EPA/635/R-06/001
www.epa.gov/iris
TOXICOLOGICAL REVIEW
OF
PHOSGENE
(CAS No. 75-44-5)
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
December 2005
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
11
-------
CONTENTS - TOXICOLOGICAL REVIEW OF PHOSGENE (CAS No. 75-44-5)
LIST OF TABLES v
LIST OF FIGURES v
LIST OF ABBREVIATIONS AND ACRONYMS vi
FOREWORD viii
AUTHORS, CONTRIBUTORS, AND REVIEWERS ix
1. INTRODUCTION 1
2. CHEMICAL AND PHYSICAL INFORMATION 3
3. TOXICOKINETICS 5
4. HAZARD IDENTIFICATION 6
4.1. STUDIES IN HUMANS - EPIDEMIOLOGY, CASE REPORTS 6
4.1.1. Acute Inhalation Exposure 6
4.1.2. Subchronic Inhalation Exposures 7
4.1.3. Occupational Epidemiology Studies 7
4.2. ACUTE/SUBCHRONIC AND CHRONIC STUDIES IN ANIMALS 9
4.2.1. Oral Exposures 9
4.2.2. Inhalation Exposures 9
4.2.2.1. Acute Exposures 9
4.2.2.2. Subchronic Exposures 10
4.3. REPRODUCTIVE/DEVELOPMENTAL TOXICITY STUDIES 17
4.4. OTHER EFFECTS 18
4.4.1. Dermal Toxicity 18
4.4.2. Ocular Toxicity 18
4.4.3. Neurotoxicity 18
4.4.4. Genotoxicity 18
4.5. SYNTHESIS AND EVALUATION OF MAJORNONCANCER EFFECTS .... 18
4.5.1. Oral 18
4.5.2. Inhalation 19
4.5.3. Mode of Action Information 19
4.6. WEIGHT-OF-EVIDENCE EVALUATION AND CANCER
CHARACTERIZATION 21
4.7. SUSCEPTIBLE POPULATIONS AND LIFE STAGES 22
4.7.1. Possible Childhood Susceptibility 22
4.7.2. Possible Gender Differences 22
4.7.3. Other 22
5. DOSE-RESPONSE ASSESSMENTS 23
5.1. ORAL REFERENCE DOSE (RfD) 23
5.2. INHALATION REFERENCE CONCENTRATION (RfC) 23
5.2.1. Choice of Principal Study and Critical Effect(s) 23
in
-------
5.2.2. Methods of Analysis for Point of Departure, Including Application of
Models (BMD, NOAEL/LOAEL, and CatReg) 24
5.2.3. BMD Approach 25
5.2.4. NOAEL/LOAEL Approach 30
5.2.5. CatReg Approach 31
5.2.6. Comparison of Approaches 32
5.2.7. RfC Derivation: Application of Uncertainty Factors 34
5.2.8. Previous RfC Assessment 37
5.3. CANCER ASSESSMENT 37
5.3.1. Oral Slope Factor 37
5.3.2. Inhalation Unit Risk 37
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF
HAZARD AND DOSE RESPONSE 38
6.1. HUMAN HAZARD POTENTIAL 38
6.2. DOSE RESPONSE 39
6.2.1. Noncancer/Oral 39
6.2.2. Noncancer/Inhalation 39
6.2.3. Cancer/Oral and Inhalation 40
7. REFERENCES 41
APPENDIX A: ACUTE EXPOSURE GUIDELINE LEVELS (AEGLs) FOR PHOSGENE A-l
APPENDIX B: BMD AND CATREG ANALYSES B-l
APPENDIX C: SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC
COMMENTS AND DISPOSITION C-l
IV
-------
LIST OF TABLES
Table 1. Histopathology incidence indicating the number of animals affected following
phosgene exposure (from Kodavanti et al., 1997) 12
Table 2. Pulmonary histopathology severity score in rats following subchronic phosgene
exposure (from Kodavanti et al., 1997) 13
Table 3. Benchmark dose results from a subchronic study in rats (Kodavanti et al., 1997) ... 26
Table 4. Results of CatReg analysis of severity-graded lung lesions reported by Kodavanti
et al. (1997) 32
Table 5. Application of uncertainty factors (UFs) for two different approaches for deriving the
RfC 37
LIST OF FIGURES
Figure 1. Increased collagen staining of terminal bronchiole/peribronchiolar region
(multistage model) 28
-------
LIST OF ABBREVIATIONS AND ACRONYMS
AEGL Acute Exposure Guideline Level
ARE Acute reference exposure
BAL Bronchio-alveolar lavage
BMD Benchmark dose
BMDL Lower-bound confidence limit on the benchmark dose
BMDS Benchmark Dose Software
BMR Benchmark response
C Concentration
CatReg Categorical regression
CL Confidence limit
CO2 Carbon dioxide
EC Effective concentration
EPA U.S. Environmental Protection Agency
ERD Extra risk dose
FVC Forced vital capacity
G6PD Glucose 6-phosphate dehydrogenase
HC1 Hydrochloric acid
HEC Human equivalent concentration
IRIS Integrated Risk Information System
LOAEL Lowest-observed-adverse-effect level
LOEL Lowest-observed-effect level
MW Molecular weight
NCTR National Center for Toxicological Research
NOAEL No-observed-adverse-effect level
NOEL No-observed-effect level
NPSH Nonprotein sulfhydryl
NTP National Toxicology Program
PBPK Physiologically based pharmacokinetic
PMN Polymorphonuclear
POD Point of departure
RfC Reference concentration
RfD Reference dose
RGDR Regional gas-dose ratio
SAR Structure-activity relationship
SMR Standard mortality ratio
T Time
UF Uncertainty factor
Units of Measure
mg/kg Milligrams per kilogram body weight
mg/m3 Milligrams per cubic meter
ng/m3 Nanograms per cubic meter
ppb Parts per billion
VI
-------
ppm Parts per million
ppt Parts per trillion
vn
-------
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 phosgene.
It is not intended to be a comprehensive treatise on the chemical or toxicological nature of
phosgene.
In Section 6, Major Conclusions in the Characterization of Hazard and Dose Response,
EPA has characterized its overall confidence in the quantitative and qualitative aspects of hazard
and dose response by addressing knowledge gaps, uncertainties, quality of data, and scientific
controversies. The discussion is intended to convey 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 IRIS Hotline at 202-566-1676 (phone), 202-566-1749 (fax), or
hotline.iris@epa.gov (email address).
Vlll
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGER
Dharm Singh
National Center for Environmental Assessment-Washington Office
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
AUTHORS
Jeff Gift
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Robert McGaughy
National Center for Environmental Assessment-Washington Office
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
Babasaheb Sonawane
National Center for Environmental Assessment-Washington Office
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
CONTRIBUTOR
Femi Adeshina, National Center for Environmental Assessment, Washington, DC
REVIEWERS
This document and the accompanying IRIS Summary have been peer reviewed by EPA
scientists and independent scientists external to EPA. Comments from all peer reviewers were
evaluated carefully and considered by the Agency during the fmalization of this assessment.
During the fmalization process, the IRIS Program Director achieved common understanding of
the assessment 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, Economics, and Innovation; Office of Children's
Health Protection; Office of Environmental Information; and EPA's regional offices.
IX
-------
INTERNAL EPA REVIEWERS
Robert Bruce, National Center for Environmental Assessment, Cincinnati, OH
David Chen, Office of Children's Health Protection, Washington, DC
James Cogliano, National Center for Environmental Assessment, Washington, DC
Julie Du, Office of Science and Technology, Office of Water, Washington, DC
Annie Jarabek, National Center for Environmental Assessment, Research Triangle Park, NC
Urmila Kodavanti, National Health and Environmental Effects Research Laboratory, Research
Triangle Park, NC
Deirdre Murphy, Office of Air Quality Planning and Standards/Office of Air and Radiation,
Research Triangle Park, NC
Bruce Rodan, National Center for Environmental Assessment, Immediate Office
Michel Stevens, National Center for Environmental Assessment, Research Triangle Park, NC
Paul White, National Center for Environmental Assessment, Washington, DC
Tracey Woodruff, Office of Policy, Economics and Innovation, Washington, DC
EXTERNAL PEER REVIEWERS
Walter Piegorsch, Ph.D., Professor of Statistics, University of South Carolina, Columbia, SC
Andrew Salmon, D. Phil., Chief, Air and Toxics Risk Assessment, Cal-EPA
Hanspeter Witschi, MD, Professor emeritus, University of California-Davis
A.M. Sciuto, Ph.D., U.S. Army Medical Research Institute of Chemical Defense, Aberdeen, MD
Summaries of the external peer reviewers' comments and public comments and the
disposition of their recommendations are provided in Appendix C.
ACKNOWLEDGMENTS
A preliminary draft of the document was prepared by Bruce Buxton, Patricia McGinnis,
and Mark Osier of Battelle Memorial Institute, Columbus, Ohio, under EPA Contract No. 68-C-
00-122.
-------
1. INTRODUCTION
This document presents background information and justification for the Integrated Risk
Information System (IRIS) Summary of the hazard and dose-response assessment of phosgene.
IRIS Summaries may include an oral reference dose (RfD), an inhalation reference concentration
(RfC), and a carcinogenicity assessment.
The RfD and RfC provide quantitative information for use in risk assessments for health
effects known or assumed to be produced through a nonlinear (possibly threshold) mode of
action. The RfD (expressed in units of mg/kg-day) is defined as 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 effects during a
lifetime. The inhalation RfC (expressed in units of mg/m3) 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).
This document does not attempt to develop concentration values protective of acute
toxicity. For reference purposes, Appendix A presents a summary of the phosgene Acute
Exposure Guideline Levels (AEGLs) that was prepared by the National Academy of Sciences.
The carcinogenicity assessment provides information on the carcinogenic hazard
potential of the substance in question and quantitative estimates of risk from oral 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. The quantitative risk estimates are derived from the application of a low-dose
extrapolation procedure, and are presented in two ways to better facilitate their use. First, route-
specific risk values are presented. The "oral slope factor" is an upper bound on the estimate of
risk per mg/kg-day of oral exposure. Similarly, a "unit risk" is an upper bound on the estimate
of risk per unit of concentration, either per |ig/L drinking water or per |ig/m3 air breathed.
Second, the estimated concentration of the chemical substance in drinking water or air when
associated with cancer risks of 1 in 10,000, 1 in 100,000, or 1 in 1,000,000 is also provided.
Development of these hazard identification and dose-response assessments for phosgene
has followed the general guidelines for risk assessment as set forth by the National Research
Council (NRC, 1983). EPA guidelines that were used in the development of this assessment
may include the following: Guidelines for the Health Risk Assessment of Chemical Mixtures
(U.S. EPA, 1986a), Guidelines for Mutagenicity Risk Assessment (U.S. EPA, 1986b), Guidelines
for Developmental Toxicity Risk Assessment (U.S. EPA, 1991), Guidelines for Reproductive
1
-------
Toxicity Risk Assessment (U.S. EPA, 1996a), Guidelines for Neurotoxicity Risk Assessment (U.S.
EPA, 1998a), Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), Supplemental
Guidance for Assessing Susceptibility from Early-Life Exposures to Carcinogens (U.S. EPA,
2005b), 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), Use of the Benchmark Dose Approach in Health Risk Assessment (U.S. EPA, 1995),
Science Policy Council Handbook: Peer Review (1st and 2nd editions) (U.S. EPA, 1998b, 2000a),
Science Policy Council Handbook: Risk Characterization (U.S. EPA, 2000b), Benchmark Dose
Technical Guidance Document (U.S. EPA, 2000c), Supplementary Guidance for Conducting
Health Risk Assessment of Chemical Mixtures (U.S. EPA, 2000d), and^4 Review of the Reference
Dose and Reference Concentration Processes (U.S. EPA, 2002).
The literature search strategy employed for this compound was based on the CASRN and
at least one common name. Any pertinent scientific information submitted by the public to the
IRIS Submission Desk was also considered in the development of this document. The relevant
literature was reviewed through March 2004.
-------
2. CHEMICAL AND PHYSICAL INFORMATION
Phosgene is also known as carbon dichloride oxide, carbonic dichloride, carbon
oxychloride, carbonyl chloride, carbonyl dichloride, and chloroformyl chloride. Some relevant
physical and chemical properties are listed below (NTP, 2001; WHO, 1997):
CAS number: 75-44-5
Structural formula: COC12
Molecular weight: 98.92
Vapor pressure: 1,180 mmHg at 20°C
Water solubility: slight, reacts with water
Boiling point: 8.2°C
Odor threshold: 0.4 to 1.5 ppm
Irritation threshold: 3 ppm
Conversion factor: 1 ppm = 4.05 mg/m3, 1 mg/m3 = 0.247 ppm (25°C, 760 mmHg)
Phosgene is primarily used in the polyurethane industry for the production of polymeric
isocyanates (WHO, 1998, 1997; U.S. EPA, 1986c, 1984). Phosgene is also used in the
polycarbonate industry and in the manufacture of carbamates and related pesticides, dyes,
perfumes, Pharmaceuticals, and isocyanates. The majority of phosgene for industrial
applications is made on site by the reaction of carbon monoxide and chlorine gas using an
activated carbon catalyst. Phosgene may also be produced as a combustion product of carbon
tetrachloride, methylene chloride, trichloroethylene, or butyl chloroformate, although these
methods are not utilized industrially. Estimated worldwide production exceeds 5 billion pounds
(WHO, 1997). Phosgene is a colorless gas at room temperature with an odor ranging from
strong and stifling when concentrated to hay-like when diluted. Phosgene is slightly soluble in
aqueous media, but, when dissolved, it is very rapidly hydrolyzed to carbon dioxide (CO2) and
hydrochloric acid (HC1), with a half-life at 37°C of approximately 0.026 seconds (Schneider and
Diller, 1989; Manogue and Pigford, 1960).
Phosgene levels have been measured in ambient air (U.S. EPA, 1983; Singh et al., 1981,
1977; Singh, 1976). Multiple samples (10-257) were taken from four locations in California
within a 24-hour period. The average level for three clean-air (rural and coastal) locations was
87 ng/m3 (21.7 ppt). In the Los Angeles basin, the average was 127.2 ng/m3 (31.8 ppt). These
values were also cited by the World Health Organization (WHO, 1997). Kelly et al. (1994) also
reported concentrations and transformations of hazardous air pollutants. They reported that the
-------
phosgene ambient concentration median was 80 ng/m3.
Inhalation is the primary exposure route for phosgene. Suspected sources of atmospheric
phosgene are fugitive emissions, thermal decomposition of chlorinated hydrocarbons, and photo-
oxidation of chloroethylenes. Although the existence of atmospheric sinks for phosgene has
been questioned, it is postulated that phosgene's removal from the atmosphere is rather slow
(Singh etal., 1977).
The American Conference of Governmental Industrial Hygienists (ACGIH, 2000)
recommends a time-weighted average of 0.1 ppm (0.4 mg/m3) to protect against irritation,
anoxia, and pulmonary edema. The National Institute for Occupational Safety and Health
(NIOSH, 2001) recommended exposure limit is 0.1 ppm, and the Occupational Safety and Health
Administration (OSHA, 1993) has promulgated an 8-hour permissible exposure limit of 0.1 ppm.
-------
3. TOXICOKINETICS
Phosgene is rapidly hydrolyzed in aqueous solution to CO2 and HC1, which are likely to
be exhaled (Schneider and Diller, 1989; Diller, 1985; Diller et al., 1979). Consequently,
phosgene is not expected to leave the pulmonary circulation following inhalation exposure, nor
is exposure by the oral route likely (WHO, 1998, 1997; U.S. EPA, 1986c, 1984). Data on
phosgene absorption are not available. Phosgene is electrophilic and undergoes attack by a
variety of nucleophiles. The predominant reaction is hydrolysis by water yielding carbon
dioxide and hydrochloric acid. It also reacts with a wide variety of nucleophiles, including
primary and secondary amines, hydroxy groups, and thiols. In addition, it also reacts with
macromolecules, such as enzymes, proteins, or other polar phospholipids, resulting in formation
of covalent adducts that can interfere with molecular functions. The loss of enzyme activity may
lead to loss of cellular function and cell death (reviewed in WHO, 1998). Studies on the
distribution and elimination of phosgene in animals or humans were not located in the published
literature.
Phosgene is thought to participate directly in acylation reactions with amino, hydroxyl, or
sulfhydryl groups (WHO, 1998, 1997; Schneider and Diller, 1989; U.S. EPA, 1986c; Diller,
1985). Formation of phosgene as a metabolite of other compounds has been hypothesized
(reviewed in U.S. EPA, 2001, 1984) but not directly measured, perhaps owing to the rapid
reaction of phosgene with tissue molecules or hydrolysis in aqueous solution. Phosgene is
believed to be the major intermediate metabolite of chloroform (oxidative metabolism) (U.S.
EPA, 200la). Despite rapid conversion of phosgene to less "toxic" end products, other systemic
effects, such as permeability-related edema (Borak and Diller, 2001) and adenosine triphosphate-
related changes (Currie et al., 1987), have been noted.
-------
4. HAZARD IDENTIFICATION
4.1. STUDIES IN HUMANS - EPIDEMIOLOGY, CASE REPORTS
As noted in Section 2, phosgene is a gas at room temperature, and aqueous phosgene
rapidly hydrolyzes to CO2, and HC1; consequently, exposure by the oral route is highly unlikely.
Diller and Zante (1982) performed an extensive literature review of human effects from
phosgene inhalation exposure and found that a great majority of data were anecdotal or rough
estimates and, thus, did not contain reliable exposure concentrations and/or durations. Many
case reports describe symptomology and postmortem results from human phosgene poisonings;
however, exposure concentrations were not reported.
4.1.1. Acute Inhalation Exposure
The acute toxicity of phosgene inhalation has been well documented in humans (WHO,
1998, 1997; U.S. EPA, 1986c, 1984; Underbill, 1919).
Inhalation of phosgene at high concentrations results in a sequence of events, including
an initial bioprotective phase, a symptom-free latent period, and a terminal phase characterized
by pulmonary edema (Schneider and Diller, 1989; Diller, 1985). In the initial phase, high
concentrations (>3 ppm) may result in a vagal reflex action that causes frequent, shallow
respiration and decreased respiratory vital capacity and volume. This, in turn, leads to a
decreased arterial CO2 pressure increase and decreased blood pH. After cessation of exposure,
the reflex syndrome shows a tendency to regress.
In the second phase, which may last for several hours postexposure, clinical signs and
symptoms are generally lacking (Schneider and Diller, 1989; Diller, 1985). However, histologic
examination reveals the beginnings of an edematous swelling, with blood plasma increasingly
entering the pulmonary interstitium and alveoli. This may result in damage to the alveolar type I
cells and a rise in hematocrit. In exposed humans, the individual is unaware of these processes;
thus, this phase is termed the "clinical latent phase." The length of this phase varies inversely
with the inhaled dose.
In the third clinical phase of phosgene toxicity (Schneider and Diller, 1989; Diller, 1985),
the accumulating fluid in the lung results in the edema becoming apparent both directly and
indirectly. The severity of the edema increases, potentially resulting in decreased gas exchange
as the fluid gradually rises from the alveoli to the proximal segments of the respiratory tract.
Agitated respiration may cause the protein-rich fluid to take on a frothy consistency. A severe
edema may result in an increased concentration of hemoglobin in the blood and congestion of the
alveolar capillaries. At sufficiently high exposure levels, the heart also may be affected,
-------
resulting in cardiac failure due to pulmonary congestion. In general, this phase peaks
approximately 24 hours after an acute exposure and, assuming lethality does not occur, recedes
over the next 3 to 5 days.
A case history of phosgene poisoning was reported in a 45-year-old welder who had
symptoms of dyspnea and weakness (Glass et al., 1971). The authors concluded that phosgene
poisoning causes reduction of forced vital capacity (FVC), airway obstruction, arterial hypoxia,
and impaired co-transfer. Some of the pulmonary events precipitated by phosgene exposure,
such as neutrophil and leukocyte infiltration, edema, and bronchial dilation, are also observed in
asthmatics in the presence of ozone and nitrous oxide. Although the mechanisms for the
phosgene-produced effect (acylation) and the ozone and nitrous oxide effect (oxidation) are
presumed to be different, the resulting health endpoint appears to be similar (Jaskot et al., 1991)
because phosgene acts as a lung irritant.
Cases of acute phosgene toxicity associated with two large-scale releases of phosgene in
Germany and Japan have been reported. In Hamburg, Germany, on May 20, 1928, 11 metric
tons (24,640 pounds) of "pure" phosgene escaped from a storage tank, resulting in a large-scale
exposure to the airborne gas (Hegler, 1928; Wohlwill, 1928, both cited in U.S. EPA, 1986c). A
total of 300 people—some located as far as 6 miles from the site—reported illness within a few
days of the release. Of those, 10 died as a result of the exposure. One hospital reported
admitting 195 victims on the night of May 20. Of those, 17 were very ill, 15 were moderately ill,
and the rest were only slightly affected. Autopsy of six of the fatalities revealed abnormalities
primarily in the lungs. Occasional lesions of the kidney, liver, and heart were observed.
In November 1966, phosgene was accidentally released from a factory in Japan
(Sakakibara et al., 1967, cited in WHO, 1997). A total of 382 people were reported poisoned, 12
of whom were hospitalized. Signs and symptoms of exposure in the 12 hospitalized patients
included headache, nausea, cough, dyspnea, fatigue, pharyngeal pain, chest tightness, chest pain,
and fever. Seven patients showed evidence of pulmonary edema, as revealed by chest x-ray 48
hours postexposure. One patient reported lacrimation and redness of the eyes.
4.1.2. Subchronic Inhalation Exposures
Galdston et al. (1947) reported six cases (four women, two men) of phosgene exposure,
with exposure ranging from 1 to 24 months. Common symptoms included rapid, shallow
breathing; high minute volume; and low oxygen extraction. The measurable changes in
pulmonary function that were consistently observed varied in type and severity, but they could
not be correlated with the severity of phosgene intoxication or with chronic symptoms.
