DRAFT EPA/635/R-06/003
DO NOT CITE OR QUOTE August 2006
External Review Draft
Evaluation of the Carcinogenicity of Ethylene Oxide
NOTICE
This document is an external review draft for review purposes only and does not
constitute U.S. Environmental Protection Agency policy. Mention of trade names
or commercial products does not constitute endorsement or recommendation for
use.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
-------
DISCLAIMER
This document is a draft for review purposes only and does not constitute U.S.
Environmental Protection Agency policy. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
08/29/06 ii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS
LIST OF TABLES v
LIST OF FIGURES vii
PREFACE viii
AUTHORS, CONTRIBUTORS, AND REVIEWERS ix
1. EXECUTIVE SUMMARY 1
2. INTRODUCTION 3
3. HAZARD IDENTIFICATION 4
3.1. EVIDENCE OF CANCER IN HUMANS 4
3.1.1. Conclusions Regarding the Evidence of Cancer in Humans 11
3.2. EVIDENCE OF CANCER IN EXPERIMENTAL ANIMALS 12
3.2.1. Conclusions Regarding the Evidence of Cancer in Experimental Animals ... 14
3.3. SUPPORTING EVIDENCE 15
3.3.1. Metabolism and Kinetics 15
3.3.2. Mutagenicity 17
3.3.2.1. Mutations 17
3.3.2.2. Sister Chromatid Exchanges 19
3.3.2.3. Chromosomal Aberrations 19
3.4. MODE OF ACTION 21
3.4.1. Analysis of the Mode of Action for Ethylene Oxide Carcinogenicity
Under EPA's Mode of Action Framework 24
3.5. HAZARD CHARACTERIZATION 26
3.5.1. Characterization of Cancer Hazard 26
3.5.2. Susceptible Lifestages and Subpopulations 28
3.5.3. Cancer Classification Conclusions of Other Agencies 28
4. CANCER DOSE-RESPONSE RELATIONSHIP FOR INHALATION EXPOSURE 29
4.1. INHALATION UNIT RISK ESTIMATES DERIVED FROM HUMAN DATA .... 29
4.1.1. Risk Estimates for Lymphohematopoietic Cancer 30
4.1.1.1. Lymphohematopoietic Cancer Results From the NIOSH Study ... 30
4.1.1.2. Prediction of Lifetime Extra Risk of Lymphohematopoietic Cancer
Mortality 31
4.1.1.3. Prediction of Lifetime Extra Risk of Lymphohematopoietic Cancer
Incidence 35
4.1.2. Risk Estimates for Breast Cancer 36
4.1.2.1. Breast Cancer Results From the NIOSH Study 36
4.1.2.2. Prediction of Lifetime Extra Risk of Breast Cancer Mortality 36
4.1.2.3. Prediction of Lifetime Extra Risk of Breast Cancer Incidence 38
4.1.3. Sources of Uncertainty in the Cancer Risk Estimates 41
4.1.4. Conclusions 49
08/29/06 iii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
4.2. INHALATION UNIT RISK DERIVED FROM EXPERIMENTAL ANIMAL
DATA 49
4.2.1. Overall Approach 49
4.2.2. Cross-Species Scaling 50
4.2.3. Dose-Response Modeling Methods 52
4.2.4. Description of Experimental Animal Studies 54
4.2.5. Results of Data Analysis of Experimental Animal Studies 55
4.3. INHALATION UNIT RISK ESTIMATES—CONCLUSIONS 56
4.4. COMPARISON WITH OTHER ASSESSMENTS 58
4.4.1. Assessments Based on Human Studies 58
4.4.2. Assessments Based on Laboratory Animal Studies 59
4.5. RISK ESTIMATES FOR OCCUPATIONAL EXPOSURES 61
APPENDIX A: CRITICAL REVIEW OF EPIDEMIOLOGIC EVIDENCE 95
APPENDIX B: REFERENCES FOR FIGURE 3 139
APPENDIX C: EXTRA RISK CALCULATION 140
APPENDIX D: MODEL PARAMETERS IN THE ANALYSIS OF ANIMAL TUMOR
INCIDENCE 142
REFERENCES 146
08/29/06 iv DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES
1. Tumor incidence data in National Toxicology Program Study of B6C3FJ
mice (NTP, 1987) 64
2. Tumor incidence data in Lynch et al. (1982, 1984) study of male F344 rats 65
3. Tumor incidence data in Snellings et al. (1984) and Garman et al. (1985) reports on
F344 rats 66
4. Cytogenetic effects in humans 67
5. Cox regression results for all lymphohematopoietic cancer mortality in males 70
6. Extra risk estimates for all lymphohematopoietic cancer mortality in males from
various levels of lifetime exposure to ethylene oxide 71
7. EC01, LEC01, and unit risk estimates for lymphohematopoietic cancer in males 72
8. Cox regression results for "lymphoid" cancer mortality in males 73
9. EC01, LEC01, and unit risk estimates for "lymphoid" cancer in males 74
10. Cox regression results for breast cancer mortality in females 75
11. Extra risk estimates for breast cancer mortality in females from various levels of
lifetime exposure to ethylene oxide 76
12. EC01, LEC01, and unit risk estimates for breast cancer mortality in females 77
13. Cox regression results for breast cancer incidence in females 78
14. EC01, LEC01, and unit risk estimates for breast cancer incidence in females -
invasive and in situ 79
15. EC01, LEC01, and unit risk estimates for breast cancer incidence in
females - invasive only 80
16. Upper-bound unit risks (per |ig/m3) obtained by combining tumor sites 81
17. Unit risk values from multistage Weibull time-to-tumor modeling of mouse tumor
incidence in the NTP (1987) study 82
18. Summary of unit risk estimates (per |ig/m3) in animal bioassays 83
08/29/06 v DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
19. Comparison of unit risk estimates 84
20. Extra risk estimates for lymphohematopoietic cancer in males for various occupational
exposure levels 85
21. Extra risk estimates for breast cancer in females for various occupational exposure
levels 86
08/29/06 vi DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
1. Metabolism of ethylene oxide 87
2. Simulated blood AUCs for EtO following a 6-hour exposure to EtO from the rat,
mouse, and human PBPK models of Fennell and Brown (2001); based on data
presented in Fennell and Brown (2001) 88
3. Display of 203 data sets, including bacteria, fungi, plants, insects, and mammals
(in vitro and in vivo), measuring the full range of genotoxic endpoints 89
4. RR estimate for lymphohematopoietic cancer in males vs. mean exposure (from
Steenland et al., 2004, Table 6, Cox regression results, except for linear regression
[see text]; log and categorical exposures with 15-year lag), unadjusted for
continuous exposure 90
5. RR estimate for "lymphoid" cancer in males vs. mean exposure (from Steenland et al.,
2004, Table 7, Cox regression results, except for linear regression [see text]; log and
categorical exposures with 15-year lag), unadjusted for continuous exposure 91
6. RR estimate for breast cancer mortality in females vs. mean exposure (from Steenland
et al., 2004, Table 8, Cox regression results, except for linear regression [see text];
with 20-year lag), unadjusted for continuous exposure 92
7. RR for breast cancer incidence in females (full cohort) vs. mean exposure (from
Steenland et al., 2003, Table 5, Cox regression results, except for linear regression
[see text]; 15-year lag), unadjusted for continuous exposure 93
8. RR for breast cancer incidence in females (subcohort with interviews) vs. mean
exposure (from Steenland et al., 2003, Table 5, Cox regression results, except for
linear regression [see text]; 15-year lag), unadjusted for continuous exposure 94
08/29/06 vii DRAFT—DO NOT CITE OR QUOTE
-------
PREFACE
This document is the U.S. Environmental Protection Agency's (EPA's) external review
draft of the evaluation of the carcinogenicity of ethylene oxide (EtO). The assessment was
prepared by the National Center for Environmental Assessment, which is the health risk
assessment program in the Office of Research and Development. The assessment broadly
supports activities authorized in the 1990 Clean Air Act and is of particular interest to EPA's
Office of Air and Radiation. However, this review also should be applicable to the needs of all
program Offices and Regions in evaluating the carcinogenicity of EtO.
EPA last published a health assessment of the potential carcinogenicity of EtO in 1985
(U.S. EPA, 1985). The current assessment reviews the more recent database on the
carcinogenicity of EtO, pertinent data from the 1985 assessment, and several reviews and
assessments issued by other organizations (IARC, 1994; Health Canada, 2001; CalEPA, 1999;
EOIC, 2001). This document was preceded by an internal review draft (NCEA-W-1341). The
scientific literature search for this assessment is generally current through June 2004, although a
few later publications are included. This assessment focuses on lifetime cancer risk from
inhalation exposure.
08/29/06 viii DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
ASSESSMENT AUTHORS
Henry D. Kahn, Chemical Manager
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
David Bayliss
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Jennifer Jinot
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Nagu Keshava
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Robert McGaughy
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Ravi Subramaniam
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Larry Valcovic
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
INTERNAL EPA REVIEWERS
Michele Burgess, OSWER
Deborah Burgin, OPEI
Kerry Dearfield, ORD/OSP
Joyce Donahue, OW
Michael Firestone, AO/OCHP
08/29/06 ix DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
Karen Hogan, ORD/NCEA-IO
Aparna Koppikar, ORD/NCEA-W
Deirdre Murphy, OAR/ESD
Steve Nesnow, ORD/NHEERL
Marian Olsen, Region 2
Brenda Perkovich-Foos, AO/OCHP
Julian Preston, ORD/NHEERL
Santhini Ramasamy, OPP/HED
Nancy Rios-Jafolla, Region 3
Tracey Woodruff, AO/NCEE
The authors would like to acknowledge Dr. Julian Preston and Mr. David Bussard for
their contributions during the drafting process and Randall Smith of the National Institute for
Occupational Safety and Health for his previous assistance with the BEIR (1988) methodology.
08/29/06 x DRAFT—DO NOT CITE OR QUOTE
-------
1 1. EXECUTIVE SUMMARY
2
3 Ethylene oxide (EtO) is a gas at room temperature. It is manufactured from ethylene and
4 used primarily as a chemical intermediate in the manufacture of ethylene glycol. It is also used
5 as a sterilizing agent for medical equipment and as a fumigating agent for spices. The largest
6 sources of human exposure are in occupations involving contact with the gas in plants (facilities)
7 and in hospitals that sterilize medical equipment. EtO can also be inhaled by residents living
8 near production or sterilizing/fumigating facilities.
9 This review should be applicable to the needs of all program Offices and Regions in
10 evaluating the carcinogenicity of EtO. EPA last published a health assessment of the potential
11 carcinogenicity of EtO in 1985 (U.S. EPA, 1985). The current assessment reviews the more
12 recent database that has developed on the carcinogenicity of EtO, pertinent data from the 1985
13 assessment, and several reviews and assessments issued by other organizations (IARC, 1994;
14 Health Canada, 2001; CalEPA, 1999; EOIC, 2001; NTP, 2000). This assessment focuses on
15 lifetime cancer risk from inhalation exposure.
16 The DNA-damaging properties of EtO have been studied since the 1940s. EtO is known
17 to be mutagenic in a large number of living organisms, ranging from bacteriophage to mammals,
18 and it also induces chromosome damage. It is carcinogenic in mice and rats, inducing tumors of
19 the lymphohematopoietic system, brain, lung, connective tissue, uterus, and mammary gland. In
20 humans employed in EtO-manufacturing facilities and in sterilizing facilities, the greatest
21 evidence of a cancer risk from exposure is for cancer of the lymphohematopoietic system.
22 Increases in the risk of lymphohematopoietic cancer have been seen in several studies,
23 manifested as an increase either in leukemia or in cancer of the lymphoid tissue. In one large
24 epidemiologic study of sterilizer workers that had a well-defined exposure assessment for
25 individuals, positive exposure-response trends for lymphohematopoietic cancer mortality in
26 males and for breast cancer mortality in females were reported (Steenland et al., 2004). The
27 positive exposure-response trend for female breast cancer was confirmed in an incidence study
28 based on the same worker cohort (Steenland et al., 2003).
29 Although the evidence of carcinogenicity from human studies was short of conclusive,
30 EtO was characterized as carcinogenic to humans based on the total weight of evidence, in
31 accordance with EPA's 2005 Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a).
32 Supporting information includes sufficient evidence of carcinogenicity in laboratory animals,
33 clear evidence that EtO is genotoxic, and evidence of chromosome damage in humans exposed
34 to EtO.
35 This document derives inhalation unit risk estimates for cancer mortality and incidence
36 based on the human data. An EC01 of 44 |ig/m3 (0.024 ppm) was calculated using a life-table
08/29/06 1 DRAFT—DO NOT CITE OR QUOTE
-------
1 analysis and linear modeling of the categorical Cox regression analysis results for excess
2 lymphohematopoietic cancer mortality in males reported in a high-quality occupational
3 epidemiologic study (Steenland et al., 2004). Linear low-dose extrapolation from the LEC01
4 yielded a lifetime extra cancer mortality unit risk estimate of 5.0 x 10"4 per jig/m3 (0.92 per ppm)
5 of continuous EtO exposure. Applying the same linear regression coefficient and life-table
6 analysis to background male lymphohematopoietic cancer incidence rates yielded an EC01 of 24
7 M-g/rn3 (0.013 ppm) and a preferred lifetime extra cancer unit risk estimate of 9.0 x 10"4 per |ig/m3
8 (1.6 per ppm). The preferred estimate is greater than the estimate of 5.0 x 10"4 per |ig/m3 (0.91
9 per ppm; EC01 = 44 jig/m3) calculated, using the same approach, from the results of a breast
10 cancer incidence study of the same worker cohort (Steenland et al., 2003), and is recommended
11 as the potency estimate for Agency use. Although there was no clear exposure-response
12 relationship for lymphohematopoietic cancer in females in the Steenland et al. (2004) study, an
13 increased risk to females cannot be ruled out. Nonetheless, the Steenland et al. results suggest
14 that if such a risk exists for females, it is likely to be lower than the risk estimated for males;
15 thus, it is expected that the risk estimate based on lymphohematopoietic cancer in males would
16 be protective of females, even if they have an increased risk for both breast cancer and
17 lymphohematopoietic cancer.
18 Because the weight of evidence supports a mutagenic mode of action for EtO
19 carcinogenicity, and in the absence of chemical-specific data on early-life susceptibility,
20 increased early-life susceptibility should be assumed and, if there is early-life exposure, the age-
21 dependent adjustment factors (ADAFs) should be applied, as appropriate, in accordance with
22 EPA's Supplemental Guidance for Assessing Susceptibility From Early-Life Exposure to
23 Carcinogens, hereinafter referred to as "EPA's Supplemental Guidance" (U.S. EPA, 2005b).
24 Applying the ADAFs to the unit risk estimate of 9.0 x 10"4 per |ig/m3 yields a full lifetime unit
25 risk estimate of 1.5 x 10"3 per i-ig/m3, and the commensurate lifetime chronic exposure level of
26 EtO corresponding to an increased cancer risk of 10"6 is 0.0007 |ig/m3. [Note that for less-than-
27 lifetime exposure scenarios (or for exposures that vary with age), the adult-based potency
28 estimate of 9.0 x 10"4 per |ig/m3 should be used, in conjunction with the ADAFs as appropriate,
29 in accordance with EPA's Supplemental Guidance.]
30 Unit risk estimates were also derived from the three chronic rodent bioassays for EtO
31 reported in the literature, without considering early-life susceptibility. These estimates, ranging
32 from 2.2 x 10"5 per |ig/m3 to 4.6 x 10"5 per i-ig/m3, are about an order of magnitude lower than the
33 estimates based on human data. The Agency takes the position that human data, if adequate data
34 are available, provide a more appropriate basis than rodent data for estimating population risks
35 (U.S. EPA, 2005a), primarily because uncertainties in extrapolating quantitative risks from
36 rodents to humans are avoided. Although there is a fairly sizable difference between the rodent-
08/29/06 2 DRAFT—DO NOT CITE OR QUOTE
-------
1 and human-based estimates, the similarity between the unit risk estimates based on the male
2 lymphohematopoietic cancer and the female breast cancer results increases confidence in the use
3 of the unit risk estimate based on the male lymphohematopoietic cancer results.
4 The unit risk estimates were developed for environmental exposure levels and are not
5 necessarily applicable to higher-level occupational exposures, which appear to be subject to a
6 different exposure-response relationship. However, occupational exposure levels are of concern
7 to EPA when EtO is used as a pesticide (e.g., fumigant for spices). Therefore, this document also
8 presents extra risk estimates for cancer for a number of occupational exposure scenarios.
9
10
11 2. INTRODUCTION
12
13 Ethylene oxide (EtO) is a gas at room temperature. It is manufactured from ethylene and
14 used primarily as a chemical intermediate in the manufacture of ethylene glycol (NTP, 2000). It
15 is also used as a sterilizing agent for medical equipment and as a fumigating agent for spices.
16 The largest sources of human exposure are in occupations involving contact with the gas in
17 plants (facilities) and in hospitals that sterilize medical equipment. EtO can also be inhaled by
18 residents living near production or sterilizing/fumigating facilities.
19 The purpose of this document is to derive the cancer inhalation unit risk estimate for
20 ethylene oxide (EtO). The document was prepared by the National Center for Environmental
21 Assessment (NCEA), Office of Research and Development, for use by EPA's Office of Air
22 Quality Planning and Standards (OAQPS), Office of Air and Radiation. EPA last published a
23 health assessment of the potential carcinogenicity of EtO in 1985 (U.S. EPA, 1985). Under
24 Section 112 of the 1990 Clean Air Act Amendments, the Agency is required to promulgate
25 national standards for source categories emitting any of the 188 currently listed hazardous air
26 pollutants (HAPs) in amounts exceeding specific emission thresholds. The initial standards were
27 technology-based emission standards, with further requirements in Section 112(f) for EPA to
28 consider the need for additional "residual risk" standards that would, if required, "provide an
29 ample margin of safety to protect public health". Ethylene Oxide is one of the 188 HAPs for
30 which the Agency is considering residual risk standards. Understanding the time frame for this
31 risk-based regulatory decision and the evidence of EtO carcinogenicity, OAQPS and NCEA
32 chose to put initial emphasis on the assessment of cancer hazard and dose-response.
33 Although OAQPS was the requesting office, this review should be applicable to the needs
34 of other program offices and regions in evaluating the carcinogenicity of EtO. The current
35 assessment reviews the more recent database that has developed on the carcinogenicity of EtO,
36 pertinent data from the 1985 assessment, and several reviews and assessments issued by other
08/29/06 3 DRAFT—DO NOT CITE OR QUOTE
-------
1 organizations (IARC, 1994; Health Canada, 2001; CalEPA, 1999; EOIC, 2001). This
2 assessment focuses on lifetime cancer risk from inhalation exposure.
3
4
5 3. HAZARD IDENTIFICATION
6
7 3.1. EVIDENCE OF CANCER IN HUMANS
8 The literature from 1988 to present contains numerous studies of the carcinogenic effects
9 of EtO in occupational cohorts; some of these cohorts were the subject of multiple reports. The
10 conclusions about the human evidence of carcinogenicity in this assessment are based on the
11 following summary of those studies, which are critically reviewed in Appendix A.
12 Two primary sources of exposures to EtO are production facilities and sterilization
13 operations. There are two types of production facilities (IARC, 1994):
14
15 1. those using the older chlorohydrin process, where ethylene is reacted with
16 hypochlorous acid and then with calcium oxide to make EtO (this method produces
17 unwanted byproducts, the most toxic of which is ethylene dichloride), and
18
19 2. those producing EtO via direct oxidation of ethylene in a pressurized vessel, which
20 involves less EtO exposure and eliminates the chemical byproducts of the
21 chlorohydrin process.
22
23 Exposure in the sterilization of medical equipment and in the direct oxidation process is
24 predominantly to EtO, whereas exposure in the chlorohydrin process is to EtO mixed with other
25 chemicals.
26 Hogstedt et al. (1986) and Hogstedt (1988) summarized findings of three Swedish
27 occupational cohorts (539 men and 170 women) exposed in a plant where hospital equipment is
28 sterilized, in a chlorohydrin production facility, and in a direct oxidation production facility. The
29 incidence of leukemia was elevated in all cohorts, although the risk was not statistically
30 significant in the cohort from the direct oxidation facility. For the three cohorts combined there
31 were statistically significantly elevated standard mortality ratios (SMRs) for leukemia (SMR =
32 9.2; 95% confidence interval ([CI] = 3.7-19), based on 7 deaths, and for stomach cancer (SMR =
33 5.5; 95% CI = 2.6-10), based on 10 deaths. Although this study produced high SMRs for
34 leukemia, stomach cancer, and total cancer, it suffers from some limitations, such as multiple
35 exposures to numerous other chemicals, lack of personal exposure information, and lack of
36 latency analysis. No gender differences were separately analyzed. No dose-response calculations
37 were possible. This study provides suggestive evidence of the carcinogenicity of EtO.
