, and there is no scientific basis for an additional adjustment for
toxicodynamics. In fact, it was suggested that if a dynamic adjustment was made it should be a
factor of less than 1 rather than greater, based upon evidence that humans are likely to be less
sensitive to cancer induction by VC than are laboratory animals. It was also pointed out that a
conservative adjustment is already included by the use of a 95% upper confidence limit.
Response: A body surface correction was not incorporated in the dose-response estimates for
cancer.
E. Is a threefold adjustment (or some other value) appropriate to account for the
possibility of tumors at sites other than the liver?
The panel felt that an additional UF of 3 to account for the possibility of tumor induction at
sites other than the liver is neither justified by the data nor supported by the discussion in the
Toxicological Review. Because nonliver tumors in epidemiological and animal studies are
sporadic and do not show statistical significance or a dose-response relationship, it is difficult to
scientifically support an additional threefold quantitative adjustment to a potency estimate that is
based on the most sensitive tumor endpoint well supported by mechanistic information. The
discussion of possible sex differences in tumor induction at sites other than the liver in rodents
does not necessarily mean differences are expected in humans. Apparent sex differences are
more pronounced in rats compared to other species, or may be related to different compartment
sizes between males and females.
One suggestion was for EPA to combine tumors for all sites and perhaps put an upper
bound on the potency estimate. The panel members understood EPA's attempt to be prudent and
E-ll
-------
protective of human health, but agreed that if EPA feels the need to retain this UF, then it should
be better supported and indicated as an expert judgment call due to database deficiencies such as
inadequate exposure information, or stated as a policy decision.
Response: The threefold uncertainty factor was eliminated.
F. Is a twofold adjustment for early life exposure appropriate?
Some of the reviewers agreed with the use of a twofold adjustment, but not for the reasons
cited in the document. The evidence for a greater sensitivity during early life is at best
suggestive and the basis is unknown. While it is possible that neonatal animals may be more
sensitive (i.e., increased response to the same tissue dose), the increased tumor incidence might
also be due to a greater internal dose for the weanling relative to the adult for the same
administered dose, increased potential exposure time, and/or increased expression time for
tumors. Use of a small adjustment might be appropriate, but needs to be articulated by EPA as a
prudent decision based on the above exposure considerations, or based on policy. It was
suggested that if an adjustment for early-life exposure were to be made, then it should be based
on dosimetric considerations.
Other reviewers felt the twofold adjustment to potency to account for additional
exposure/expression time for exposed children does not appear to be justified on the basis of the
marginal quality of the studies cited, metabolic differences in young vs. adult animals, and large
species-specific differences in developmental biology and hormonal signaling.
Response: Although the evidence is not definitive, it appears that laboratory species are more
sensitive to cancer induction by VC when exposed very early in life. Such data are not directly
translatable to humans because rats and mice are less mature at birth. Metabolic pathways, DNA
repair mechanisms, etc. may be less developed. Nevertheless such data are strongly suggestive
of early-life sensitivity. Furthermore, dose per unit body weight will be greater in the very
young. For example, infants have a very high intake of liquids. Finally, exposure during early
life allows more time for tumor development. For these reasons it was considered prudent to
retain the twofold adjustment. The justification, however, was changed to include dosimetry
considerations.
G. Do the epidemiological studies published prior to the CMA (1998b) study provide an
adequate basis for the cancer dose-response assessment?
The majority of the panel agreed that epidemiology studies published prior to the CMA
study do not present adequate exposure characterization for dose-response assessment. Further,
results are equivocal for a causal association with tumors at sites other than the liver. The
Simonato et al. study had a reasonable number of liver cancer deaths where exposure was able to
be estimated. This study, however, included workers from several European factories, with
possible wide variations in exposure levels. In the Fox and Collier study, there were only four
men with liver cancer and estimated exposure; the Jones et al. study included seven deaths from
E-12
-------
liver cancer with information about duration of exposure, but the small number of cases resulted
in wide uncertainty bands for relative risk.
The animal data are adequate and reliable as the basis for the dose-response. The site
concordance between humans and animals for the major tumor site and the relative consistency
between the human exposure-response and animal dose-response predictions increase the
confidence in the overall risk assessment. An update of the Simonato study (European cohort) is
expected near the time of completion of this document. It was not known whether there will be
additional information to update the previous exposure-response analysis. More studies are
needed in the area of human exposure assessment.
Most of the panel members felt the human data could be used to quantitatively provide
bounds for risk estimates for hepatocellular tumors and angiosarcomas of the liver (i.e., a
"reality" check), and to support the animal data as presented in the Toxicological Review (i.e.,
Appendix B). Even though the estimated exposure histories may only be accurate to within an
order of magnitude, and considerable variation is expected in exposure levels among jobs,
plants, and employment periods, it is possible to compare the estimated exposure-response in
male workers (liver cancer/ angiosarcoma) with the dose-response seen in animal studies. The
agreement between the VC animal data and epidemiological data is extremely important and
helps evaluate the plausibility of the risks predicted on the basis of the animal data. However,
acknowledged problems of estimating exposures would tend to limit the usefulness of the human
data without the available animal data. The wide bounds for the relative risk estimates, because
of the small number of liver cancer cases in most cohorts, magnify the level of uncertainty.