4.1.3. Occupational Epidemiology Studies
-------
The effect of occupational exposure to phosgene was examined in workers employed
from 1943 to 1945 at a uranium processing plant in the United States (Polednak and Hollis,
1985; Polednak, 1980). In the initial report (Polednak, 1980), a comparison was made between a
group of 699 male workers exposed daily to phosgene and 9,352 male controls employed during
the same time period but not exposed to phosgene. The duration of exposure was generally 2
months to 1 year; the followup period was 30 years. Exposure levels were not reported but were
instead described as "low" (undetectable), with the level exceeding 1 ppm four to five times
daily. Standard mortality ratios (SMRs) for respiratory diseases were not significantly different
between controls (SMR = 113, 95% confidence limit [CL] = 98-130) and exposed workers
(SMR = 78, 95% CL = 31-161) relative to cause- and age-specific death rates for white males in
the United States. Likewise, no differences in the SMRs for lung cancer were found between
controls (SMR =113, 95% CL = 97-131) and exposed workers (SMR = 127, 95% CL =
66-222). No significant differences were found between controls and exposed workers for any
other cause of death.
Interestingly, it should be noted that approximately 30 years after exposure, this cohort
showed no statistically significant increases in mortality from overall cancer, from cancers at
specific anatomical sites, or from diseases of the respiratory system or in overall mortality.
However, the exposure period covered by the study was short, exposed groups were small, and
exposure levels were not well documented. Consequently, evidence presented in this study is
inadequate to assess the chronic toxicity or carcinogenicity of phosgene.
In the followup study (Polednak and Hollis, 1985), the number of subjects had decreased
to 694 male workers who were exposed daily to phosgene and 9,280 male controls who were
employed in the same plant but not exposed to phosgene. The SMRs for respiratory diseases
were not significantly different between controls (SMR = 119, 95% CL = 106-133) and exposed
workers (SMR = 107, 95% CL = 59-180). Likewise, no differences in the SMRs for lung cancer
were found between controls (SMR =118, 95% CL = 105-133) and exposed workers (SMR =
122, 95% CL = 72-193). No significant differences were found between controls and exposed
workers for any other cause of death. The authors pointed out, however, that because of the small
sample sizes, only large differences in mortality rates would have been detected in these studies.
Polednak and Hollis (Polednak and Hollis, 1985; Polednak, 1980) also examined a
subgroup of 106 men who were exposed to high levels of phosgene (thought to be 50 ppm-min
or greater) as a result of accidental workplace exposures. The reported overall SMR for all
causes for exposed workers was 109 (95% CL = 73-157) in the 1980 study and 121 (95% CL =
86-165) in 1985 follow-up study. In the respiratory disease category, the SMR increased from
219 (3 deaths reported, 1.37 expected, 95% CL not reported) in the 1980 study to 266 (95% CL
= 86-622) in the 1985 study; however, several of these cases reported using tobacco, making the
-------
role of phosgene in the deaths uncertain. None of these values reached statistical significance.
An attempt was made in the 1985 study to analyze a similar cohort of 91 female workers also
exposed to approximately 50 ppm-min, but ascertainment of deaths and followup was less
certain for this group and prevented a full analysis. Approximately 35 years after exposure to
phosgene, no increase in overall mortality or mortality from cancer or respiratory disease was
noted in this cohort.
4.2. ACUTE/SUBCHRONIC AND CHRONIC STUDIES IN ANIMALS
4.2.1. Oral Exposures
No animal studies on the toxicity of phosgene following oral exposure were identified.
4.2.2. Inhalation Exposures
No chronic studies in experimental animals on the effects of inhaled phosgene were
located in the published literature. The majority of studies of phosgene are of acute duration,
spanning from minutes to several hours. However, several studies (Kodavanti et al., 1997;
Franch and Hatch, 1986; Clay and Rossing, 1964; Rossing, 1964) examined the effects of
repeated short-term, "acute" exposures over 2 to 12 weeks. These studies are described below.
4.2.2.1. Acute Exposures
A number of studies have examined the effects of acute phosgene exposure in animals. A
similar spectrum of effects was seen across the many species examined. Exposures were limited
to concentrations between 0.5 and 40 ppm (2 to 160 mg/m3) for intervals ranging from 5 minutes
to 8 hours.
Animals exposed to phosgene for a short duration show changes in breathing, including
decreased tidal volume and minute volume, increased breathing frequency (Lehnert, 1992), and
increased heart rate (Meek and Eyster, 1920). Exposed animals also show decreased body
weight relative to air-exposed animals (Lehnert, 1992). An increase in lung weight also has been
observed (Sciuto, 1998; Jaskot et al., 1991, 1989). After exposure to phosgene, lungs appear
voluminous and heavy, contain considerable amounts of pale yellow fluid, and show signs of
edema and emphysema (Ardran, 1950; Durlacher and Bunting, 1947). Exposure also results in
changes in bronchio-alveolar lavage (BAL) parameters, including increased protein (Jugg et al.,
1999; Sciuto, 1998; Jaskot et al., 1989; Slade et al., 1989; Hatch et al., 1986), phospholipid
content (Jugg et al., 1999), and enzyme levels (Lehnert, 1992; Jaskot et al., 1991), as well as
increases in the numbers of inflammatory cells (Lehnert, 1992). It has been reported that prior
acute exposure to phosgene is protective against the effects of a later acute exposure (Ohio and
Hatch, 1996; Box and Cullumbine, 1947).
-------
Histopathologic examination of the lungs of phosgene-exposed animals reveals exposure-
dependent edema and a progressive bronchiolar inflammatory response, with an infiltration of
polymorphonuclear cells and lymphocytes and the presence of extravasated erythrocytes (Jugg et
al., 1999; Lehnert, 1992; Keeler et al., 1990; Gross et al., 1965; Durlacher and Bunting, 1947;
Meek and Eyster, 1920). This condition progresses with increasing exposure, causing alveolar
hyperplasia, a progressive fibrotic response, and the gorging of capillaries with blood cells.
Following phosgene exposure, an initial increase in blood volume occurs, followed by a
significant decrease. With the resulting increase in hemoglobin concentration (Meek and Eyster,
1920), it is thought that the volume decrease is the result of fluid entering the lungs during
edema formation.
Acute exposure to phosgene also has been shown to result in a decreased immune
response in animals, as evidenced by an increased susceptibility to in vivo bacterial and injected
tumor cells (Belgrade et al., 1989) and viral (Ehrlich and Burleson, 1991) infections as well as a
decreased in vitro virus-killing and T-cell response (Burleson and Keyes, 1989; Ehrlich et al.,
1989). Belgrade et al. (1989) reported that a single 4-hour exposure to phosgene concentrations
as low as 0.025 ppm significantly enhanced mortality due to streptococcal infection in mice.
Furthermore when the exposure time was increased from 4 to 8 hours, a significant increase in
susceptibility to streptococcus was also seen at an exposure concentration of 0.01 ppm. The
authors attempted to establish a mechanism for these findings by measuring alveolar macrophage
activity. With intratracheal administration of bacteria, which delivers a much larger amount of
bacteria than the inhalation route used in the earlier experiments, phosgene concentrations of
0.25 ppm and higher, which is 10-fold higher than the lowest observable effect, had little or no
effect on alveolar macrophage phagocytic activity and little or no effect on total cells recovered,
viability, or differential cell counts in lavage fluid obtained shortly after exposure. The
mechanism(s) responsible for increased sensitivity to bacterial infection are unclear.
4.2.2.2. Subchronic Exposures
Kodavanti et al. (1997) exposed groups of male F344 rats to phosgene levels designed to
provide equal products of concentration times time (C x T) for all groups but the lowest
exposure concentration. Groups of eight rats were exposed to clean air (control) or phosgene for
6 hours per day as follows: to 0.1 ppm (0.4 mg/m3) for 5 days per week, to 0.2 ppm (0.8 mg/m3)
for 5 days per week, to 0.5 ppm (2 mg/m3) for 2 days per week, or to 1 ppm (4 mg/m3) for 1 day
per week for 4 or 12 weeks. Groups of similarly exposed rats were allowed clean air recovery
for 4 weeks after 12 weeks of exposure. The measured 12-week average concentrations were as
follows (mean ± SD range): 0.1 ppm group was 0.101 ± 0.002 (0.098-0.113); 0.2 ppm group
was 0.201 ± 0.002 (0.196-0.207); 0.5 ppm group was 0.505 ± 0.008 (0.495-0.536); 1 ppm group
10
-------
was 0.976 ± 0.03 (0.912-1.009). At the end of the exposure or recovery period, animals were
sacrificed, and the lungs were weighed and processed for histologic examination. The 0.5 ppm
histology samples were inadvertently lost, but other analyses were performed (e.g., BAL, lung
volume, and biochemical parameters).
No mortality was reported for any exposure level or time examined. However, small but
statistically significant decreases in body weight gain were reported in the 0.5 and 1 ppm rats at
both 4 and 12 weeks of exposure. A concentration-dependent increase in relative lung weight
was seen following both 4 and 12 weeks of exposure (statistically significant at 0.2 ppm or
greater). The phosgene concentration at either time point seemed to drive this response rather
than the C x T product. The effect appeared to be more severe at the end of 4 weeks than after
12 weeks of exposure. Phosgene also increased the lung displacement volume (an index of total
lung volume) in all exposed groups at 4 weeks and at 0.2 ppm or greater at 12 weeks of
exposure.
Histologic examination of animals exposed for 4 weeks revealed changes in the
bronchiolar regions, with a small but apparent thickening and mild inflammation seen at 0.1 ppm
that progressed in severity with concentration to a severe inflammation and thickening of the
terminal bronchiolar regions and alveolar walls at 1 ppm (Tables 1 and 2). An increase in
collagen staining, using Masson's trichrome stain, was seen in the 0.2 and 1 ppm animals,
although no elevation of pulmonary hydroxyproline, a measure of collagen deposition, was
observed.
11
-------
Table 1. Histopathology incidence indicating the number of animals affected following phosgene exposure"
(from Kodavanti et al., 1997)
Phosgene concentration (ppm)
Number of animals examined
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Terminal bronchiole,
peribronchiolar alveolus,
epithelial alteration
Terminal bronchiole/
peribronchiolar, increased
collagen staining
4 weeks
0.0
(12)
0
0
0
1
2
2
1
O.ld
(8)
0
2
1
2
3
4
1
0.2d
(8)
0
5b
2
2
8b
5b
8b
1.0e
(6)
2
6b
3
3
6b
6b
6b
12 weeks
0.0
(12)
0
0
0
0
1
0
2
O.ld
(8)
0
2
0
0
3
1
2
0.2d
(8)
0
4b
0
0
8b
7b
8b
1.0e
(8)
1
8b
1
3
8b
8b
8b
16 weeks c
0.0
(9)
2
0
0
0
0
0
2
O.ld
(6)
0
0
0
0
0
2
1
0.2d
(7)
0
0
0
0
1
3
7b
1.0e
(5)
0
0
0
0
1
2
5b
a Number in each column indicates number of animals affected (of total numbers used in analysis).
b Statistically significant compared with unexposed (0.0 ppm) group, pO.Ol (pairwise Fisher's exact test).
0 Indicates 12-week exposure, followed by a 4-week recovery period.
d Phosgene exposure was 6 hours per day, 5 days per week.
e Phosgene exposure was 6 hours per day, 1 day per week.
-------
Table 2. Pulmonary histopathology severity score in rats following subchronic phosgene exposure" (from
Kodavanti et al., 1997)
Phosgene concentration (ppm)
Number of animals examined
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Terminal bronchiole, peribronchiolar
alveolus, epithelial alteration
Terminal bronchiole/peribronchiolar,
increased collagen staining
4 weeks
0.0
(12)
0
0
0
0.08
0.17
0.17
0.08
0.1
(8)
0
0.25
0.13
0.25
0.38
0.5
0.13
0.2
(8)
0
0.63
0.50
0.4
1.00
0.63
1.00
1.0
(6)
0.33
1.83
0.33
0.83
3.00
2.50
1.00
12 weeks
0.0
(12)
0
0
0
0
0.08
0
0.17
0.1
(8)
0
0.25
0
0
0.38
0.13
0.25
0.2
(8)
0
0.5
0.13
0
1.13
0.88
1.0
1.0
(8)
0.13
2.13
0.25
0.13
2.13
2.38
2.0
16 weeks b
0.0
(9)
0.4
0
0
0
0
0
0.2
0.1
(6)
0
0
0
0
0
0.33
0.17
0.2
(7)
0
0
0
0
0.14
0.43
1.0
1.0
(5)
0
0
0
0
0.16
0.60
1.0
a Severity scores given to individual animals from a complete pathological examination are 0, not remarkable; 1, minimal; 2, slight/mild; 3, moderate; 4,
moderately severe; and 5, severe, based on relative evaluation of lesions. Based on severity scoring, a maximum score of 3 was assigned for some of the lesion
types at highest phosgene concentrations. Severity scores for each animal within a group were added, and an average score per animal was calculated; this is
shown in the table.
b Indicates 12-week exposure, followed by a 4-week recovery period.
-------
Similar changes were seen following 12 weeks of exposure; the lesions did not appear to
have progressed beyond those seen at 4 weeks. Both pulmonary prolyl hydroxylase activity and
pulmonary desmosine were elevated at both 4 and 12 weeks of exposure in the 1 ppm animals
only. The intensity of collagen staining in the bronchiolar region was elevated (higher than in
controls) in the 0.2 and 1 ppm groups. The pulmonary hydroxyproline level was significantly
elevated only in the 1 ppm animals after 12 weeks of exposure.
Following 4 weeks of clean air recovery, body weights were significantly reduced only in
the 1 ppm rats, with absolute lung weights also significantly increased only in the 1 ppm
animals. The displacement volumes returned to control levels regardless of phosgene
concentration. Histopathology following 4 weeks of recovery showed considerable, although not
complete, recovery of the bronchiolar lesions and inflammation. Both prolyl hydroxylase
activity and desmosine levels had returned to normal postrecovery, but hydroxyproline levels in
the 0.5 and 1 ppm groups were significantly higher than in controls. Collagen staining remained
at the same level of intensity as seen in the 12-week groups dosed at 0.2 and 1 ppm. Phosgene-
induced changes in collagen staining were not reversible within the 4-week recovery period, and
the severity of lesions in the 12- week exposure group was dependent on concentration, not on
the C x T product.
As a followup to the same study, Hatch et al. (2001) pointed out that hydroxyproline
content and collagen staining are standard measures of lung fibrosis and can be considered good
markers of chronic injury. Fibrosis is accompanied by decreased lung compliance and diffusion
capacity. The critical toxic effect for purposes of defining the point of departure in the RfC
derivation is collagen staining, which is indicative of irreversible lung fibrosis. As Table 1
shows, the effect is not statistically significant at 0.1 ppm, but it is significant at 0.2 ppm, not
only for the 4- and 12-week exposure groups but also for the 16-week recovery group.
Kodavanti et al. (1997) found that, at 0.1 ppm, the lung displacement volume was statistically
significantly elevated in the 4-week exposure group but not in the 12-week exposure group or
the 16-week recovery group. This effect is not considered an adverse effect of chronic exposure
because it diminished with longer exposure (12 vs. 4 weeks), and it disappeared after the 4-week
recovery. Taking the pathology incidence findings as indications of chronic toxicity, a lowest-
observed-adverse-effect level (LOAEL) of 0.2 ppm (0.8 mg/m3) for collagen staining, indicative
of irreversible lung fibrosis, can be identified. The no-observed-adverse-effect level (NOAEL)
for this effect was 0.1 ppm in this study.
Rossing (1964) exposed 14 mongrel dogs to phosgene for 30 minutes at concentrations of
between 24 and 40 ppm (96 and 160 mg/m3); pretest values for each animal served as its control.
The dogs were exposed three times per week until a definite rise was seen in their airway
resistance; at that time, the frequency of exposure was reduced to once or twice a week.
14
-------
Exposures were performed for 10 to 12 weeks. During the fifth and sixth weeks, the
experimental schedules were disrupted.
Phosgene exposure resulted in no apparent discomfort to the animals. Seven of the 14
animals died within the first 3 weeks of exposure, and 3 additional animals were sacrificed at the
end of 3 weeks. Animals that died during exposure or were sacrificed were autopsied, and their
lungs were fixed and examined. The dynamic elasticity rose very quickly, reaching a maximum
mean value of four times the control in the first week of exposure. It fell slightly during the next
3 weeks but remained significantly elevated above that of controls. After the disruption of
exposure, elastance returned to the week 4 levels (approximately twice those of controls) until
the ninth week, when it increased again. Mean lower airway resistance followed a similar
pattern, with a rise for the first 4 weeks, a recovery period during the disruption of exposure, and
then another rise once exposure had resumed.
During the first 2 to 3 weeks, the animals were often tachypneic and breathed with
reduced tidal volume. After the first 3 weeks, the breathing pattern was similar to that seen in
patients with obstructive airway disease: the animals had a slow respiratory rate and, frequently,
active respiratory effort, as suggested by active contraction of the abdominal muscles. In two
animals that were allowed to survive beyond the exposure period, elastance dropped rapidly to
normal. Histologic examination revealed bronchiolitis with peribronchiolar edema, hemorrhage,
and inflammation at earlier time points (3 weeks or less). In animals surviving to the fourth
week and beyond, the inflammatory reaction was still present but less intense, despite continuing
exposure. Owing to inadequate reporting of exposure levels, poor experimental design, and
inadequate number of animals per treatment group, no NOAEL or LOAEL could be identified
from this study.
Clay and Rossing (1964) also described lung histopathology for a separate group of dogs
as described in Rossing (1964). They exposed groups of mongrel dogs (sex not specified) to
phosgene at levels of between 24 and 40 ppm (96 and 160 mg/m3) for 30 minutes for one to three
exposures per week. Group 1 animals (n = 2) consisted of unexposed controls; group 2 dogs (n
= 7) were exposed one or two times and sacrificed 1-2 days postexposure; group 3 animals (n =
7) were exposed 4-10 times and sacrificed up to 7 days postexposure; group 4 animals (n = 5)
were exposed 15-25 times and sacrificed immediately or up to 2 weeks postexposure; and group
5 animals (n = 4) were exposed 30-40 times and sacrificed immediately or up to 12 weeks after
the final exposure. The lungs of the sacrificed animals were inflated with fixative and dried.
Both histologic sections and 1-mm-thick macrosections of the dried lungs were examined for all
groups.
Both micro- and macroscopic examination revealed progressive pulmonary changes with
increasing number of exposures. Microscopic changes revealed acute bronchiolitis and
15
-------
peribronchiolitis that affected only scattered sections of the lung at the lowest number of
exposures. With increasing number of exposures, a progression to a chronic obliterative
bronchi olitis was seen, with fibrotic changes that affected the majority, but not all, of the lung
tissue. Macrosections similarly revealed little or no changes in animals exposed one or two
times, with a progressing fibrosis and emphysema seen with increasing number of exposures,
resulting in severe dilation of the respiratory bronchioles and increased alveolar pore size in
animals exposed 30-40 times. Owing to the poor design of the study and the inadequate number
of experimental animals and dose level tested, no NOAEL or LOAEL values could be identified.
Franch and Hatch (1986) performed a series of experiments examining the effects of
inhaled phosgene in male Sprague-Dawley rats. In the first exposure regimen, groups of rats
(4-10 per group) were exposed to 0 or 1 ppm (4.05 mg/m3) of phosgene for 4 hours and then
sacrificed immediately after exposure or at 1, 2, 7, 14, or 38 days postexposure. Body weights
were decreased to 13% below those of controls (p<0.01) on the first day postexposure and then
rose toward control values, reaching 3% below control values on day 14 of recovery. Food
intake was also significantly decreased in exposed animals on days 1-3 postexposure before
returning to nearly normal values. Lung wet weights were significantly elevated in exposed rats
immediately after exposure and remained elevated through day 7 postexposure. No change in
nonprotein sulfhydryl (NPSH) content was seen immediately postexposure, but it showed an
upwardly increasing trend thereafter. G6PD activity was elevated over that of controls from
days 1-14 postexposure.
The second regimen consisted of a single 7-hour exposure during which one rat per
group (control, exposed) was sacrificed each hour; the experiment was replicated three times.
Lung weights were significantly increased 4 hours into the exposure and beyond, whereas NPSH
content was decreased. No significant change in G6PD activity was seen.
In their third exposure regimen, Franch and Hatch (1986) exposed groups of male
Sprague-Dawley rats to 0.125 (0.5 mg/m3) or 0.25 ppm (1 mg/m3) of phosgene for 4 hours per
day, 5 days per week, for 17 total exposures over 4 weeks. Lung weight was significantly
increased at exposure day 7 and later in the 0.25 ppm group and at day 17 in the 0.125 ppm
group. Pooled over all time points, the 0.25 ppm group had higher NPSH content than did the
0.125 ppm group, and it was significantly greater than in controls. In animals allowed to recover
postexposure, lung weights and NPSH levels returned to near control levels. Histology of the
lungs after 17 days of exposure to 0.25 ppm of phosgene revealed moderate multifocal
mononuclear-cell accumulations in the walls of the terminal bronchioles and a minimal type-II
cell hyperplasia; lesions in the 0.125 ppm groups were minimal.
Selgrade et al. (1995) administered Streptococcus zooepidemicus bacteria via an aerosol
spray to the lungs of male Fischer 344 rats immediately after phosgene exposure and measured
16
-------
the subsequent clearance of bacteria. They also evaluated the immune response of uninfected
rats similarly exposed to phosgene, as measured by an increase in the percentage of
polymorphonuclear leukocytes (PMN) in lung lavage fluid. The exposure regimen was similar
to that of Kodavanti et al. (1997); animals inhaled phosgene at concentrations of 0, 0.1, or 0.2
ppm, 6 hours per day, 5 days per week, and 0.5 ppm, 6 hours per day, 2 days per week, for 4 and
12 weeks. For each of the 12-week exposure regimen, additional groups of animals were
assessed for all endpoints at 4 weeks postexposure.
Within 24 hours after bacterial infection, the lungs of unexposed animals cleared the
bacteria, but in animals exposed to phosgene for both 4 and 12 weeks at all three concentrations
the clearance was impaired. After 4 weeks of recovery following 12 weeks of phosgene
exposure, the bacteria clearance was comparable to that of unexposed animals. In uninfected
rats, the % PMN cells was statistically significantly higher than in the unexposed group in all 4-
week phosgene exposure groups, and it was also higher at the highest concentration (0.5 ppm) in
the 12-week exposure group. In the 4-week recovery animals, no difference in % PMN cells was
observed between the exposed and control groups. This experiment shows that all phosgene
concentrations from 0.1 to 0.5 ppm impaired resistance to bacterial infection and that the
immune response is stimulated by phosgene exposure. After 4 weeks following exposure,
bacterial resistence is back to normal and there is no immune response in excess of unexposed
controls.