08/29/06 4 DRAFT—DO NOT CITE OR QUOTE
-------
1 Coggon et al. (2004) reported the results of a followup study of a cohort originally studied
2 by Gardner et al. (1989). The cohort included workers in three EtO production facilities (two
3 using both chlorohydrin and direct oxidation processes and the third using direct oxidation only);
4 in a fourth facility that used EtO in the manufacture of other chemicals; and in eight hospitals
5 that used EtO in sterilizing units. The total cohort comprised 1,864 men and 1,012 women. No
6 statistically significant excesses were observed for any cancer site. Slight increases, based on
7 small numbers, were observed for the various lymphohematopoietic cancers: Hodgkin's disease
8 (2 vs. 1 expected), non-Hodgkin's lymphoma (NHL) (7 vs. 4.8), multiple myeloma (3 vs. 2.5),
9 and leukemia (5 vs. 4.6). The increases were concentrated in the 1,471 chemical-manufacturing
10 workers, of whom all but 1 were male. In the chemical-manufacturing workers with "definite"
11 exposure, 4 leukemias were observed (1.7 expected) and 9 lymphohematopoietic cancers were
12 observed (4.9 expected). A slight deficit in the risk of breast cancer deaths (11 vs. 13.2) was
13 observed in the cohort. No individual exposure measurements were obtained from cohort
14 members, and no exposure measurements were available before 1977. Multiple exposures to
15 other chemicals, small numbers of deaths, and lack of individual EtO measurements make this
16 study only suggestive of a higher risk of leukemia from exposure to EtO.
17 A series of retrospective mortality studies of 2,174 male workers in a Union Carbide
18 Corporation (UCC) EtO production facility in West Virginia (Greenberg et al., 1990; Teta et al.,
19 1993, 1999; Benson and Teta, 1993) has been published. The chlorohydrin process was used
20 from 1925 to 1957, and the direct oxidation process was used from 1937 to 1971. The cohort
21 was observed from 1940 through 1978 in the original study (Greenberg et al., 1990) and through
22 1988 in the latter three studies. A large-scale industrial hygiene survey and monitoring of EtO
23 concentrations was carried out in 1976.
24 Greenberg et al. (1990) found elevated but not statistically significant risks of pancreatic
25 cancer (SMR = 1.7) and leukemia (SMR = 2.3) (each based on seven cases) in the entire cohort;
26 most of the cases occurred in the chlorohydrin production unit. Limitations to this study
27 included multiple exposures to many different chemicals in the facility through the years and lack
28 of exposure measurements to EtO prior to 1976. Exposure levels prior to 1976 were assumed to
29 be similar to those found in 1976. Three categories of exposure were established for
30 analysis—low, intermediate, and high—based on the likelihood of the occurrence of
31 dermatologic or other medical problems and the results of the 1976 survey. No significant
32 findings of a dose-response relationship were discernable. No quantitative estimates of
33 individual exposure were made in this study. With no latency analysis, there exists only the
34 suggestion of an increased risk of cancer in this study. Furthermore, EtO is not the only chemical
35 to which this excess mortality could be attributed.
08/29/06 5 DRAFT—DO NOT CITE OR QUOTE
-------
1 A followup study (Teta et al., 1993) that extended the observation of direct oxidation
2 workers only (eliminating the 278 chlorohydrin workers) for an additional 10 years to 1988,
3 found no significant risk of total cancer; there was a slight trend in the risk of leukemia with
4 increasing duration of assignment to departments using or processing EtO, but it was not
5 significant and was based on five cases (see Table A-2). The same problems of exposure
6 ascertainment exist for this study as for that of Greenberg et al. (1990). For estimates of
7 exposure prior to 1976, levels were assumed to be the same as those at other "similar" plants. As
8 with Greenberg et al. (1990), no latency analysis was conducted in this study, although the
9 average followup was reported to be 27 years. Essentially, the study did not support the earlier
10 studies of cancer in EtO workers; however, it was limited by most of the same problems as the
11 larger Greenberg et al. study and thus could not assist in determining whether exposure to EtO
12 was causally related to cancer.
13 In a parallel followup study through 1988 of only the chlorohydrin employees, Benson
14 and Teta (1993) found that pancreatic cancer and hematopoietic cancer cases continued to
15 accumulate and that the SMRs were statistically significant for pancreatic cancer (SMR = 5; Obs
16 =8,p<0.05) and for hematopoietic cancer (SMR = 3; Obs = 8,p<0.05). They interpreted these
17 excesses as possibly due to ethylene dichloride, a byproduct in the chlorohydrin process. Again,
18 this small study of only 278 workers was limited by the same problems as the Greenberg et al.
19 (1990) study and the Teta et al. (1993) study. No individual estimates of exposure are available
20 and the workers were potentially exposed to many different chemicals (Table A-3). Hence this
21 study is marginal in determining the carcinogenicity of EtO.
22 In a later analysis, Teta et al. (1999) included an update of their earlier study of UCC
23 workers (Teta et al., 1993), and they fitted dose-response models to the updated UCC data and to
24 data from a study by the National Institute for Occupational Safety and Health (NIOSH)
25 (described below). They reported that latency and lagging of exposure did not appreciably affect
26 the fitted Poisson regression models to these data, which the authors assumed to be the best
27 models for evaluating dose-response relationships. Because Teta et al. (1999) did not present
28 aggregate risk ratios in the categories used to model dose-response relationships, the only
29 comparison that can be made between the UCC and NIOSH data is based on the fitted models.
30 These models are almost identical for leukemia, but for the lymphoid category, the
31 risk—according to the fitted model for the UCC data—decreased as a function of exposure,
32 whereas the risk for the modeled NIOSH data increased as a function of exposure. It is possible
33 that the difference is due to geographic differences in the coding of death certificates for specific
34 types of leukemia. The UCC workers were in West Virginia only, whereas workers in the
35 NIOSH study were from multiple states.
08/29/06 6 DRAFT—DO NOT CITE OR QUOTE
-------
1 In a study of 2,658 male workers at eight chemical plants where EtO is produced
2 (manufacturing process not stated), Kiesselbach et al. (1990) found slightly increased SMRs for
3 cancers of the stomach, esophagus, and lung. A latency analysis was done only for stomach
4 cancer and total mortality. The investigators considered 71.6% of the cohort to be "weakly"
5 exposed; only 2.6% were "strongly exposed." No data were provided to explain how these
6 exposure categories were derived. The workers were followed for a median 15.5 years. Without
7 additional information on exposure to EtO, this study is of little help at this time in determining
8 the carcinogenicity of EtO.
9 NIOSH conducted an industrywide study of 18,254 workers (45% male and 55% female)
10 in 14 plants where EtO was used (Steenland et al., 1991; Stayner et al., 1993; Steenland et al.,
11 2004). Most of the workers were exposed while sterilizing medical supplies and treating spices
12 and in the manufacture and testing of medical sterilizers. The procedures for selecting the
13 facilities and defining the cohort are described in Steenland et al. (1991), and the exposure model
14 and verification procedures are described in Greife et al. (1988) and Hornung et al. (1994).
15 Results of the original followup study are presented in Steenland et al. (1991) and Stayner et al.
16 (1993). To qualify for the study, each of the 14 plants had to achieve at least 400 person-years
17 of risk before 1978, and to be included in the cohort, a worker had to have been exposed for at
18 least 3 months. The average year of first exposure was 1970. The cohort averaged 26.8 years of
19 followup in the extended followup study (Steenland et al., 2004). The age at entry is not
20 provided, nor is an age breakdown available. By the extended cutoff date on December 31,
21 1998, 16% of the cohort had died. Individual exposure estimates were derived for workers from
22 13 of the 14 plants.
23 The overall SMR for cancer was 0.98, based on 860 deaths (Steenland et al., 2004). The
24 SMR for (lympho)hematopoietic cancer was 1, based on 79 cases. Exposure-response analyses,
25 however, revealed exposure-related increases in hematopoietic cancer mortality risk, although
26 the effect was limited to males. In categorical life-table analysis, men with >13,500 ppm-days of
27 cumulative exposure had an SMR of 1.46, n = 13. In internal Cox regression analyses with
28 exposure as a continuous variable, statistically significant trends in males for all hematopoietic
29 cancer (p=0.02) and for "lymphoid" cancers (NHL, lymphocytic leukemia, and myeloma;
30 /?=0.02) were observed using log cumulative exposure (ppm-days) with a 15-year lag. In internal
31 categorical analyses, statistically significant odds ratios (ORs) were observed in the highest
32 cumulative exposure quartile (with a 15-year lag) in males for all hematopoietic cancer (OR =
33 3.42; 95% CI = 1.09-10.73) and "lymphoid" cancer (OR = 3.76; 95% CI = 1.03-13.64). The
34 exposure metrics of duration of exposure, average concentration, and maximum (8-hour time-
35 weighted average [TWA]) concentration did not predict the hematopoietic cancer results as well
36 as did the cumulative exposure metric.
08/29/06 7 DRAFT—DO NOT CITE OR QUOTE
-------
1 Although the overall SMR for female breast cancer was 0.99, based on 102 deaths, the
2 NIOSH mortality follow-up study reported a significant excess of breast cancer mortality in the
3 highest cumulative exposure quartile using a 20-year lag period compared to the U.S. population
4 (SMR = 2.07; 95% CI = 1.10-3.54; n = 13). Internal exposure-response analyses also noted a
5 significant positive trend for breast cancer using the log of cumulative exposure and a 20-year
6 lag time (p=0.01). In internal categorical analyses, a statistically significant OR was observed in
7 the highest cumulative exposure quartile with a 20-year lag (OR = 3.13; 95% CI = 1.42-6.92).
8 In summary, although the overall external comparisons did not demonstrate increased
9 risks, the NIOSH investigators found significant internal exposure-response relationships
10 between exposure to EtO and cancers of the hematopoietic system, as well as breast cancer
11 mortality. [Internal comparisons are considered superior to external comparisons in occupational
12 epidemiology studies because internal comparisons help control for the healthy worker effect
13 and other factors that might be more comparable within a study's worker population than
14 between the workers and the general population.] Exposures to other chemicals in the workplace
15 were believed to be minimal or nonexistent. This study is the most useful of the epidemiologic
16 studies in terms of carrying out a quantitative risk assessment. It possesses more attributes than
17 the others for performing risk analysis (e.g., better estimates of individual exposure, lack of
18 exposure to other chemicals, and a large and diverse distribution of workers).
19 It should be noted that Steenland et al. (2004) used Cox regression models, which are
20 loglinear relative rate models, thus providing some low-dose sublinear curvature for doses
21 expressed in terms of cumulative exposure. However, the best-fitting model for both male
22 lymphoid and all hematopoietic cancers was for dose expressed in terms of log cumulative
23 exposure, indicating supralinearity of the low-dose data. This is in contrast to the reported
24 results of Kirman et al. (2004) based on the Teta et al. (1999) analysis combining the 1993 UCC
25 leukemia data with the 1993 NIOSH leukemia data, which are claimed by the authors to provide
26 empirical evidence supporting a quadratic dose-response relationship. The 2004 NIOSH dose-
27 response data for hematopoietic cancers clearly do not provide empirical evidence in support of a
28 quadratic dose-response relationship. On the contrary, the NIOSH data suggest a supralinear
29 dose-response relationship in the observable range.
30 Wong and Trent (1993) investigated the same cohort as Steenland et al. (1991) but added
31 474 new unexplained subjects and increased the followup period by one year. They incremented
32 the total number of deaths by 176 and added 392.2 more expected deaths. The only positive
33 finding was a statistically significantly increased risk of NHL among men (SMR = 2.5; Obs =
34 16;/><0.05). However, there was a deficit risk of NHL among women. For breast cancer, there
35 was no trend of increasing risk by duration of employment or by latency. This study has major
36 limitations, not the least of which is a lack of detailed employment histories, making it
08/29/06 8 DRAFT—DO NOT CITE OR QUOTE
-------
1 impossible to quantify individual exposures and develop dose-response relationships.
2 Furthermore, the addition of more than twice as many expected deaths as observed deaths makes
3 the analysis by the authors questionable.
4 In a mortality study of 1,971 male chemical workers in Italy, 637 of whom were licensed
5 to handle EtO but not other toxic gases, Bisanti et al. (1993) reported statistically significant
6 excesses of hematopoietic cancers (SMR = 7.1, Obs = 5,/><0.05). The study was limited by the
7 lack of exposure measurements and by the young age of the cohort. Although this study
8 suggests that exposure to EtO leads to a significant excess of hematopoietic cancer, the lack of
9 personal exposure measurements and the fact that members were potentially exposed to other
10 chemicals in the workplace lessen its usefulness for establishing dose-response relationships.
11 Hagmar et al. (1991, 1995) studied cancer incidence in 2,170 Swedish workers (861 male
12 and 1,309 female) in two medical sterilizing plants. They determined concentrations in six job
13 categories and estimated exposure (ppm-years) for each worker. They found hematopoietic
14 cancers in 6 individuals versus 3.4 expected (SMR = 1.8) and a nonsignificant doubling in the
15 risk when a 10-year latency period was considered. Even though the cohort was young, the
16 followup time was short, and only a small fraction of the workers was highly exposed, the report
17 is suggestive. The risk of breast cancer was less than expected (standardized incidence ratio
18 [SIR] = 0.5, Obs = 5). In the latent category of 10 years or more, the risk was even lower (SIR =
19 0.4, Obs = 2).
20 In a large chemical manufacturing plant in Belgium (number of employees not stated),
21 Swaen et al. (1996) performed a nested case-control study of Hodgkin's disease to determine
22 whether a cluster of 10 cases in the active male work force was associated with any particular
23 chemical. They found a significant association for benzene and EtO. This study is limited by
24 the exclusion of inactive workers and the potential confounding effect of other chemicals besides
25 EtO, and it is not useful for quantitative risk assessment.
26 Olsen et al. (1997) studied 1,361 male employees working at four EtO chlorohydrin
27 process plants in the United States. Although they found a nonsignificant positive trend between
28 duration of employment as ethylene chlorohydrin workers and hematopoietic cancer, they
29 concluded that there was no appreciable risk in these workers, contradicting the findings of
30 Benson and Teta (1993). The small cohort size and the lack of data on EtO exposures limit the
31 usefulness of this study in inferring risks due to EtO.
32 Norman et al. (1995) studied 1,132 workers (204 male and 928 female) in a medical
33 sterilizing plant in the United States. In the women, there was a significant excess incidence of
34 breast cancer (SIR = 2.6, Obs = 12,/><0.05); no other cancer sites were elevated. The risk of
35 breast cancer was not noted to be excessive in the few previous studies where adequate numbers
36 of females were included and analyzed for breast cancer; however, only one of these was also an
08/29/06 9 DRAFT—DO NOT CITE OR QUOTE
-------
1 incidence study. The followup time was too short to draw meaningful conclusions at this time.
2 This study lacks the power to determine whether risks for cancers other than breast cancer are
3 statistically significantly elevated. It has no information regarding historical exposure and some
4 breast cancer victims had worked for less than one month.
5 Tompa et al. (1999) reported a cluster of 8 breast cancers and 8 other cancers in 98 nurses
6 exposed to EtO in a hospital in Hungary; however, the expected number of cases cannot be
7 identified.
8 The NIOSH investigators used the NIOSH cohort to conduct a study of breast cancer
9 incidence and exposure to EtO (Steenland et al., 2003). The researchers identified 7,576 women
10 from the initial cohort who had been employed in the commercial sterilization facilities for at
11 least 1 year (76% of the original cohort). Breast cancer incidence was determined from
12 interviews (questionnaires), death certificates, and cancer registries. Interviews were obtained
13 for 5,139 women (68% of the study cohort). The main reason for nonresponse was inability to
14 locate the study subject (22% of cohort). The average duration of exposure for the cohort was
15 10.7 years. For the full study cohort, 319 incident breast cancer cases were identified, including
16 20 cases of carcinoma in situ. Overall, the SIR was 0.87 (0.94 excluding the in situ cases) using
17 SEER reference rates for comparison. Results with the full cohort are expected to be
18 underestimated, however, because of case underascertainment in the women without interviews.
19 A significant exposure-response trend was observed for SIR across cumulative exposure
20 quintiles, using a 15-year lag time (p=0.002). In internal Cox regression analyses, with exposure
21 as a continuous variable, a significant trend was obtained for log cumulative exposure with a 15-
22 year lag (p=0.05), taking age, race, and year of birth into account. Using duration of exposure,
23 lagged 15 years, provided a slightly better fit (p=0.02), while models with maximum or average
24 exposure did not fit as well. In the Cox regression analysis with categorical exposures and a 15-
25 year lag, the top cumulative exposure quintile had a statistically significant OR of 1.74 (95% CI
26 = 1.16-2.65).
27 In the subcohort with interviews, 233 incident breast cancer cases were identified.
28 Information on various risk factors for breast cancer was also collected in the interviews, but
29 only parity and breast cancer in a first-degree relative turned out to be important predictors of
30 breast cancer incidence. In internal analyses with continuous exposure variables, the model with
31 duration of exposure (lagged 15 years) again provided the best fit (p=0.006). Both the
32 cumulative exposure and log cumulative exposure models also yielded significant regression
33 coefficients with a 15-year lag (p=0.02 and/>=0.03, respectively), taking age, race, year of birth,
34 parity, and breast cancer in a first-degree relative into account. In the Cox regression analysis
35 with categorical exposures and a 15-year lag, the top cumulative exposure quintile had a
36 statistically significant OR of 1.87 (95% CI = 1.12-3.10).
08/29/06 10 DRAFT—DO NOT CITE OR QUOTE
-------
1 Steenland et al. (2003) suggest that their findings are not conclusive of a causal
2 association because of inconsistencies in exposure-response trends, possible biases due to
3 nonresponse, and an incomplete cancer ascertainment. Although that conclusion seems
4 appropriate, those concerns do not appear to be major limitations. As noted by the authors, it is
5 not uncommon for positive exposure-response trends not to be strictly monotonically increasing,
6 conceivably due to random fluctuations or imprecisions in exposure estimates. Furthermore, the
7 consistency of results between the full study cohort, which is less subject to nonresponse bias,
8 and the subcohort with interviews, which should have full case ascertainment, alleviates some of
9 the concerns about those potential biases.
10 In a study of 299 female workers employed in a hospital in Hungary where gas sterilizers
11 were used, Kardos et al. (2003) observed 11 cancer deaths, including 3 breast cancer deaths,
12 compared with slightly more than 4 expected total cancer deaths. Site-specific expected deaths
13 are not available in this study, so it cannot be determined whether there is an excess risk of any
14 site-specific cancer.
15
16 3.1.1. Conclusions Regarding the Evidence of Cancer in Humans
17 Most of the human studies suggest a possible increased risk of lymphohematopoietic
18 cancers, but the total weight of the epidemiological evidence does not provide conclusive proof
19 of causality. Of the seven criteria of causality envisioned by Hill (1965), temporality, coherence,
20 and biological plausibility are clearly satisfied. There is also evidence of consistency in the
21 response, of a dose-response relationship (biological gradient), and of specificity when the
22 loosely defined blood malignancies are combined under the rubric "cancer of the hematopoietic
23 system." On the other hand, there is little strength in the magnitude of most of the estimates of
24 risk.
25 The NIOSH study (Steenland et al., 1991, 2004; Stayner et al., 1993) of workers at 14
26 chemical plants around the country provides the strongest evidence of carcinogenicity. A
27 positive trend is evident in the risk of lymphohematopoietic neoplasms with increasing
28 cumulative exposure to EtO, although only in males. Despite limitations in the data, most other
29 epidemiologic studies have also found elevated risks of lymphohematopoietic cancer from
30 exposure to EtO. Furthermore, when the exposure is relatively pure, such as in sterilization
31 workers, there is an elevated risk of lymphohematopoietic cancer that cannot be attributed to the
32 presence of confounders such as those that could potentially appear in the chlorohydrin process.
33 In addition, the studies that do not report a significant lymphohematopoietic cancer effect from
34 exposure to EtO suffer from severe limitations, such as small numbers of cases and inadequate
35 exposure information (see Table A-3).