One panel member thought that if the human exposure estimates were not considered
accurate, they should not be used for comparing exposure-response with those derived from
using animal data.
Response: EPA agrees with the majority opinion of the panel and continues to rely on animal
data for recommended risk estimates.
H. Does the CMA study provide adequate exposure characterization to quantitate
humans' risk for liver cancer?
A brief review of the CMA epidemiological study (1998b) was presented to the panel.
The re-analysis and update of the analysis of the cohort mortality were well conducted, the
discussion of the results was complete, and there were no major differences in opinion with the
author's interpretation of the results. Although the study results add to the overall weight of
evidence, there were no exposure data (analyses were based on relative duration of
employment), precluding use of this study for dose-response assessment. The study confirms the
strong causal association of vinyl chloride with liver cancers, especially angiosarcoma of the
liver. Results suggest a possible association with brain cancer and connective and soft tissue
cancers. However, caution was urged in the interpretation of these two associations. A casual
interpretation is not possible with the current epidemiological evidence. The study found no
excess in mortality for lung cancer, cancers of the lymphatic and hematopoietic system, or
E-13
-------
emphysema and pneumoconioses and other respiratory diseases (the category that includes
chronic obstructive pulmonary disease); each of these causes showed mortality deficits. One
limitation of this study, as well as of most other VC epidemiology studies, is the inability to
assess breast cancer risks for women. Few women have been occupationally exposed to VC.
The study did consider breast cancer risks in men; for the period 1942-1995, two deaths were
attributed to breast cancer (SMR = 190; 95% CI=23-687). This SMR is statistically unstable
and provides no useful information to evaluate breast cancer risks in women.
Response: EPA is in agreement with the panel findings.
(3) Additional Comments on Issues Deemed Important by the Panelists
A. It was suggested that use of a tenfold UF for intra-individual variation (sensitive human
subpopulations) is overly conservative because differences in rate of hepatic blood flow (~2-
fold) and metabolic enzyme activity (e.g., CYP2E1) do not vary by that magnitude. For
example, a simulation of a fivefold increase in enzyme induction (Vm) led to only a 50%
increase in VC metabolite formation ("RISK") in the liver. Thus, a threefold (or less) UF should
be sufficiently protective.
Response: The Agency considers this analysis and proposal to have merit. However, based on
the lack of direct or indirect information on which to confirm this action, the tenfold UF will
remain. The Agency notes this suggestion and shall pursue this venue in subsequent
assessments.
(4) Panel Recommendations
The panel felt the document was of high quality and contained most of the relevant
information, and that EPA should be commended for this Toxicological Review. Based on their
reading and analysis of the information provided, the panel agreed that their overall
recommendation for the IRIS materials is: acceptable with minor revision.
(5) Observer Comments
A. The statement that the epidemiologic evidence is suggestive for cancer of the brain
and lymphopoietic system is an overinterpretation of the studies.
Excess brain cancers or other tumors, with the exception of liver angiosarcomas, are not
related to VC exposure. According to reviews by Doll and by Blair (NCI), VC is associated only
with angio sarcomas. These tumors are due to high occupational exposure levels that have not
existed for a couple of decades.
Response: The discussion was modified to eliminate "highly suggestive," but there is still some
concern because, while increases in relative risk were considerably smaller than for liver cancer,
they were still statistically significant in the CMA cohort.
E-14
-------
B. The use of an arbitrary threefold UF for cancer potency is not warranted because
animal and human potencies are similar.
It is not justified on a scientific basis and it is not consistent with EPA's current practice to
use UFs to adjust potency.
Response: The threefold UF has been eliminated.
C. Exposure data can be estimated from the human studies to within an order of
magnitude.
More can be done with the human data in the Toxicological Review to bound the cancer
estimates. There is greater uncertainty in using the animal data.
Response: EPA epidemiologists as well as epidemiologists on both expert panels concluded that
the epidemiologic data were too uncertain to base a quantitative risk estimate on. Although
exposure may be estimated within an order of magnitude, the uncertainty in relative risk
estimates, due to small numbers of liver tumors, will magnify the uncertainty.
D. Calculations in the Toxicological Review need to be checked when the twofold
uncertainty factor is applied.
It appears as if these calculations have been applied twice in the summary sheet.
Response: All the risk estimates have been recalculated because of the change in the km value.
Any errors in the twofold factor have been corrected.
E. There may be some justification for use of other tumors to estimate an upper bound.
Response: Induction of liver tumors, both hepatocellular tumors and angiosarcomas, has been
the most consistent response to VC exposure. Because the liver is the most sensitive site and VC
is activated by the liver, there is little justification for use of other tumors to estimate an upper
bound. The panel indicated that protection against liver cancer will also protect against induction
of other tumors.
E-15
------- |