In an earlier experiment measuring the same effects with a single 6-hour exposure to
phosgene concentrations of 0.1 and 0.2 ppm, Yang et al. (1995) also reported a decrease in
bacterial clearance in the lungs at 24 hours post infection, but over a period of 72 hours post-
infection it returned to normal in the 0.1 ppm group. In comparison with single exposures, the
multiple daily exposures extending to 4 and 12 weeks in the Belgrade et al. (1995) report showed
a slight enhancement of effect in the 0.1 ppm group at 24 hours post-infection, but no
"adaptation," or lessening of the effect. Yang et al. (1995) found that if the bacteria are
administered 18 hours after single phosgene exposures rather than immediately, the clearance is
normal, indicating that recovery from the toxic effect of phosgene is rapid. Belgrade et al.
(1995) suggested that in a repeated cycle of intermittent exposures there is an increased chance
for infection to occur during and immediately after each exposure period.
4.3. REPRODUCTIVE/DEVELOPMENTAL TOXICITY STUDIES
No epidemiological studies examining the effects of phosgene on reproduction or
development for any exposure duration or route in humans were located in the published
literature. A case report by Gerritsen and Buschmann (1960) describes a 7-months-pregnant
woman who survived severe phosgene-induced pulmonary edema and subsequently delivered a
17
-------
normal, full-term infant. No experimental animal studies have been reported on the effects of
phosgene on reproductive and developmental organ systems. Therefore, the data from which to
draw any conclusions about potential reproductive/developmental effects of phosgene in humans
or animals are inadequate.
4.4. OTHER EFFECTS
4.4.1. Dermal Toxicity
Skin contact with phosgene has been known to cause severe skin burns in humans.
Vapor contact with moist or wet skin can lead to irritation and erythema (WHO, 1997). No
dermal toxicity studies in experimental animals have been conducted.
4.4.2. Ocular Toxicity
In humans, low vapor concentration exposure to phosgene gas can cause conjunctival
inflammation, and high vapor concentration exposure can lead to corneal opacifications and
perforation (Grant and Schuman, 1993).
4.4.3. Neurotoxicity
Phosgene-induced hypoxia and hypotension may cause anoxic injury to the brain (Borak
and Diller, 2001).
4.4.4. Genotoxicity
The in vivo cytogenetic effects of phosgene inhalation were investigated in mice at 5, 10,
or 15 ppm for 6 hours. No evidence was found that phosgene is clastogenic, aneuploidogenic, or
capable of inducing sister chromatid exchanges and micronuclei (Klingerman et al., 1994).
Furthermore, Reichert et al. (1983) reported that phosgene was negative under the conditions of
the Ames bacterial mutagenicity assay with and without metabolic activation. The authors
concluded that the negative result was likely due to phosgene reacting rapidly in the test
medium. Additional in vitro testing would be subject to similar technical limitations imposed by
the water reactivity of phosgene. As discussed, the physical and chemical properties of
phosgene preclude a valid in vivo test of genetic toxicity.
4.5. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS
4.5.1. Oral
No published studies on the toxicity of phosgene following oral exposure in animals were
found. The lack of oral studies reflects the fact that phosgene is a gas at room temperature, and
that aqueous phosgene rapidly hydrolyzes to CO2 and HC1.
18
-------
4.5.2. Inhalation
No relevant published studies are available with which to evaluate the nonpulmonary
effects of inhaled phosgene. Therefore, the discussion in this section focuses on pulmonary
effects. The acute toxicity of phosgene inhalation in humans and animals has been well
documented (Underbill, 1919, as reviewed in WHO, 1998, 1997; U.S. EPA, 1986c, 1984).
Acute inhalation of phosgene results in a sequence of events, including an initial bioprotective
phase, a symptom-free latent period, and a terminal phase characterized by pulmonary edema
(Schneider and Diller, 1989; Diller, 1985).
Phosgene is not expected to leave the pulmonary circulation following inhalation
exposure. The effects of inhaled phosgene reported in human and animal studies have been
attributed to a direct effect on the respiratory tissues or to secondary consequences resulting from
damage to the respiratory system. The toxicity of phosgene is thought to result from its ability to
directly participate in acylation reactions with amino, hydroxyl, or sulfhydryl groups (WHO,
1998, 1997; Schneider and Diller, 1989, as discussed in U.S. EPA, 1986c; Diller, 1985).
4.5.3. Mode of Action Information
Fibrosis is a common consequence of various exogenous insults to a variety of
parenchymal tissues in the lung. The underlying mechanism of the induction and progression of
fibrosis—at both the molecular and cellular levels—is not well understood. Fibrosis is
characterized by a dense, hard mass in the lung; it may be diffuse and interstitial in character
rather than nodular. Phosgene-induced pulmonary inflammation and fibrosis in the experimental
animals provides a good model for chronic pulmonary inflammation and fibrosis in humans.
Connective tissue may develop both interstitial and intra-alveolar fibrosis following short-term
exposure. Hydroxyproline content and the activities of prolyl hydroxylase and galactosyl-
hydroxy-lysyl glucotransferase were increased in the lungs of rats exposed to phosgene. These
observations were reported by Kodavanti et al. (1997) and later by Hatch et al. (2001), who also
indicated that lung fibrosis can be considered a good marker for chronic injury from exposure to
phosgene. Borak and Diller (2001) reviewed the biochemical mechanisms that lead to adult
respiratory distress syndrome due to phosgene exposure; they are summarized below. In
addition, several other postulated biochemical mechanism studies are also reviewed that relate to
phosgene mode of action.
Phosgene is a highly reactive gas capable of damaging a variety of biological
macromolecules in an oxidant-like fashion. This activity potentially results from at least two
separate chemical reactions: acylation and hydrolysis.
Acylation, the more important and rapid mechanism, results from the reaction of
phosgene with nucleophilic moieties, such as the amino, hydroxyl, and sulfhydryl groups of
19
-------
tissue macromolecules. Acylation causes destruction of proteins and lipids, irreversible
alterations of membrane structures, and disruption of enzyme and other cell functions. Exposure
to phosgene depletes lung nucleophiles, particularly glutathione, and restoration of glutathione
seems to protect against phosgene-induced injury (Sciuto and Moran, 1999; Sciuto et al., 1998,
1995; Schroeder and Gurtner, 1992; Jaskot et al., 1991; Sciuto and Gurtner, 1989). For several
days after acute phosgene exposure, tissue levels of antioxidant enzymes, such as glutathione
reductase and superoxide dismutase, increase as part of the lungs' response to injury (Jaskot et
al., 1991).
In addition to acylation, phosgene is hydrolyzed to HC1 as shown below:
COC12 + H2O^CO2 + 2HC1
The formation of HC1 occurs on moist membranes and may cause irritation and tissue damage
(Diller, 1985). Because of the limited water solubility of phosgene, it is unlikely that large
quantities of HC1 could result from the exposure of biological tissues. However, small amounts
do form and may contact moist membranes of the eye, nasopharynx, and respiratory tract.
Hydrolysis to HC1 is the probable cause of immediate inflammation and discomfort after
phosgene exposure at concentrations greater than 3 ppm (>12 mg/m3).
Pulmonary cellular glycolysis and oxygen uptake following phosgene exposure are
depressed and, thus, leads to a corresponding decrease in the levels of intracellular adenosine
triphosphate and cyclic adenosine monophosphate (Sciuto et al., 1996; Kennedy et al., 1989;
Currie et al., 1985). This is associated with increased water uptake by epithelial, interstitial, and
endothelial cells (Helm, 1980). The semipermeability of the blood-air barrier becomes gradually
compromised as a result of fluid entering the interstitial and alveolar spaces. Later, the blood-air
barrier disrupts, opening channels for the flooding of alveoli (Diller et al., 1969; Schulz, 1959).
Compression of pulmonary microvasculature leads to the opening of arteriovenous shunts
(Schocimerich et al., 1975). The onset of pulmonary edema correlates temporally with the
decrease in adenosine triphosphate levels (Currie et al., 1985). Interventions that increase
intracellular cyclic adenosine monophosphate, such as treatment with phosphodiesterase
inhibitors (e.g., aminophylline), beta-adrenergic agonists (e.g., isoproterenol), or cyclic
adenosine monophosphate analogs, markedly reduce pulmonary edema formation in animals
exposed to phosgene (Sciuto et al., 1998, 1997, 1996; Kennedy et al., 1989).
Phosgene exposure also has been shown to cause lipid peroxidation in lungs. In mice and
guinea pigs, phosgene exposure of 22 ppm via inhalation for 20 minutes increased levels of lipid
peroxidation products, such as thiobarbituric acid-reactive substances in tissue and bronchio-
alveolar lavage fluid (Sciuto et al., 1998).
20
-------
The effects of phosgene on pulmonary arachidonic-acid metabolism were studied both in
vivo and in vitro (Madden et al., 1991). Male Wistar rats were exposed to 0.05, 0.10, 0.25, 0.50,
or 1 ppm phosgene for 4 hours. Lung lavage fluid total and differential cellularity and viability
were determined at 0, 4, 20, or 44 hours after exposure. Furthermore, the lavage fluid was
analyzed for prostaglandin E2 (PGE2), leukotriene B4 (LTB4), and leukotriene C4, leukotriene
D4 plus leukotriene E4 (LTCDE4). Phosgene at 1 ppm significantly decreased lavage fluid cell
viability at all time points but resulted in transient decrease at 0.1 ppm only at 4 and 20 hours.
The decreases in PGE2, LTB4, and LTCDE4 induced by the 0.1 and 0.25 ppm exposures
returned to, or exceeded, the control values at 44 hours postexposure. Phosgene did not affect
the PGE2 and LTCDE4 formation in rat macrophages. The authors concluded that phosgene-
induced alterations in arachidonic-acid metabolism may be involved in its toxicity. Guo et al.
(1990) investigated the role of arachidonate mediators in phosgene-induced lung toxicity in male
New Zealand rabbits. The authors concluded that phosgene stimulated the synthesis of
lipooxygenase products of arachidonic-acid metabolism, which appear to contribute to
pulmonary edema.
Increased thromboxane production occurred in human pulmonary microvascular
endothelial cells after phosgene exposure in vitro (Cheli et al., 1995). Neutrophils migrated to
the lung surface in large numbers following phosgene exposure in several animal species
(Robinson, 1994; Schroeder and Gurtner, 1992). Pre-exposure injections of cyclophosphamide,
which significantly reduced circulating neutrophil counts, also decreased neutrophil migration to
the lungs and limited phosgene-induced edema and mortality (Ghio et al., 1991).
Acyltransferase activity in alveolar type II cell microsomes (which is necessary for the
synthesis of pulmonary surfactant) was shown to be inhibited in rabbits after edematrogenic
doses of phosgene (Frosolono and Passarelli, 1978).
The above studies shed some light on the postulated mechanisms of phosgene toxicity;
however, they are inadequate to define modes of action at the cellular level.
4.6. WEIGHT-OF-EVIDENCE EVALUATION AND CANCER CHARACTERIZATION
Based on the criteria in the Guidelines for Carcinogen Risk Assessment (U.S. EPA,
2005a), the available toxicity and mode(s) of action data provide inadequate information to
assess the carcinogenic potential of phosgene. A single epidemiology study of phosgene-exposed
workers (Polednak and Hollis, 1985; Polednak, 1980) was not considered adequate for
evaluating carcinogenic potential in humans. Furthermore, no animal cancer bioassays of
phosgene have been conducted to evaluate carcinogenic potential in experimental animals.
Phosgene has been identified as a reactive intermediate in the metabolism of a number of
chemical carcinogens, including chloroform (Pohl et al., 1981, 1977); however, its role in the
21
-------
carcinogenesis of these compounds is not clearly understood. The reactive metabolites of
chloroform covalently bind to proteins and lipids but only minimally to DNA and nucleic acids.
The failure of the reactive species (e.g., phosgene, trichloromethyl free radical, and other
metabolites) to significantly bind to DNA has been ascribed to their short half-lives and to their
lack of nuclear penetration (as cited in U.S. EPA, 200 la).
There is some concern for the carcinogenic potential of phosgene on the basis of SAR
analysis because the two chlorines (linked to the carbonyl group) are highly reactive; however,
phosgene rapidly hydrolyzes into CO2 and HC1, such that exposure to phosgene might not result
in a reaction with nuclear DNA. However, no data exist regarding DNA alkylation as a result of
exposure to phosgene. Covalent binding of phosgene with cellular macromolecules has been
proposed as a mechanism of chloroform-induced hepatic and renal toxicity (Pohl et al., 1980a,
b), and it is generally accepted that the carcinogenic activity of chloroform resides in its highly
reactive intermediate metabolites, such as phosgene. Irreversible binding of reactive chloroform
metabolites to cellular macromolecules supports several theoretical concepts as a mechanism for
possible phosgene's carcinogenicity (as discussed in the Toxicological Review of Chloroform;
U.S. EPA, 200la).
4.7. SUSCEPTIBLE POPULATIONS AND LIFE STAGES
4.7.1. Possible Childhood Susceptibility
No published studies are available to evaluate the effects of phosgene exposure on
children or young experimental animals.
4.7.2. Possible Gender Differences
No published studies have directly compared the effects of phosgene inhalation exposure
in males and females.
4.7.3. Other
No published experimental animal or human epidemiological studies are available to
evaluate the effects of phosgene in the geriatric population or in individuals with compromised
disease conditions, such as asthmatics or those with respiratory impairments.
22
-------
5. DOSE-RESPONSE ASSESSMENTS
5.1. ORAL REFERENCE DOSE (RfD)
No published studies on the toxicity of phosgene following oral exposure in humans or
experimental animals were located. Phosgene is a gas at room temperature, and aqueous
phosgene rapidly hydrolyzes to CO2 and HC1. Therefore, exposure by the oral route is unlikely
and the lack of data precludes derivation of an RfD.
5.2. INHALATION REFERENCE CONCENTRATION (RfC)
For effects other than cancer, the risk from exposure via the inhalation route is assessed
by deriving an inhalation RfC. The RfC 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 appreciable risk of deleterious effects during a lifetime. Like the RfD,
the RfC is based on the assumption that a threshold exists for certain toxic effects. Exposure to
phosgene for a short period of time can have serious acute effects (NAS, 2002). Therefore, the
RfC cannot be directly compared to average air concentrations without also examining available
benchmarks regarding acute effects from the inhalation of phosgene (see Appendix A).
In this assessment, the RfC was estimated using three different approaches: the standard
NOAEL/LOAEL approach, which has been used extensively in the past (U.S. EPA, 1994b); the
benchmark dose (BMD) approach, which is currently being used by the Agency and has several
advantages over the NOAEL/LOAEL approach (U.S. EPA, 2000c); and a categorical regression
(CatReg) approach, which is suited to the analysis of severity-graded data and makes use of
recently developed EPA CatReg software (U.S. EPA, 2000e). Use of these approaches has
the potential to add multiple dimensions of information that include the slope of the dose-
response curve and the severity of effect.
5.2.1. Choice of Principal Study and Critical Effect(s)
In the selection of principal studies for identifying critical endpoints of phosgene toxicity,
two studies are relevant for deriving the RfC: Selgrade et al. (1995) and Kodavanti et al. (1997).
These are subchronic inhalation studies with periods of recovery following exposure. Both
studies have limitations, not being of chronic duration; however, they have similar exposure
protocols and used the same experimental animal strain (F344 rats) to measure two different
endpoints (immune response and pulmonary damage).
The subchronic study in rats reported by Kodavanti et al. (1997) and the followup study
by Hatch et al. (2001) are considered to be suitable for development of an RfC. Results of the
23
-------
study are summarized in Tables 1 and 2. The most sensitive target organ following chronic
inhalation exposure to phosgene appeared to be the lungs. The investigators observed
statistically nonsignificant terminal bronchiolar changes and interstitial thickening of the
alveolar walls, inflammatory cell influx, and epithelial alterations of the terminal bronchioles at
0.1 ppm after 4 and 12 weeks of exposure and a statistically significant increase in lung
displacement volume in all exposed groups after 4 weeks of exposure. The incidence and
severity of these effects increased in the 0.2 ppm and 1 ppm exposed groups. These effects were
not statistically significantly increased after a 4-week recovery period; they may be adverse, but
they are not persistent. Other effects noted at the 0.2 and 1 ppm exposure levels were as follows:
a statistically significant increase in collagen staining within thickened terminal bronchioles that
was more intense at 1 ppm and persisted after a 4-week recovery period; and a statistically
significantly increase in lung displacement volume and lung/body weight changes, which
returned to control levels after the 4-week recovery period.
Collagen staining was considered by Kodavanti et al. (1997) as a marker of chronic lung
damage, and an increase in lung displacement volume was considered as a sensitive indicator of
structural changes in the lung. These changes and the other histopathological changes noted in
Table 1 are considered adverse effects and, consequently, BMD and Cat Reg modeling were
done, as described in the next section.
Belgrade et al. (1995) concluded that phosgene exposure at 0.1, 0.2, and 0.5 ppm
impaired resistance to bacterial infection in rats. However, at the 0.5 ppm concentration, an
immune response was observed in noninfected animals. Phosgene is toxic to the immune cells
that are in the lungs, but after phosgene exposure stops, the cells repopulate the lung from
elsewhere in the body and no permanent damage to immune system cells is evident. It appears
that concentration rather than exposure duration is the more critical factor for the extent of toxic
response to phosgene, even at these low concentrations. A concentration of 0.1 ppm is
considered as a LOAEL for this effect in this study.
5.2.2. Methods of Analysis for Point of Departure, Including Application of Models (BMD,
NOAEL/LOAEL, and CatReg)
This assessment makes use of two dose-response modeling software suites developed by
EPA, the Benchmark Dose Software (BMDS) (U.S. EPA, 2001b) and CatReg (U.S. EPA,
2000e). BMD assessment methods (U.S. EPA, 2000c, 1995) and supporting software were
developed to improve upon the NOAEL/LOAEL approach by taking into account the quality of
the study and the complete dose-response, and the CatReg software was developed to allow for
the evaluation of categorically graded responses over time. The following sections describe how
three assessment methods (BMD, NOAEL/LOAEL, and CatReg) were used to analyze the
24
-------
critical effects identified from the Kodavanti et al. (1997) rat subchronic inhalation study to
obtain a point of departure (POD) for use in derivation of an RfC for phosgene. These sections
also describe attempts to use the BMD approach for the Belgrade et al. (1995) study and the
decision to use the NOAEL/LOAEL approach instead for that study.
5.2.3. BMD Approach
Kodavanti et al.. 1997
Following subchronic inhalation exposure of phosgene, Kodavanti et al. (1997) found
that the most sensitive target organ in rats is the lung, as discussed in Section 5.2.1. Lung
hydroxyproline content and trichrome staining for collagen are standard methods for measuring
lung fibrosis and can be considered reliable chronic injury markers. Support for this is found in
the present study, which showed lack of reversibility of the collagen accumulation and possibly
even a progression during the 4-week recovery period, terminal bronchiolar thickening and
inflammatory cell influx, and an increase in the lung displacement volume. Measurements of
hydroxyproline in the whole lung, which is considered to be a chemical manifestation of fibrosis,
were statistically increased in the high-dose group (1 ppm) only and were persistent after the
recovery period. Concentration seems to be more important than duration in determining this
pathology response. Collagen staining increased slightly at 4 weeks and increased markedly at
12 weeks in both the 0.2 and 1 ppm groups, the effect at 1 ppm being more intense. For the
BMD approach taken in this assessment, it is assumed that the administered concentration is an
appropriate dose-metric. Although this assumption is uncertain (see discussion in paragraph 2
below), there is no reasonable alternative assumption. The Kodavanti et al. (1997) data at 1.0
ppm is not used for the BMD modeling because the exposure duration (once per week) differs
markedly from the 0.1 and 0.2 ppm groups (5 times per week) and from continuous exposure.
The BMD approach attempts to fit curves to the dose-response data for a given endpoint. It has
the advantage of taking most of the dose-response data into account when determining the POD
as well as estimating the lowest dose for which an adverse effect may have a specific probability
of occurring. This approach is used when a biologically based dose-response model cannot be
formulated.
A benchmark analysis was performed for lung effects considered to be adverse, as
discussed in Sections 4.2.2 and 5.2.1. An overall summary of this analysis is provided in
Appendix B, Table B-l. A summary of the results most relevant to the development of a POD
for quantification of phosgene noncancer risk is provided in Table 3 for 4- and 12-week
exposures. The lower-bound confidence limit values reported in Table 3 represent the 95%
BMDL (lower-bound confidence limit on the benchmark dose) on the estimated ppm exposure
associated with a 10% extra risk (dichotomous endpoints) or a one-standard-deviation change
25
-------
from the estimated control mean (continuous endpoints, lung volume change).1 The 10%
response level was chosen as the point of departure (POD) for several reasons: (1) The small size
of the treated and control animals (only eight were exposed) does not allow the detection of
statistically significant effects below 10%. (2) A POD of larger than 10% cannot be justified
because the collagen staining is considered to be a toxicological significant finding of fibrosis.
(3) A POD of 10% is the default assumption used in the RfC methodology. Although 4-week
data are not used to derive the POD for an RfC, they are provided in Table 3 for comparison
purposes.
Table 3. Benchmark dose results from a subchronic study in rats (Kodavanti
et al., 1997)
Effects3
Interstitial thickening of the alveolus
Inflammatory cell influx to terminal bronchiole/alveolus
Epithelial alteration of terminal bronchiole/peribronchiolar alveolus
Increased collagen staining of terminal bronchiole/peribronchiolar
Displacement volume, left lung (mL/kg body weight * 100)
BMD/BMDL" (ppm)
12-week
exposure
0.044/0.025
0.083/0.031
0.078/0.026
0.10/0.018
0.081/0.059C
4-week
exposure
0.026/0.015
0.087/0.031
0.031/0.017
0.11/0.053
0.083/0.060C
a Only endpoints for which a dose-response could be modeled are listed.
bEPA's Benchmark Dose Software (BMDS), versions 1.3 and 1.4 were used to estimate the BMDLs. For
dichotomous endpoints, BMDLs are the 95% BMDL on the ppm exposure for a 10% extra risk. More details on
the BMD analysis, including data analyzed, models used, and options employed, are presented in Appendix B.