08/29/06 11 DRAFT—DO NOT CITE OR QUOTE
-------
1 In addition, there is evidence of an increase in the risk of both breast cancer mortality and
2 incidence in women who are exposed to EtO. Two studies have reported increases in the risk of
3 breast cancer in women employees of commercial sterilization plants (Steenland et al., 2003,
4 2004; Norman et al., 1995) as well as in Hungarian hospital workers exposed to EtO (Kardos et
5 al., 2003). In several other studies where exposure to EtO would be expected to have occurred
6 among female employees, no elevated risks were seen (Hagmar et al., 1991; Hogstedt, 1988;
7 Hogstedt et al., 1986; Coggon et al., 2004). However, these studies had far fewer cases to
8 analyze than the NIOSH studies, did not have individual exposure estimates, and relied on
9 external comparisons. The Steenland et al. (2003, 2004) studies, on the other hand, used the
10 largest cohort of women potentially exposed to EtO and clearly show significantly increased
11 risks of breast cancer incidence and mortality based upon internal exposure-response analyses.
12 In summary, the most compelling evidence of a cancer risk from exposure to EtO is for
13 cancer of the lymphohematopoietic system. Increases in the risk of lymphohematopoietic cancer
14 are present in most of the studies, manifested as an increase in either leukemia and/or cancer of
15 the lymphoid tissue. The evidence of lymphohematopoietic cancer is strongest in the one study
16 (the NIOSH study) that appears to possess the fewest limitations. In this large study, a
17 significant dose-response relationship was evident with cumulative exposure to EtO. However,
18 this effect was observed only in males and the magnitude of the effect was not large. Similarly,
19 in most of the other studies, the increased risks are not great, and other chemicals in some of the
20 workplaces cannot be ruled out as possible confounders. Thus, the findings of increased risks of
21 lymphohematopoietic cancer in the NIOSH and other studies cannot conclusively be attributed to
22 exposure to EtO. The few studies that fail to demonstrate any increased risks of cancer do not
23 have those strengths of study design that give confidence to the reported lack of an exposure-
24 related effect.
25 There is also evidence of an elevated risk of breast cancer from exposure to EtO in a few
26 studies. The strongest evidence again comes from the NIOSH studies, which found positive
27 exposure-response relationships for both breast cancer incidence and mortality. Hopefully,
28 future studies will shed more light on this recent finding.
29
30 3.2. EVIDENCE OF CANCER IN EXPERIMENTAL ANIMALS
31 The International Agency for Research on Cancer (IARC) monograph (IARC, 1994) has
32 summarized the rodent studies of carcinogenicity, and Health Canada (2001) has used this
33 information to derive the levels of concern for human exposure. EPA concludes that the IARC
34 summary of the key studies is valid and is not aware of any animal cancer bioassays that have
35 been published since 1994. The Ethylene Oxide Industry Council (EOIC) (EOIC, 2001) also
36 reviewed the same studies and did not cite additional studies. The qualitative results are
08/29/06 12 DRAFT—DO NOT CITE OR QUOTE
-------
1 described here and the incidence data are tabulated in the unit risk derivation section of this
2 document.
3 One study of oral administration in rats has been published; there are no oral studies in
4 mice. Dunkelberg (1982) administered EtO in vegetable oil to groups of 50 female Sprague-
5 Dawley rats by gastric intubation twice weekly for 150 weeks. There were two control groups
6 (untreated and oil gavage) and two treated groups (7.5 and 30 mg/kg-day). A dose-dependent
7 increase in the incidence of malignant tumors in the forestomach was observed in the treated
8 groups (8/50 and 31/50 in the low- and high-dose groups, respectively). Of the 39 tumors, 37
9 were squamous cell carcinomas, and metastases to other organs were common in these animals.
10 This study was not evaluated quantitatively because oral risk estimates are beyond the scope of
11 this document.
12 One inhalation assay was reported in mice (NTP, 1987) and two inhalation assays were
13 reported in rats (Lynch et al, 1982, 1984, in males; Snellings et al, 1984; Garman et al, 1985,
14 1986, in both males and females). In the National Toxicology Program (NTP) mouse bioassay
15 (NTP, 1987), groups of 50 male and 50 female B6C3Fj mice were exposed to EtO via inhalation
16 at concentrations of 0, 50, and 100 ppm for 6 hours per day, 5 days per week, for 102 weeks.
17 Mean body weights were similar for treated and control animals, and there was no decrease in
18 survival associated with treatment. A concentration-dependent increase in the incidence of
19 tumors at several sites was induced in both sexes. These data are summarized in Table 1. Males
20 had carcinomas and adenomas in the lung. Females had carcinomas and adenomas in the lung,
21 malignant lymphomas, adenocarcinomas in the uterus, and adenocarcinomas in the mammary
22 glands. The NTP also reports that both sexes had dose-related increased incidences of
23 cystadenomas of the Harderian glands, but these are benign lesions and are not considered further
24 here.
25 In the Lynch et al. (1982, 1984) bioassay in male Fischer 344 (F344) rats, groups of 80
26 animals were exposed to EtO via inhalation at concentrations of 0, 50, and 100 ppm for 7 hours
27 per day, 5 days per week, for 2 years. Mean body weights were statistically significantly
28 decreased in both treated groups compared with controls (p<0.05). Increased mortality was
29 observed in the treated groups, and the increase was statistically significant in the 100-ppm
30 exposure group (p<0.01). Lynch et al. (1984) suggest that survival was affected by a pulmonary
31 infection alone and in combination with EtO exposure. Concentration-dependent increases in the
32 incidence of mononuclear cell leukemia in the spleen, peritoneal mesothelioma in the testes, and
33 glioma in the brain were observed (Table 2). The fact that the increased incidence of
34 mononuclear cell leukemia was statistically significant in the low-exposure group but not in the
35 high-exposure group is probably attributable to the increased mortality in the high-exposure
08/29/06 13 DRAFT—DO NOT CITE OR QUOTE
-------
1 group. The increased incidence in just the terminal kill rats in the 100-ppm group was
2 statistically significant compared with controls.
3 In the bioassay conducted by Snellings et al. (1984), 120 male and 120 female F344 rats
4 in each sex and dose group were exposed to EtO via inhalation at concentrations of 0 (2 control
5 groups of 120 rats of each sex were used), 10, 33, and 100 ppm for 6 hours per day, 5 days per
6 week, for 2 years, with some scheduled kills at 6 (10 rats per group), 12 (10 rats per group), and
7 18 (20 rats per group) months. Significant decreases in mean body weight were observed in the
8 100-ppm exposure group in males and in the 100-ppm and 33-ppm exposure groups in females.
9 During the 15th month of exposure, an outbreak of viral sialodacryoadenitis occurred, resulting
10 in the deaths of 1-5 animals per group. Snellings et al. claim that it is unlikely that the viral
11 outbreak contributed to the EtO-associated tumor findings. After the outbreak, mortality rates
12 returned to pre-outbreak levels and were similar for all groups until the 20th or 21st month, when
13 cumulative mortality in the 33-ppm and 100-ppm exposure groups of each sex remained above
14 control values. By the 22nd or 23rd months, mortality was statistically significantly increased in
15 the 100-ppm exposure groups of both sexes.
16 In males, concentration-dependent increases in the incidence of mononuclear cell
17 leukemia in the spleen and peritoneal mesothelioma in the testes were observed, and in females
18 an increase in mononuclear cell leukemia in the spleen was seen. These data are summarized in
19 Table 3. Note that these investigators observed the same types of tumors (splenic leukemia and
20 peritoneal mesothelioma) seen by Lynch et al. (1982, 1984). Snellings et al. (1984) only report
21 incidences (of incidental and nonincidental primary tumors for all exposure groups) for the 24-
22 month (terminal) kill. However, in their paper they state that significant findings for the
23 mononuclear cell leukemias were also obtained when all rats were included and that a mortality-
24 adjusted trend analysis yielded positive findings for the EtO-exposed females (p<0.005) and
25 males (p<0.05). Similarly, Snellings et al. report that when male rats with unscheduled deaths
26 were included in the analysis of peritoneal mesotheliomas, it appeared that EtO exposure was
27 associated with earlier tumor occurrence, and a mortality-adjusted trend analysis yielded a
28 significant positive trend (p<0.005). In later publications describing brain tumors (Garman et al.,
29 1985, 1986), both males and females had a concentration-dependent increased incidence of brain
30 tumors (see Table 3). Garman et al. report incidences including all rats from the 18- and 24-
31 month kills and found dead or killed moribund. The earliest brain tumors were observed in rats
32 killed at 18 months.
33
34 3.2.1. Conclusions Regarding the Evidence of Cancer in Experimental Animals
35 In summary, there is strong and sufficient evidence that EtO causes cancer in
36 experimental animals. After inhalation exposure to EtO, statistically significant increased
08/29/06 14 DRAFT—DO NOT CITE OR QUOTE
-------
1 incidences of cancer have been observed in both rats and mice, in both males and females, and in
2 multiple tissues of both epithelial origin (lung, mammary gland, uterus) and mesothelial origin
3 (lymph tissue, blood, brain, tunica vaginalis testis). In addition, one oral study in rats has been
4 conducted, and a significant dose-dependent increase in carcinomas of the forestomach was
5 reported.
6
7 3.3. SUPPORTING EVIDENCE
8 3.3.1. Metabolism and Kinetics
9 Information on the kinetics and metabolism of EtO has been derived primarily from
10 studies conducted with laboratory animals exposed via inhalation, although some limited data
11 from humans have been identified. Details are available in several reviews (Brown et al., 1996,
12 1998; Csanady et al., 2000; Fennell and Brown, 2001).
13 Following inhalation, EtO is absorbed efficiently into the blood and rapidly distributed to
14 all organs and tissues. EtO is metabolized primarily by two pathways (see Figure 1): (1)
15 hydrolysis to ethylene glycol (1,2-ethanediol), with subsequent conversion to oxalic acid, formic
16 acid, and carbon dioxide; and (2) glutathione conjugation and the formation of S-(2-
17 hydroxyethyl)cysteine and N-acetylated derivatives (WHO, 2003). From the available data, the
18 route involving conjugation with glutathione appears to predominate in mice; in larger species
19 (including humans), the conversion of EtO is primarily via hydrolysis through ethylene glycol.
20 Because EtO is an epoxide capable of reacting directly with cellular macromolecules, both
21 pathways are considered to be detoxifying.
22 Among rodent species, there are clear quantitative differences in metabolic rates. The
23 rate of clearance of EtO from the blood, brain, muscle, and testes was measured by Brown et al.
24 (1996, 1998). Clearance rates were nearly identical across blood and other tissues. Following a
25 4-hour inhalation exposure to 100 ppm EtO in mice and rats, the average blood elimination half-
26 lives ranged from 2.4 to 3.2 minutes in mice and 11 to 14 minutes in rats. The elimination half-
27 life in humans is 42 minutes (Filser et al., 1992), and the half-life in salt water is 4 days (IARC,
28 1994).
29 In a more detailed study in mice, Brown et al. (1998) measured EtO concentrations in
30 mice after 4-hour inhalation exposures at 0, 50, 100, 200, 300, or 400 ppm. They found that
31 blood EtO concentration increased linearly with inhaled concentrations of less than 200 ppm, but
32 above 200 ppm the blood concentration increased more rapidly than linearly. In addition,
33 glutathione levels in liver, lung, kidney, and testes decreased as exposures increased above 200
34 ppm. The investigators interpreted this, along with other information, to mean that at low
35 concentrations glutathione conjugation is responsible for the metabolism and disappearance of
36 EtO, but at higher concentrations, when tissue glutathione begins to be depleted, the elimination
08/29/06 15 DRAFT—DO NOT CITE OR QUOTE
-------
1 occurs via a slower non-enzymatic hydrolysis process, leading to a greater-than-linear increase
2 in blood EtO concentration.
3 Fennell and Brown (2001) constructed physiologically based pharmacokinetic (PBPK)
4 models of uptake and metabolism in mice, rats, and humans, based on previous studies. They
5 reported that the models adequately predicted blood and tissue EtO concentrations in rats and
6 mice, with the exception of the testes, and blood EtO concentrations in humans. Modeling 6-
7 hour inhalation exposures yielded simulated blood peak concentrations and areas under the curve
8 (AUCs) that are similar for mice, rats, and humans (human levels are within about 15% of rat
9 and mouse levels; see Figure 2). In other words, exposure to a given EtO concentration in air
10 results in similar predicted blood EtO AUCs for mice, rats, and humans.
11 These studies show that tissue concentrations in mice, rats, and humans exposed to a
12 particular air concentration of EtO are approximately equal and that they are linearly related to
13 inhalation concentration, at least in the range of exposures used in the rodent cancer bioassays
14 (i.e., 100 ppm and below).
15 EtO forms DNA and hemoglobin adducts with tissues throughout the body (Walker,
16 1992a, b). In experiments with rats and mice exposed to EtO at concentrations of 0, 3, 10, 33,
17 100, or 300 (rats only) ppm for 6 hours per day, 5 days per week, for 4 weeks, Walker et al.
18 (1992b) measured 7-(2-hydroxyethyl)guanine (HEG) in the DNA of lung, brain, kidney, spleen,
19 liver, and testes. At 100 ppm, the adduct levels for all tissues except testis were similar (within a
20 factor of 3), despite the fact that not all of these tissues are targets for toxicity. The study's data
21 on the persistence of the DNA adducts indicate that DNA repair rates differ in different tissues.
22 In a companion paper, Walker et al. (1992a) reported measurements of hemoglobin
23 adducts and showed how the concentration of these adducts changes according to the dynamics
24 of red blood cell turnover. Formation of hemoglobin adducts has been used as a measure of
25 exposure to EtO. Walker et al. (1992a) measured hemoglobin adduct formation in mice and rats
26 exposed to 0, 3, 10, 33, 100, and 300 (rats only) ppm of EtO (6 h/day, 5 days/wk, for 4 weeks).
27 Response was linear in both species up to 33 ppm, after which the slope significantly increased.
28 The dose-related decrease in glutathione concentration in liver reported by Brown et al. (1998) is
29 a plausible explanation for the higher incidence of hemoglobin adducts.
30 In humans, hemoglobin adducts can be used as biomarkers of recent exposure to EtO
31 (IARC, 1994), and several studies have reported exposure-response relationships between
32 hemoglobin adduct levels and EtO exposure levels (e.g., Schulte et al., 1992; Van Sittert et al.,
33 1993). Hemoglobin adducts are good indicators of exposure because they are stable (DNA
34 adducts, on the other hand, may be repaired or fixed as mutations and hence cannot be used as
35 reliable measures of exposure). However, because levels of 7V-(2-hydroxyethyl)valine (HEVal)
36 were approximately twofold greater in persons with a null GSTT1 genotype than in those with
08/29/06 16 DRAFT—DO NOT CITE OR QUOTE
-------
1 positive genotypes (Yong et al., 2001), adjustments for genotype would be necessary for accuracy
2 in inferring recent exposure from hemoglobin adduct data.
3 Alkylating agents may induce a dozen different DNA alkylation products (Beranek, 1990)
4 with varying proportions, depending primarily on the electrophilic properties of the agent. The
5 predominant DNA adduct formed by EtO and other SN2-type alkylating agents is HEG. In a
6 study in rats, Zhao et al. (1997) reported three main adducts, HEG, 3-hydroxyethyladenine, and
7 O-6 hydroxyethylguanine, in the ratios 200:8.8:1. In DNA extracted from the lymphocytes of
8 unexposed individuals, mean background levels of HEG ranged from 2 to 8.5 pmol/mg DNA
9 (Bolt, 1996). Because EtO is formed during the metabolism of ethylene, a natural body
10 constituent, endogenous as well as exogenous sources of ethylene and EtO contribute to
11 background alkylation of proteins such as hemoglobin and albumin as well as DNA (Bolt, 1996).
12
13 3.3.2. Mutagenicity
14 Since the first report of EtO induction of sex-linked recessive lethals in drosophila
15 (Rapoport, 1948), numerous papers have been published on the positive mutagenic activity in
16 biological systems, spanning the whole range of assay systems, from bacteriophage to higher
17 plants and animals. Figure 3 shows the 203 test entries in the EPA Genetic Activity Profile
18 (GAP) database. In prokaryotes and lower eukaryotes, EtO induced DNA damage and gene
19 mutations in bacteria, yeast, and fungi and gene conversions in yeast. In mammalian cells, EtO-
20 induced effects include unscheduled DNA synthesis, gene mutations, sister chromatid exchanges
21 (SCEs), micronuclei, and chromosomal aberrations. Several publications contain details of
22 earlier genetic toxicity studies (Thier and Bolt, 2000; Natarajan et al., 1995; Preston et al., 1995;
23 Dellarco et al., 1990; Walker et al., 1990; Ehrenberg and Hussain, 1981). This review focuses on
24 recently published studies that provide information on the mode of action of EtO.
25
26 3.3.2.1. Mutations
27 As a direct-acting alkylating agent, EtO has invariably yielded positive results in in vitro
28 mutation assays from bacteriophage, bacteria, fungi, yeast, insects, plants, and mammalian cell
29 cultures (including human cells). The results of in vivo studies on the genotoxicity of EtO have
30 also been consistently positive following ingestion, inhalation, or injection (Tates et al., 1999).
31 Increases in the frequency of gene mutations in the lung (lad locus) (Sisk et al., 1997), in T-
32 lymphocytes (hprt locus) (Walker et al., 1997), and bone marrow and testes (Recio et al., 2004)
33 have been observed in transgenic mice exposed to EtO via inhalation at concentrations similar to
34 those in carcinogenesis bioassays with this species (NTP, 1987).
08/29/06 17 DRAFT—DO NOT CITE OR QUOTE
-------
1 In male Big Blue (lad transgenic) B6C3Fj mice exposed to 0, 50, 100, or 200 ppm (0, 92,
2 183, or 366 mg/m3) EtO for 6 hours per day, 5 days per week, for 4 weeks, the observed mean
3 frequency of mutation at the hprt locus in splenic T-lymphocytes was 2.2, 3.8, 6.8, and 14.1 x
4 10"6, respectively (Sisk et al., 1997). The frequency of hprt mutations in splenic T-lymphocytes
5 was increased (compared with unexposed controls) 5- to 5.6-fold in male F344 rats as well as in
6 (nontransgenic) male B6C3Fj mice exposed to 200 ppm (366 mg/m3) EtO for 6 hours per day, 5
7 days per week, for 4 weeks (Walker et al., 1997). Similarly, the frequency of lad mutations in
8 the lungs, bone marrow, and spleen was increased in male Big Blue (lad transgenic) B6C3Fj
9 mice exposed to 0 or 200 ppm (0 or 366 mg/m3) EtO (Recio et al., 1999; Sisk et al., 1997).
10 In a later study by Recio et al. (2004), male Big Blue (lad transgenic) B6C3Fj mice were
11 exposed to 0, 25, 50, 100, or 200 ppm EtO (6 hours per day, 5 days per week) for 12, 24, and 48
12 weeks (Recio et al., 2004). Clear mutagenic response in the bone marrow was observed only
13 after 48 weeks, with lad mutant frequencies of 7.3, 11.3, 9.3, 14.1, and 30.3 x 10'5. Mutant
14 frequencies from testes (seminiferous tubules) were significantly greater than in controls at 25,
15 50, and 100 ppm (48-week exposure). The lad mutant frequency after 48 weeks of 200 ppm
16 EtO exposure was not different from that of controls. The authors suggested that this was
17 probably due to testicular toxicity. Mutation spectrum analysis of induced mutations in bone
18 marrow indicated a decrease in mutations at G:C base pairs and an increase at A:T base pairs,
19 exclusively in A:T^T:A transversions. The mutation spectrum in EtO-induced mutations from
20 testes was similar to the spectrum from untreated animals. The authors suggested that the
21 difference in mutation spectrum between the two tissues may be due to differential repair of
22 DNA adducts.
23 In a study of workers in an EtO production facility (Tates et al., 1995), hprt mutations
24 were measured in three exposed groups and one unexposed group (seven workers per group). No
25 significant differences in mutant frequencies were observed between the groups; however, the
26 authors stated that about 50 subjects per group would have been needed to detect a 50% increase.
27 Major et al. (2001) measured hprt mutations in female nurses employed in hospitals in
28 Eger and Budapest, Hungary. This study and an earlier study measuring effects on chromosomes
29 (see Table 4) were conducted to examine a possible causal relationship between EtO exposure
30 and a cluster of cancers (mostly breast) in nurses exposed to EtO in the Eger hospital. The
31 Budapest hospital was chosen because there was no apparent increase in cancer among nurses
32 exposed to EtO. Controls were female hospital workers in the respective cities, and nurses in
33 Eger with known cancers were excluded. Mean peak levels of EtO were 5 mg/m3 (2.7 ppm) in
34 Budapest and 10 mg/m3 (5.4 ppm) in Eger. Hprt variant frequencies in both controls and EtO-
35 exposed workers in the Eger hospital were higher than either group in the Budapest hospital, but
36 there was no significant increase among the EtO-exposed workers in either hospital when
08/29/06 18 DRAFT—DO NOT CITE OR QUOTE
-------
1 compared with the respective controls. The authors noted that the hprt variant frequencies
2 among smoking EtO-exposed nurses in Eger were significantly higher than among smokers in
3 the Eger controls; however, the fact that the hprt variant frequency was almost three times higher
4 in nonsmokers than in smokers in the Eger hospital control group raises questions about the basis
5 of the claimed EtO effect.
6
7 3.3.2.2. Sister Chromatid Exchanges
8 The genotoxicity of EtO was demonstrated in humans as early as 1979. Table 4
9 summarizes the cytogenetic effects of EtO on human exposures. Garry et al. (1979) analyzed
10 SCEs in lymphocytes cultured from exposed individuals as well as comparable controls.
11 Significant increase in SCE was observed at three weeks and at eight weeks following exposure
12 to EtO. Although this study does not describe the exact exposure estimates, EtO was recognized
13 as a mutagenic or genotoxic agent.
14 SCE frequency in workers exposed to high levels of EtO in a hospital sterilization service
15 was studied by Laurent et al. (1984). Blood samples were obtained retrospectively from a group
16 of 25 subjects exposed to high levels of EtO for a period of two years. A significant increase in
17 SCE rate was observed in the exposed group when compared with the control group. The
18 authors concluded that the effect of exposure to EtO was sufficient to produce a cumulative—
19 and in some cases a persistent—genetic change.
20 Peripheral blood lymphocytes of nurses exposed to low and high doses of EtO were
21 studied by Major et al. (1996). SCEs were slightly elevated in the low-dose exposure group but
22 were significantly increased in the high-dose exposure group. Several studies by Sarto et al.
23 (1984, 1987, 1990, 1991) showed significant increases in chromosomal aberrations, SCEs, and
24 micronucleus formation (Table 4).
25
26 3.3.2.3. Chromosomal Aberrations
27 Clare et al. (1985) conducted chromosomal analysis on lymphocytes from 33 workers
28 employed in the manufacture of EtO. A slightly higher frequency of chromatid aberrations was
29 observed in the cells of the EtO workers than in those of controls. Further, a positive correlation
30 between length of employment in the EtO group and the number of aberrations was observed
31 (Table 4). Galloway et al. (1986) studied chromosomal aberration frequencies in 61 employees
32 potentially exposed to EtO. Three work sites with different historical ambient levels of EtO were
33 chosen for study. Blood samples were drawn several times over a 24-month period and
34 aberrations were analyzed in 100 cells per sample after culture for 48-5 Ihours. At work sites I
35 and II, no consistent differences in aberration frequencies were found. However, at work site III,
36 aberration frequencies in potentially exposed individuals were significantly increased when
08/29/06 19 DRAFT—DO NOT CITE OR QUOTE
-------
1 compared with controls. A previous study by the same group (Stolley et al., 1984) showed an
2 association between SCE frequency and EtO exposure. When the aberrations were compared
3 with the levels of SCEs, the authors found a weak overall association (Table 4).
4 Further, Lerda and Rizzi (1992) showed a significant increase in chromosomal aberration
5 frequencies in EtO-exposed individuals when compared with controls. Major et al. (1996)
6 studied hospital nurses exposed to low doses and high doses of EtO for changes in structural and
7 numerical chromosomal aberrations. Chromosomal aberrations were found to be significantly
8 elevated in both the low-dose and the high-dose exposure groups. Deletions—and to a lesser
9 extent, chromatid exchanges and dicentrics—were detected in the low-dose exposure group;
10 however, in the high-dose group, in addition to the increased number of deletions, the
11 frequencies of dicentrics and rings showed a significant excess when compared with controls.
12 The authors acknowledged that an unexpected, significant increase in dicentrics and ring
13 frequencies was detected among the controls. When analyzed for confounding factors, a possible
14 active confounding factor was natural radioactivity from the local tap water. Several other
15 studies by Sarto et al. (1984, 1987, 1990, 1991) showed significant increase in chromosomal
16 aberrations after exposure to EtO. In a study of 28 EtO-exposed sterilizer workers and 20
17 unexposed controls, Hogstedt et al. (1983) reported a statistically significant increase in
18 micronuclei, but not chromosomal breaks or gaps, in bone marrow cells (erythroblasts and
19 polychromatic erythrocytes) in the exposed workers, adjusted for age, smoking, drug intake, and
20 exposure to ionizing radiation.
21 The above data clearly indicate that EtO is a genotoxic agent that causes a variety of types
22 of genetic damage.
23 As discussed by Preston (1999), a variety of cytogenetic assays can be used to measure
24 induced chromosome damage. However, most of the assays commonly employed measure
25 events that are detectable only in the first (or in some cases the second) metaphase after exposure
26 and require DNA synthesis to convert DNA damage into a chromosomal aberration. In addition,
27 DNA repair is operating in peripheral lymphocytes to repair induced DNA damage. The events
28 measured include all types of chromosomal aberrations, micronuclei, SCE, and numerical
29 chromosomal changes. Thus, for acute exposures, the timing of sampling is of great importance.
30 For chronic studies, the endpoints measure only the most recent exposures, and if the time
31 between last exposure and sampling is long, any induced DNA damage not converted to a stable
32 genotoxic alteration is certain to be missed. Stable chromosomal aberrations include reciprocal
33 translocations, inversions, and some fraction of insertions and deletions as well as some
34 numerical changes. However, until the development of fluorescent in situ hybridization (FISH),
35 chromosome banding techniques were needed to detect these types of aberrations.
08/29/06 20 DRAFT—DO NOT CITE OR QUOTE
-------
1 Preston (1999) has provided an exhaustive review of the cytogenetic effects of EtO.
2 Table 4 shows the details of the studies that were reviewed. In addition to summarizing the
3 available cytogenetic studies of exposed workers, Preston discussed the basic guidelines for
4 cytogenetic assays, noting especially that because most cytogenetic assays measure damage that
5 is either repaired rapidly or is nontransmissible, they are not valuable in chronic exposures in
6 either laboratory animals or humans. The situation in humans is further confounded because of
7 the dependence of chromosomal aberrations on age, smoking status, medical condition (including
8 radiation exposure), and other life style variables. Based on the review, Preston (1999)
9 concluded that high exposures to EtO can be detected as increases in unstable chromosomal
10 aberrations or SCE in peripheral lymphocytes but that chronic or low-level acute exposures are
11 not detectable using routine measures.
12 The overall available data from in vitro studies, animal models, and epidemiological
13 studies indicate that EtO is both a mutagen and a genotoxicant. It has been recognized that stable
14 translocations seen in human leukemias can arise from similar DNA adducts that produce
15 chromosome breaks, micronuclei, SCEs, and even gene mutations observed in peripheral
16 lymphocytes.
17
18 3.4. MODE OF ACTION
19 EtO is an alkylating agent that has consistently been found to produce numerous
20 genotoxic effects in a variety of biological systems ranging from bacteriophage to occupationally
21 exposed humans. It is carcinogenic in mice and rats, inducing tumors of the
22 lymphohematopoietic system, brain, lung, connective tissues, uterus, and mammary gland. In
23 addition, epidemiological studies have shown an increased risk of various types of human
24 cancers (Table A-5), in particular lymphohematopoietic and breast cancers. Target tissues for
25 EtO carcinogenicity in laboratory animals are both epithelial and mesothelial in origin and are
26 not clearly attributable to any specific type of genetic alteration. The lymphomas in mice
27 induced by EtO are considered to be generally similar in characteristics to leukemias in humans
28 (U.S. EPA, 1997). Although the precise mechanism(s) by which the multi-site carcinogenicity in
29 mice, rats, and humans occurs cannot be well established, currently available information
30 indicates that genotoxicity plays an important role in EtO-induced carcinogenicity.
31 Exposure of cells to DNA-reactive agents results in the formation of carcinogen-DNA
32 adducts. The formation of DNA adducts results from a sequence of events involving absorption
33 of the agent, distribution to different tissues, and accessibility of the molecular target (Swenberg
34 et al., 1990). Alkylating agents may induce several different DNA alkylation products (Beranek,
35 1990) with varying proportions, depending primarily on the electrophilic properties of the agent.
36 The predominant DNA adduct formed by EtO is HEG. Zhao et al. (1997) reported three main
08/29/06 21 DRAFT—DO NOT CITE OR QUOTE
-------
1 adducts in rat liver and lymphocytes: HEG, 3-hydroxyethyladenine, and O-6
2 hydroxyethylguanine in the ratios 200:8.8:1. The various adducts are processed by different
3 repair pathways, and the subsequent genotoxic response elicited by unrepaired DNA adducts is
4 dependent on a wide range of variables. HEG adducts result in various types of cytogenetic
5 damage, including gene mutations, which have been observed in exposed mice and rats. The
6 predominant fraction of mutations detected by sequencing were base pair changes that involved
7 adenine as well as guanine adducts (Walker and Skopek, 1993; Walker et al., 1999; Recio et al.,
8 2004). Further, DNA adducts, SCEs, and hprt mutations resulting from EtO exposure in rats
9 have shown statistically significant linear dose-response functions (van Sittert et al., 2000).
10 The events involved in the formation of chromosome damage from HEG are complex.
11 N-alklylated bases are removed from DNA by base excision repair pathways. A review by
12 Memisoglu and Samson (2000) notes that the action of DNA glycosylase and apurinic
13 endonuclease creates a DNA single-strand break, which can in turn lead to DNA double-strand
14 breaks (DSBs). DSBs can also be produced by normal cellular functions, such as during V(D)J
15 recombination in the development of lymphoid cells or topoisomerase II-mediated cleavage at
16 defined sites. A recent review of mechanisms of DSB repair indicates that the molecular
17 mechanisms are not fully understood (Pfeiffer et al., 2000). This review provides a thorough
18 discussion of both sources (endogenous and exogenous) of DSBs and the variety of repair
19 pathways that have evolved to process the breaks. Although homology-directed repair generally
20 restores the original sequence, during nonhomologous end-joining (NHEJ), the ends of the
21 breaks are frequently modified by addition or deletion of nucleotides.
22 Leukemias, like all other cancers, are believed to be a consequence of an accumulation of
23 genetic and epigenetic changes involving multiple genes and chromosomal alterations. Although
24 it is clear that chromosome translocations are common features of hematopoietic cancers, there is
25 evidence that mutations in p53 or N-ras are involved in some types of leukemia (U.S. EPA,
26 1997). It should also be noted that therapy-related leukemias exhibiting reciprocal translocations
27 are generally seen in patients who have previously been treated with chemotherapeutic agents
28 that act as topoisomerase II inhibitors (U.S. EPA, 1997). In NHL, the BCL6 gene is frequently
29 activated by translocation with breakpoints within the gene (Chaganti et al., 1998) as well as by
30 mutations within the gene coding sequence (Losses and Levy, 2000). Preudhomme et al. (2000)
31 observed point mutations in the AML1 gene in 9 of 22 patients with the Mo type of acute
32 myeloid leukemia (AML), and Harada et al. (2003) identified AML1 point mutations in cases of
33 radiation-associated and therapy-related myelodysplastic syndrome/AML. In both reports, point
34 mutations within the coding sequence were found in patients with normal karyotypes as well as
35 some with translocations or other chromosomal abnormalities. Several models have been
36 developed to integrate these various types of genetic alterations. One recent model suggests that
08/29/06 22 DRAFT—DO NOT CITE OR QUOTE
-------
1 two general types of therapy-related AML can be subdivided into at least eight genetic pathways
2 that have different etiologies and different biologic characteristics (Pedersen-Bjergaard et al.,
3 2002).
4 A mode-of-action-motivated modeling approach based solely on chromosome
5 translocations has been proposed by Kirman et al. (2004). The authors suggested a nonlinear
6 dose-response for EtO and leukemia, based on a consideration that "chromosomal aberrations are
7 the characteristic initiating events in chemically induced acute leukemia and gene mutations are
8 not characteristic initiating events." They thus proposed that EtO must be responsible for two
9 nearly simultaneous DNA adducts, yielding a dose-squared relationship between EtO exposure
10 and leukemia risk. However, as discussed above, there is evidence that does not support the
11 assumption that chromosomal aberrations represent the sole initiating event. In fact, these
12 aberrations or translocations could be a downstream event resulting from genomic instability. In
13 addition, if two reactions with DNA resulting in chromosomal aberrations or translocations are
14 early-occurring events in some EtO-induced lymphohematopoietic cancers, it is not necessary
15 that both events be associated with EtO exposure (e.g., background error repair rates or exposure
16 to other alkylating agents may be the cause). Furthermore, EtO could also produce translocations
17 indirectly by forming DNA or protein adducts that affect the normally-occurring recombination
18 activities of lymphocytes or the repair of spontaneous double-strand breaks.
19 Of course, with chronic exposure, the observed stable translocations are the combined
20 effect of induced translocations and their persistence. Clearly, with respect to leukemia related to
21 EtO exposure, the accumulations of translocations maybe of mechanistic relevance. Dose-
22 response curves for translocation frequencies following protracted or chronic radiation exposure
23 have shown significant linear but not quadratic terms in experimental studies in mice (Sorensen
24 et al., 2000). Tucker et al. (1997) reported translocation frequencies linear with dose (the
25 quadratic coefficient was negative) among 81 radiation workers at the Sellafield Nuclear Facility.
26 In summary, EtO induces a variety of types of genetic damage. It directly interacts with
27 DNA, resulting in gene mutations and chromosome damage. Depending on a number of
28 variables, the predominant DNA adduct (HEG) (1) may be repaired, (2) may result in a base-pair
29 mutation during replication, or (3) may be converted to a DSB, which also may be repaired or
30 result in unstable (micronuclei) or stable (translocation) cytogenetic damage. All of the available
31 data are strongly supportive of a mutagenic mode of action involving gene mutations and
32 chromosomal aberrations (translocations, deletions, or inversions) that critically alter the function
33 of oncogenes or tumor suppressor genes. Although it is clear that chromosome translocations are
34 common features of many hematopoietic cancers, there is evidence that mutations in p53, AML1
35 or N-ras are involved in other leukemias. The current scientific consensus is that there is very
36 good correspondence between ability of an agent to cause mutations, as does EtO, and
08/29/06 23 DRAFT—DO NOT CITE OR QUOTE
-------
1 carcinogenicity. All of the above scientific evidence points to a mutagenic mode of action with
2 linear dose response functions, and therefore, application of linearity of the dose-response
3 relationship seems appropriate.
4
5 3.4.1. Analysis of the Mode of Action for Ethylene Oxide Carcinogenicity Under EPA's
6 Mode of Action Framework
7 In this section, the mode of action evidence for EtO carcinogenicity is analyzed under the
8 mode of action framework in EPA's 2005 Guidelines for Carcinogen Risk Assessment (U.S. EPA,
9 2005a, Section 2.4.3).
10 The hypothesis is that EtO carcinogenicity has a mutagenic mode of action. This
11 hypothesized mode of action is presumed to apply to all of the tumor types.
12 The key events in the hypothesized mutagenic mode of action are DNA adduct formation
13 by EtO, which is a direct-acting alkylating agent, and the resulting genetic damage, including the
14 formation of point mutations as well as translocations, which can also result in mutations at the
15 translocation sites. Mutagenicity is a well-established cause of carcinogenicity.
16
17 1. Is the hypothesized mode of action sufficiently supported in the test animals?
18
19 Numerous studies have demonstrated that EtO forms protein and DNA adducts, in mice
20 and rats (see Sections 3.3.1 and 3.4 and Figure 2). For example, Walker et al. (1992a, b)
21 demonstrated that EtO forms protein adducts with hemoglobin in the blood and DNA adducts
22 with tissues throughout the body, including in the lung, brain, kidney, spleen, liver, and testes.
23 In addition, there is incontrovertible evidence that EtO is mutagenic (see Section 3.2.2).
24 The evidence is strong and consistent; EtO has invariably yielded positive results in in vitro
25 mutation assays from bacteriophage, bacteria, fungi, yeast, insects, plants, and mammalian cell
26 cultures. The results of in vivo studies on the mutagenicity and genotoxicity of EtO have also
27 been consistently positive following ingestion, inhalation, or injection. Increases in the frequency
28 of gene mutations in the lung, in T-lymphocytes, in bone marrow, and in testes have been
29 observed in transgenic mice exposed to EtO via inhalation at concentrations similar to those in the
30 mouse carcinogenesis bioassays.
31 Ethylene oxide induces a variety of mutagenic and genotoxic effects, including
32 chromosome breaks, micronuclei, SCEs, and gene mutations; however, the more general effect of
33 mutagenicity/genotoxicity is specific and occurs in the absence of cytotoxicity or other overt
34 toxicity. A temporal relationship is also clearly evident, with adducts and mutagenicity observed
35 in subchronic assays.
08/29/06 24 DRAFT—DO NOT CITE OR QUOTE
-------
1 Dose-response relationships have been observed between EtO exposure in vivo and
2 hemoglobin adducts (e.g., Walker et al., 1992a), as well as DNA adducts, SCEs, and hprt
3 mutations (e.g., van Sittert et al., 2000) (see also Sections 3.3 and 3.4). A mutagenic mode of
4 action for EtO carcinogenicity also clearly comports with notions of biological plausibility and
5 coherence because EtO is a direct-acting alkylating agent. Such agents are generally capable of
6 forming DNA adducts, which in turn have the potential to cause genetic damage, including
7 mutations; and mutagenicity, in its turn, is a well-established cause of carcinogenicity. This chain
8 of key events is consistent with current understanding of the biology of cancer.
9 In addition to the clear evidence supporting a mutagenic mode of action in test animals,
10 we are not aware of any alternative or additional hypothesized modes of action for EtO
11 carcinogenicity.
12
13 2. Is the hypothesized mode of action relevant to humans?
14
15 The evidence discussed above demonstrates that EtO is a systemic mutagen in test
16 animals; thus, there is the presumption that it would also be a mutagen in humans. Moreover,
17 there is human evidence directly supporting a mutagenic mode of action for EtO carcinogenicity.
18 Several studies of humans have reported exposure-response relationships between hemoglobin
19 adduct levels and EtO exposure levels (e.g., Schulte et al., 1992; van Sittert et al., 1993; see
20 Section 3.3.1), demonstrating the ability of EtO to bind covalently in systemic human cells, as it
21 does in rodent cells.
22 In addition, EtO has yielded positive results in in vitro mutagenicity studies of human
23 cells (see Figure 2). There is also clear evidence from a number of human studies that EtO causes
24 chromosomal aberrations, SCEs, and micronucleus formation in peripheral blood lymphocytes
25 (see Sections 3.3.2.2 and 3.3.2.3 and Table 4). At least one study suggested an exposure-response
26 relationship for the formation of SCEs in peripheral blood lymphocytes (Major et al., 1996).
27 Another study reported a statistically significant increase in micronuclei in bone marrow cells in
28 EtO-exposed workers (Hogstedt et al., 1983).
29 Finally, there is strong evidence that EtO causes cancer in humans, including cancer types
30 observed in rodent studies (i.e., lymphohematopoietic cancers and breast cancer), providing
31 further weight to the relevance of the aforementioned events to the development of cancer in
32 humans (see Sections 3.1 and 3.5.1).
33
34 In conclusion, the weight of evidence supports a mutagenic mode of action for EtO
35 carcinogenicity.
36
08/29/06 25 DRAFT—DO NOT CITE OR QUOTE
-------
1 3. Which populations or lifestages can be particularly susceptible to the hypothesized mode of
2 action?
3
4 The mutagenic mode of action is considered relevant to all populations and lifestages.
5 According to EPA's Supplemental Guidance (U.S. EPA, 2005b), there may be increased
6 susceptibility to early-life exposures for carcinogens with a mutagenic mode of action. Therefore,
7 because the weight of evidence supports a mutagenic mode of action for EtO carcinogenicity, and
8 in the absence of chemical-specific data to evaluate differences in susceptibility, increased early-
9 life susceptibility should be assumed and, if there is early-life exposure, the age-dependent
10 adjustment factors (ADAFs) should be applied, in accordance with the Supplemental Guidance.
11 In addition, as discussed in Section 3.5.2, people with DNA repair deficiencies or genetic
12 polymorphisms conveying a decreased efficiency in detoxifying enzymes may have increased
13 susceptibility to EtO carcinogenicity.