0 For this continuous endpoint, the BMDL represents a one-standard-deviation change from the estimated control
mean. The means and standard deviations for this endpoint were obtained in an e-mail dated October 22, 2001,
from Dr. Urmila Kodavanti, U.S. EPA/NHEERL, to Dr. Jeff Gift, U.S. EPA/NCEA.
An element of the BMD approach is the use of several models to determine which one
best fits the data.2 The model that best fits the experimental data is used when the mode of
JBMD analyses at 15, 5, and 1% were also performed and are reported in Appendix B for the 4- and
12-week exposure durations. However, the exposure group size of eight rats per exposure group is not conducive to
obtaining response estimates below 10%. One indication of this is the fact that as the BMR% goes down (i.e., x =
15%, 10% to 5% to 1%) the BMDx/BMDLx ratio goes up (i.e., 3.9, 5.6 to 10.5 to 44.1) for the collagen staining,
multistage model. This ratio indicates that although the BMDLx's are all 95% confidence intervals, in a certain
sense the "variability and/or reliability" of the models is considerably worse at BMRs below 10%.
2EPA's BMD Software (BMDS), versions 1.3 and 1.4, were usedforthis effort. BMDS can be
downloaded from the Internet at www.epa.gov/ncea/bmds.htm. BMDS facilitates the application of BMD methods
by providing simple data-management tools and an easy-to-use interface to run multiple models on the same
dose-response data set. At this time, BMDS offers nine different models appropriate for the analysis of dichotomous
(quantal) data (Gamma, Logistic, Log-Logistic, Multistage, Probit, Log-Probit, Quantal-Linear, Quantal-Quadratic,
26
-------
action is not known and, consequently, there is no theoretical basis for choosing a particular
model. As described in EPA's BMD technical guidance (U.S. EPA, 2000c), this is done by
measures of fit. In this case, the multistage model provided the best fit of all the dichotomous
models (see Appendix B) to the endpoint characterized as increased collagen staining of terminal
bronchioles. The dose-response data for the incidence of collagen staining and the multistage
model fitting these data are shown in Figure 1, which graphically shows the BMD10 and the
BMDL10. The BMDL10 for this effect is 0.018 ppm.
Weibull), continuous data (Linear, Polynomial, Power, Hill), and four nested models appropriate for developmental
toxicology data (NLogistic, NCTR, Rai, and Van Ryzin). Results from all models include a reiteration of the model
formula and model run options chosen by the user, goodness-of-fit information, the benchmark concentration, and
the BMDL.
27
-------
Multistage Model with 0.95 Confidence Level
0.8
T3
-------
dosing with phosgene, were dependent on concentration and not on the product of C x T. The
hydroxyproline data for this experiment are given in Hatch et al. (2001); Kodavanti et al. (1997)
show the same data in graphical form. A more detailed examination of these data reveals that
hydroxyproline concentration in the 12-week study increased with both C (at fixed T) and T (at
fixed C), and it also increased with the product of C x T. Therefore, the proper dose-metric is a
combination of these factors; perhaps it is C x Ta, where "a" is a fractional power of duration.
The experimental data are not definitive enough to derive a numerical description of the dose-
response surface. There are no data for collagen staining or for hydroxyproline resulting from
daily exposures in the range from 1 hour per day to 24 hours per day. However, two studies
employing continuous exposure show that toxic effects are proportional to the C x T product. In
the range of 0.5 minutes to 2 hours (5 to 500 mg/m3 [0.74 to 74 ppm], C>
-------
account volume breathed per day and the surface area of the thoracic region of the rat lung
versus the human lung. This is the standard procedure for dose conversions from animals to
humans for Category 1 gases, which are completely and irreversibly absorbed by the lung (U.S.
EPA, 1994b). The thoracic region, which consists of both the pulmonary and tracheobronchial
regions of the lungs, was chosen for three reasons. First, some of these lesions have been
classified as pulmonary lesions. Second, some of the assays measured would not make a
distinction between the two lung regions (e.g., whole-lung prolyl hydroxylase and
hydroxyproline as an index of collagen synthesis, volume displacement measurements). Third,
some lesions appear to occur in both regions (bronchus inflammation, alveolar interstitial
thickening).
The RGDR for the thoracic region of the respiratory tract (RGDR^) is used to adjust for
differences between rat and human ventilation rates and thoracic surface areas and is calculated
as follows (values used in this derivation were taken from U.S. EPA, 1988):
RGDRra = (MVa/Sa)/(MVh/Sh) =1.51
where:
MVa (minute ventilation for F344 rats) = 0.19 nrVday,
Sa (thoracic surface area for F344 rats) = 3,423 cm2,
MVh (minute ventilation for humans) = 20 m3/day, and
Sh (thoracic surface area for humans) = 543,200 cm2.
The BMDLjjEc was calculated by multiplying the BMDLADJ by the RGDRra:
BMDLjjEc = 0.0182 mg/m 3 x 1.51 = 0.0275 mg/m3
Selgrade et al. (1995)
Application of the benchmark dose approach to the Selgrade et al. (1995) data is
problematic because of the difficulties establishing what level of bacterial resistance adversely
affects the overall health and survival of the animals. The extent, duration, and health
consequences of impaired bacterial resistance from phosgene exposure is highly dependent on
secondary factors such as the exposure scenario involved, the health status of the exposed
individual, and the type of infection. Since the quantitative relevance of the rat model of
bacterial resistance to humans is unknown, it would be inappropriate to use these results in a
benchmark dose determination of the RfC.
5.2.4. NOAEL/LOAEL Approach
30
-------
The impairment of bacterial resistance observed by Belgrade et al. (1995) occurred at a
lower concentration (0.1 ppm) than the morphological lung damage observed by Kodavanti et al.
(1997). Therefore, the POD for the NOAEL/LOAEL approach for deriving an RfC is 0.1 ppm.
As was done for the BMD approach described above, RfC default methods for lung toxicity
caused by gaseous exposures (U.S. EPA, 1994a, b), in the absence of a relevant PBPK model,
were used to derive the HEC corresponding to the LOAEL of 0.1 ppm. Using the same default
procedures described in Section 5.2.3, aLOAELj^ of 0.15 mg/m3 is estimated (0.1 ppm x 4.05
= 0.405 mg/m3; 0.405 x 6/24 = 0.101 mg/m3; 0.101 x 1.51 =0.153 mg/m3).
5.2.5. CatReg Approach
As described in Appendix B-2, a CatReg analysis was performed using the individual
animal scores3 that resulted in the severity grade averages reported in Table 2. For purposes of
the CatReg analysis, and to ensure that the scores assigned by Kodavanti were appropriately
weighted according to the severity of the various endpoints, the scores assigned to endpoints that
did not significantly regress or disappear during the 4-week recovery period (epithelial alteration
and collagen staining of the terminal bronchioles) were increased by 1 severity grade, and the
scores of endpoints deemed to have recognized and serious long-term consequences (collagen
staining) were increased by an additional severity grade (see Table B-2a). Thus, reversible
lesions scored as "minimal" received a severity grade of 1, reversible lesions scored as
"slight/mild" and potentially irreversible lesions scored as "minimal" received a severity grade
of 2, and potentially irreversible lesions scored as "slight/mild" or any occurrence of a lesion
considered to have long-term consequences (collagen staining) received a severity grade of 3.
The data at 1.0 ppm is not used in the CatReg analysis because the exposure duration (once per
week) differs markedly from the 0.1 and 0.2 ppm groups (5 times per week) and from continuous
exposure. CatReg analysis was used to approximate ppm exposure levels that would result in a
10, 20, and 30% extra risk4 of attaining a severity grade 1, 2, or 3 level of lung effect. As
discussed in Appendix B-2, the analysis did not indicate that time was an explanatory variable,
so the results presented in Table 4 are for the 4- and 12-week data combined. The combined
analysis is preferred because time does not appear to be a significant factor for most of the
endpoints reported and because of the small number of animals involved in this study.
The CatReg analysis cannot be compared directly to the NOAEL and BMDL analyses at
this time and is not appropriate for use in the derivation of an RfC because EPA has not
3Personal communication between Dr. Jeff Gift and Dr. U.P. Kodavanti.
4As described in Appendix B-2, extra risk is what is generally used in a BMD analysis and is defined as the
estimated increased risk over background (Pd-P0) divided by the maximum risk with background excluded (1-P0).
31
-------
published guidance for the application of CatReg to Agency risk assessments and because the
CatReg software does not provide an estimate of the lower bound confidence limit on the extra
risk dose (ERD) that would be comparable to the BMDL.5 However, CatReg does provide a
maximum likelihood estimate, the ERD, that is comparable to the BMD and, because all of the
observations from the critical study can be severity graded and used in CatReg, it provides an
estimate of the ERD that is informed by more of the response data. Therefore, the ERD10
estimate of 0.05 ppm for a severity grade 3 effect (Table 4), which would include collagen
staining, can be compared to the BMD10 estimates for collagen staining in Table B-ld of
Appendix B-l. This comparison reveals that the CatReg ERD10 estimate for severity grade 3 is
similar to the BMD10 estimate of 0.1 ppm for collagen staining alone and falls between the
BMD10 and the BMDL10 estimate of 0.018 ppm.
Table 4. Results of CatReg analysis of severity-graded lung lesions reported
by Kodavanti et al. (1997) [estimates of the exposures that would cause a 10, 20,
and 30% probability of an effect equal to or greater than severity grade 1, 2, and 3
(ERD10, ERD20, and ERD30]
Severity
grade
1
2
3
ERD10
(ppm)
0.021
0.031
0.050
ERD20
(ppm)
0.038
0.052
0.077
ERD30
(ppm)
0.051
0.068
0.096
Model
Cumulative Odds
model
Link function
CLogLog
5.2.6. Comparison of Approaches
Each approach considered for determining the POD has strengths and limitations;
however, combining the three approaches yields a consistent and more robust determination of
the POD for the phosgene RfC. The NOAEL/LOAEL approach allows for a crude comparison
of results between multiple species and the target species. This approach is less dependent on
having the same experimental paradigms and results for comparison (e.g., a NOAEL/LOAEL
can be determined experimentally with less dependence on characterization of other points on
the dose-response curve). Using the NOAEL/LOAEL approach, the LOAEL for impairment of
host defenses against bacterial infection of the lung is 0.1 ppm for male rats (Belgrade et al.,
1995). This value was converted to a LOAELj^c of 0.15 mg/m3, about fivefold higher than the
In particular, CatReg does not provide an estimate of the standard error associated with its estimate of
background risk, which is used in derivation of the ERD (see Appendix B-2 for details).
32
-------
c estimate of 0.03 mg/m3.
A major disadvantage of the NOAEL approach is that NOAELs and LOAELs are
restricted to the set of doses used. BMD estimates are the results of interpolation using sound
statistical principles and so can take on any value in a range of doses. This results in the BMD
approach having the following advantages over NOAELs for dose-response assessment:
• BMDs can be more consistent among different studies.
• BMDs, because they are based on interpolation, do not depend on sample size,
whereas NOAELs, because they are essentially based on statistical testing, depend on
sample size, such that, for the same dose-response, smaller sample sizes yield larger
potential NOAELs.
• The statistical uncertainty of a BMD estimate can be calculated and weighed in a risk
assessment, whereas the uncertainty of a NOAEL calculation cannot.
CatReg requires the user to classify each effect within a study, or combination of effects,
into severity levels. Duration of exposure, as well as concentration, is included in CatReg
because it affects the probability of achieving the various severity levels. "Duration" can be
omitted, however, which is convenient when all subjects are exposed for the same duration or, as
is the case for this assessment, when duration is not an important explanatory variable.
CatReg fits a cumulative probability distribution to the combined data from all treatment
groups using the method of maximum likelihood estimation. From the probability distribution,
with parameters replaced by their estimates (i.e., the fitted model), the estimated probability of
any specified severity level or worse (e.g., mild adverse or worse) can be determined at any
specified concentration and duration. Viewed as an "exposure-response curve" (or "exposure-
response relationship"), an "exposure" is a combination of concentration and duration and
"response" is the probability of an adverse effect (of specified severity or worse) occurring at
that exposure.
Although the BMD approach has several advantages over the NOAEL approach, neither
it nor the NOAEL approach is capable of incorporating severity grades into a quantitative
assessment. In this case, for certain endpoints, such as inflammatory cell influx to terminal
bronchiole/alveolus and increased collagen staining of terminal bronchiole/peribronchiolar,
incidence data did not indicate a response at the low dose that was significantly different from
that of controls (Kodavanti et al., 1997) (Table 1), yet severity score data (see Table 2) indicate
that there may be some level of response at the low dose. This illustrates how BMD and
NOAEL analyses are sometimes not reflective of a changing profile of severity of response and
emphasizes the usefulness of a CatReg analysis that does account for differences in severity of
33
-------
response. Hence, a CatReg analysis that can explicitly account for severity-graded lung effects
was performed to supplement the BMD analysis (Appendix B-2).
As mentioned above, when the dose-response data for all severity grade effects are
considered together in a CatReg analysis, the ERD10 estimate for a severity grade 3 effect of 0.05
ppm is similar to, but about half that of, the BMD10 estimates obtained from the application of
several models to the collagen staining data of Kodavanti et al. (1997) (Table B-ld of Appendix
B-l). The multistage model used in the derivation of the 0.018 ppm BMDL10 point of departure
discussed above was the only model whose 95% lower bound confidence interval encompassed
(is lower than) the 0.05 ppm ERD10 estimate. This provides additional justification for the
choice of a relatively flexible multistage model for derivation of the BMDL point of departure.
5.2.7. RfC Derivation: Application of Uncertainty Factors
Uncertainty factors6 (UFs) are applied to account for recognized uncertainties in
extrapolation from experimental conditions to the assumed human scenario (i.e., chronic
exposure over a lifetime). Historically, UFs are applied as values of 10 in a multiplicative
fashion (Dourson and Stara, 1983). Recent EPA practice, however, also includes use of a partial
UF of 101/2 (3.162) (U.S. EPA, 2002) on the assumption that the actual values for the UFs are
6RfDs apply to lifetime human environmental exposure and include sensitive subgroups. Differences
between study conditions and conditions of human environmental exposure may make a dose that appears to be safe
in an experiment not safe in the environment. UFs account for differences between study conditions and conditions
of human environmental exposure. These differences include the following:
a. Variation from average humans to sensitive humans: RfDs apply to the human population, including
sensitive subgroups, but studies rarely target sensitive humans. Sensitive humans could be adversely
affected at doses lower than those in a general study population; consequently, general-population
NOAELs are reduced to cover sensitive humans.
b. Uncertainty in extrapolating from animals to humans: If an RfD is developed from animal studies, the
animal NOAEL is reduced to reflect pharmacokinetic and pharmacodynamic factors that may make
humans more sensitive than animals.
c. Uncertainty in extrapolating from subchronic NOAELs to chronic NOAELs: RfDs apply to lifetime
exposure, but sometimes the best data come from shorter studies. Lifetime exposure can have effects
that do not appear in a shorter study; consequently, a safe dose for lifetime exposure can be less than
the safe dose for a shorter period. If an RfD is developed from less-than-lifetime studies, the less-than-
lifetime NOAEL is adjusted to estimate a lifetime NOAEL.
d. Uncertainty in extrapolating from LOAELS to NOAELs: RfDs estimate a dose that is without
appreciable risks, but sometimes adverse effects are observed at all study doses. If an RfD is
developed from a dose where there are adverse effects, that dose is adjusted to estimate a NOAEL.
e. Other factors reflecting professional assessment of scientific uncertainties not explicitly treated above,
including completeness of the overall database, minimal sample size, or poor exposure
characterization.
34
-------
log-normally distributed. In the assessments, when a single partial UF is applied, the factor is
rounded to 3, such that the total factor for a UF of 3 and 10, for example, would be 30 (3 x 10).
When two partial UFs are evoked, however, they are not rounded, such that a UF of 3, 3, and 10
would result in total uncertainty of 100 (actually 101/2 x 101/2 x 10). UFs applied for this RfC
assessment and the justification for their use are as follows:
1. Human variation: UFH = 10. This factor is used to account for the variation in
susceptibility within the human population and for the possibility that the data
available are not representative of sensitive subgroups and lifestages, including
children (U.S. EPA, 2002). For phosgene, two studies are suitable for derivation of
the RfC, and, because they are in animals, they cannot be regarded as representative
of sensitive humans. Therefore the default value of 10 is appropriate.
2. Animal-to-human uncertainty: UFA = 3. Use of an RGDR to estimate an FtEC is
deemed to largely account for the pharmacokinetic portion of this uncertainty. A
threefold UF is retained to account for uncertainties regarding pharmacodynamic
differences between animals and humans.
3. Subchronic-to-chronic uncertainty: UFS = 3. The PODs are based on adverse
effects in two subchronic inhalation studies. The full factor of 10 is not appropriate
because the lung damage observed by Kodavanti et al. (1997) and the impairment in
bacterial resistence observed by Belgrade et al. (1995) are not likely to progress
significantly with further exposure. However, a partial factor of 3 is still necessary
because of the remaining uncertainty in predicting full lifetime effects from both 12-
week studies.
4. LOAEL-to-NOAEL uncertainty: UFL = 3 in the NOAEL/LOAEL approach; UFL
= 1 in the BMD approach. A partial uncertainty factor of 3 rather than the full factor
of 10 is used in the NOAEL/LOAEL approach because the impairment of lung
immunological function in the Belgrade et al. (1995) study at the LOAEL of 0.1 ppm
is considered to be a minimal effect. The effect is local to the lung, resulting in the
impairment of the bacterial clearance process; the impairment occurs only during the
exposure and it does not persist after phosgene exposure stops. No uncertainty factor
is applied to the 0.018 ppm BMDL derived from collagen staining in the Kodavanti et
35
-------
al. (1997) study because this POD is consistent with the NOAEL7 of 0.1 ppm given
the small group sizes in this study and because it represents minimal severity of lung
damage.
5. Database: UFD = 1. In general, a database UF is needed to account for the potential
for deriving an underprotective RfC as a result of an incomplete characterization of
the toxicity (U.S. EPA, 2002). This includes areas where there is a complete lack of
information as well as areas where existing data indicate that further information on a
particular subject has the potential for demonstrating effects at lower exposures.
Because phosgene is a chemically reactive agent with an extremely short half-life in
water and in lung tissue, its effects when inhaled are not likely to be observed outside
the lung, and no such effects have been observed to date. While it is recognized that
the investigation of systemic effects following phosgene exposure has not been the
focus of existing studies, there is no reason to expect that reproductive,
developmental, or other systemic effects would occur, and no UF is needed for the
absence of data on these effects. In view of the Belgrade et al. (1989) finding of
increased sensitivity to bacterial infection in mice due to short-term (4 and 8 hour)
phosgene exposures (section 4.2.2.1) at lower concentrations than in the sub-chronic
rat experiments (Belgrade et al., 1995), there is a possibility that a longer-term study
in mice might show effects at a lower concentration than in rats. That possibility
would be a rationale for a data base uncertainty factor of greater than one. However,
the species difference between the response in mice and rats is small and adequately
accounted for in the subchronic-to-chronic factor of 3 and the animal-to-human factor
of 3, and a separate data base uncertainty factor is not necessary.
The PODs derived using the NOAEL/LOAEL and BMD approaches are compared in
Table 5. A POD of 0.03 mg/m3, derived from the BMD analysis of collagen-staining lesions in
terminal bronchioles, is chosen for derivation of the RfC. The BMD approach is preferred
because it is based on the entire dose-response data. Using the BMD approach, the RfC is
calculated as follows:
RfC = 0.03 mg/m3 - 100 = 3E-4 mg/m3
7This does not mean that the 0.1 ppm level is deemed to be a true no-effect level. It is recognized that
responses at putative NOAELs can be as high as 20% (U.S. EPA, 2000c).
36
-------
Table 5. Application of uncertainty factors (UFs) for two different
approaches for deriving the RfC
Factor
POD (me/m3)
UFH
UFA
UFS
UFL
UFn
UFfrotaD
RfC (mg/m3)
NOAEL/LOAEL3
Approach
0.15
10
3
3
3
1
300
5E-4
BMDLb
Approach
0.03
10
3
3
1
1
100
3E-4
3 The LOAELHEC, based on impairment of resistance to bacterial infection in the Selgrade et al. (1995) study, was the
point of departure (POD). It is shown for comparison purposes only.
b The BMDLHEC, based on collagen staining in the Kodavanti et al. (1997) study, was the POD used to derive the
RfC.
5.2.8. Previous RfC Assessment
The health effects data for phosgene were evaluated in the IRIS database in 1990 and
were determined to be inadequate for derivation of an inhalation RfC.
5.3. CANCER ASSESSMENT
5.3.1. Oral Slope Factor
No studies on the carcinogenicity of phosgene following oral exposure in humans or
animals were located. Therefore, the lack of data precludes the derivation of an oral slope factor
for phosgene.
5.3.2. Inhalation Unit Risk
No studies on the carcinogenicity of phosgene in animals and no carcinogenicity studies
which adequately characterized inhalation exposure in humans were located. The Polednak
(1980) study of mortality among men occupationally exposed to phosgene was considered
inadequate to derive cancer unit risk. Thus, the lack of relevant data precludes the derivation of
an inhalation unit risk for phosgene.
37
-------
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF
HAZARD AND DOSE RESPONSE
6.1. HUMAN HAZARD POTENTIAL
Phosgene (CAS No. 75-44-5) has a chemical formula of COC12 and a molecular weight
of 98.92. At room temperature, it is a colorless gas with an aroma of moldy hay that may be
stifling at high concentrations. Phosgene is poorly soluble in water and is rapidly hydrolyzed to
CO2 and HC1 in aqueous solution. Industrially, phosgene is used as a chemical intermediate,
primarily in the polyurethane industry. The majority of phosgene used industrially is produced
by the reaction of carbon monoxide and chlorine gas using an activated charcoal catalyst, and it
is used at the production site.
Data on the effects of phosgene following exposure by the oral route are lacking.
Because phosgene is a gas at room temperature and because it is highly reactive and hydrolyzes
rapidly in water to CO2 and HC1, exposure to phosgene by the oral route is unlikely to occur.