14
15 3.5. HAZARD CHARACTERIZATION
16 3.5.1. Characterization of Cancer Hazard
17 In humans there is evidence that EtO exposure may be causally associated with
18 lymphohematopoietic cancer, but the evidence is not strong enough to be conclusive. The
19 strongest evidence comes from a high-quality study of a large NIOSH cohort. Of the seven
20 relevant Hill criteria for causality (Hill, 1965), temporality, coherence, and biological plausibility
21 are largely satisfied. There is evidence of consistency between studies with respect to cancer of the
22 lymphohematopoietic system as a whole. There is some evidence of a dose-response relationship
23 (biological gradient) in males but not in females. There is little strength in the magnitude of most
24 of the risk estimates.
25 Most of the relevant studies focus on examining risks of cancer associated with
26 subcategories of the lymphohematopoietic organ system. These cancers include leukemia and its
27 various forms (i.e., myeloid or lymphocytic) and also Hodgkin's disease, NHL, reticulosarcoma,
28 and myeloma. One study has focused on "lymphoid cancer," which is a combination of
29 lymphocytic leukemia, NHL, and myeloma. No other study has examined the risk of this
30 particular combination. In this study, risk of cancer of the lymphoid tissue was significantly
31 elevated in subgroups of the workforce likely to have received the highest exposures to EtO.
32 Elevated risks of other subcategories of the hematopoietic system—either singly or in
33 combination—have sometimes, but not always, appeared in other studies.
34 In most of these studies, when all the subcategories are combined, an enhanced risk of
35 cancer of the lymphohematopoietic system is evident, and in some studies, it is significant. Hence
36 there is some specificity with respect to the lymphohematopoietic system, but the specificity
08/29/06 26 DRAFT—DO NOT CITE OR QUOTE
-------
1 criterion is not expected to be satisfied by agents, such as EtO, that are not only widely distributed
2 in all tissues but are also directly acting chemicals.
3 There is also recent evidence of an increased breast cancer risk in females from exposure
4 to EtO. This evidence comes predominantly from high-quality studies of the large NIOSH cohort,
5 in which positive exposure-response relationships for both breast cancer incidence and mortality
6 were observed. The criteria of temporality, coherence, and biological plausibility are also
7 satisfied. On the other hand, the magnitudes of the risk were not large, and none of the other
8 studies had enough breast cancer cases to be very informative.
9 Stomach cancer was noted in the earlier Hogstedt studies but is not found in recent studies.
10 Pancreatic cancer was observed in some studies and not others, and some studies observed no
11 EtO-related cancer risks.
12 The experimental animal evidence for carcinogenicity is concluded to be "sufficient"
13 based on findings of tumors at multiple sites, by both oral and inhalation routes of exposure, and
14 in both sexes of both rats and mice. Tumor types resulting from inhalation exposure included
15 mononuclear cell leukemia in male and female rats and malignant lymphoma and mammary
16 carcinoma in female mice, suggesting some site concordance with the lymphohematopoietic and
17 breast cancers observed in humans, also exposed by inhalation.
18 The evidence of genotoxicity is very strong, and there is no doubt that EtO is mutagenic.
19 EtO is a direct-acting alkylating agent and has invariably tested positive in in vitro mutation
20 assays from bacteriophage, bacteria, fungi, yeast, insects, plants, and mammalian cell cultures
21 (including human cells). In mammalian cells, EtO-induced genotoxic effects include unscheduled
22 DNA synthesis, gene mutations, SCEs, micronuclei, and chromosomal aberrations. The results of
23 in vivo genotoxicity studies of EtO have also been consistently positive following ingestion,
24 inhalation, or injection. Increases in frequencies of gene mutations have been reported in the
25 lung, T-lymphocytes, bone marrow, and testes. Several studies of humans occupationally exposed
26 to EtO have reported increased levels of SCEs, chromosomal aberrations, and micronuclei in
27 lymphocytes of exposed workers, and one study has reported increased levels of micronuclei in
28 bone marrow cells (erythroblasts and polychromatic erythrocytes) in EtO-exposed workers. In
29 addition to the various genotoxic effects of EtO, it is well established that EtO induces
30 hemoglobin adducts in rodents and in humans.
31 In the framework of EPA's 2005 Guidelines for Carcinogen Risk Assessment (U.S. EPA,
32 2005a), the statement can be made that EtO is carcinogenic to humans because the evidence
33 satisfies the conditions required in the absence of conclusive epidemiologic evidence establishing
34 cause and effect: (1) there is evidence of cancer in humans associated with EtO exposure—
35 strong evidence for lymphohematopoietic cancers and some evidence for breast cancer in EtO-
36 exposed workers—although less than conclusive; (2) there is extensive evidence of EtO-induced
08/29/06 27 DRAFT—DO NOT CITE OR QUOTE
-------
1 carcinogenicity in laboratory animals, including lymphohematopoietic cancers in rats and mice
2 and mammary carcinomas in mice; (3) EtO is a direct-acting alkylating agent whose mutagenic
3 and genotoxic capabilities have been well established in a variety of experimental systems, and
4 the mode of carcinogenic action in animals involves the key event of DNA adduct formation,
5 which may result in mutation, chromosome breaks, or chromosome translocation; and (4)
6 chromosome damage, including chromosomal aberrations, SCEs, and micronuclei, has been
7 observed in human populations exposed to EtO. EtO is regarded as carcinogenic via the oral
8 route as well as by inhalation (where nearly all of the evidence has accumulated) because it causes
9 tumors in laboratory animals by both routes of exposure and the mode of action of EtO is
10 independent of the route of exposure.
11
12 3.5.2. Susceptible Lifestages and Subpopulations
13 There are no data on the relative susceptibility of children and the elderly when compared
14 with adult workers, in whom the evidence of hazard has been gathered, but because EtO does not
15 have to be metabolized before binding to DNA and proteins, the maturing of enzyme systems in
16 very young children is thought not to be a predominant factor in its hazard, at least for activation.
17 However, the immaturity of detoxifying enzymes in very young children may increase children's
18 susceptibility because they may clear EtO at a slower rate than adults. As discussed in Section
19 3.3.1, EtO is metabolized (i.e., detoxified) primarily by hydrolysis in humans but also by
20 glutathione conjugation. Both hydrolytic activity and glutathione-S-transferase activity apparently
21 develop after birth (Clewell et al., 2002); thus, very young children might have a decreased
22 capacity to detoxify EtO compared to adults.
23 People with DNA repair deficiencies such as xeroderma pigmentosum, Bloom's
24 syndrome, Fanconi anemia, and ataxia telangiectasia (Gelehrter and Collins, 1990) are expected to
25 be especially sensitive to the damaging effects of EtO exposure. Paz-y-Mino et al. (2002) have
26 recently identified a specific polymorphism in the excision repair pathway gene hMSH2. The
27 polymorphism was present in 7.5% of normal individuals and in 22.7% of NHL patients,
28 suggesting that this polymorphism may be associated with an increased risk of developing NHL.
29 In addition, Yong et al. (2001) measured approximately twofold greater EtO-hemoglobin adduct
30 levels in occupationally exposed persons with a null GSTT1 genotype than in those with positive
31 genotypes.
32
33 3.5.3. Cancer Classification Conclusions of Other Agencies
34 Organizations other than EPA have come to similar conclusions regarding the
35 carcinogenicity of EtO. Health Canada (2001) concluded that "EtO is highly likely to be
36 carcinogenic in humans, based on consideration of cytogenetic changes in people with
08/29/06 28 DRAFT—DO NOT CITE OR QUOTE
-------
1 occupational exposures and overwhelming evidence of the biological plausibility of its
2 carcinogenicity, which is based on animal carcinogenesis bioassays and convincing evidence of
3 genotoxicity in all systems tested. This evidence adds to the less-than-convincing evidence of
4 carcinogenicity in human studies." The World Health Organization's Concise International
5 Chemical Assessment Document (CICAD) on EtO, which is based largely on Health Canada's
6 material, reached the same conclusion that EtO is "highly likely to be carcinogenic to humans"
7 (WHO, 2003). Similarly, in its ninth report on carcinogens, the National Toxicology Program
8 (NTP, 2000) has placed EtO in the category "known to be a human carcinogen." The EOIC
9 (2001), using the descriptors in EPA's 1999 proposed carcinogen risk assessment guidelines (U.S.
10 EPA, 1999), concluded that EtO is "likely to be carcinogenic" to humans, a classification that
11 implies more uncertainty than the highest-degree-of-evidence category of "carcinogenic" to
12 humans. IARC (1994) concluded that EtO is carcinogenic to humans and assigned it to its Group
13 1 category.
14
15
16 4. CANCER DOSE-RESPONSE RELATIONSHIP FOR INHALATION EXPOSURE
17
18 4.1. INHALATION UNIT RISK ESTIMATES DERIVED FROM HUMAN DATA
19 The NIOSH retrospective cohort study of more than 18,000 workers in 13 sterilizing
20 facilities (Steenland et al., 2003, 2004) provides the most appropriate data sets for deriving
21 quantitative cancer risk estimates in humans for several reasons: (1) exposure estimates were
22 derived for the individual workers, (2) the cohort was large and diverse (e.g., 55% female), and
23 (3) there was little reported exposure to chemicals other than EtO. The early exposures for which
24 no measurements were available were determined by consultations with plant industrial hygienists
25 and the use of regression modeling to estimate exposures to each individual as a function of
26 facility, exposure category, and time period. The investigators were then able to estimate the
27 cumulative exposure (ppm x days) for each individual worker by multiplying the estimated
28 exposure for each job (exposure category) held by the worker by the number of days spent in that
29 job and summing over all the jobs held by the worker. Steenland et al. (2004) present followup
30 results for the cohort mortality study previously discussed by Steenland et al. (1991) and Stayner
31 et al. (1993). Positive findings in the current followup include increased rates of
32 (lympho)hematopoietic cancer mortality in males and of breast cancer mortality in females.
33 Steenland et al. (2003) present results of a breast cancer incidence study of a subcohort of 7,576
34 women from the NIOSH cohort.
35 The other major occupational study (Teta et al., 1993) described risks in a single EtO-
36 manufacturing facility, but this study is less useful for estimating quantitative cancer risks for a
08/29/06 29 DRAFT—DO NOT CITE OR QUOTE
-------
1 number of reasons. Because of the more indirect method of determining exposures (inferring
2 exposure to the West Virginia cohort from other, possibly dissimilar, plants in Texas and
3 Minnesota), the resulting estimates of individual exposure are less reliable. Furthermore, this
4 cohort is of smaller size, and there is the possibility of co-exposure to other chemicals, such as
5 ethylene dichloride.
6 The derivation of unit risk estimates, defined as the lifetime risk of cancer from chronic
7 inhalation of EtO per unit of air concentration, for lymphohematopoietic cancer mortality and
8 incidence in males and for breast cancer mortality and incidence in females, based on results of
9 the recent analyses of the NIOSH cohort, is presented in the following subsections.
10
11 4.1.1. Risk Estimates for Lymphohematopoietic Cancer
12 4.1.1.1. Lymphohematopoietic Cancer Results From the NIOSH Study
13 Steenland et al. (2004) investigated the relationship between EtO exposure and mortality
14 from cancer at a number of sites using life-table analyses with the U.S. population as the
15 comparison population. Categorical analyses were also done by quartiles of cumulative exposure.
16 Then, to further investigate apparent exposure-response relationships observed for
17 (lympho)hematopoietic cancer and breast cancer, internal exposure-response analyses were
18 conducted using Cox proportional hazards models, which have the form
19
20 Relative rate (RR) = epx,
21
22 where P represents the regression coefficient and X is the exposure. A nested case-control
23 approach was used, with age as the time variable used to form the risk sets. Risk sets were
24 constructed with 100 controls randomly selected for each case from the pool of those surviving to
25 at least the age of the index case. According to the authors, use of 100 controls per case has been
26 shown to result in ORs virtually identical to the RR estimates obtained with full cohorts. Cases
27 and controls were matched on race (white/nonwhite), sex, and date of birth (within 5
28 years). Exposure was the only covariate in the model, so the/? value for the model also serves as a
29 p value for the regression coefficient, P, as well as for a test of exposure-response trend.
30 The exposure-response analyses focused on cumulative exposure and (natural) log
31 cumulative exposure, with various lag periods. A lag period defines an interval before death, or
32 end of followup, during which any exposure is disregarded because it is not considered relevant to
33 the outcome under investigation. One ppm x day was added to cumulative exposures in lagged
34 analyses to avoid taking the log of 0. Steenland et al. found that, for all lymphohematopoietic
35 cancers combined, positive exposure-response trends were seen only in males. Some of the
36 results for males are presented in Table 5. Steenland et al. also analyzed a subcategory of
08/29/06 30 DRAFT—DO NOT CITE OR QUOTE
-------
1 lymphohematopoietic cancers that they called "lymphoid" cancers; these included NHL,
2 myeloma, and lymphocytic leukemia. Positive trends were also observed for these cancers, again
3 concentrated in males, but the model fits were not notably better than for all lymphohematopoietic
4 cancers, and the "lymphoid" category did not include Hodgkin's lymphoma, which also exhibited
5 evidence of exposure-response trends, although based on few cases. In addition, misclassification
6 or nonclassification of tumor type is more likely to occur for subcategories of
7 lymphohematopoietic cancer (e.g., 4 of the 25 leukemias in the analyses were classified as "not
8 specified"). Finally, the exposure-response modeling results for lymphohematopoietic cancer are
9 based on more cases (37 deaths in males) than are the results for "lymphoid" cancer (n = 27).
10 Risk estimates based on "lymphoid" cancer results are presented for comparison. Although
11 myeloid and lymphoid cells develop from different progenitor cells, these progenitor cells derive
12 from the same pluripotent stem cells. Thus, without a clearer reason for subcategorizing the
13 lymphohematopoietic cancers and given the misclassification issues and the smaller numbers of
14 cases in the "lymphoid" category, risk estimates based on the more comprehensive results for all
15 lymphohematopoietic cancers are preferred in this assessment. Other EtO exposure metrics
16 (duration of exposure, average exposure, and peak exposure) were also examined, but models
17 using these metrics did not generally predict lymphohematopoietic cancer as well as models using
18 cumulative exposure. For additional details and discussion of the Steenland et al. (2004) study,
19 see Appendix A.
20
21 4.1.1.2. Prediction of Lifetime Extra Risk of Lymphohematopoietic Cancer Mortality
22 The results of internal exposure-response analyses of lymphohematopoietic cancer
23 presented by Steenland et al. (2004) and summarized in Table 5 were used for predicting the extra
24 risks of lymphohematopoietic cancer mortality in males from continuous environmental exposure
25 to EtO. Extra risk is defined as
26
Extra risk = (Rx-Ro)/(l-Ro),
1 where Rx is the lifetime risk in the exposed population and Ro is the lifetime risk in an unexposed
2 population (i.e., the background risk). These risk estimates were calculated using the P regression
3 coefficients and an actuarial program that accounts for competing causes of death.1
This program is an adaptation of the approach previously used by the Committee on the Biological Effects
of Ionizing Radiation (BEIR, 1988). The same methodology was also used in EPA's 1,3-butadiene health risk
assessment (U.S. EPA, 2002). A spreadsheet illustrating the extra risk calculation for the derivation of the LEC01 for
lymphohematopoietic cancer incidence in males (see Section 4.1.1.3) is presented in Appendix C.
08/29/06 31 DRAFT—DO NOT CITE OR QUOTE
-------
1 U.S. age-specific all-cause mortality rates for 1999 for males of all race groups combined
2 (NCHS, 2002) were used to specify the all-cause background mortality rates in the actuarial
3 program. The National Center for Health Statistics 1997-2001 cause-specific background
4 mortality rates for all lymphohematopoietic cancers in males were obtained from a Surveillance,
5 Epidemiology, and End Results (SEER) report (NCI, 2004a). The risks were computed up to age
6 85 for continuous exposures to EtO. Conversions between occupational EtO exposures and
7 continuous environmental exposures were made to account for differences in the number of days
8 exposed per year (240 vs. 365 days) and in the amount of air inhaled per day (10 vs. 20 m3; U.S.
9 EPA, 1994). An adjustment was also made for the lag period. The reported standard errors for
10 the regression coefficients from Table 5 and from the weighted linear regression calculation
11 described below were used to compute the 95% upper confidence limits (UCLs) for the relative
12 rates, based on a normal approximation.
13 The best-fitting model presented by Steenland et al. (2004) for all lymphohematopoietic
14 cancer mortality in males was for log cumulative exposure with a 15-year lag (/?=0.02). However,
15 using the log cumulative exposure model to estimate the risks from low environmental exposures
16 is problematic because this model, which is intended to fit the full range of occupational
17 exposures in the study, is highly supralinear (i.e., risk increases steeply with increasing exposures
18 in the low exposure range and then plateaus), and results are unstable for low exposures (i.e.,
19 small changes in exposure correspond to large changes in risk; see Figure 4). Consideration was
20 thus given to the cumulative exposure model (/?=0.12), which is typically used and which is stable
21 at low exposures. However, the Cox regression model with cumulative exposure is sub linear for
22 low exposures and does not reflect the apparent supralinearity of the data.
23 It was determined that the best way to reflect the data in the lower exposure region, which
24 is the region of interest for low-exposure extrapolation, was to do a weighted linear regression of
25 the results from the model with categorical cumulative exposure and a 15-year lag. In addition,
26 the highest exposure group was not included in the regression to alleviate some of the
27 "plateauing" in the exposure-response relationship at higher exposure levels and to provide a
28 better fit to the lower exposure data. (Linear modeling of categorical epidemiologic data and
29 elimination of the highest exposure group(s) to obtain a better fit of low-exposure data are both
30 standard techniques used in EPA risk assessments [U.S. EPA, 2005a].) The weights used for the
31 ORs were the inverses of the variances, which were calculated from the confidence intervals.
32 Mean and median exposures for the cumulative exposure groups were kindly provided by Dr.
33 Steenland (e-mail dated April 21, 2004, from Kyle Steenland, Emory University, to Jennifer Jinot,
08/29/06 32 DRAFT—DO NOT CITE OR QUOTE
-------
1 U.S. EPA).2 The mean values were used for the weighted linear regression because the
2 (arithmetic) mean exposures best represent the model's linear relationship between exposure and
3 cancer risk. If the median values had been used, a slightly larger regression coefficient would
4 have been obtained, resulting in slightly larger risk estimates.
5 Using this approach,3 a regression coefficient of 0.000347 per ppm x day (SE = 0.000251
6 per ppm x day) was obtained from the weighted linear regression of the categorical results. These
7 parameter estimates were used as the basis for the primary risk calculations for
8 lymphohematopoietic cancer in males. Risk estimates using the model results for the continuous
9 exposure variables (i.e., cumulative exposure and log cumulative exposure; see Table 5) are also
10 derived for comparison.
11 Point estimates and one-sided 95% UCLs for the extra risk of all lymphohematopoietic
12 cancer mortality in males associated with varying levels of environmental exposure to EtO based
13 on the Steenland et al. (2004) model results were determined by the actuarial program; the results
14 are presented in Table 6 (point estimates only are presented from the continuous variable
15 comparison models). The models based on cumulative exposure yield extra risk estimates that are
16 fairly linear for exposures below about 0.1 to 1 ppm but not above 1 ppm. The risk estimates
17 based on log cumulative exposure, on the other hand, are not (low-dose) linear even in the range
18 of 0.0001 to 0.001 ppm, the lowest exposure concentrations evaluated (i.e., not linear with respect
19 to low-exposure extrapolation to zero exposure).
20 Consistent with EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), the
21 same data and methodology were also used to estimate the exposure level (ECX; "effective
22 concentration") and the associated 95% lower confidence limit (LECX) corresponding to an extra
23 risk of 1% (x = 0.01). A 1% extra risk level is commonly used for the determination of the point
24 of departure (POD) for low-dose extrapolation from epidemiological data; higher extra risk levels,
25 such as 10%, would be an upward extrapolation for these data. Based on the actuarial program,
26 the risk ratio (i.e., Rx/Ro) for an extra risk of 1% for all lymphohematopoietic cancers is 1.5,
27 which better corresponds with the ORs reported by Steenland et al. (2004) than the higher risk
28 ratio of 5.9 that would be associated with a 10% extra risk. Thus, 1% extra risk was selected for
29 determination of the POD, and, consistent with the Guidelines for Carcinogen Risk Assessment,
30 the LEC value corresponding to that risk level was used as the actual POD. For the linear model
Mean exposures for males with a 15-year lag for the exposure categories in Table 5 were 442, 2,191,
7,105, and 60,269 ppm x days. Median values were 360, 2,093, 6,230, and 43,212 ppm x days. These values are
for the risk sets but should provide a good approximation to the full cohort values.
Equations for this weighted linear regression approach are presented in Rothman (1986), pp. 343-344.
08/29/06 33 DRAFT—DO NOT CITE OR QUOTE
-------
1 that was selected, the unit risk is independent of the benchmark risk level used to determine the
2 POD; however, selection of a benchmark risk level is useful for comparisons across models.
3 Because EtO is clearly mutagenic, a linear low-dose extrapolation was performed, also in
4 accordance with EPA's Guidelines for Carcinogen Risk Assessment. The EC01, LEC01, and
5 inhalation unit risk estimates for the different lymphohematopoietic cancer mortality models
6 examined are presented in Table 7. As discussed above, the unit risk estimate based on linear
7 regression of the Steenland et al. categorical results using cumulative exposure with a 15-year lag
8 (i.e., 0.917 per ppm) is the preferred estimate for environmental exposure levels. Estimates from
9 the continuous variable Cox regression models are presented for comparison only. As one can
10 see, the continuous variable cumulative exposure model, with its extreme sub linearity in the lower
11 exposure region, yields a substantially lower unit risk estimate (0.0146 per ppm), while the log
12 cumulative exposure model, with its extreme supralinearity in the lower exposure region, yields a
13 substantially higher unit risk estimate (80.6 per ppm). Converting the units, 0.917 per ppm
14 corresponds to a unit risk of 5.01 x 10"4 per |ig/m3 for lymphohematopoietic cancer mortality in
15 males.
16 As discussed in Section 4.1.1.1, risk estimates based on Steenland et al.'s (2004)
17 "lymphoid" cancer results are also derived, for comparison. The same methodology presented
18 above for the lymphohematopoietic cancer results was used for the "lymphoid" cancer risk
19 estimates. Age-specific background mortality rates for the relevant subcategories (NHL [C82-
20 C85 of 10th revision of ICD], multiple myeloma [C88,C90], and lymphoid leukemia [C91]) of
21 lymphohematopoietic cancer in males for the year 2001 were obtained from the NCHS Data
22 Warehouse website (http://www.cdc.gov/nchs/datawh/statab/unpubd/mortabs.htm). The
23 Steenland et al. Cox regression results for the cumulative exposure models for "lymphoid" cancer
24 in males are presented in Table 8. As for lymphohematopoietic cancer, the best-fitting model was
25 for log cumulative exposure with a 15-year lag (/?=0.02), and the cumulative exposure model did
26 not appear to fit the data well for the lower exposure groups (Figure 5). Thus, a linear regression
27 of the categorical results was again performed, dropping the highest exposure group, as described
28 above. The exposure categories, and corresponding average exposures, are the same as for
29 lymphohematopoietic cancer. The linear regression yielded a regression coefficient of 0.000279
30 per ppm x day (SE = 0.000269 per ppm x day).
31 The EC01, LEC01, and inhalation unit risk estimates for the different "lymphoid" cancer
32 mortality models examined are presented in Table 9. The results are similar to those based on
33 lymphohematopoietic cancer mortality (i.e., within 35%; see Table 7), especially for the two
34 continuous variable models.
35
36
08/29/06 34 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1.3. Prediction of Lifetime Extra Risk of Lymphohematopoietic Cancer Incidence
2 EPA cancer risk estimates are typically derived to represent an upper bound on increased
3 risk of cancer incidence, as from experimental animal incidence data. Cancer data from
4 epidemiologic studies are more generally mortality data, as is the case in the Steenland et al.
5 (2004) study. For tumor sites with low survival rates, mortality-based estimates are reasonable
6 approximations of cancer incidence risk; however, for many lymphohematopoietic cancers, the
7 survival rate is substantial, and incidence-based risks are preferred because EPA endeavors to
8 protect against cancer occurrence, not just mortality (U.S. EPA, 2005a).
9 Therefore, another calculation was done using the same regression coefficients presented
10 above (Section 4.1.1.2), but with age-specific male lymphohematopoietic cancer incidence rates
11 for 1996-2000 from SEER (NCI, 2003; Tables IX, XVIII, XVII, and XIII: all races) in place of
12 the lymphohematopoietic cancer mortality rates in the actuarial program. SEER collects good-
13 quality cancer incidence data from a variety of geographical areas in the United States. The
14 incidence data used here are from "SEER 9," a registry of nine states and cities covering about
15 10% of the U.S. population.
16 The incidence-based calculation assumes that lymphohematopoietic cancer incidence and
17 mortality have the same exposure-response relationship for the relative rate of effect from EtO
18 exposure and that the incidence data are for first occurrences of primary lymphohematopoietic
19 cancer or that relapses and secondary lymphohematopoietic cancers provide a negligible
20 contribution. (The latter assumption is probably sound; the former assumption is more potentially
21 problematic. Because various lymphohematopoietic subtypes with different survival rates are
22 included in the categorization of all lymphohematopoietic cancers, if the relative rates of the
23 subtypes differ and if the relative rate-weighted survival rates for all lymphohematopoietic cancers
24 are different from those for the combined subtypes, a bias could occur, resulting in either an
25 underestimation or overestimation of the extra risk for lymphohematopoietic cancer incidence.)
26 The incidence-based calculation also relies on the fact that the lymphohematopoietic cancer
27 incidence rates are small when compared with the all-cause mortality rates. The resulting EC01,
28 LEC01, and inhalation unit risk estimates for lymphohematopoietic cancer incidence in males from
29 the various models examined are presented in Table 7.
30 The unit risk estimates for cancer incidence range from about 65% (cumulative exposure
31 model) to 120% (log cumulative exposure model) higher than the corresponding mortality-based
32 estimates. The incidence estimate from the categorical results is about 80% higher than the
33 mortality-based estimate.
34 The preferred estimate for the unit risk for lymphohematopoietic cancer in males is the
35 estimate of 1.64 per ppm (8.99 x 10"4 per |ig/m3) derived, using incidence rates for the cause-
08/29/06 35 DRAFT—DO NOT CITE OR QUOTE
-------
1 specific background rates, from the categorical results of Steenland et al. 's (2004) internal
2 exposure-response modeling.
3 As discussed in Section 4.1.1.1, risk estimates based on Steenland et al.'s (2004)
4 "lymphoid" cancer results are also derived, for comparison. The same methodology presented
5 above for the lymphohematopoietic cancer results was used for the "lymphoid" cancer incidence
6 risk estimates. Age-specific SEER incidence rates for the relevant subcategories (NHL, myeloma,
7 and lymphocytic leukemia) of lymphohematopoietic cancer in males for the years 1997-2001 were
8 used (NCI, 2004a). The EC01, LEC01, and inhalation unit risk estimates for "lymphoid" cancer
9 incidence from the different "lymphoid" cancer mortality models examined are presented in Table
10 9. The results are similar to those for lymphohematopoietic cancer incidence (i.e., within 30%;
11 see Table 7), especially for the two continuous variable models.
12
13 4.1.2. Risk Estimates for Breast Cancer
14 4.1.2.1. Breast Cancer Results From the NIOSH Study
15 The Steenland et al. (2004) study discussed above in Section 4.1.1.1 also presents results
16 from exposure-response analyses for breast cancer mortality in female workers. Steenland et al.
17 (2003) present results of a breast cancer incidence study of a subcohort of the female workers
18 from the NIOSH cohort.
19
20 4.1.2.2. Prediction of Lifetime Extra Risk of Breast Cancer Mortality
21 The Cox regression modeling results presented by Steenland et al. (2004) and summarized
22 in Table 10 were used for predicting the extra risks for breast cancer mortality in females from
23 continuous environmental exposure to EtO, applying the methodologies described in Section
24 4.1.1.
25 U.S. age-specific all-cause mortality rates for 1999 for females of all race groups
26 combined (NCHS, 2002) were used to specify the all-cause background mortality rates in the
27 actuarial program. The National Center for Health Statistics 1997-2001 cause-specific
28 background mortality rates for invasive breast cancers in females were obtained from a
29 Surveillance, Epidemiology, and End Results (SEER) report (NCI, 2004a). The risks were
30 computed up to age 85 for continuous exposures to EtO, conversions were made between
31 occupational EtO exposures and continuous environmental exposures, and 95% UCLs were
32 calculated for the relative rates, as described above.
33 The best-fitting Cox regression model presented by Steenland et al. (2004) for breast
34 cancer mortality in females was for log cumulative exposure with a 20-year lag (/?=0.01).
35 However, as for the lymphohematopoietic cancers in Section 4.1.1, using the log cumulative
36 exposure model to estimate the risks from low environmental exposures is problematic because
08/29/06 36 DRAFT—DO NOT CITE OR QUOTE
-------
1 this model is highly supralinear and results are unstable for low exposures (see Figure 6). The
2 (continuous variable) cumulative exposure model, which is typically used and which is stable at
3 low exposures, did not provide a good fit to the breast cancer mortality data (p=0.34). In
4 addition, the continuous variable Cox regression model with cumulative exposure is sublinear for
5 low exposures and would not reflect the apparent supralinearity of the data.
6 It was again determined that the best way to reflect the data in the lower exposure region,
7 which is the region of interest for low-exposure extrapolation, was to do a weighted linear
8 regression of the results from the model with categorical cumulative exposure (and a 20-year lag),
9 excluding the highest exposure group. The weights used for the ORs were the inverses of the
10 variances, which were calculated from the confidence intervals. Mean and median exposures for
11 the cumulative exposure groups were kindly provided by Dr. Steenland (e-mail dated May 26,
12 2004, from Kyle Steenland, Emory University, to Jennifer Jinot, U.S. EPA).4 The mean values
13 were used for the weighted regression analysis because the cancer response is presumed to be a
14 function of cumulative exposure, which is expected to be best represented by mean exposures. If
15 the median values had been used, a slightly larger regression coefficient would have been
16 obtained, resulting in slightly larger risk estimates.
17 Using this approach, a regression coefficient of 0.000201 per ppm x day (SE = 0.000120
18 per ppm x day) was obtained from the weighted linear regression of the categorical results. These
19 parameter estimates were used for the primary risk calculations for breast cancer mortality in
20 females. Risk calculations using the Cox model results for the continuous log cumulative
21 exposure variable (Table 10) were also performed for comparison. No risk estimates were derived
22 based on the continuous cumulative exposure model results because of the poor fit for this model
23 (/?=0.34).
24 Point estimates and one-sided 95% UCLs for the extra risk of breast cancer mortality in
25 females associated with varying levels of continuous environmental exposure to EtO based on the
26 Steenland et al. (2004) model results were calculated using the actuarial program; the results are
27 presented in Table 11 (point estimates only are presented from the continuous log cumulative
28 exposure variable comparison model). The model based on (categorical) cumulative exposure
29 yields extra risk estimates that are fairly linear for exposures below 0.1 ppm. The risk estimates
30 based on log cumulative exposure, on the other hand, are not (low-dose) linear even in the range
31 of 0.0001 to 0.001 ppm, the lowest exposure concentrations evaluated (i.e., not linear through
32 zero exposure).
Mean exposures for females with a 20-year lag for the categorical exposure quartiles in Table 10 were
276; 1,453; 5,869; and 26,391 ppm x days. Median values were 250; 1,340; 5,300; and 26,676 ppm x days. These
values are for the risk sets but should provide a good approximation to the full cohort values.
08/29/06 37 DRAFT—DO NOT CITE OR QUOTE
-------
1 The same data and methodology were also used to estimate the exposure level (ECX;
2 "effective concentration") and the associated 95% lower confidence limit (LECX) corresponding to
3 an extra risk of 1% (x = 0.01). As discussed in Section 4.1.1, a 1% extra risk level is amore
4 reasonable response level for defining the POD for these epidemiologic data than 10%. Based on
5 the actuarial program, the risk ratio for an extra risk of 1% for breast cancer mortality is 1.4,
6 which better corresponds with the ORs reported by Steenland et al. (2004) than the higher risk
7 ratio of 4.9 that would be associated with a 10% extra risk. In fact, for the linear model that was
8 selected, the unit risk is independent of the benchmark risk level used to determine the POD;
9 however, selection of a benchmark risk level is useful for comparisons across models.
10 Because EtO is clearly mutagenic, a linear low-dose extrapolation was performed, in
11 accordance with EPA's Guidelines for Carcinogen Risk Assessment. The EC01, LEC01, and
12 inhalation unit risk estimates for the different breast cancer mortality models examined are
13 presented in Table 12. As discussed above, the unit risk estimate based on linear regression of the
14 Steenland et al. (2004) categorical results using cumulative exposure with a 20-year lag (i.e.,
15 0.513 per ppm) is the preferred estimate for breast cancer mortality risks from environmental
16 exposures. Estimates from the continuous log cumulative exposure variable model are presented
17 for comparison only. The log cumulative exposure model, which is supralinear in the lower
18 exposure region, yields a substantially higher unit risk estimate (45.7 per ppm). In this case, the
19 supralinear model appears to better approximate the lower exposure categorical data than for
20 lymphohematopoietic cancer (See Figures 4 and 6), but it is difficult to obtain stable unit risk
21 estimates from this model. Furthermore, there is uncertainty around the RR estimates; thus, in the
22 absence of mechanistic data to support a supralinear model, it seems appropriate to use the results
23 of the linear (regression) model, as EPA typically does for categorical epidemiologic data.
24 Converting the units, 0.513 per ppm corresponds to a unit risk of 2.80 x 10"4 per |ig/m3.
25
26 4.1.2.3. Prediction of Lifetime Extra Risk of Breast Cancer Incidence
27 As discussed in Section 4.1.1.3, risk estimates for cancer incidence are preferred to
28 estimates for cancer mortality. In the case of female breast cancer in the NIOSH cohort, there is a
29 corresponding incidence study (Steenland et al., 2003) with exposure-response results for breast
30 cancer incidence, so we can estimate cancer incidence risks directly rather than estimate them
31 from mortality data. The incidence study used a subcohort of 7,576 (76%) of the female workers
32 from the original cohort. Subcohort eligibility was restricted to the female workers who had been
33 employed at 1 of the 14 plants for at least 1 year, owing to cost considerations and the greater
34 difficulties in locating workers with short-term employment. Completed questionnaires were
35 received for 5,139 (68%) of the 7,576 women in the subcohort. The investigators also attempted
36 to acquire breast cancer incidence data for the entire subcohort from cancer registries (available
08/29/06 38 DRAFT—DO NOT CITE OR QUOTE
-------
1 for 9 of the 11 states in which the plants were located) and death certificates; thus, results are
2 presented for both the full (sub)cohort (n = 7,576) and the subcohort of women with completed
3 questionnaires (n = 5,139). For additional details and discussion of the Steenland et al. (2003)
4 study, see Appendix A.
5 Steenland et al. (2003) identified 319 incident cases of breast cancer in the cohort through
6 1998. Interview (questionnaire) data were available for 73% (233 cases). Six percent were
7 carcinoma in situ (20 cases). Steenland et al. (2003) performed internal exposure-response
8 analyses similar to those described in their 2004 paper and in Section 4.2.1 above. Controls for
9 each case were selected from the cohort members without breast cancer at the age of diagnosis of
10 the case. Cases and controls were matched on race. Of the potential confounders evaluated for
11 those with interviews, only parity and breast cancer in a first-degree relative were important
12 predictors of breast cancer, and only these variables were included in the final models for the
13 subcohort analyses. In situ cases were included with invasive breast cancer cases in the analyses;
14 however, the in situ cases represent just 6% of the total, and excluding them reportedly did not
15 greatly affect the results.
16 For internal analyses using the full cohort, the best-fitting model with exposure as a
17 continuous variable was for (natural) log cumulative exposure, lagged 15 years (/?=0.05).
18 Duration of exposure, lagged 15 years, provided a slightly better fitting model. Models using
19 maximum or average exposure did not fit as well. In addition, use of a threshold model did not
20 provide a statistically significant improvement in fit. For internal analyses using the subcohort
21 with interviews, the cumulative exposure and log cumulative exposure models, both lagged 15
22 years, and the log cumulative exposure model with no lag all fit almost equally well, and the
23 duration of exposure (also lagged 15 years) model fit slightly better. Results of the Cox
24 regression analyses for the cumulative and log cumulative exposure models, with 15-year lags, are
25 shown in Table 13, and these are the results used for the unit risk calculations. The models using
26 duration of exposure are less useful for estimating exposure-related risks, duration of exposure
27 and cumulative exposure are correlated, and the fits for these models are only marginally better
28 than those with cumulative exposure. The log cumulative exposure model with no lag was
29 considered less biologically realistic than the corresponding model with a 15-year lag because
30 some lag period would be expected for the development of breast cancer. Furthermore, although
31 risk estimates based on the full cohort results are calculated for comparison, the preferred
32 estimates are those based on the subcohort with interviews because the subcohort should have
33 complete case ascertainment and has additional information available on potential breast cancer
34 confounders.
35 For the actuarial program, U.S. age-specific all-cause mortality rates for 1999 for females
36 of all race groups combined (NCHS, 2002) were used to specify the all-cause background
08/29/06 39 DRAFT—DO NOT CITE OR QUOTE
-------
1 mortality rates. Because breast cancer incidence rates are not negligible compared to all-cause
2 mortality rates, the all-cause mortality rates in the life-table analysis were adjusted to reflect
3 women dying or acquiring incident cases of breast cancer in a given age interval. All-cause
4 mortality rates and breast cancer incidence rates were summed, and breast cancer mortality rates
5 were subtracted so that those dying of breast cancer were not counted twice (i.e., as deaths and as
6 incident cases of breast cancer). The National Center for Health Statistics 1997-2001 mortality
7 rates for invasive breast cancer in females were obtained from a SEER report (NCI, 2004a). The
8 SEER report also provided SEER-9 incidence rates for invasive and in situ breast cancer. The
9 Cox regression results reported by Steenland et al. (2003) are for invasive and in situ breast
10 cancers combined. It is consistent with EPA policy to combine these two tumor types because the
11 in situ tumors can progress to invasive tumors. Thus, the primary risk calculations in this
12 assessment use the sum of invasive and in situ breast cancer incidence rates for the cause-specific
13 background rates. Comparison calculations are performed using just the invasive breast cancer
14 incidence rates for the cause-specific rates; this issue is further discussed in Section 4.1.3 on
15 sources of uncertainty. The risks were computed up to age 85 for continuous exposures to EtO,
16 conversions were made between occupational EtO exposures and continuous environmental
17 exposures, and 95% UCLs were calculated for the relative rates, as described above.
18 For breast cancer incidence in both the full cohort (Figure 7) and the subcohort with
19 interviews (Figure 8), the categorical data suggest a more linear exposure-response relationship
20 than that obtained with either the continuous variable log cumulative exposure (supralinear) or
21 cumulative exposure (sublinear) models. Thus, as with the lymphohematopoietic cancer and the
22 breast cancer mortality results above, it was determined that the best way to reflect the data in the
23 lower exposure region, which is the region of interest for low-exposure extrapolation, was to do a
24 weighted linear regression of the results from the model with categorical cumulative exposure
25 (with a 15-year lag). In addition, the highest exposure group was not included in the regression to
26 provide a better fit to the lower exposure data. The weights used for the ORs were the inverses of
27 the variances, which were calculated from the confidence intervals. Mean and median exposures
28 for the cumulative exposure groups for the full cohort were kindly provided by Dr. Steenland (e-
29 mail dated April 21, 2004, from Kyle Steenland, Emory University, to Jennifer Jinot, U.S. EPA).5
30 The mean values were used for the weighted regression analysis because the (arithmetic) mean
31 exposures best represent the model's linear relationship between exposure and cancer response.
32 Differences between means and medians were not large for the females, especially for the lower
Mean exposures for females with a 15-year lag for the exposure categories in Table 3 were 280; 1,241;
3,304; 8,423; and 36,022 ppm x days. Median values were 253; 1,193; 3,241; 7,741; and 26,597 ppm x days.
These values are for the risk sets but should provide a good approximation to the full cohort values.
08/29/06 40 DRAFT—DO NOT CITE OR QUOTE
-------
1 four quintiles. If the median values had been used, a slightly larger regression coefficient would
2 have been obtained, resulting in slightly larger risk estimates. Although the exposure values are
3 for risk sets from the full cohort, they should be reasonably close to the values for the subcohort
4 with interviews.
5 Using this approach, a regression coefficient of 0.0000264 per ppm x day (SE =
6 0.0000269 per ppm x day) was obtained from the weighted linear regression for the full cohort,
7 and a regression coefficient of 0.0000517 per ppm x day (SE = 0.0000369 per ppm x day) was
8 obtained for the subcohort of women with interviews. The regression coefficients and standard
9 errors from the weighted linear regression of the categorical results were used for the primary risk
10 calculations for breast cancer incidence. Risk estimates based on the model results for the
11 continuous exposure variables (i.e., cumulative exposure and log cumulative exposure; see Table
12 13) are also derived for comparison.