The acute effects of phosgene inhalation have been well studied. Short-term exposure
results in the development of pulmonary edema and an increased concentration of hemoglobin in
the blood resulting from a decreased blood volume. At relatively high concentrations (>3 ppm,
or 12 mg/m3), irritation of the eyes and alterations in respiratory parameters may occur.
Symptoms of acute exposure (>0.5 ppm, or 2 mg/m3) increase in severity with both
concentration (C) and time (T), as described by Haber's Law. At sufficiently high C x T levels,
death may occur as a result of hypoxia or cardiac failure, both believed to be secondary
responses resulting from the severe pulmonary edema associated with high levels of inhaled
phosgene.
Inhalation of phosgene for subchronic or chronic durations is less well studied; there are
limited human data and very few animal studies. Available studies point to the respiratory tract
as the target for subchronic phosgene toxicity. The lungs are identified as the primary target
organ in all species. In addition, immunotoxicity has been observed in a few animal studies; the
importance of these findings to human hazard cannot be addressed at this time. U.S. and
international health and safety institutions have determined that 0.1 ppm (0.4 mg/m3) phosgene is
an exposure limit that offers some protection in occupational settings. According to the
Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), there is inadequate information
to assess the carcinogenic potential of phosgene.
38
-------
6.2. DOSE RESPONSE
6.2.1. Noncancer/Oral
Phosgene is a gas at room temperature and rapidly hydrolyzes to CO2 and HC1 in aqueous
solution; exposure by the oral route is unlikely. Available data for humans and animals are
inadequate to assess the potential chronic toxicity of phosgene following oral exposure. Data on
the effects of phosgene on reproductive function or on the developing organism are not
available.
6.2.2. Noncancer/Inhalation
The RfC was derived using the BMD approach to estimate a lower confidence limit on
the toxic lung effects observed in the 12-week study by Kodavanti et al. (1997). Five measures
of toxicity were modeled at doses of 0.1 and 0.2 ppm (0.4 and 0.8 mg/m3): (1) epithelial
alteration of terminal bronchioles and peribronchiolar alveoli, (2) increased collagen staining of
terminal bronchioles, (3) interstitial thickening of alveoli, (4) influx of inflammatory cells into
the terminal bronchioles and alveoli, and (5) lung volume. The BMD and BMDL values for
various benchmark response levels (15, 10, 5, and 1%) were calculated for each of the five
responses using seven different dose-response models. The 10% response level was chosen as
the lowest level of response that can be reliably modeled. The results giving the lowest value of
the BMD10 for each response were selected, and the lowest collagen-staining data were chosen to
characterize the BMD10 and BMDL10 for the entire study. Using this procedure, a BMDL10 of
0.018 ppm was derived. This value from the rat data was adjusted for continuous human
exposure by using the RGDR and exposure duration data. The resulting POD is 0.03 mg/m3.
The RfC was derived by dividing by a composite UF of 100 (10 for human variability and 3 each
for animal-to-human uncertainty in pharmacodynamics and subchronic-to-chronic animal data).
UFs of 3 are actually 10'/2, so when two factors of 3 are present, the combined UF is 10.
Therefore, the RfC is 0.03/100 = 3E-4 mg/m3.
Two additional alternative approaches were considered (the LOAEL/NOAEL approach
and the CatReg approach). The NOAEL/LOAEL approach uses the LOAEL of 0.1 ppm in the
Belgrade et al. (1995) study as the POD. This was adjusted for continuous human exposure by
using the exposure duration and RGDR to give a POD of 0.15 mg/m3. The total UF was 300 (10
for human variation and 3 each for animal-to-human, subchronic-to-chronic and LOAEL-to-
NOAEL). The resulting RfC using this approach is 0.15/300 = 5E-4 mg/m3.
The CatReg approach uses the EPA CatReg model applied to graded severity of lung
responses. The model estimated from the Kodavanti et al. (1997) study that the exposure
concentration associated with a 10% extra risk of attaining a severity grade 3 effect in rats
(e.g./'minimal" or more severe collagen staining) would be 0.05 ppm. This result cannot be
39
-------
compared directly to the BMDL10 or NOAEL but provides justification for the choice of the
Multistage model for derivation of the BMDL point of departure. The BMDL10 POD of 0.018
ppm obtained from the Multistage model is well below the CatReg ERD10 estimate and the 0.05
to 0.07 ppm range of BMDL10 estimates provided by the other BMDS models.
6.2.3. Cancer/Oral and Inhalation
Available data in humans are inadequate to assess the potential carcinogenicity of
phosgene. In addition, chronic animal bioassays are not available.
40
-------
7. REFERENCES
ACGIH (American Conference of Governmental Industrial Hygienists). (2000) Threshold limit values for chemical
substances and physical agents and biological exposure indices 2000. Cincinnati, OH.
Ardran, GM. (1950) The pulmonary effects of toxic gases and smokes: an experimental radiographic investigation.
Brit JRadiol 23:107-115.
Borak, J; Diller, WF. (2001) Phosgene exposure of mechanisms of injury and treatment strategies. J Occup Environ
Med 43:110-119.
Box, GEP; Cullumbine, H. (1947) The effect of exposure to sub-lethal doses of phosgene on the subsequent L(Ct)50
for rats and mice. Brit J Pharmacol 2:38-55.
Burleson, GR; Keyes, LL. (1989) Natural killer activity in Fischer-344 rat lungs as a method to assess pulmonary
immunocompetence: immunosuppression by phosgene inhalation. Immunopharmacol Immunotoxicol
ll(2-3):421-443.
CEH (Committee on Environmental Health). (1993) Ambient air pollution—respiratory hazard to children.
Pediatrics 91(6): 1210-1213.
Cheli, C; Davidson, C; Anderson, DJ; et al. (1995) Phosgene exposure induces throboxane synthesis in human
pulmonary microvascular endothelial cells. Abstract presented at the 91st International Conference of the American
Thoracic Society, Seattle, WA, 1995. Abstract no. 317491.
Clay, JR; Rossing, RG. (1964) Histopathology of exposure to phosgene: an attempt to produce pulmonary
emphysema experimentally. ArchPathol 78:544-551.
Currie, WD; Hatch, GE; Frosolono, MF. (1987) Changes in lung ATP concentration in rat after low level phosgene
exposure. J Bioch Toxicol 2:105-114.
Currie, WD; Pratt, PC; Frosolono, MF. (1985) Response of pulmonary energy metabolism to phosgene. Toxicol Ind
Health 1:17-27.
Dietert, RR. (2000) Workshop to identify critical window of exposure. Children's Health: Immune and Respiratory
System Workgroup Summary. Environ Health Perspec (suppl 3) 483-490.
Diller, WF. (1985) Pathogenesis of phosgene poisoning. Toxicol Ind Health 1(2):7-15.
Diller, WF; Zante, R. (1982) Dosis-wirkungs-beziehungen bei phosgen-einwirkung auf mensch und tier
(literaturstudie) [Dose-effect relation in the effects of phogene on man and animals (literature study)]. Zentralbl
Arbeitsmed Arbeitsschutz Prophyl Ergonomie 32:360-368. (Cited in U.S. EPA, 1986c.)
Diller, WF; Schnellbaecher, F; Wuestefeld, E. (1979) Pulmonale spaetfolgen nach phosgenvergiftungbzw.
Inhalationstoxischem lungenoedem [Late pulmonary sequelae of phosgene poisoning or inhalation-toxic pulmonary
edema]. Zentralbl Arbeitsmed Arbeitsschutz Prophyl 29:5-16. (Cited in U.S. EPA, 1986c.)
Diller, WF; Bils, RF; Kimmerle, G; et al. (1969) Die fruehphase der phosgenvergiftung im lichtmikroskopischen,
elektronenmikroskopischen, roetgenologischen und klinischen bild [The early phases of phosgene poisoning in light
microscopic, electron-microscopic and roentgenological studies]. Virchows Arch Abt A Path Anat 348:230-348.
(Cited in U.S. EPA, 1986c.)
41
-------
Dourson, ML; Stara, JF. (1983) Regulatory history and experimental support of uncertainty (safety) factors. Regul
Toxicol Pharmacol 3:224-238.
Durlacher, SH; Bunting, H. (1947) Pulmonary changes following exposure to phosgene. Am J Pathol 23:679-693.
Ehrlich, JP; Burleson, GR. (1991) Enhanced and prolonged pulmonary influenza virus infection following phosgene
inhalation. J Toxicol Environ Health 34:259-273.
Ehrlich, JP; Gunnison, AF; Burleson, GR. (1989) Influenza virus-specific cytotoxic T-lymphocyte activity in Fischer
344 rat lungs as a method to assess pulmonary immunocompetence: effect of phosgene inhalation. Inhal Toxicol
1:129-138.
Flury, F. (1921) Ueber kampfasgasvergiftungen: 1. Ueber Reizgase [Combat gas poisonings. 1. Irritant gases].
(Cited in EPA 1986c).
Franch, S; Hatch, GE. (1986) Pulmonary biochemical effects of inhaled phosgene in rats. J Toxicol Environ Health
19:413-423.
Frosolono, MF; Passarelli, LM. (1978) Inhibitions of rabbit lung microsomal acyl transferase after in vivo exposure
to phosgene. Am Rev Respir Dis 117:234.
Galdston, M; Luetscher, JA; Longcope, WT; et al. (1947) A study of the residual effects of phosgene poisoning in
human subjects: I. after acute exposure. J Clin Invest 26:145-168.
Gerritsen, WB; Buschmann, CH. (1960) Phosgene poisoning caused by the use of chemical paint removers
containing methylene chloride in ill-ventilated rooms heated by kerosene stoves. Brit J Ind Med 17:187-189.
Ohio, AJ; Hatch, GE. (1996) Tolerance to phosgene is associated with a neutrophilic influx into the rat lung. Am J
Respir Crit Care Med 153(3): 1064-1071.
Ohio, AJ; Kennedy, TP; Hatch, GE; et al. (1991) Reduction of neutrophil influx diminishes lung injury and mortality
following phosgene inhalation. J Appl Physiol 71:657-665.
Glass, WI; Harris, EA; Whitlock, RML. (1971) Phosgene poisoning. Case report. New Eng J Med 74:386-389.
Grant, WM; Schuman, JS. (1993) Toxicology of the eye. 4th ed. Springfield, IL: Charles C. Thomas; p. 733.
Gross, P; Rinehart, WE; Hatch, T. (1965) Chronic pneumonitis caused by phosgene. Arch Environ Health
10:768-775.
Guo, Y-L; Kennedy, TP; Michael, JR; et al. (1990) Mechanism of phosgene-induced lung toxicity: role of
arachidonate mediators. J Appl Toxicol 69(5):1615-1622.
Haber, FR. (1924) Zur geschichte des gaskrieges [On the history of the gas war]. In: Fuenf vortraege aus denjahren
1920-23 [Five lectures from the years 1920-1923]. Berlin, Germany: Verlag von Julius Springer; pp. 76-92. (Cited
in U.S. EPA, 1986c.)
Hatch, GE; Kodavanti, U; Crissman, K; et al. (2001) An "injury-time integral" model for extrapolating from acute to
chronic effects of phosgene. Toxicol Ind Health 17:285-293.
Hatch, GE; Slade, R; Stead, AG; etal. (1986) Species comparison of acute inhalation toxicity of ozone and
phosgene. J Toxicol Environ Health 19:43-53.
42
-------
Hegler, C. (1928) Ueber eine massenvergifung durch phosgengas in Hamburg. 1. Klinische berobachungen [On the
mass poisoning by phosgene in Hamburg. 1. Chemical observations]. Dtsch Med Wochenschr 54:1551-1553.
(Cited in U.S. EPA, 1986c.)
Helm, UK. (1980) Toxisches lungenoedm. In: Rebentisch, E. ed. Wehrmedizin. Baltimore, MD: pp. 28-284.
Jaskot, RH; Grose, EC; Stead, AG. (1989) Increase in angiotensin-converting enzyme in rat lungs following
inhalation of phosgene. Inhal Toxicol 1:71-78.
Jaskot, RH; Grose, BC; Richards, JH; et al. (1991) Effects of inhaled phosgene on rat lung antioxidant systems.
Fundam Appl Toxicol 17:666-674.
Jugg, B; Jenner, J; Rice, P. (1999) The effects of perfluoroisobutene and phosgene on rat lavage fluid surfactant
phospholipids. HumExp Toxicol 18(ll):659-668.
Keeler, JR; Hurt, HH; Nold, JB; et al. (1990) Phosgene-induced lung injury in sheep. Inhal Tox 2:391-406.
Kelly, TJ; Mukund, R; Spicer, C; etal. (1994) Concentrations and transformations of hazardous air pollutants.
Environ Sci Tech 28(8)379-387.
Kennedy, TP; Michael, JR; Hoidal, JR; et al. (1989) Dibutyryl cAMP, aminophylline, and B-adrenergic agonists
protect against pulmonary edema caused by phosgene. J Appl Physiol 67:2542-2552.
Klingerman, AD; Bryant, MF; Doerr, CL; et al. (1994) Cytogenetic effects of phosgene inhalation by rodents. I:
Acute 6-hour exposure of mice. Environ Molec Mutagenesis 23(3): 186-189.
Kodavanti, UP; Costa, DL; Giri, SN; et al. (1997) Pulmonary structural and extracellular matrix alterations in
Fischer 344 rats following subchronic phosgene exposure. Fundam Appl Toxicol 37:54-63.
Lehnert, BE. (1992) Acute inhalation toxicity of pyrolysis products of Halon 1301: kinetic course of lung injury,
degradation in work performance, and exercise potentiation of lung injury after phosgene exposure. Government
Reports Announcements and Index, No. 12. NTIS/AD-A260 873/5, 96 p.
Madden, MC; Friedman, M; Keyes, LL; et al. (1991) Effects of phosgene exposure on lung arachidonic acid
metabolism. Inhal Toxicol 3:73-90.
Manogue, WH; Pigford, RL. (1960) The kinetics of the absorption of phosgene into water and aqueous solutions.
AIChE J 6:494-500. (Cited in U.S. EPA, 1986c.)
Meek, WJ; Eyster, JAE. (1920) Experiments on the pathological physiology of acute phosgene poisoning. Am J
Physiol 51:303-320.
NAS (National Academy of Sciences). (2002) Acute exposure guideline levels for selected airborne chemicals.
Washington, DC: National Academy Press; pp. 15-70.
NIOSH (National Institute for Occupational Safety and Health). (2001) NIOSH pocket guide to chemical hazards.
Index by CASRN, examined January 2001. Available from: .
NRC (National Research Council). (1983) Risk assessment in the Federal Government: managing the process.
Washington, DC: National Academy Press.
NTP (National Toxicology Program). (2001) Chemical health and safety data. Available from:
.
43
-------
OSHA (Occupational Safety and Health Administration). (1993) 29 CFR Part 1910. Air Contaminants; Rule.
Federal Register 58(124):35338-35351.
Pohl, LR; Bhooshan, B; Whittaker, NF; et al. (1977) Phosgene: a metabolite of chloroform. Biochem Biophys Res
Commun 79:684-691.
Pohl, LR; Branchflower, RV; Highet, RJ; et al. (1981) The formation of diglutathionyl dithiocarbonate as a
metabolite of chloroform, bromotrichloromethane, and carbon tetrachloride. Drug Metab Dispos 9:334-338.
Pohl, LR; Martin, JL; George, JW. (1980a) Mechanism of metabolic activation of chloroform by rat liver
microsomes. Biochem Pharmacol 29:3271-3276.
Pohl, LR; Martin, JL; Taburet, AM; et al. (1980b) Oxidative bioactivation of haloforms into hepatotoxins. In: Coon,
MJ, ed. Microsomes, drug oxidations, and chemical carcinogenesis, Vol. 2. New York: Academic Press; pp.
881-884.
Polednak, AP. (1980) Mortality among men occupationally exposed to phosgene in 1943-1945. Environ Res
22:357-367.
Polednak, AP; Hollis, DR. (1985) Mortality and causes of death among workers exposed to phosgene in 1943-1945.
Toxicol Ind Health 1(2): 137-151.
Reichert, D; Neudecker, T; Spengler V; et al. (1983) Mutagenicity of dichloroacetylene and its degradation products.
Mutation Research 117:21-29.
Rinehard, WE; Hatch, T. (1964) Concentration-time product (CT) as an expression of dose in sublethal exposures to
phosgene. Am Ind Hyg Assoc J 25:545-553 (Cited in EPA, 1986c).
Robinson, NP. (1994) Mechanism of acute lung injury. Cah Med Trav 31:11-14.
Rossing, RG. (1964) Physiologic effects of chronic exposure to phosgene in dogs. Am J Physiol 207:265-272.
Sakakibara, H; Shiiki, Y; Okamoto, N; et al. (1967) Mass poisonings with phosgene. Shindan Chiryo 55:1433-1437.
(Cited in WHO, 1997.)
Schneider, W; Diller, W. (1989) Phosgene. In: Encyclopedia of industrial chemistry. 5th ed. Vol. A, 19:411-420.
Weinheim, Germany: VCH Verlag.
Schocimerich, P; Schuster, HP; Schoenboro, H; et al. (1975) Internintensivmedizin. Stuttgart, Germany: Thieme
Verlag, 259-260.
Schroeder, S; Gurtner, GH. (1992) Evidence for a species difference in susceptibility and mechanisms of phosgene
toxicity between rabbits and dogs. Am Rev Respir Dis 145:A606.
Schulz, H. (1959) Die submikroskopische anatomic und pathologie der lunge. Heidelberg, Germany: Springer
Berlin-Goettingel.
Sciuto, AM. (1998) Assessment of early acute lung injury in rodents exposed in phosgene. Arch Toxicol
72:283-288.
Sciuto, AM; Gurtner GH. (1989) Tracheal but not intravascular administration of N-actyl cysteine attenuates
phosgene induced lung injury in rabbits. Am Rev Respir Dis 138:A419.
Sciuto, AM; Moran, TS. (1999) A diet enhances the survival of mice exposed to phosgene: the effect of BUA on
glutathione levels in the lung. Inhal Toxicol 11:855-871.
44
-------
Sciuto AM; Strickland, PT; Kennedy, TP; et al. (1995) Protective effect of N-acetylcysteine treatment after phosgene
exposure in rabbits. Am J Respir Cris Case Med 151:768-772.
Sciuto, AM; Strickland, PT; Kennedy, TP; et al. (1996) Intratracheal administration of DbcAMP attenuates edema
formation in phosgene-induced acute lung injury. J Appl Physiol 80:149-157.
Sciuto, AM; Strickland, PT; Kennedy, TP; et al. (1997) Postexposure treatment with aminophylline protects against
phosgene-induced acute lung injury. Exp Lung Res 23:317-332.
Sciuto, AM; Strickland, PT; Gurtner, GH. (1998) Post-exposure treatment with isoproterenol attenuates pulmonary
edema in phosgene-exposed rabbits. J Appl Toxicol 18:321-329.
Selgrade, MK; Gilmore, MI; Yang, YG; et al. (1995) Pulmonary host defenses and resistance to infection following
subchronic exposure to phosgene. Inhal Toxicol 7:1257-1268.
Selgrade, MK; Starnes, DM; Illing, JW; et al. (1989) Effects of phosgene exposure on bacterial, viral, and neoplastic
lung disease susceptibility in mice. Inhal Toxicol 1:243-259.
Singh, HB. (1976) Phosgene in the ambient air. Nature 264:428-429.
Singh, HB; Salas, L; Shigeishi, H; et al. (1977) Urban-nonurban relationships of halocarbons, SF6, N2O, and other
atmospheric trace constituents. Atmos Environ 11:819-828.
Singh, HB; Sales, LJ; Smith, AJ; et al. (1981) Measurement of some potentially hazardous organic chemicals in the
urban environment. Atmos Environ 15:601-612.
Slade, R; Highfill, JW; Hatch, GE. (1989) Effects of depletion of ascorbic acid or nonprotein sulfhydrals on the
acute inhalation toxicity of nitrogen dioxide, ozone, and phosgene. Inhal Toxicol 1:261-271.
Underbill, FP. (1919) The physiology and experimental treatment of poisoning with the lethal war gases. Arch Int
Med 23:753-770 (Cited in WHO, 1997, 1998).
U.S. EPA (Environmental Protection Agency). (1983) Volatile organic chemicals in the atmosphere: an assessment
of available data. Environmental Sciences Research Laboratory, Research Triangle Park, NC, EPA/600/8-90/057F.
Available from: National Technical Information Service, Springfield, VA. PB83-195503.
U.S. EPA. (1984) Health and environmental effects profile for phosgene. Draft. Prepared by the Office of Health
Environmental Assessment, Environmental Criteria Assessment Office, Cincinnati, OH, for the Office of Solid
Waste and Emergency Response, Washington, DC. ECAO-CIN-P043.
U.S. EPA. (1986a) Guidelines for the health risk assessment of chemical mixtures. Federal Register
51(185):34014-34025. Available from: .
U.S. EPA. (1986b) Guidelines for mutagenicity risk assessment. Federal Register 51(185):34006-34012. Available
from: .
U.S. EPA. (1986c) Health assessment document for phosgene draft. Office of Health and Environmental
Assessment, Environmental Criteria Assessment Office, Cincinnati, OH. EPA/600/8-86/022A.
U.S. EPA. (1986d) The risk assessment guidelines of 1986. National Center for Environmental Assessment,
Washington DC. EPA/600/8-87/045.
U.S. EPA. (1988) Recommendations for and documentation of biological values for use in risk assessment. EPA
600/6-87/008. Available from: National Technical Information Service, Springfield, VA. NTIS PB88-179874/AS,
February 1988.
45
-------
U.S. EPA. (1991) Guidelines for developmental toxicity risk assessment. Federal Register 56(234):63798-63826.
Available from: .
U.S. EPA. (1994a) Interim policy for particle size and limit concentration issues in inhalation toxicity: notice of
availability. Federal Register 59(206):53799.
U.S. EPA. (1994b) Methods for derivation of inhalation reference concentrations and application of inhalation
dosimetry. EPA/600/8-90/066F.
U.S. EPA. (1995) Use of the benchmark dose approach in health risk assessment. National Center for Environmental
Assessment, Washington, DC. EPA/630/R-94/007.
U.S. EPA. (1996a) Guidelines for reproductive toxicity risk assessment. Federal Register 61(212):56274-56322.
Available from: .