13 The resulting EC01, LEC01, and inhalation unit risk estimates for breast cancer incidence in
14 females from the various models examined are presented in Tables 14 (invasive and in situ) and
15 15 (invasive only). The primary calculation for invasive and in situ breast cancer incidence based
16 on the linear regression coefficient from the categorical results (without the highest exposure
17 group) in the subcohort of women with interviews yielded an EC01 of 0.024 ppm, an LEC01 of
18 0.011 ppm, and a unit risk estimate of 0.909 per ppm. The comparable unit risk estimate for the
19 full cohort is about 40% lower. One would expect this value to be lower because of incomplete
20 case ascertainment in the full cohort. The corresponding unit risk estimate derived based on the
21 subcohort results but using invasive breast cancer only for the background incidence rates is about
22 17% lower, reflecting the difference between incidence rates for invasive breast cancer only and
23 for combined in situ and invasive breast cancer. The unit risk estimate of 0.909 per ppm is the
24 preferred estimate for female breast cancer risk because it is based on incidence data versus
25 mortality data, it is based on more cases (n = 233) than the mortality estimate (n = 103), and
26 information on personal breast cancer risk factors obtained from the interviews is taken into
27 account. Converting the units, 0.909 per ppm corresponds to a unit risk of 4.97 x 10"4 per |ig/m3.
28
29 4.1.3. Sources of Uncertainty in the Cancer Risk Estimates
30 The two major sources of uncertainty in quantitative cancer risk estimates are generally
31 from interspecies extrapolation and high-dose to low-dose extrapolation. The risk estimates
32 derived from the Steenland et al. (2003, 2004) analyses are not subject to interspecies uncertainty
33 because they are based on human data. Furthermore, the human-based estimates suffer less from
34 high-dose to low-dose extrapolation than do rodent-based estimates and, thus, uncertainty from
35 that source is reduced somewhat. For example, the average exposure in the NIOSH cohort was
36 more than 10 times lower than the lowest exposure level in a rodent bioassay after adjustment to
08/29/06 41 DRAFT—DO NOT CITE OR QUOTE
-------
1 continuous lifetime exposure. Nonetheless, some uncertainty remains in the extrapolation from
2 occupational exposures to lower environmental exposures. Although the actual exposure-
3 response relationship at low exposure levels is unknown, the clear evidence of EtO mutagenicity
4 supports the linear low-dose extrapolation that was used (U.S. EPA, 2005a).
5 Other sources of uncertainty emanate from the epidemiologic studies and their analyses
6 (Steenland et al., 2003, 2004), including the retrospective estimation of EtO exposures in the
7 cohort, the modeling of the epidemiologic exposure-response data, the proper dose metric for
8 exposure-response analysis, and potential confounding or modifying factors. Although these are
9 areas of inevitable uncertainties in epidemiologic studies, they were generally well addressed in
10 the NIOSH studies.
11 Regarding exposure estimation, the NIOSH investigators conducted a detailed
12 retrospective exposure assessment to estimate the individual worker exposures. They used
13 extensive data from 18 facilities, spanning a number of years, to develop a regression model
14 (Greife et al., 1988; Hornung et al., 1994). The model accounted for 85% of the variation in
15 average EtO exposure levels. Detailed work history data for the individual workers were
16 collected for the 1987 followup (Steenland et al., 1991). For the extended followup (Steenland et
17 al., 2003, 2004), additional information on the date last employed was obtained for those workers
18 still employed and exposed at the time of the original work history collection for the plants still
19 using EtO (25% of the cohort). It was then assumed that exposure for these workers continued
20 until the date of last employment and that their exposure level stayed the same as that in their last
21 job held at the time of the original data collection. Thus, there would be more exposure
22 misclassification in the extended followup. However, when the investigators compared
23 cumulative exposures estimated with and without the extended work histories, they found little
24 difference because exposure levels were very low by the mid-1980s and, therefore, had little
25 impact on cumulative exposure (Steenland et al., 2003, 2004). While the NIOSH regression
26 model performed very well in estimating exposures in validation tests (Hornung et al., 1994),
27 there is, nonetheless, considerable uncertainty associated with any retrospective exposure
28 assessment, and this can affect the ability to discriminate among exposure-response models.
29 With respect to the male lymphohematopoietic cancer response (Steenland et al., 2004), it
30 must be noted that all attempts at exposure-response modeling are limited by the small number of
31 cases (n = 37). The Cox proportional hazards model used by Steenland et al. is commonly used
32 for this type of analysis because exposure can be modeled as a continuous variable, competing
33 causes of mortality can be taken into account, and potential confounding factors can be controlled
34 for in the regression. Normally, model dependence should be minimized by the practice, under
35 EPA's 2005 Guidelines for Carcinogen Risk Assessment, of modeling only in the observable
36 range and then performing a linear extrapolation from the "POD" (in this case the LEC01).
08/29/06 42 DRAFT—DO NOT CITE OR QUOTE
-------
1 However, the log cumulative exposure Cox regression model with 15-year lag, which provides the
2 best fit to the overall data, is too steep in the low-exposure region and then plateaus rapidly,
3 making it difficult to derive stable risk estimates (i.e., estimates that are not highly dependent on
4 the POD). And the alternative (continuous variable) cumulative exposure model, though typically
5 used for epidemiologic data, is too sublinear in the low-exposure region for these data, which
6 exhibit supralinearity. Thus, a weighted linear regression of the categorical cumulative exposure
7 (with 15-year lag) results (dropping the highest exposure group) was used in this assessment to
8 better represent the data in the lower exposure region. Dropping the highest two exposure groups
9 was not done because it is desirable to use as much of the data as reasonably possible and because
10 the second-highest exposure group is not so far from potential environmental exposure levels
11 (unlike the highest exposure group, which reflects much higher exposures). It should also be
12 noted that the various models gave similar risk estimates for the subcategory of "lymphoid"
13 cancer in males as were obtained for all lymphohematopoietic cancer.
14 Although linear regression of the categorical results seems to be a reasonable approach for
15 best reflecting the exposure-response results at the lower end of the exposure range, clearly there
16 is uncertainty regarding the exposure-response model, as suggested by the range of risk estimates
17 resulting from the different models (Table 7). The best-fitting continuous exposure model (the
18 log cumulative exposure model) yields a higher unit risk estimate, but the unit risk based on the
19 linear regression of the categorical results is preferred because it is more stable. Linear modeling
20 of categorical epidemiological results is a standard technique used by EPA (U.S. EPA, 2005a).
21 Another area of uncertainty related to the exposure-response modeling is the lag period.
22 The best-fitting models presented by Steenland et al. (2004) for lymphohematopoietic cancer had
23 a 15-year lag (lag periods of 0, 5, 10, 15, and 20 years were explored). Lymphohematopoietic
24 cancers are thought to generally (but not always) have a relatively short latency period, but a 15-
25 year lag period means that exposures in the 15 years prior to death or the end of followup are not
26 taken into account. In other words, in the best-fitting models, relevant exposures for the
27 development of the lymphohematopoietic cancers occurred over 15 years before death. Yet, the
28 analyses of the investigators indicate that the regression coefficient for cumulative exposure might
29 have decreased with followup, suggesting that the higher exposure levels encountered by the
30 workers in the more distant past are having less of an impact on current risk. The regression
31 coefficient was 1.12 x 10"5 per ppm x day, for both sexes with a 10-year lag, in the 1987 followup
32 (Stayner et al., 1993) versus 4 x 10"6 per ppm x day, for males with no lag, in the 1998 followup
33 (Steenland et al., 2004). The earlier value was for males and females, but even then the
34 lymphohematopoietic cancer response was observed predominantly in males, so a regression
35 coefficient for males only would have been even greater. On the other hand, with no lag period,
08/29/06 43 DRAFT—DO NOT CITE OR QUOTE
-------
1 as for the more recent coefficient for cumulative exposure, the earlier value would have been
2 diminished somewhat.
3 The life-table analysis used in this risk assessment accrues exposure over the full lifetime
4 for the cumulative exposure metric. If, in fact, exposures in the distant past cease to have a
5 meaningful impact on risk of lymphohematopoietic cancer, this approach would tend to
6 overestimate the unit risk. Thus, a comparison analysis was conducted to evaluate the impact of
7 ignoring exposures over 55 years in the past in the life-table analysis. The actual value of such a
8 cut-point, if warranted, is unknown. A value less than 55 years might not be appropriate because
9 exposures for some of the workers began in 1943, so any diminution of potency for past exposures
10 occurring since 1943 is already reflected in the regression coefficient with followup through 1998,
11 at least for those workers, although it is unknown what proportion of workers had such early
12 exposures and how long they survived. The comparison analysis yielded an LEC01 of 0.00815
13 ppm and a unit risk estimate of 1.23 per ppm, which is 25% less than the estimate obtained from
14 the unrestricted life-table analysis. Because the appropriate cut-point for excluding past exposures
15 is unknown and the unit risk estimate from the linear regression is already substantially less than
16 that obtained from the best-fitting model, the estimate from the full life-table analysis is preferred.
17 Several dose metrics (cumulative exposure, duration of exposure, maximum [8-hour
18 TWA] exposure, and average exposure) were analyzed by the investigators, and cumulative
19 exposure was the best predictor of mortality from lymphohematopoietic cancers. Cumulative
20 exposure is considered a good measure of total exposure because it integrates exposure (levels)
21 overtime.
22 Also, the important potential modifying/confounding factors of age, sex, race, and
23 calendar time were taken into account in their analysis, and the plants included in this cohort were
24 specifically selected for the absence of any known confounding exposures (Stayner et al., 1993).
25 The linear regression discussed above is based on the categorical data for the males only. A
26 concern about the results of this study is the observation of increased risks of
27 lymphohematopoietic cancers in males but not in females. In females, increased ORs were
28 observed in the second exposure quartile for lymphohematopoietic cancer and in the second and
29 third quartiles for "lymphoid" cancer compared to the lowest quartile, but there is no evidence of a
30 positive exposure-response relationship overall. Average exposures in males (mean 37.8 ppm,
31 standard deviation 87.6 ppm, median 7.6 ppm) were higher than in females (mean 18.2 ppm,
32 standard deviation 38.2 ppm, median 4.6 ppm), but, according to Steenland et al. (2004), there
33 was enough variation in the exposures in females to have observed an exposure-response
34 relationship if one existed. There is no known biological basis for this observed discrepancy in
35 the lymphohematopoietic cancer response between males and females, and no such sex-specific
36 effect is suggested by the rodent data. In the NTP (1987) mouse bioassay, only females had EtO-
08/29/06 44 DRAFT—DO NOT CITE OR QUOTE
-------
1 induced lymphohematopoietic cancer (malignant lymphoma), whereas in the Snellings et al.
2 (1984) rat study both males and females had EtO-induced lymphohematopoietic cancer
3 (mononuclear cell leukemia), and the risk in females was slightly higher than that in males
4 (Table 3).
5 As for the other epidemiological studies that included females, they generally had too few
6 lymphohematopoietic cancer cases to reveal much about the possible risks to females. In the
7 Hogstedt et al. (1988) study, there were 3 leukemias in 170 women versus 0.2 expected. In the
8 Hagmar et al. (1991, 1995) study of 1309 women and 861 men, there was an SIR of 1.78 for
9 lymphohematopoietic cancer and an SIR of 2.44 for leukemia, but there was no breakdown by
10 sex. In the Norman et al. (1995) study, one leukemia case was observed in 928 women versus
11 0.41 expected. In the Kardos et al. (2003) study of 233 females, there were 11 cancer deaths
12 versus 4.03-4.38 expected; one of these deaths was from lymphoid leukemia, but expected deaths
13 by cancer type were not provided. In the Coggon et al. (2004) study of 1012 women and 1864
14 men, there were 17 lymphohematopoietic cancer deaths versus 12.9 expected. No breakdown by
15 sex was provided; however, 12 of these 17 deaths were in chemical workers (vs. 7.9 expected, i.e.,
16 all of the observed excess was in the chemical workers, as opposed to the hospital workers), and
17 all but one of the chemical workers were male. Coggon et al. claim that it seems unlikely that
18 either peak or average exposures to EtO were markedly higher in the chemical workers than in the
19 hospital workers, but they acknowledge that this claim is based on limited exposure data. Thus,
20 the Coggon et al. results are consistent with a higher risk of lymphohematopoietic cancer in males
21 than in females; however, the data are insufficient to infer anything.
22 With respect to the breast cancer mortality response (Steenland et al., 2004), the exposure-
23 response modeling was based on 103 deaths. As for the lymphohematopoietic cancer response in
24 males, the exposure-response data for breast cancer mortality are fairly supralinear, especially for
25 the low-exposure groups. Thus, the same approach of using a weighted linear regression of the
26 categorical results with the highest exposure group excluded was taken to obtain a regression
27 coefficient for the life-table analysis. As shown in Table 12, the unit risk estimate obtained from
28 this approach is substantially lower than that obtained from the log cumulative exposure model.
29 Nonetheless, the linear regression approach is considered appropriate because the log cumulative
30 exposure model generates unstable risk estimates, there is uncertainty in the RR models, and a
31 linear exposure-response model is the default model typically used by EPA for categorical human
32 cancer data (U.S. EPA, 2005a). For the lag period, the best-fitting model had a lag of 20 years,
33 which was longest lag period investigated. This is a commonly used lag period for solid tumors,
34 which typically have longer latency periods than lymphohematopoietic cancers. It is unknown
35 whether a lag period longer than 20 years would have provided a better model fit. The Steenland
36 et al. (2004) analysis took into account age, race, and calendar time. Other risk factors for breast
08/29/06 45 DRAFT—DO NOT CITE OR QUOTE
-------
1 cancer could not be included in the mortality analysis, but many of these factors were considered
2 in the breast cancer incidence study (Steenland et al., 2003), as discussed below.
3 Steenland et al. (2003) conducted an incidence study for breast cancer; therefore, it was
4 not necessary to calculate unit risk estimates for breast cancer incidence indirectly from the
5 mortality data as was done for lymphohematopoietic cancer. Further advantages to using the
6 results from the incidence study are that more cases were available for the Cox regression
7 modeling (319 cases) and that the investigators were able to include data on potential confounders
8 in the modeling for the subcohort with interviews (233 cases). For the full cohort, the cumulative
9 exposure model providing the best fit to the data was again the log cumulative exposure model.
10 With breast cancer incidence, a 15-year lag provided the best model fits. For the subcohort, the
11 cumulative exposure and log cumulative exposure models fit nearly equally well. For both
12 groups, the categorical data suggest that a linear model lying between the supralinear log
13 cumulative exposure model and the sub linear cumulative exposure model would better represent
14 the low-exposure data than either of the two presented continuous variable models (Figures 6 and
15 7). Thus, for both groups, a linear regression was fit through the categorical results, dropping the
16 highest exposure group to provide a better fit to the lower exposure data.
17 As can be seen in Tables 14 and 15, there is substantial variation in the risk estimates
18 obtained from these different models. The categorical data for breast cancer incidence do not
19 display the supralinearity in the lower exposure groups seen in the cases discussed above (with the
20 inclusion of the highest exposure group, some supralinearity is evident); thus, the difference
21 between the unit risk estimates from the cumulative exposure model and the linear regression
22 model are not as dramatic as seen in those cases (the linear regression results are less than
23 sevenfold higher).
24 An area of uncertainty in the life-table analysis for breast cancer incidence pertains to the
25 rates used for the cause-specific background rate. The regression coefficients presented by
26 Steenland et al. (2003) represent invasive and in situ cases combined, where 6% of the cases are
27 in situ, and the preferred unit risk estimates in this assessment are calculated similarly using
28 background rates for invasive and in situ cases combined. The regression coefficients for invasive
29 and in situ cases combined should be good approximations for a regression coefficients for
30 invasive cases alone; however, it is uncertain how well they reflect the exposure-response
31 relationships for in situ cases alone. Diagnosed cases of in situ breast cancer would presumably
32 be remedied and not progress to invasive breast cancer, so double-counting is unlikely to be a
33 significant problem. Carcinoma in situ is a risk factor for invasive breast cancer; however, this
34 observation is most likely explained by the fact that these two types of breast cancer have other
35 breast cancer risk factors in common, some of which have been considered in the subcohort
36 analysis. One might hypothesize that EtO exposure could cause a more rapid progression to
08/29/06 46 DRAFT—DO NOT CITE OR QUOTE
-------
1 invasive tumors; however, there is no specific evidence that this occurs. On the other hand, there
2 is some indication that in situ cases in the incidence study might have been diagnosed at relatively
3 low rates in comparison to the invasive cases. Steenland et al. (2003) reported that 6% of the
4 cases in their study are in situ; according to the National Cancer Institute; however, ductal
5 carcinoma in situ accounted for about 18% of newly diagnosed cases of breast cancer in 1998
6 (NCI, 2004b).
7 There are several possible explanations for this difference. One is that it reflects
8 differences in diagnosis with calendar time because the rate of diagnosis of carcinoma in situ has
9 increased over time with increased use of mammography. Another is that the difference is
10 partially a reflection of the age distribution in the cohort because the proportion of new cases
11 diagnosed as carcinoma in situ varies by age. A third possible explanation is that the low
12 proportion of in situ cases is at least partially a consequence of underascertainment of cases
13 because in situ cases will not be reported on death certificates, although, even if all 20 in situ
14 cases were in the subcohort with interviews, that would still be only 8.6% of the cases. In any
15 event, this is a relatively minor source of uncertainty, and a comparison of the unit risk estimates
16 in Tables 14 (invasive + in situ) and 15 (invasive only) shows that the preferred estimate of 0.909
17 per ppm is less than 20% higher than the corresponding estimate using only invasive breast cancer
18 background rates.
19 The results for the subcohort with interviews are used for the primary breast cancer unit
20 risk calculations because, in addition to including the data on potential confounders, the subcohort
21 is considered to have full ascertainment of the breast cancer cases, whereas the full cohort for the
22 incidence study has incomplete case ascertainment, as illustrated by the fact that death certificates
23 were the only source of case ascertainment for 14% of the cases. Thus, risk estimates based on
24 the full cohort would be underestimated; nevertheless, these estimates were calculated for
25 comparison with the subcohort estimates. As can be seen in Table 14, the preferred unit risk
26 estimate of 0.909 per ppm is about 60% higher than the corresponding estimate from the full
27 cohort.
28 With respect to dose metrics for breast cancer incidence, models using duration provided
29 better model fits than those using cumulative exposure; however, duration is less useful for
30 estimating unit risks and the cumulative exposure models also provided a statistically significant
31 fit to the data, thus the cumulative exposure metric was used for the quantitative risk estimates.
32 Models using peak or average exposure did not fit as well. Regarding potential
33 confounders/modifying factors, analyses for the full cohort were adjusted for age, race, and
34 calendar time, and exposures to other chemicals in these plants were reportedly minimal. For the
35 subcohort with interviews, a number of specific breast cancer risk factors were investigated,
36 including body mass index, breast cancer in a first-degree relative, parity, age at menopause, age
08/29/06 47 DRAFT—DO NOT CITE OR QUOTE
-------
1 at menarche, socioeconomic status, and diet; however, only parity and breast cancer in a first-
2 degree relative were determined to be important predictors of breast cancer and were included in
3 the final models.
4 Some additional sources of uncertainty are not so much inherent in the exposure-response
5 modeling or in the epidemiologic data themselves but, rather, arise in the process of obtaining
6 more general Agency risk estimates from the epidemiologic results. EPA cancer risk estimates
7 are typically derived to represent an upper bound on increased risk of cancer incidence for all sites
8 affected by an agent for the general population. From experimental animal studies, this is
9 accomplished by using tumor incidence data and summing across all the tumor sites that
10 demonstrate significantly increased incidences, customarily for the most sensitive sex and species,
11 to be protective of the general human population. However, in estimating comparable risks from
12 the NIOSH epidemiologic data, certain limitations are encountered. First, the study reported by
13 Steenland et al. (2004) is a retrospective mortality study, and cancer incidence data are not
14 available for lymphohematopoietic cancer (for breast cancer, a separate incidence study
15 [Steenland et al., 2003] was performed). Second, these occupational epidemiology data represent
16 a healthy-worker cohort. Third, the epidemiologic study may not have sufficient statistical power
17 and followup time to observe associations for all the tumor sites that may be affected by EtO.