U.S. EPA. (1996b) Air quality criteria for ozone and related photochemical oxidants. ORDEPA/600/P-93/004CF.
U.S. EPA. (1998a) Guidelines for neurotoxicity risk assessment. Federal Register 63(93):26926-26954. Available
from: .
U.S. EPA. (1998b) Science policy council handbook: peer review. First edition. Prepared by the Office of Science
Policy, Office of Research and Development, Washington, DC. EPA-100-B-98-001.
U.S. EPA. (2000a) Science policy council handbook: peer review. Second edition. Prepared by the Office of
Science Policy, Office of Research and Development, Washington, DC. EPA 100-B-OO-OOl.
U.S. EPA. (2000b) Science policy council handbook: risk characterization. Prepared by the Office of Science
Policy, Office of Research and Development, Washington, DC. EPA 100-B-00-002.
U.S. EPA. (2000c) Benchmark dose technical guidance document [external review draft]. Office of Research and
Development, Risk Assessment Forum, Washington, DC. EPA/630/R-00/001. Available from:
.
U.S. EPA. (2000d) Supplementary guidance for conducting health risk assessment of chemical mixtures. Office of
Research and Development, Risk Assessment Forum, Washington, DC. EPA/630/R-00/002.
U.S. EPA. (2000e) CatReg software documentation. Office of Research and Development, Washington, DC.
EPA/600/R-98/053F. Available from: .
U.S. EPA. (2001a) Toxicological review of chloroform. Integrated Risk Information System (IRIS), National Center
for Environmental Assessment, Washington, DC; EPA/63 5/R-01/001. Available from: National Technical
Information Service, Springfield, VA, and .
U.S. EPA. (2001b) Help manual for benchmark dose software version 1.3. National Center for Environmental
Assessment. EPA/600/R-00/014F. Available from: .
U.S. EPA. (2002) A review of the reference dose concentration and reference concentration processes [final report].
Risk Assessment Forum, Washington, DC. Available from: .
U.S. EPA. (2004) Integrated Risk Information System (IRIS). National Center for Environmental Assessment,
Washington, DC. Available from: .
U.S. EPA. (2005a) Guidelines for carcinogen risk assessment. Risk Assessment Forum, Washington, DC;
EPA/630/P-03/001B. Available from:.
U.S. EPA. (2005b) Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens.
46
-------
Risk Assessment Forum, Washington, DC; EPA/630/R-03/003F. Available from:.
WHO (World Health Organization). (1997) Environmental health criteria monograph on phosgene. Monograph 193.
International Programme on Chemical Safety. Geneva, Switzerland.
WHO. (1998) Phosgene health and safety guide. Health and Safety Guide No. 106. International Programme on
Chemical Safety. Geneva, Switzerland.
Wohlwill, F. (1928) II. Zur pathologischen anatomic der phosgenvergiftung [II. Pathological findings of phosgene
poisoning]. Arch Exp Pathol Pharmakol 181:198-206. (Cited in U.S. EPA, 1986c.)
Yang, YG; Gilmore, MI; Lange, R; et al. (1995) Effects of acute exposure to phosgene on pulmonary host defenses
and resistance to infection. Inhal Toxicol 7:393-404.
Zwart, A; Arts, JHE; Klokman-Houweling, JM; et al. (1990) Determination of concentration-time-mortality
relationship to replace LC50 values. Inhal Toxicol 2:105-117.
47
-------
APPENDIX A: ACUTE EXPOSURE GUIDELINE LEVELS (AEGLs)
FOR PHOSGENE
The development and application of AEGLs was first described in NAS (2002). They
represent threshold exposure limits for the general public and are applicable to emergency
exposure periods ranging from 10 minutes to 8 hours. AEGL-2 and AEGL-3, and AEGL-1
levels as appropriate, are developed for each of five exposure periods (10 minutes, 30 minutes, 1
hour, 4 hours, and 8 hours) and are distinguished by varying degrees of severity of toxic effects.
It is believed that the recommended exposure levels are applicable to the general population,
including infants and children and other individuals who may be susceptible. The three AEGLs
are defined as follows.
AEGL-1 is the airborne concentration (expressed as ppm or mg/m3) of a substance above
which it is predicted that the general population, including susceptible individuals, could
experience notable discomfort, irritation, or certain asymptomatic, nonsensory effects. However,
the effects are not disabling and are transient and reversible upon cessation of exposure.
AEGL-2 is the airborne concentration (expressed as ppm or mg/m3) of a substance above
which it is predicted that the general population, including susceptible individuals, could
experience irreversible or other serious, long-lasting adverse health effects or an impaired ability
to escape.
AEGL-3 is the airborne concentration (expressed as ppm or mg/m3) of a substance above
which it is predicted that the general population, including susceptible individuals, could
experience life-threatening health effects or death.
Airborne concentration below AEGL-1 represent exposure levels that could produce mild
and progressively increasing but transient and nondisabling odor, taste, and sensory irritation or
certain asymptomatic, nonsensory effects. With increasing airborne concentrations above each
AEGL, there is a progressive increase in the likelihood of occurrence and the severity of effects
described for each corresponding AEGL. Although the AEGL values represent threshold levels
for the general public, including susceptible subpopulations such as infants, children, the elderly,
persons with asthma, and those with other illnesses, it is recognized that individuals subject to
unique or idiosyncratic responses could experience the effects described at concentrations below
the corresponding AEGL.
Appropriate data were not available for deriving AEGL-1 values for phosgene. Odor
cannot be used as a warning for potential exposure. The odor threshold is reported to be between
0.5 and 1.5 ppm, a value above or approaching AEGL-2 and AEGL-3 values, and tolerance to
the pleasant odor of phosgene occurs rapidly. Furthermore, following odor detection and minor
irritation, serious effects may occur after a clinical latency period of 24 hours.
A-l
-------
AEGL-2 values were based on chemical pneumonia in rats (2 ppm for 90 min) (Gross et
al., 1965). An uncertainty factor (UF) of 3 was applied for interspecies extrapolation because
little species variability is observed with both lethal and nonlethal endpoints after exposure to
phosgene. A UF of 3 was applied to account for sensitive human subpopulations because of the
steep concentration-response curve and because the mechanism of phosgene toxicity (binding
to macromolecules and irritation) is not expected to vary greatly between individuals. Therefore,
the total UF is 10. The 1.5-hour value was then scaled to the 30-minute value and the 1-, 4-, and
8-hour AEGL exposure periods using Cn x T = k, where n = 1 (Haber's Law), because Haber's
Law has been shown to be valid for phosgene within certain limits. Haber's Law was originally
derived from phosgene data (Haber, 1924). The 30-minute value is also adopted as the 10-
minute value because extrapolation would yield a 10-minute AEGL-2 value approaching
concentrations producing alveolar edema in rats; Diller et al. (1985) observed alveolar
pulmonary
edema in rats exposed to 5 ppm phosgene for 10 minutes. Applying a total of UF of 10 to this
data point yields a supporting 10-minute value of 0.5 ppm.
The 30-minute and 1-, 4-, and 8-hour AEGL-3 values were based on the highest
concentration causing no mortality in the rat after a 30-minute exposure (15 ppm) (Zwart et al.,
1990). A UF of 3 was applied for interspecies extrapolation because little species variability is
observed with both lethal and nonlethal endpoints after exposure to phosgene. A UF of 3 was
applied to account for sensitivity of phosgene toxicity (binding to macromolecules and
irritation), which is not expected to vary greatly between individuals. Therefore, the total UF is
10. The value was then scaled to the 1-, 4-, and 8-hour AEGL periods using Cn x T = k, where n
= 1 (Haber's Law), because Haber's Law has been shown to be valid for phosgene within certain
limits. Haber's Law was originally derived from phosgene data (Haber, 1924). The 10-minute
AEGL-3 value was based on the highest concentration causing no mortality in the rat or mouse
(36 ppm) after a 10-minute exposure (Zwart et al., 1990). A UF of 3 was applied for interspecies
extrapolation because little species variability is observed with both lethal and nonlethal
endpoints after exposure to phosgene. A UF of 3 was applied to account for sensitive human
subpopulations because of the steep concentration-response curve and because the mechanism of
phosgene toxicity (binding to macromolecules and irritation) is not expected to vary greatly
between individuals (total UF = 10). The calculated values are listed in Table A-l.
A-2
-------
Table A-l. Summary of AEGL values for phosgene [ppm (mg/m3)]
Classification
AEGL-1
Nondisabling
AEGL-2
Disabling
AEGL-3
Lethal
10-minute
value
NA
0.60
(2.5)
3.6
(15)
30-minute
value
NA
0.60
(2.5)
1.5
(6.2)
1-hour
value
NA
0.30
(1.2)
0.75
(3.1)
4-hour
value
NA
0.08
(0.33)
0.20
(0.82)
8-hour
value
NA
0.04
(0.16)
0.09
(0.34)
Endpoint (reference)
NA
Chemical pneumonia
rats (Gross etal.,
1965)
Highest concentration
causing no mortality
in the rat after a 30-
minute or 10-minute
exposure (Zwart et al,
1990)
Source: NAS (2002).
Available from: http://www.epa.gov/oppt/aegl/results7.htm
A-3
-------
APPENDIX B: BMD AND CATREG ANALYSES
APPENDIX B-l: SUMMARY OF BENCHMARK DOSE (BMD) ANALYSIS
EPA's Benchmark Dose Software (BMDS), versions 1.3 and 1.4, were used to perform a
benchmark analysis of lung effects reported in the subchronic study by Kodavanti et al. (1997).
The BMD approach involves the use of several models to determine which one best fits the data.
As described in EPA's BMD technical guidance (U.S. EPA, 2000c), this is done by comparing
both the graphical fit and the statistical measures of fit. Tables B-la through B-le summarize
the textual output of the model runs, including AIC andp value measures of statistical fit.
Following the tables, complete output files, including graphical plots, are included for model
runs used in the derivation of each endpoint BMDL10 value.
B-l
-------
Table B-la. Kodavanti et al. (1997): Interstitial thickening of the alveolus in male rats
1)1)111
0
0.1
0.2
NOAEL (ppm)
NOAEL(HEC) (ms/mY
BMDs and BMDLs"
Model
Multistage0
Weibull"
Gamma"
Logistic
Log-Logistic'
Probit
Log-Probitf
Selected BMD/BMDL1(lvalue
BMDL,n(HEC) (mg/rn3)"
n
12
8
8
10%
15%
5%
1%
10%
15%
5%
1%
10%
15%
5%
1%
10%
15%
5%
1%
10%
15%
5%
1%
10%
15%
5%
1%
1%
15%
5%
1%
(ppm)g
4-week exnosure 12-week exnosure
0
2
5
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
22.00 0.82 0.026 0.015 1.73333
23.58 1.0 0.056 0.015 3.73333
23.58 1.0 0.06 0.015 4
24.46 0.43 0.079 0.046 1.71739
23.58 1.0 0.062 0.010 6.2
24.16 0.51 0.075 0.043 1.74419
23.58 1.0 0.065 0.028 2.32143
0.026 0.015
0.016
0
2
4
0.1
0.15
AIC P BMD
22.13 0.98 0.033
0.050
0.016
0.0031
24.09 1.0 0.045
0.064
0.026
0.0071
24.09 1.0 0.05
0.065
0.027
0.0085
25.29 0.34 0.085
0.11
0.056
0.016
24.09 1.0 0.05
0.067
0.031
0.011
25.00 0.40 0.08
0.10
0.052
0.016
22.09 1.0 0.055
0.070
0.038
0.019
0.044
BMD/
BMDL BMDL
0.018 1.8333
0.027 1.8519
0.0086 1.8605
0.00168 1.8452
0.018 2.5
0.027 2.3704
0.0086 3.0233
0.0017 4.1765
0.018 2.7778
0.027 2.4074
0.0086 3.1395
0.0017 5
0.05 1.7
0.068 1.6176
0.028 2
0.0063 2.5397
0.012 4.1667
0.019 3.5263
0.0055 5.6364
0.0011 10
0.047 1.7021
0.065 1.5385
0.026 2
0.0059 2.7119
0.032 1.7188
0.041 1.7073
0.022 1.7273
0.011 1.7273
0.025
0.038
1 Human equivalent concentration (HEC) calculated via EPA methods (U.S. EPA, 1994b). The regional gas-dose ratio for the
thoracic region of the respiratory tract (RGDRTH ) was used (see details in Section 5.2.3 of the main text).
b BMDL estimates are 95% lower confidence limits on the dose that would elicit 10,15, 5, or 1% extra risk.
c Betas restricted to >0 in multistage/polynomial models; parsimony used to select polynomial order.
d Power always restricted to > 1 in Weibull and Power models.
e Power always restricted to > 1 in Gamma and Hill models.
f Slope always restricted to > 1 in Log-Logistic and Log-Probit models.
8 Selected models (results in bold) were chosen based on statistical (AlC,p, scaled residual values) and visual assessment per
BMD guidance (U.S. EPA, 2000c). May represent an average of more than one BMDL if model fits are similar.
NOAEL = no-observed-adverse-effect level determined via statistical comparison with control response; NA = not applicable or
not applied; ZD = zero degrees of freedom; p value could not be calculated.
B-2
-------
Table B-lb. Kodavanti et al. (1997): Inflammatory cell influx to terminal
bronchiole/alveolus in male rats
ppm n
0 12
0.1 8
0.2 8
NOAEL (ppm)
NOAEL(HEC) (me/m1)"
BMDs and BMDLs"
Model
Multistage0 10%
15%
5%
1%
Weibuir1 10%
15%
5%
1%
Gamma6 10%
15%
5%
1%
Logistic 10%
15%
5%
1%
Log-Logistic' 10%
15%
5%
1%
Probit 10%
15%
5%
1%
Log-Probif 10%
15%
5%
1%
Selected BMD/BMDL10 value (ppm)g
BMDL,n(HEC) (mg/m3)a
4-week exposure
2
3
8
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
25.40 0.98 0.082 0.013 6.30769
27.40 ZD 0.084 0.033 2.54545
25.49 0.83 0.084 0.040 2.1
28.00 0.16 0.031 0.018 1.72222
25.40 0.99 0.094 0.050 1.88
27.88 0.16 0.028 0.017 1.64706
27.40 ZD 0.093 0.050 1.86
0.087 0.027
0.041
12-week exposure
1
3
8
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
21.47 1.0 0.077 0.012 6.4167
0.084 0.018 4.6667
0.067 0.0057 11.754
0.048 0.0011 43.636
23.47 ZD 0.079 0.029 2.7241
0.085 0.037 2.2973
0.069 0.019 3.6316
0.051 0.0069 7.3913
21.52 0.87 0.081 0.035 2.3143
0.086 0.043 2
0.073 0.025 2.92
0.061 0.012 5.0833
23.02 0.28 0.041 0.022 1.8636
0.053 0.031 1.7097
0.025 0.012 2.0833
0.0066 0.0026 2.5385
21.47 0.99 0.092 0.046 2
0.095 0.054 1.7593
0.089 0.036 2.4722
0.081 0.021 3.8571
23.03 0.27 0.035 0.020 1.75
0.047 0.029 1.6207
0.021 0.011 1.9091
0.0051 0.0023 2.2174
23.47 ZD 0.091 0.046 1.9783
0.094 0.053 1.7736
0.087 0.037 2.3514
0.081 0.025 3.24
0.079 0.024
0.037
1 Human equivalent concentration (HEC) calculated via EPA methods (U.S. EPA, 1994b). The regional gas-dose ratio for the
thoracic region of the respiratory tract (RGDRTH ) was used (see details in Section 5.2.3 of the main text).
b BMDL estimates are 95% lower confidence limits on the dose that would elicit 10, 15, 5, or 1% extra risk.
c Betas restricted to >0 in multistage/polynomial models; parsimony used to select polynomial order.
d Power always restricted to > 1 in Weibull and Power models.
e Power always restricted to > 1 in Gamma and Hill models.
f Slope always restricted to > 1 in Log-Logistic and Log-Probit models. The Log-logistic model was deemed to be inappropriate
for this endpoint because of the large impact that a maximum high-dose response has on its shape in the low-dose region.
8 Selected models (results in bold) were chosen based on statistical (AIC, p, scaled residual values) and visual assessment per
BMD guidance (U.S. EPA, 2000c). May represent an average of BMDLs from selected models.
NOAEL = no-observed-adverse-effect level determined via statistical comparison with control response; NA = not applicable or
not applied; ZD = zero degrees of freedom; p value could not be calculated.
B-3
-------
Table B-lc. Kodavanti et al. (1997): Epithelial alteration of terminal
bronchiole/peribronchiolar alveolus, male rat
ppm n
0 12
0.1 8
0.2 8
NOAEL (ppm)
NOAEL(HEC) (mg/m3)a
BMDs and BMDLs"
Model
Multistage0 10%
15%
5%
1%
Weibuir1 10%
15%
5%
1%
Gamma6
10%
15%
5%
1%
Logistic 10%
15%
5%
1%
Log-Logistic' 10%
15%
5%
1%
Probit 10%
15%
5%
1%
Log-Probitr
10%
15%
5%
1%
Selected BMD/BMDL10 value (ppm)g
BMDL10(HEC) (mg/rn3)"
4-week exposure
2
4
5
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
36.56 0.79 0.024 0.012 2
36.56 0.79 0.024 0.012 2
36.56 0.79 0.024 0.012 2
36.87 0.53 0.042 0.026 1.61538
36.49 0.94 0.017 0.0064 2.65625
36.85 0.55 0.041 0.026 1.57692
36.66 0.69 0.043 0.023 1.86957
0.031 0.017
0.026
12-week exposure
0
1
7
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
14.48 0.82 0.078 0.026 3
0.090 0.039 2.3077
0.061 0.013 4.6923
0.035 0.0025 14
16.06 1.0 0.094 0.044 2.1364
0.11 0.056 1.9643
0.079 0.030 2.6333
0.052 0.012 4.3333
16.06 1.0 0.096 0.050 1.92
0.10 0.061 1.6393
0.084 0.036 2.3333
0.065 0.018 3.6111
16.12 0.85 0.096 0.054 1.7778
0.11 0.068 1.6176
0.078 0.034 2.2941
0.041 0.0091 4.5055
16.06 1.0 0.096 0.055 1.7455
0.10 0.065 1.5385
0.084 0.041 2.0488
0.062 0.021 2.9524
16.06 0.95 0.095 0.050 1.9
0.11 0.64 0.1719
0.079 0.032 2.4688
0.050 0.0084 5.9524
16.06 1.0 0.096 0.057 1.6842
0.10 0.066 1.5152
0.086 0.046 1.8696
0.070 0.030 2.3333
0.078 0.026
0.040
Human equivalent concentration (HEC) calculated via EPA methods (U.S. EPA, 1994b). The regional gas-dose ratio for the
thoracic region of the respiratory tract (RGDRTH ) was used (see details in Section 5.2.3 of the main text).
b BMDL estimates are 95% lower confidence limits on the dose that would elicit 10, 15, 5, or 1% extra risk.
c Betas restricted to >0 in multistage/polynomial models; parsimony used to select polynomial order.
d Power always restricted to > 1 in Weibull and Power models.
e Power always restricted to > 1 in Gamma and Hill models.
f Slope always restricted to > 1 in Log-Logistic and Log-Probit models.
8 Selected models (results in bold) were chosen based on statistical (AIC, p, scaled residual values) and visual assessment per
BMD guidance (U.S. EPA, 2000c). May represent an average of BMDLs from selected models.
NOAEL = no-observed-adverse-effect level determined via statistical comparison with control response; NA = not applicable or
not applied; ZD = zero degrees of freedom; p value could not be calculated.
B-4
-------
Table B-ld. Kodavanti et al. (1997): Increased collagen staining of terminal
bronchiole/peribronchiolar, male rats
ppm n
0 12
0.1 8
0.2 8
NOAEL (ppm)
NOAEL(HEC) (ms/rn3)"
BMDs and BMDLs"
Model
Multistage0 10°/
15°/<
5°/<
l°/<
WeibuH" 10°/<
15°/<
5°/<
l°/<
Gamma6
10%
15°/<
5°/<
l°/<
Logistic 10°/<
Log-Logistic' 10°/
15°/<
5°/<
l°/<
Probit 10%
Log-Probif
10%
15°/<
5°/<
l°/<
Selected BMD/BMDL10 value (ppm)h
BMDL,n(HEC) (mg/m3)11
4-week exposure 12-week exposure
1
1
8
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
16.91 0.99 0.11 0.027 4.0740'
18.91 ZD 0.11 0.068 1.6176f
17.66 0.48 0.096 0.071 1.35211
indeterminate8
16.91 0.98 0.1 0.079 1.26582
indeterminate8
18.91 ZD 0.10 0.079 1.26582
0.11 0.027
0.041
2
2
8
0.1
0.15
BMD/
AIC P BMD BMDL BMDL
23.81 1.0 0.10 0.018 5.5556
0.11 0.028 3.9286
0.091 0.0087 10.46
0.075 0.0017 44.118
25.81 ZD 0.10 0.053 1.8868
0.11 0.062 1.7742
0.091 0.040 2.275
0.073 0.021 3.4762
24.17 0.63 0.092 0.059 1.5593
0.098 0.067 1.4627
0.083 0.048 1.7292
0.069 0.031 2.2258
indeterminate8
23.81 0.99 0.10 0.068 1.4706
0.10 0.075 1.3333
0.096 0.057 1.6842
0.088 0.038 2.3158
indeterminate8
25.81 ZD 0.10 0.068 1.4706
0.10 0.74 0.1351
0.096 0.059 1.6271
0.090 0.045 2
0.10 0.018
0.028
1 Human equivalent concentration (HEC) calculated via EPA methods (U.S. EPA, 1994b). The regional gas-dose ratio for the
thoracic region of the respiratory tract (RGDRTH ) was used (see details in Section 5.2.3 of the main text).
b BMDL estimates are 95% lower confidence limits on the dose that would elicit 10, 15, 5, or 1% extra risk.
c Betas restricted to >0 in multistage/polynomial models; parsimony used to select polynomial order.
d Power always restricted to > 1 in Weibull and Power models.
e Power always restricted to > 1 in Gamma and Hill models.
f Slope always restricted to > 1 in Log-Logistic and Log-Probit models. The Log-logistic model was deemed to be inappropriate
for this endpoint because of the large impact that a maximum high-dose response has on its shape in the low-dose region.
8 Inadequate model fit, p<0.1.
h Selected models (results in bold) were chosen based on statistical (AlC,p, scaled residual values) and visual assessment per
BMD guidance (U.S. EPA, 2000c). May represent an average of BMDLs from selected models.
NOAEL = no-observed-adverse-effect level determined via statistical comparison with control response; NA = not applicable or
not applied; ZD = zero degrees of freedom; p value could not be calculated.
B-5
-------
Table B-le. Kodavanti et al. (1997): Volume displaced, left lung
(mL/kg body weight x 100)
ppm
0
0.1
0.2
NOAELs
NOAEL (ppm)
NOAEL(HEC) (mg/m3)a
BMDs and BMDLs"
Model STD
Polynomial0 1
1.5
0.5
0.1
Power" 1
1.5
0.5
0.1
HilP 1
1.5
0.5
0.1
Selected BMD/BMDL10 value (ppm)g
BMDL10(HEC) (mg/m3)11
4-week exposure n S.D.
1.0715 12 0.12
1.21687 8 0.1614
1.3531 8 0.0767
0.1
0.15
4-week exposure
BMD/
AIC P BMD BMDL BMDL
-86.24 0.93 0.083 0.060 1.38333
-84.24 ZD 0.083 0.060 1.38333
indeterminate5
0.083 0.060
0.092
12-week exposure n S.D.
1.03736 11 0.0937
1.13614 7 0.0725
1.2882 8 0.1428
0.1
0.15
12-week exposure
BMD/
AIC P BMD BMDL BMDL
-87.16 0.55 0.081 0.059 1.3729
0.12 0.88 0.1364
0.041 0.029 1.4138
0.0081 0.0059 1.3729
-85.52 ZD 0.10 0.060 1.6667
0.14 0.090 1.5556
0.060 0.030 2
0.018 0.0060 3
indeterminate5
0.081 0.059
0.090
1 Human equivalent concentration (HEC) calculated via EPA methods (U.S. EPA, 1994b). The regional gas-dose ratio for the
thoracic region of the respiratory tract (RGDRTH ) was used (see details in Section 5.2.3 of the main text).
b BMDL estimates are 95% lower confidence limits on the dose that would result in a response equal to 1, 1.5, 0.5, or 0.1
standard deviations from the control mean.
c Betas restricted to >0 in multistage/polynomial models; parsimony used to select polynomial order.
d Power always restricted to > 1 in Weibull and Power models.
e Power always restricted to > 1 in Gamma and Hill models.
f Inadequate model fit, p<0.1.
8 Selected models (results in bold) were chosen based on statistical (AlC,p, scaled residual values) and visual assessment per
BMD guidance (U.S. EPA, 2000c). May represent an average of BMDLs from selected models.
NOAEL = no-observed-adverse-effect level determined via statistical comparison with control response; NA = not applicable or
not applied; ZD = zero degrees of freedom; p value could not be calculated.
B-6
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Interstitial Thickening of the Alveolus in Male Rats (12 Weeks) - Multistage Model
Multistage Model with 0.95 Confidence Level
0.8
0.7
0.6
T3
CD
I °'5
< 0.4
I 0-3
LL
0.2
0.1
0
Multistage
PMDL
BMD
0.05
0.1
dose
0.15
0.2
15:41 08/032004
Multistage Model. SRevision: 2.1 $ SDate: 2000/08/21 03:38:21 $
Input Data File: F:\BMDS\DATA\PHOSGENE\AIT-12WK-MUL.(d)
Gnuplot Plotting File: F:\BMDS\DATA\PHOSGENE\AIT-12WK-MUL.plt
ThuOct25 15:31:042001
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (l-background)*[l-EXP(-betal*doseAl)]
The parameter betas are restricted to be positive
Dependent variable = AIT-12wk
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-7
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Interstitial Thickening of the Alveolus in Male Rats (12 Weeks) - Multistage Model
Background = 0
Beta(l) = 3.46574
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(l)
Beta(l) 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0 NA
Beta(l) 3.24026 2.09548
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -10.0439
Fitted model -10.067 0.0462843 2 0.9771
Reduced model -14.5482 9.00875 2 0.01106
AIC: 22.134
Goodness of Fit
Dose Est._Prob. Expected Observed Size ChiA2 Res.
i: 1
0.0000
0.1000
0.2000
0.0000
0.2768
0.4769
0.000
2.214
3.815
0
2
4
12
8
8
0.000
-0.134
0.092
Chi-square = 0.05 DF = 2 P-value = 0.9774
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0325161
BMDL = 0.0175799
B-8
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Interstitial Thickening of the Alveolus in Male Rats (12 Weeks) - Log Probit Model
Probit Model with 0.95 Confidence Level
0.9
0.8
0.7
-o °-6
H 0.5
t
c 0.4
o
ro 0.3
0.2
0.1
0
Probit
BMDL
BMD
0.05
0.1
dose
0.15
0.2
15:4808/032004
Probit Model SRevision: 2.1 $ SDate: 2000/02/26 03:38:53 $
Input Data File: F:\BMDS\DATA\PHOSGENE\AIT-12WK-LOG-PRO.(d)
Gnuplot Plotting File: F:\BMDS\DATA\PHOSGENE\AIT-12WK-LOG-PRO.plt
TnuOct25 15:31:282001
BMDS MODEL RUN
The form of the probability function is:
P[response] = Background
+ (l-Background) * CumNorm(Intercept+Slope*Log(Dose)),
where CumNormQ is the cumulative normal distribution function
Dependent variable = AIT-12wk
Independent variable = Dose
Slope parameter is restricted as slope >= 1
Total number of observations = 3
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial (and Specified) Parameter Values
B-9
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Interstitial Thickening of the Alveolus in Male Rats (12 Weeks) - Log Probit Model
Background = 0
Intercept = 1.6281
Slope = 1
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -background -slope
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept
intercept 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0 NA
Intercept 1.61798 0.326388
Slope 1 NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -10.0439
Fitted model -10.0443 0.000811983 2 0.9996
Reduced model -14.5482 9.00875 2 0.01106
AIC: 22.0885
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 12 0
0.1000 0.2468 1.974 2 8 0.02101
0.2000 0.5034 4.027 4 8 -0.01928
Chi-square = 0.00 DF = 2 P-value = 0.9996
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.055049
BMDL = 0.0323742
B-10
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
Multistage Model
Multistage Model with 0.95 Confidence Level
T3
0
0.8
0.6
I 0-4
0.2
Multistage
BMDL
BMD
0.05
0.1
dose
0.15
0.2
15:5208/032004
Multistage Model. SRevision: 2.1 $ SDate: 2000/08/21 03:38:21 $
Input Data File: F:\BMDS\DATA\PHOSGENE\INF-12WK-MUL.(d)
Gnuplot Plotting File: F:\BMDS\DATA\PHOSGENE\INF-12WK-MUL.plt
ThuOct25 16:17:322001
BMDS MODEL RUN
Observation # < parameter # for Multistage model.
The form of the probability function is:
P[response] = background + (l-background)*[l-EXP(-betal*doseAl-beta2*doseA2-beta3*doseA3-beta4*doseA4-beta5*doseA5)]
The parameter betas are restricted to be positive
Dependent variable = INF-12wk
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 6
Total number of specified parameters = 0
Degree of polynomial = 5
B-ll
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
Multistage Model
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 1
Beta(l) = 5e+020
Beta(2) = 2.5e+021
Beta(3) = 1.25e+022
Beta(4) = 6.25e+022
Beta(5) = 3.125e+023
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -Beta(l) -Beta(2) -Beta(3) -Beta(4)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(5)
Background 1 -0.56
Beta(5) -0.56 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.0833264 0.275955
Beta(l) 0 NA
Beta(2) 0 NA
Beta(3) 0 NA
Beta(4) 0 NA
Beta(5) 38308.2 53667
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -8.73454
Fitted model -8.73457 6.96773e-005 1 0.9933
Reduced model -19.1214 20.7738 2 <.0001
AIC: 21.4691
Goodness of Fit
Dose Est._Prob. Expected Observed Size ChiA2 Res.
i: 1
0.0000
i: 2
0.1000
0.2000
0.0833
0.3751
1.0000
1.000
3.000
8.000
1
3
8
12
8
8
0.000
-0.000
1.000
Chi-square = 0.00 DF = 1 P-value = 0.9953
B-12
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
Multistage Model
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0772462
BMDL = 0.0117632
B-13
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
Gamma Model
Gamma Multi-Hit Model with 0.95 Confidence Level
0.8
•a
a>
CO
0.4
0.2
Gamma Multi-Hit
BMDL
BMD
0.05
0.1
dose
0.15
0.2
15:5308/032004
SRevision: 2.2 $ SDate: 2001/03/14 01:17:00 $
Input Data File: F:\BMDS\DATA\PHOSGENE\INF-12WK-GAM.(d)
Gnuplot Plotting File: F:\BMDS\DATA\PHOSGENE\INF-12WK-GAM.plt
ThuOct25 16:33:032001
BMDS MODEL RUN
The form of the probability function is:
P[response]=background+(l-background)*CumGamma[slope*dose,power],
where CumGammaQ is the cumulative Gamma distribution function
Dependent variable = INF-12wk
Independent variable = Dose
Power parameter is restricted as power >=1
Total number of observations = 3
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.115385
Slope = 7.44381
Power =1.3
B-14
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
Gamma Model
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.28
Slope -0.28 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.0823419 0.0785952
Slope 158.604 20.2326
Power 18 NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -8.73454
Fitted model -8.75909 0.049106 1 0.8246
Reduced model -19.1214 20.7738 2 <.0001
AIC: 21.5182
Goodness of Fit
Dose Est. Prob.
0.0000 0.0823
0.1000 0.3830
0.2000 0.9971
Expected Observed
0.988 1
3.064 3
7.977 8
Scaled
Size Residual
12 0.01249
8 -0.04677
8 0.153
Chi-square = 0.03 DF = 1 P-value = 0.8725
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0808407
BMDL = 0.0352477
B-15
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
LogLogistic Model
Log-Logistic Model with 0.95 Confidence Level
0.8
0.8
o
'•5 0.4
CD
0.2
Log-Logistic
BMDL
BMP
0.05
0.1
dose
0.15
0.2
15:5508/032004
Logistic Model SRevision: 2.1 $ SDate: 2000/02/26 03:38:20 $
Input Data File: F:\BMDS\DATA\PHOSGENE\INF-12WK-LOG-LOG.(d)
Gnuplot Plotting File: F:\BMDS\DATA\PHOSGENE\INF-12WK-LOG-LOG.plt
ThuOct25 16:29:082001
BMDS MODEL RUN
The form of the probability function is:
P[response] =background+(l-background)/[l+EXP(-intercept-slope*Log(dose))]
Dependent variable = INF-12wk
Independent variable = Dose
Slope parameter is restricted as slope >= 1
Total number of observations = 3
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial Parameter Values
Background = 0.0833333
Intercept = 11.1814
B-16
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Inflammatory Cell Influx to Terminal Bronchiole/Alveolus, Male Rats (12 Weeks) -
LogLogistic Model
Slope = 5.187
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -slope
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Intercept
Background Intercept
1 -0.3
-0.3
1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.0833321 0.0797846
Intercept 40.6845 0.903084
Slope 18 NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -8.73454
Fitted model -8.7346 0.000119887 1 0.9913
Reduced model -19.1214 20.7738 2 <.0001
AIC: 21.4692
Goodness of Fit
Dose Est._Prob. Expected Observed Size
1 12
Scaled
Residual
0.0000 0.0833
0.1000 0.3750
0.2000 1.0000
1.000
3.000
8.000
1.567e-005
-6.786e-005
0.007742
Chi-square = 0.00 DF = 1 P-value = 0.9938
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0923365
BMDL = 0.0464352
B-17
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Epithelial Alteration of Terminal Bronchiole/Peribronchial Alveolus (12 Weeks) -
Multistage Model
Multistage Model with 0.95 Confidence Level
0.8
T3
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Epithelial Alteration of Terminal Bronchiole/Peribronchial Alveolus (12 Weeks) -
Multistage Model
Default Initial Parameter Values
Background = 0
Beta(l) = 0
Beta(2) = 100
Beta(3) = 250
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -Background -Beta(l) -Beta(2)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(3)
Beta(3) 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0 NA
Beta(l) 0 NA
Beta(2) 0 NA
Beta(3) 225.422 104.079
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -6.02832
Fitted model -6.24181 0.426967 2 0.8078
Reduced model -16.7515 21.4465 2 <.0001
AIC: 14.4836
Goodness of Fit
Dose Est._Prob. Expected Observed Size ChiA2 Res.
i: 1
0.0000
i: 2
0.1000
i: 3
0.2000
0.0000
0.2018
0.8353
0.000
1.615
6.682
0
1
7
12
8
8
0.000
-0.477
0.289
Chi-square = 0.38 DF = 2 P-value = 0.8249
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0776057
BMDL = 0.0261034
B-19
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Increased Collagen Staining of Terminal Bronchiole/Peribronchiolar (12 Weeks) -
Multistage Model
T3
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Increased Collagen Staining of Terminal Bronchiole/Peribronchiolar (12 Weeks) -
Multistage Model
Default Initial Parameter Values
Background = 1
Beta(l) = 5e+020
Beta(2) = 2.5e+021
Beta(3) = 1.25e+022
Beta(4) = 6.25e+022
Beta(5) = 3.125e+023
Beta(6)= 1.5625e+024
Beta(7) = 7.8125e+024
Asymptotic Correlation Matrix of Parameter Estimates
(*** The model parameter(s) -Beta(l) -Beta(2) -Beta(3) -Beta(4) -Beta(5) -Beta(6)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Beta(7)
Background
1
-0.6
Beta(7)
-0.6
1
Parameter Estimates
Variable
Background
Beta(l)
Beta(2)
Beta(3)
Beta(4)
Beta(5)
Beta(6)
Beta(7) 1
Estimate
0.166651
0
0
0
0
0
0
.05428e+006
Std. Err.
0.261628
NA
NA
NA
NA
NA
NA
5.0641 le+006
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance
Full model -9.90542
Fitted model -9.90542 1.84535e-005
Reduced model -19.1214 18.432
Test DF P-value
0.9966
<.0001
AIC: 23.8108
Goodness of Fit
Dose Est._Prob. Expected Observed Size ChiA2 Res.
i: 1
0.0000
0.1000
i: 3
0.2000
Chi-square
0.1667
0.2500
1.0000
= 0.00
2.000
2.000
8.000
DF = 1
2
2
8
P-value
12
8
8
= 0.9976
0.000
-0.000
1.000
B-21
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Increased Collagen Staining of Terminal Bronchiole/Peribronchiolar (12 Weeks) -
Multistage Model
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0999908
BMDL = 0.018414
B-22
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Volume Displaced, Left Lung (mL/kg body weight x 100) (12 Weeks) - Polynomial Model
Linear Model with 0.95 Confidence Level
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Volume Displaced, Left Lung (mL/kg body weight x 100) (12 Weeks) - Polynomial Model
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
alpha 0.0102218 0.002835 0.00466525 0.0157783
beta_0 1.03121 0.0286698 0.975016 1.0874
betaj 1.24267 0.234088 0.783862 1.70147
Asymptotic Correlation Matrix of Parameter Estimates
alpha betaj) beta_l
alpha 1 5.1e-008 -9.7e-008
betaj) 5.1e-008 1 -0.72
beta_l -9.7e-008 -0.72 1
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev ChiA2 Res.
0 11 1.04 0.0937 1.03 0.101 0.202
0.1 7 1.14 0.0725 1.16 0.101 -0.506
0.2 8 1.29 0.143 1.28 0.101 0.237
Model Descriptions for Likelihoods Calculated
Model Al: Yij = Mu(I) + e(ij)
Var{e(ij)} = SigmaA2
Model A2: Yij = Mu(I) + e(ij)
Var{e(ij)} = Sigma(I)-"2
Model R: Yi = Mu + e(I)
Var{e(I)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) d.f AIC
Al 46.759659 4 -85.519318
A2 48.576042 6 -85.152085
fitted 46.582087 3 -87.164174
R 36.527257 2 -69.054514
Test 1: Does response and/or variances differ among dose levels (A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Testl 24.0976 4 <.0001
Test 2 3.63277 2 0.1626
Test3 0.355144 1 0.5512
The p-value for Test 1 is less than .05. There appears to be a difference between response and/or variances among the dose levels. It seems
appropriate to model the data
The p-value for Test 2 is greater than .05. A homogeneous variance model appears to be appropriate here
The p-value for Test 3 is greater than .05. The model chosen appears to adequately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
B-24
-------
Model Runs Used in the Derivation of Table B-l Selected BMDL10 Values
Volume Displaced, Left Lung (mL/kg body weight x 100) (12 Weeks) - Polynomial Model
Confidence level = 0.95
BMD = 0.0813596
BMDL = 0.058627
B-25
-------
APPENDIX B-2: EXPLANATION OF CATREG ANALYSIS
CatReg is a computer program developed to support toxicologists and health scientists in
the conduct of exposure-response analyses. Prior to performing categorical regression using
CatReg, effects observed in toxicological studies must be assigned to ordinal severity categories
(e.g., no effect, adverse effect, severe effect) and associated with the exposure conditions (e.g.,
concentration and duration) under which the effects occurred. CatReg executes a regression
analysis of the severity scores and exposure parameters. The categorization of observed
responses allows expression of dichotomous, continuous, and descriptive data in terms of effect
severity and supports the analysis of the data from single studies or a combination of similar
studies. CatReg is designed to work with S-PLUS® Professional version 3 or higher, and the
user must have access to this software to execute the CatReg program.1 Although familiarity
with S-PLUS® may assist the user, an understanding of the S-PLUS® programming language is
not required.
In the case of dichotomous data for phosgene, the lung is assessed in terms of presence or
absence of an effect, such as the collagen-staining effect observed following phosgene exposure
(Table B-ld, Appendix B-l). The mathematical form of the models that were used to compute
the benchmark dose-probability relationship is a function that is nondecreasing as x increases
and takes only values between 0 and 1 (properties required of a cumulative probability function).
As is the case in the Kodavanti et al. (1997) report, experimental effects may be reported
in more detail than simply "absence" or "presence," which allows for more detailed data analysis
using CatReg. For example, the number of animals in each treatment group and the number with
varying degrees of inflammation may be classified into severity levels, such as "no adverse
effect," "mild adverse effect," or "moderate/severe effect." After classifying the observations
into the severity levels (which are now the "data" to be input to CatReg), the data are ordinal
(ranked in terms of severity) and categorical (each animal in each treatment group is in a
severity category or a range of severity categories). Duration of exposure, as well as
concentration, is included in CatReg because it affects the probability of achieving the various
severity levels. "Duration" can be omitted, however, which is convenient when all subjects were
exposed for the same duration, or if duration is not considered to be an explanatory variable (as
has been determined to be the case for this analysis).
In the CatReg analysis of data from Kodavanti et al. (1997), the scores assigned by the
JA version of CatReg that will run in the freely distributable R statistical platform is
under development, but has not been released to date.
B-26
-------
study authors were weighted according to the severity of the various endpoints as follows for
endpoints that did not significantly regress or disappear during the 4-week recovery period
(epithelial alteration and collagen staining of the terminal bronchioles) were increased by 1
severity grade, and the scores of endpoints deemed to have recognized and serious long-term
consequences (collagen staining) were increased by an additional severity grade (see Tables B-
2a-l and B-2a-2). Thus, the absence of a lesion was scored as a severity grade of 0, reversible
lesions scored as "minimal" by the study authors received a severity grade of 1, reversible
lesions scored as "slight/mild" and potentially irreversible lesions scored as "minimal" by the
study authors received a severity grade of 2, and potentially irreversible lesions scored by study
authors as "slight/mild" or any occurrence of a lesion considered to have long-term
consequences (collagen staining) received a severity grade of 3. Table B-2b is the input file for
CatReg that was generated from the data for the 4-week, 12-week, and combined exposure
periods.
B-27
-------
Table B-2a-l. Adjusted severity grades for CatReg analysis of lung lesions in
rats exposed to phosgene for 4 weeks (Kodavanti et al., 1997)
4-Week Controls
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Alveolus, epithelial alteration
Increased collagen staining
Assigned Severity Grade
0.1 ppm Exposure Group
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Alveolus, epithelial alteration
Increased collagen staining
Assigned Severity Grade
0.2 ppm Exposure Group
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Alveolus, epithelial alteration
Increased collagen staining
Assigned Severity Grade
1
1
1
1
1
2
2
1
1
3
3
2
0
2
1
2
2
2
1
1
2
3
3
3
0
3
1
1
3
1
2
1
1
2
3
3
4
0
4
0
4
1
1
2
3
3
5
0
5
1
1
2
2
5
1
2
1
1
2
3
3
6
2
2
6
0
6
1
3
3
7
1
1
7
1
1
1
3
3
7
1
3
3
8
0
8
2
2
8
1
1
2
3
3
9 10 11 12
1
2
o
J
0003
B-28
-------
Table B-2a-2. Adjusted severity grades for CatReg analysis of lung lesions in
rats exposed to phosgene for 12 weeks (Kodavanti et al., 1997)
12-Week Controls
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Alveolus, epithelial alteration
Increased collagen staining
Assigned Severity Grade
0.1 ppm Exposure Group
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Alveolus, epithelial alteration
Increased collagen staining
Assigned Severity Grade
0.2 ppm Exposure Group
Alveolar effusion
Alveolus, interstitial thickening
Bronchus, epithelial alteration
Bronchus, inflammation
Terminal bronchiole/alveolus,
inflammatory cell influx
Alveolus, epithelial alteration
Increased collagen staining
Assigned Severity Grade
1
0
1
0
1
1
1
2
3
3
2
o
J
3
2
2
2
2
1
1
2
3
3
3
0
3
1
3
3
3
1
2
3
3
4
1
o
J
3
4
1
1
1
4
1
2
2
3
3
5
0
5
1
1
5
1
1
2
3
3
6
0
6
1
3
3
6
1
1
3
3
7
0
7
0
7
1
2
3
3
8
0
8
0
8
1
2
3
3
9 10 11 12
0000
B-29
-------
Table B-2b. CatReg input data file for 4-week, 12-week, and combined
analysis
ppm
0
0
0
0
0.1
0.1
0.1
0.1
0.2
0
0
0.1
0.1
0.1
0.1
0.2
0
0
0
0
0.1
0.1
0.1
0.1
0.2
Weeks
4
4
4
4
4
4
4
4
4
12
12
12
12
12
12
12
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Combined
Target
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Lung
Nsub
12
12
12
12
8
8
8
8
8
12
12
8
8
8
8
8
24
24
24
24
16
16
16
16
16
GpSize
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Incid
8
2
1
1
2
1
4
1
8
10
2
3
2
1
2
8
18
2
1
3
5
o
J
5
o
J
16
SevLo
0
1
2
3
0
1
2
o
3
o
3
0
o
3
0
1
2
3
o
J
0
1
2
3
0
1
2
o
J
3
SevHi
0
1
2
3
0
1
2
3
3
0
3
0
1
2
3
3
0
1
2
3
0
1
2
3
3
CatReg was used to fit a cumulative probability distribution to the 4-week and 12-week
data and the 4-week and 12-week data combined. Given a concentration C and duration T,
CatReg models the distribution of Y as
C,T} = H[(as
b2*/2(7)]
for ordinal scores s = l,...,S. This is the probability of attaining a severity score of 1 or higher
B-30
-------
at some specified concentration and duration of exposure. The parameter ax is the intercept
for severity equals 1, and as is an increment in the intercept for severity equals s versus severity
equals 1, for s = 2,...,S. bx and b2 determine the relationship between the response and
concentration and duration, respectively; unrestricted odds models can be chosen such that b's
may depend on the severity level of interest. CatReg can restrict the slopes so that they are the
same across all severity levels; thus, the slopes are simply bx and b2.
H is a probability function that takes values between 0 and 1. The inverse ofH is
called the link function, which is used to obtain the parameter estimates. There are several
possible choices for the inverse ofH: logit, probit, and log-log functions (see U.S. EPA, 2000e,
p. 4). Transformation functions, /; andf2, transform D and Tio another scale, usually a base-10
logarithm.
The intercept term at a specified severity level determines the probability of occurrence
when concentration and duration are both zero (i.e., the probability of a "background" response
at that severity level). The coefficients of concentration and duration (along with the model
used) determine how rapidly the probability of response (at a specified severity level) increases
as concentration and duration increase. The larger the coefficient of concentration, the more
rapidly the probability increases as concentration increases, and similarly for duration; the
smaller the coefficient of a variable, the less sensitive the probability is to a change in the
variable. The coefficient of duration for all model runs of the 4-week and 12-week data was a
very small number, approximately -0.04 (SE = 0.05). In addition, a Z-test of the null hypothesis
that there is no time effect yielded ap value of 0.425, well above thep value of 0.05 that would
cause us to reject the hypothesis. For these reasons, the results of the CatReg analyses of the
combined 4-week and 12-week data are used.
The primary goal of this analysis is to determine the CatReg estimate that is most
comparable to the BMD10 estimates for collagen staining that are presented in Appendix B-l
(Table B-ld). This was determined to be the 10% extra risk dose (ERD10) for a severity score
equal to 3, using the combined 4-week and 12-week data.2 The ERD10 for a severity score of 3 is
defined as the dose d* that satisfies
Pr(Y $3 I D = d*^> - Pr(Y $3 I D = 0) = 0.1
1 - Pr(Y $3 | D = 0)
Thus, the ERD10 is the dose that is associated with a 10% relative change from the
background response at severity level 3 or greater. If the probability of a severity level 3
2Other extra risk values and severity grades were evaluated and are presented for comparison purposes.
B-31
-------
response or greater were 0 at 0 concentration, the ERD10 would be equal to the 10%
effective concentration (EC10), a value that the currently released version of CatReg can estimate
directly. Because this probability is not 0 at 0 in this case, additional calculations are necessary.3
To determine the best model for finding the extra risk doses, the appropriate link function
and model had to be identified.4 Next, the data needed to be evaluated to determine whether the
slopes of the response curves associated with each severity level could be assumed to be parallel,
allowing for the use of a simplified model. The more complex, unrestricted model, in which the
severity levels are assumed to have individual slopes, was run first. The CatReg parallel test
(partest) was then used to determine whether slope parameters estimated in a model run for each
severity level are equivalent. When the slopes were found to not significantly differ from each
other, the simpler model that assumes a common slope (parallel curves) was employed. The test
was insignificant (simpler model was justified) for most endpoints.
Table B-2c shows the results of the model test. The cumulative odds model (cloglog link
function) was chosen because it resulted in the lowest deviance values. The unrestricted form of
this model was not deemed necessary because the parallel test of CatReg did not find that the
loss of 2 degrees of freedom that this would require was justified. The results for the application
of this model to the combined 4-week and 12-week severity graded data are shown in Table B-
2d. The ERD10 for a severity score equal to 3, using the combined 4-week and 12-week data, is
estimated to be approximately 0.05 ppm.
3The current Splus version of CatReg does not give extra risk doses directly at this time; however, it does
provide the information needed to derive them. To calculate the background risk (Pr(Y > 3 | D = 0 in the above
equation) one must use the severity 3 intercept parameter estimate (ocs) provided by CatReg to solve the function of
interest (e.g., the logit function, Pr(Y > 3| D) = exp[as+ p !*D]/(1+ exp[as+ p !*D])), where D = 0. Then solve:
Pr(Y>3 D = d*) = 0.1*(l- Pr(Y>3 | D = 0)) + Pr(Y>3 D = 0)
CatReg can then be used to calculate the dose (EC) associated with Pr(Y > 3 D = d*). This is the desired d*, the
10% extra risk dose for a severity 3 effect.
4See Section 2.2 of the CatReg documentation (U.S. EPA, 2000e) for the method for determining the best
link function.
B-32
-------
Table B-2c. Selection of model for use in derivation of ERD values
Scale
None
None
None
None
Log
Log
Log
Log
Log
Log
Link
Probit
Probit
CLogLog
CLogLog
Logit
Logit
Probit
Probit
CLogLog
CLogLog
Model
Cumulative Odds
model
Unrestricted
Cumulative model
Cumulative Odds
model
Unrestricted
Cumulative model
Cumulative Odds
model
Unrestricted
Cumulative model
Cumulative Odds
model
Unrestricted
Cumulative model
Cumulative Odds
model
Unrestricted
Cumulative model
Deviance
92.5099
91.98678
87.81308
87.42469
109.2512
108.7628
109.6004
108.8286
108.7272
108.348
df
5
3
5
3
5
3
5
3
5
3
Chi-
sauare
0.5952022
0.3337553
0.588963
0.9155587
0.3245027
df
2
2
2
2
2
0-value
0.74260
0.84630
0.74492
0.63269
0.85023
Parallel test result
This is generally
considered not
significant, indicating
that it would be more
appropriate to use the
simpler restricted
cumulative model.
This is generally
considered not
significant, indicating
that it would be more
appropriate to use the
simpler restricted
cumulative model.
This is generally
considered not
significant, indicating
that it would be more
appropriate to use the
simpler restricted
cumulative model.
This is generally
considered not
significant, indicating
that it would be more
appropriate to use the
simpler restricted
cumulative model.
This is generally
considered not
significant, indicating
that it would be more
appropriate to use the
simpler restricted
cumulative model.
ERD = Extra risk dose.
B-33
-------
Table B-2d. Results of CatReg analysis of severity-graded lung lesions reported by
Kodavanti et al. (1997) [estimates of the exposures that would cause a 10, 20, and 30%
extra risk of an effect equal to or greater than severity grade 1, 2, and 3 (ERD10, ERD20,
and ERD30)]
Severity
grade
1
2
3
Model
Cumulative
Odds model
Link function
CLogLog
ERD10
(ppm)
0.02122672
0.03084543
0.04991843
ERD20
(ppm)
0.03756418
0.05159007
0.07659892
ERD30
(ppm)
0.05128771
0.0678657
0.09580849
ERD = Extra risk dose.
B-34
-------
APPENDIX C: SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC
COMMENTS AND DISPOSITION
The Toxicological Review and IRIS summary for phosgene have undergone both internal
peer review performed by scientists within EPA and a more formal external peer review
performed by scientists in accordance with EPA guidance on peer review (U.S. EPA, 1998b,
2000a). Comments made by the internal reviewers were addressed prior to submitting the
documents for external peer review and are not part of this appendix. Three 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 significant comments made by the external reviewers and EPA's response to these
comments follows. EPA also received scientific comments from the public. These comments
and EPA's response are included in a separate section of this appendix.
The reviewers considered the overall quality of the Toxicological Review and IRIS
summary to be good. A number of editorial suggestions were offered. These included
identification of typographical errors, insufficiently clear text, incomplete descriptions of
complex calculations, and use of inconsistent concentration units. Revisions and corrections
were incorporated in the document as appropriate.
Comments from External Peer Review
A. RfC Derivation
1. Principal Study
Comment: All three reviewers agreed with the Agency's selection of Kodavanti et al. (1997) as
the principal study for the derivation of the RfC. One reviewer suggested that the Franch and
Hatch (1986) study be used to support the findings of the Kodavanti et al. study.
Response: The Agency disagrees that the findings of Franch and Hatch (1986) should be used as
support for the derivation of the RfC based on Kodavanti et al. (1997). Although the doses were
similar in the two studies, different lung effects were observed.
C-l
-------
2. Methods of Analysis
Comment 1: The reviewers generally supported the Agency's use of three methods for derivation
of the RfC (LOAEL/NOAEL, BMD and categorical regression). Two reviewers expressed the
least preference for the LOAEL/NOAEL approach, considering it to be the least statistically
robust method. One of these reviewers expressed more confidence in the BMD method than the
other two methods. The second reviewer considered the decision to use the benchmark dose
analysis with the categorical regression analysis in a supporting role to be appropriate.
Response 1: The Toxicological Review provides a comparison of the three approaches, including
a discussion of the strengths and limitations of each. The Agency believes that the use of three
approaches yields a more robust determination of the point of departure for the phosgene RfC.
Comment 2: Two reviewers offered comments on the BMD approach. One reviewer
recommended a 5% response level as the point of departure for quantal endpoints in the BMD
approach rather than a 10% level because the former was considered a closer approximation of
the NOAEL. A second reviewer observed that for continuous data, the Agency used one
standard deviation to define the benchmark response, but recommended that EPA consider two
or more standard deviations from the mean based on Kodell and West (Risk Analysis, 1993).
One reviewer recommended that the appendix with documentation of the BMD approach include
a complete description of the actual source data and the methods and assumptions used, and that
this appendix be appended to the IRIS summary as well as the Toxicological Review.
Response 2: A 5% level of response was rejected because a group size of only eight rats did not
provide sufficient statistical power to derive a 5% response level. The second reviewer referred
to the continuous lung displacement volume data; the BMR selected for this continuous endpoint
was one standard deviation. The Agency notes that the use of one standard deviation for
continuous data is standard Agency practice as described in the Benchmark Dose Technical
Guidance Document (U.S. EPA, 2000c). The Toxicological Review and IRIS summary were
revised to provide a better description and interpretation of the principal study (Kodavanti et al.,
1997); citations for the published references, which provide the source data, are included in the
appendix. It is not current practice to include detailed data in the IRIS summary; the
Toxicological Review for phosgene now contains a complete description of the actual source
data and detailed assumptions for the BMD modeling.
Comment 3: One reviewer recommended modifying the calculation in the categorical regression
analysis to include the lower 95% confidence bound on the CatReg point of departure.
C-2
-------
Response 3: EPA has not published guidance for the use of CatReg for the determination of a
point of departure for use in Agency risk assessments. The primary purpose of the CatReg
analysis in this assessment is to show how a categorical regression analysis that takes into
account the different severity grades of responses reported by Kodavanti et al. (1997) compares
with the BMD analysis. The maximum likelihood estimate, ERD, reported by CatReg is
comparable to the BMD estimate reported by BMDS. The lower bound confidence limit on the
ERD is not reported because the existing version of the Agency's CatReg software does not
provide an estimate of the lower bound confidence limit on the ERD that is comparable to the
BMDL reported by BMDS.
3. NOAEL/LOAEL Approach
Comment: Three reviewers commented on the C x T assumption (Haber's Law) used to adjust
from the intermittent exposure received by animals in the Kodavanti et al. (1997) study
(6 hours/day, 5 days/week) to continuous exposure. All three acknowledged the complexity of
the issues associated with application of the C x T assumption to the Kodavanti et al. data for
phosgene. One reviewer agreed with the Agency's use of the default C x T procedure given
insufficient evidence to justify departure from the default. A second reviewer recommended not
leaving the concentration unadjusted, and offered no suggestions for a better approach. The third
reviewer indicated he was not opposed to the C x T approach as presented, but recommended
that a more thorough discussion of the default assumption be provided.
Response: A discussion of the applicability of Haber's Law in making the exposure adjustment
from intermittent dosing in the animal experiments to continuous exposures has been added. It
says essentially that although Haber's Law applies to acute exposures lasting on the order of
minutes to several hours, it does not apply to intermittent exposures where there is a daily
recovery period. In the latter situation, the data indicate that concentration is the appropriate
dose metric. In the absence of data on continuous exposure for longer than several hours, the
document assumes that continuous dosing for periods ranging from 6 to 24 hours would cause a
progressive increase in lung damage; therefore, the C x T dose metric is appropriate for that
period of time.
4. Uncertainty Factors
Comment 1: One reviewer noted that the appropriate uncertainty factors were applied and
adequately explained. Two other reviewers agreed with the uncertainty factors applied by the
EPA for variation in human susceptibility (UFjj), animal to human extrapolation, and subchronic
to chronic extrapolation. One reviewer commented that the uncertainty factor for subchronic to
C-3
-------
chronic extrapolation appeared reasonable, but that the evidence advanced to support the choice
of uncertainty factors was not adequate.
Response 1: The justifications for the uncertainty factors were expanded.
Comment 2: One reviewer questioned whether a LOAEL to NOAEL uncertainty factor (UFL) as
large as 10 was supported by the dose-response data. A second reviewer considered the UFL,
which was applied to the LOAEL/NOAEL and categorical regression approaches to be
reasonable, but further noted that this uncertainty factor was not applied in the benchmark dose
approach. This reviewer recommended that an UFL of 10 be applied in the benchmark dose
approach because he did not consider a BMDL10 as similar to an unqualified NOAEL. This
reviewer offered as an alternative but less desirable option to use the BMDL10 with a UFL of 3.
Response 2: The subchronic study by Belgrade et al. (1995) was used for the NOAEL/LOAEL
analysis in place of the Kodavanti et al. (1997) study because it identified a lower LOAEL. A
partial uncertainty factor of 3 rather than the full factor of 10 was used in the NOAEL/LOAEL
approach because the impairment of lung immunological function in the Belgrade et al. (1995)
study at the LOAEL of 0.1 ppm was considered to be a minimal effect. The effect was local to
the lung, resulting in the impairment of the bacterial clearance process; the impairment occurred
only during the exposure and it was not persistent after phosgene exposure stopped. EPA
disagrees with the recommendation to include a UFL for the BMD method. Because the data on
collagen staining were considered to represent minimal severity of lung damage, the BMDL10
associated with a 10% response was considered an appropriate point of departure. Further, it is
current Agency practice not to apply an additional uncertainty factor for LOAEL to NOAEL
extrapolation when using the BMD approach. Accordingly, no additional uncertainty factor for
LOAEL to NOAEL extrapolation was included in the BMD approach.
Comment 3: Two reviewers questioned the inclusion of a database uncertainty factor of 3 for
lack of chronic reproductive and developmental studies, noting the lack of consistency with
statements elsewhere in the assessment that, due to the reactivity of phosgene, the primary toxic
impacts are expected to occur at the point of first contact. One of the two reviewers supported
removing the uncertainty factor based on lack of reproductive/developmental studies. The
second reviewer observed that any effects on reproduction or development secondary to
respiratory impairment of mothers or newborn offspring would be avoided where the health-
protective level was low enough to avoid such damage to the respiratory system. This reviewer
raised the possible concern that impacts on the respiratory system might exacerbate or even
C-4
-------
cause asthma in some individuals, including children, although no evidence was presented to
imply that such differential impacts would exceed the default UFH of 10. The reviewer
recommended that further consideration be given to whether additional uncertainty factors are
needed specifically to protect children from this or other health impacts.
Response 3: The database uncertainty factor was changed from 3 to 1. As noted by the external
reviewers, effects outside the respiratory system are not expected because of the short half-life of
phosgene in the respiratory system, and it would not be expected that phosgene will migrate to
the systemic circulation in concentrations large enough to cause reproductive or developmental
effects. Justification for a database uncertainty factor of 1 was provided.
B. Cancer Weight of Evidence
Comment: Two reviewers agreed with the Agency conclusion that the available evidence for
phosgene are inadequate to draw conclusions about the carcinogenicity of the chemical. The third
reviewer offered no comments on cancer weight of evidence.
Response: No response is required.
C. Other Comments
Comment 1: One reviewer observed that the characterization of the exposure duration in the
Kodavanti et al. (1997) study on page 9 of the Toxicological Review as a "repeated acute
exposure" was inaccurate, and that it was more correctly characterized as a subchronic study.
Response 1: EPA recognizes that a 12-week exposure is not an acute study. The Toxicological
Review (Section 4.2.2.1) referred to "acute" exposures (in quotes) as the characterization that the
authors used in the description of their study.
Comment 2: One reviewer recommended that consistent concentration units be used throughout
the document.
Response 2: The assessment was revised to provide equivalent concentration units to facilitate the
conversion between units (e.g., from ppm to mg/m3).
Comment 3: One reviewer observed that the CatReg point estimate of 0.088 ppm is sometimes
C-5
-------
expressed as 0.09 and sometimes as 0.1, but this value should be consistent throughout the
document.
Response 3: For the calculation of the HEC at the bottom of page 26 of the external review draft
of the Toxicological Review, the CatReg value of 0.088 ppm, which reflects the unadjusted
exposure concentration from the animal study, was rounded to 0.09 ppm. The human equivalent
concentration (HEC) was calculated as 0.098 mg/m3, which was rounded to 0.1 mg/m3. Therefore,
the two values are correct as presented in the assessment.
Comments from the Public
Comment 1: A reviewer commented that the application of Haber's Law to chronic effects of a
non-cumulative substance like phosgene is not well-founded. The sharp dose-response in the
Kodavanti et al. (1997) study indicates the critical importance of exposure intensity rather than
exposure duration, as illustrated by the results of Henderson et al. (1993) on effects of ozone
exposure.
Response 1: The Agency agrees that it is inappropriate to use Haber's Law as the sole rationale
for extrapolating the intermittent animal exposures to continuous animal dosing, as was done in
the draft document. This is the default assumption used in most EPA evaluations. Since
Kodavanti et al. (1997) found that variation in intermittent phosgene exposure durations in the
range of 1 to 5 days per week had little influence on the events leading to lung fibrosis, the
revised document deletes the 5/7 factor used in the default procedure. However, the 6/24 factor
for partial day exposure was retained on the assumption that Haber's Law is valid for continuous
exposures in the range of a few hours to one day, as shown by earlier continuous acute exposures
of phosgene. Henderson's paper dealt with continuous exposures from 3 to 24 hours, and
therefore, does not contribute to understanding the effect of intermittent phosgene exposures.
Comment 2: A reviewer observed that exposure intensity, in addition to incidence, should be
taken into account in the agency's analysis. The agency should address not only whether the
chronic effects are driven by (repeated) acute effects, but also that the acute threshold may be
valid for chronic exposure. This suggested approach would be based on exposure intensity.
Response 2: In agreement with the commentor, the EPA did consider phosgene concentration
(which is equivalent to exposure intensity) to be the dose metric critical for lung damage leading
C-6
-------
to fibrosis. Acute phosgene experiments do not show a threshold, but do show a response
dependent on C x T . The threshold as analyzed in the Henderson et al. (1993) paper was actually
what EPA would call a "point of departure," which is a response level that exceeds that of the
controls, rather than a threshold dose below which there is no response.
Comment 3: A reviewer commented that the draft review did not describe in sufficient detail the
anatomical location of the lung lesions leading to fibrosis, the stains used to identify collagen
deposition or the scoring criteria used for the pathological analysis.
Response 3: The information on the stain used and a better description of the criteria that were
used to judge the severity of the lesions was added to the document.
Comment 4: A reviewer stated that the uncertainty factor for the CatReg analysis (xlO) is not
sufficiently justified and is overly conservative.
Response 4: The CatReg analysis has been rewritten (see above discussions regarding the
relation of CatReg to the BMD and NOAEL approaches).
Comment 5: A reviewer commented that the draft does not distinguish between measured and
nominal concentrations used in the various experiments and does not state the temperature and
humidity conditions under which the experiments were carried out, nor the analytical procedures
used.
Response 5: The major studies used in the evaluation (Kodavanti et al., 1997; Hatch et al., 2001;
Belgrade et al., 1995) were conducted in the same laboratory under the same temperature and
humidity conditions, and the measured concentrations closely matched the targeted concentration.
Therefore, there is no need to make concentration adjustments before comparing results between
the major studies. The measured concentrations for the Kodavanti et al. (1997) study were added
to the document. The deviation from targeted values was generally within 10% of the mean.
Comment 6: A reviewer considered the derived RfC to be overly conservative from two points of
view: (1) the BMD approach results in a lower RfC than the NOAEL approach, and although it is
preferred, a better understanding of the role of exposure intensity and duration is likely to lead to
a less conservative RfC, and (2) the RfC may be in the range of rural levels (non-anthropogenic
sources) of phosgene.
C-7
-------
Response 6: The agency agrees that the BMD approach is more valid because it uses more of the
experimental data than the NOAEL approach, and has stated preference for that approach.
However the agency is unwilling to speculate about whether better understanding of the role of
exposure intensity and duration would eventually lead to a higher or lower RfC.
The RfC derived in this assessment is 3E-4 mg/m3 = 75 ppt. The current ambient levels of
phosgene are not known with any degree of confidence but the 1977 phosgene concentration in
clean air (rural and seacoast) locations averaged 22 ppt, which is about 0.3 times the RfC. Since
the main precursors of phosgene in air are tetrachloroethylene and trichloroethylene, which have
steadily decreased over the last 15 years, the current ambient air levels of phosgene are probably
much less than 0.3 times the RfC. There is an obvious need for current ambient air measurements
in order to answer this comment.
C-8
------- |