18 The first limitation was addressed quantitatively in the life-table analysis for the
19 lymphohematopoietic cancer risk estimates. Although assumptions are made in using incidence
20 rates for the cause-specific background rates, as discussed in Section 4.1.1.3, the resulting
21 incidence-based estimates are believed to be better estimates of cancer incidence risk than are the
22 mortality-based estimates. The healthy-worker effect is often an issue in occupational
23 epidemiology studies, but the internal exposure-response analyses conducted by these
24 investigators help address this concern, at least partially. In terms of representing the general
25 population, the NIOSH study cohort was relatively diverse. It contained both female (55%) and
26 male workers, and the workers were 79% white, 16% black, and 5% "other".
27 With respect to other possible tumor sites of concern, the rodent data suggest that
28 lymphohematopoietic cancers are the major (or a close second) tumor type associated with EtO
29 exposure in female mice and in male and female rats. Thus, it is reasonable to expect that this
30 might be a tumor type of concern in humans too. Likewise, the mouse data suggest an increased
31 risk of mammary gland tumors from EtO exposure, and evidence of that can be seen in the
32 Steenland et al. (2003, 2004) study. However, the rodent data suggest associations between EtO
33 exposure and other tumor types as well, and, although site concordance across species is not
34 generally assumed, it is possible that the NIOSH study, despite its relatively large size and long
35 followup (mean length of followup was 26.8 years), had insufficient power to observe small
36 increases in risk in certain other sites. For example, the tumor site with the highest potency
08/29/06 48 DRAFT—DO NOT CITE OR QUOTE
-------
1 estimate in both male and female mice was the lung. In the NIOSH study, one cannot rule out a
2 small increase in the risk of lung cancer, which has a high background rate.
3 Despite these uncertainties, the inhalation cancer unit risk estimates of 1.64 per ppm for
4 male lymphohematopoietic cancer incidence and 0.909 per ppm for female breast cancer
5 incidence have the advantages of being based on human data from a high-quality epidemiologic
6 study with individual exposure estimates for each worker. In addition, the similarity of the
7 estimates for these two different tumor responses, one in males and one in females, provides
8 support for the use of these estimates to represent the risk in humans. The male
9 lymphohematopoietic cancer incidence risk estimate was also similar to the estimate of 1.20 per
10 ppm based on the subcategory of "lymphoid" cancers in males.
11
12 4.1.4. Conclusions
13 An EC01 of 0.024 ppm was calculated using a life-table analysis and linear modeling of the
14 categorical Cox regression analysis results for excess lymphohematopoietic cancer mortality in
15 males reported in a high-quality occupational epidemiology study. Linear low-dose extrapolation
16 from the LEC01 yielded a lifetime extra cancer mortality unit risk estimate of 0.92 per ppm of
17 continuous EtO exposure. Applying the same linear regression coefficient and life-table analysis
18 to background lymphohematopoietic cancer incidence rates yielded an EC01 of 0.013 ppm and a
19 preferred lifetime extra cancer unit risk estimate of 1.64 per ppm (9.0 x 10"4 per |ig/m3). Use of
20 this unit risk estimate is also expected to be protective of females, based on the lifetime extra
21 cancer unit risk estimate of 0.909 per ppm (5.0 x 10"4 per |ig/m3; EC01 = 0.024 ppm) calculated
22 using the same approach from the results of a breast cancer incidence study of the same worker
23 cohort. (An increased risk of EtO-associated lymphohematopoietic cancer in females cannot be
24 ruled out; however, the Steenland et al. [2004] results suggest that any increased risk to females is
25 likely to be lower than the risk estimated for males, and it is not expected that the combined
26 increased risk to breast cancer and lymphohematopoietic cancer in females would exceed the unit
27 risk estimated for lymphohematopoietic cancer in males.)
28
29 4.2. INHALATION UNIT RISK DERIVED FROM EXPERIMENTAL ANIMAL DATA
30 4.2.1. Overall Approach
31 Lifetime animal cancer bioassays of inhaled EtO have been carried out in three
32 laboratories, as described in Section 3.2. The data from these reports are presented in Tables 1
33 through 3. These studies have also been reviewed by the IARC (1994) and Health Canada (2001).
34 Health Canada calculated the ED05 for each data set using the benchmark dose methodology. The
35 EOIC report (EOIC, 2001) tabulated only lymphatic tumors because they constituted the
36 predominant risk.
08/29/06 49 DRAFT—DO NOT CITE OR QUOTE
-------
1 The overall approach in this derivation is to find a unit risk for each of the bioassays—
2 keeping data on males and females separate—from data on the incidence of all tumor types and
3 then to use the maximum of these values as the summary measure of the unit risk from animal
4 studies (i.e., the unit risk represents the most sensitive species and sex). The unit risk for the
5 animals in these bioassays is converted to a unit risk in humans by first determining the
6 continuous exposures in humans that are equivalent to the rodent bioassay exposures and then by
7 assuming that the lifetime incidence in humans is equivalent to lifetime incidence in rodents, as is
8 commonly accepted in interspecies risk extrapolations. For cross-species scaling of exposure
9 levels (see Section 4.2.2 below), an assumption of ppm equivalence is used; thus, no interspecies
10 conversion is needed for the exposure concentrations. Bioassay exposure levels are adjusted to
11 equivalent continuous exposures by multiplying by (hours of exposure/24 hours) and by (5/7) for
12 the number of days exposed per week. The unit risk in humans (risk per unit air concentration) is
13 then assumed to be numerically equal to that in rodents (after adjustment to continuous
14 exposures); the calculations from the rodent bioassay data are shown in Tables 1 through 3.
15
16 4.2.2. Cross-Species Scaling
17 In the absence of chemical-specific information, EPA's 1994 inhalation dosimetry
18 methods (U.S. EPA, 1994) provide standard methods and default scaling factors for cross-species
19 scaling. Under EPA's methodology, EtO would be considered a Category 2 gas because it is
20 reactive and water soluble and has clear systemic distribution and effects. Dosimetry equations
21 for Category 2 gases are undergoing EPA re-evaluation and are not being used at this time. For
22 cross-species scaling of extrarespiratory effects, current practice is to treat Category 2 gases as
23 Category 3 gases. For Category 3 gases, ppm equivalence is assumed (i.e., responses across
24 species are equivalent on a ppm exposure basis), unless the ainblood partition coefficient for the
25 experimental species is less than the coefficient for humans (U.S. EPA, 1994, p. 4-61). In the
26 case of EtO, measured ainblood partition coefficients are 78 in the mouse (Fennell and Brown,
27 2001), 64 in the rat (Krishnan et al, 1992), and 61 in the human (Csanady et al, 2000); thus, ppm
28 equivalence for cross-species scaling to humans can be assumed for extrarespiratory effects
29 observed in mice and rats. The assumption of ppm equivalence is further supported by the PBPK
30 modeling of Fennell and Brown (2001), who reported that simulated blood AUCs for EtO after 6
31 hours of exposure to concentrations between 1 ppm and 100 ppm were similar for mice, rats, and
32 humans and were linearly related to the exposure concentration (see Section 3.3.1 and Figure 2).
33 This modeling was validated against measured blood EtO concentrations for rodents and humans.
34 In addition, EOIC (2001) and Health Canada (2001) used ppm equivalence in their risk
35 assessments for EtO; Health Canada applied the equivalence to both extrarespiratory and
36 respiratory effects, whereas EOIC modeled only lymphohematopoietic cancers.
08/29/06 50 DRAFT—DO NOT CITE OR QUOTE
-------
1 For Category 2 gases with respiratory effects, there is no clear guidance on an interim
2 approach. One suggested approach is to do cross-species scaling using both Category 1 and
3 Category 3 gas equations and then decide which is most appropriate. In this document, the
4 preferred approach was to assume ppm equivalence was also valid for the lung tumors in mice
5 because of the clear systemic distribution of EtO (e.g., see Section 3.1). Treating EtO as a
6 Category 1 gas for cross-species scaling of the lung tumors would presume that the lung tumors
7 are arising only from the immediate and direct action of EtO as it comes into first contact with the
8 lung. In fact, some of the EtO dose contributing to lung tumors is likely attributable to
9 recirculation of systemic EtO through the lung.
10 If one were to treat EtO as a Category 1 gas for the cross-species scaling of the lung tumor
11 response as a bounding exercise, EPA's 1994 inhalation dosimetry methods present equations for
12 estimating the RGDRPU, i.e., the regional gas dose ratio for the pulmonary region, which acts as an
13 adjustment factor for estimating human equivalent exposure concentrations from experimental
14 animal exposure concentrations (adjusted for continuous exposure) (U.S. EPA, 1994, pp. 4-49 to
15 4-51). These equations rely on parameters describing mass transport of the gas (EtO) in the
16 extrathoracic and tracheobronchial regions for both the experimental animal species (mouse) and
17 humans. Without experimental data for these parameters, it seems reasonable to estimate
18 RGDRPU using a simplified equation and the adjusted alveolar ventilation rates of Fennell and
19 Brown (2001). Fennell and Brown adjusted the alveolar ventilation rates to reflect limited
20 pulmonary uptake of EtO, a phenomenon commonly observed for highly water-soluble gases
21 (Johanson and Filser, 1992). The adjusted ventilation rates were then used by Fennell and Brown
22 in their PBPK modeling simulations, and good fits to blood concentration data were reported for
23 both the mouse and human models. In this document, the adjusted alveolar ventilation rates were
24 used to estimate the RGDRPU as follows:
RGDRPU = (RGDPU)m/(RGDPU)h = (Qalv/SAPU)m/(Qalv/SAPU)h,
25 where:
26 RGDPU = regional gas dose to the pulmonary region,
27 Qalv = (adjusted) alveolar ventilation rate,
28 SAPU = surface area of the pulmonary region, and
29 the subscripts "m" and "h" denote mouse and human values.
30
31 Then, using adjusted alveolar ventilation rates from Fennell and Brown (2001) and surface area
32 values from EPA (U.S. EPA, 1994, p. 4-26),
33
08/29/06 51 DRAFT—DO NOT CITE OR QUOTE
-------
RGDRPU = ((0.78 L/h)/(0.05 m2))/((255 L/h)/(54.0 m2) = 3.3.
1 Using this value for the RGDRPU would increase the human equivalent concentration about
2 threefold, resulting in a decreased risk for lung tumors of about threefold, as a lower bound. The
3 true value of the RGDRPU is expected to be between 1 and 3, and any adjustment to the lung
4 tumor risks would still be expected to result in unit risk estimates roughly within the range of the
5 rodent unit risk estimates derived later in Section 4.2 under the assumption of ppm equivalence.
6
7 4.2.3. Dose-Response Modeling Methods
8 In this document we proceed with the following steps:
9 1. Extract the incidence data presented in the original studies. Our procedure differs in
10 only minor respects from the reviews by IARC (1994) and Health Canada (2001). In the IARC
11 monograph the incidence of brain tumors in the Garman et al. (1985) study was for gliomas alone,
12 whereas we have included all primary brain tumors. In the Health Canada report there was a
13 numerical error in the mid-dose incidence of mononuclear cell leukemia in the Lynch et al. (1982,
14 1984) study. From the NTP (1987) study Health Canada tabulated only carcinomas of the lung,
15 whereas we included lung adenomas as well as carcinomas. In order to crudely adjust for early
16 mortality in the analysis of the NTP (1987) data, we have corrected the incidence data for a
17 specific tumor type by eliminating the animals that died prior to the occurrence of the first tumor
18 or prior to 52 weeks, whichever was earlier. It was not possible to make this adjustment with the
19 other studies where data on individual animals were not available. With these exceptions, the
20 tumor incidence data in Tables 1 through 3 match the original data, the Health Canada data, and
21 IARC incidence data.
22 2. Fit the multistage model to the dose-response data using the Tox_Riskprogram.
23 The likelihood-ratio test was used to determine the lowest value of the multistage polynomial
24 degree that provided the best fit to the data while requiring selection of the most parsimonious
25 model. In this procedure, if a good fit to the data in the neighborhood of the POD is not obtained
26 with the multistage model because of a nonmonotonic reduction in risk at the highest dose tested
27 (as sometimes occurs when there is early mortality from other causes), that data point is
28 eliminated and the model is fit again to the remaining data. Such a deletion was found necessary
29 in two cases (mammary tumors in the NTP study and mononuclear cell leukemia in the Lynch
30 study). The goodness-of-fit measures for the dose-response curves and the parameters derived
31 from them are shown in Appendix D.
32 In the NTP bioassay, where the individual animal data were available, a time-to-tumor
33 analysis was undertaken to account for early mortality. The general model used in this analysis is
34 the multistage Weibull model:
08/29/06 52 DRAFT—DO NOT CITE OR QUOTE
-------
P(d,t) = 1 - exp[-(q0 + qid + q2d2 + ... + qkdk)*(t - t0)z],
4 where P(d,t) represents the probability of a tumor by age t (in bioassay weeks) for dose d (i.e.,
5 human equivalent exposure), and the parameter ranges are restricted as follows: z > 1, t0 >0, and q;
6 >0 for I = 0, 1, ..., k. The parameter t0 represents the time between when a potentially fatal tumor
7 becomes observable and when it causes death. The analyses were conducted using the
8 computer software Tox_Risk version 3.5, which is based on methods developed by Krewski et al.
9 (1983). Parameters are estimated in Tox_Risk using the method of maximum likelihood.
10 Tumor types can be categorized by tumor context as either fatal or incidental. Incidental
11 tumors are those tumors thought not to have caused the death of an animal, whereas fatal tumors
12 are thought to have resulted in animal death. Tumors at all sites were treated as incidental
13 (although it was recognized that this may not have been the case, the experimental data are not
14 detailed enough to conclude otherwise). The parameter t0 was set equal to 0 because there were
15 insufficient data to reliably estimate it.
16 The likelihood-ratio test was used to determine the lowest value of the multistage
17 polynomial degree k that provided the best fit to the data while requiring selection of the most
18 parsimonious model. The one-stage Weibull (i.e., k = 1) was determined to be the most optimal
19 value for all the tumor types analyzed.
20 3. Select the POD and calculate the unit risk for each tumor site. The effective
21 concentration that causes a 10% extra risk for tumor incidence, EC10, and the 95% lower bound of
22 that concentration, LEC10, are derived from the dose-response model. The LEC10 is then used as
23 the POD for a linear low-dose extrapolation, and the unit risk is calculated as 0.1/LEC10. This is
24 the procedure specified in the EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA,
25 2005a) for agents such as EtO that have direct mutagenic activity. See Section 3.4 for a
26 discussion of the mode of action for EtO. Tables 1 through 3 present the unit risk estimates for
27 the individual tumor sites in each bioassay.
28 4. Develop a unit risk estimate based on the incidence of all tumors combined. This
29 method assumes that occurrences of tumors at multiple sites are independent and, further, that the
30 risk estimate for each tumor type is normally distributed. Then, at a given exposure level, the
31 maximum likelihood estimates (MLEs) of extra risk due to each tumor type are added to obtain
1 the MLE of total cancer risk. The variances corresponding to each tumor type are added to give
2 the variance associated with the sum of the MLEs. The one-sided 95% UCL of the MLE for the
3 combined risk is then calculated as:
95% UCL = MLE + 1.645s.e.,
08/29/06 53 DRAFT—DO NOT CITE OR QUOTE
-------
4 where s.e. is the standard error and is the square root of the summed variance. (Note that as a
5 precursor to this step, when Tox _Risk is used to fit the incidence of a single tumor type, it
6 provides the MLE and 95% UCL of extra risk at a specific dose. The standard error in the MLE is
7 determined using the above formula). The calculation is repeated for a few exposure levels, and
8 the exposure yielding a value of 0.1 for the upper bound on extra risk is determined by
9 interpolation. The unit risk is then the slope of the linear extrapolation from this POD. The
10 results are given in Table 14.
11
12 4.2.4. Description of Experimental Animal Studies
13 NTP (1987) exposed male and female B6C3Fj mice to concentrations of 0, 50, and 100
14 ppm for 6 hours per day, 5 days per week, for 102 weeks. An elevated incidence of lung
15 carcinomas was found in males, and elevated lung carcinomas, malignant lymphomas, uterine
16 adenocarcinomas, and mammary carcinomas were found in females. These data are shown in
17 Table 1.
18 Lynch et al. (1982, 1984) exposed male F344 rats to 0, 50, and 100 ppm for 7 hours per
19 day, 5 days per week, for 2 years. They found excess incidence of tumors at three sites:
20 mononuclear cell leukemia in the spleen, testicular peritoneal mesothelioma, and brain glioma. In
21 this study the survival in the high-dose group (19%) was less than that of controls (49%), which
22 reduced the incidence of leukemias. In the animals in the high-dose group that survived to
23 the termination of the experiment, the incidence of leukemias was statistically significantly higher
24 than for controls (/?<0.01). The incidence data are shown in Table 2, unconnected for the high-
25 dose-group mortality. If the individual animal data were available to perform the correction, the
26 incidence would be higher. Therefore, using these data results in an underestimate of risk.
27 Snellings et al. (1984) exposed male and female F344 rats to 0, 10, 33, and 100 ppm for 6
28 hours per day, 5 days per week, for 2 years and described their results for all sites except the brain.
29 In two subsequent publications for the same study, Garman et al. (1985, 1986) described the
30 development of brain tumors in a different set of animals. The Snellings et al. publication
31 reported an elevated incidence of splenic mononuclear cell leukemia and peritoneal mesothelioma
32 in males and an elevated incidence of splenic mononuclear cell leukemia in females. The
1 mortality was higher in the 100 ppm groups than the other three groups for both males and
2 females. The incidence in the animals killed after 24 months is shown in Table 3a. The two
3 Garman et al. publications describe brain tumors in males and females (Table 3b). The brain
4 tumor incidence was much lower than that of the other tumors.
5
6
7
08/29/06 54 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.2.5. Results of Data Analysis of Experimental Animal Studies
2 The unit risks calculated from the individual site-sex-bioassay data sets are presented in
3 Tables 1 through 3. The highest unit risk of any individual site is 3.23 x 10~5 per |ig/m3, and it is
4 for mononuclear cell leukemia in the female rats of the Snellings et al. (1984) study.
5 Table 17 presents the results of the time-to-tumor method applied to the individual
6 animals in the NTP bioassay, compared with the results from the dose group incidence data in
7 Table 1. This comparison was done for each tumor type separately. The time-to-tumor method of
8 analyzing the individual animals results in generally higher unit risk estimates than does the
9 analysis of dose group data, as shown in Table 17. The ratio is not large (less than 2.2) across the
10 tumor types. (In the case of mammary tumors this ratio is actually less than 1. It must be noted
11 that the incidence at the highest dose [where the incidence was substantially less than at the
12 intermediate dose] was deleted from the analysis of grouped data, whereas it was retained in the
13 time-to-tumor analysis. Therefore, the comparison for the mammary tumors is not a strictly valid
14 comparison of methods.) The results also show the extent to which a time-to-tumor analysis of
15 individual animal data increases the risk estimated from data on dose groups. It is expected that if
16 individual animal data were available for the Lynch et al. (1982, 1984) and the Snellings et al.
17 (1984) bioassays, then the time-to-tumor analysis would also result in higher estimates because
18 both those studies also showed early mortality in the highest dose group.
19 The results of combining tumor types are summarized in Table 16. The sums of the
20 individual unit risks tabulated in Tables 1 to 3 are given in the second row of Table 16. Note that
21 as expected they are greater than the unit risks computed from the upper bound on the sum of
22 risks for all data sets except for the Lynch et al. (1982, 1984) data. The reason for this exception
23 is not known, but the differences are small. It is likely that the problem arises from the
24 methodology used to combine the risks across tumor sites. In an attempt to be consistent with the
25 new two-step methodology (i.e., modeling in the observable range to a POD and then doing a
26 linear extrapolation to 0 extra risk at 0 exposure), the exposure concentration at which the sum of
27 the independent tumor site risks yielded a 95% upper bound on 10% extra risk was estimated and
28 used as the POD. Summing risks in this way results in a POD for the combined tumor risk that is
29 different (lower) than the points of departure for each individual tumor site risk. Thus, the risk
30 estimate for the sum is not strictly comparable to the individual risks that constitute it. These
31 tumor-site-specific risks were based on points of departure individually calculated to correspond
32 with a 10% extra risk. In any event, adding the upper bound risks of individual tumor sites should
33 overestimate the upper bound of the sum, and the latter is the preferred measure of the total cancer
34 risk since it avoids the overestimate. However, for the exceptional Lynch et al. (1982, 1984) data,
35 the sum of upper bounds, 3.66 x 10"5 per |ig/m3, is already an overestimate of the total risk, and
36 this value is preferred over the anomalously high value of 4.17 x 10"5 per |ig/m3 corresponding to
08/29/06 55 DRAFT—DO NOT CITE OR QUOTE
-------
1 the upper bound on the sum of risks. The latter value is considered to be an excessive
2 overestimate and is therefore not carried over into the summary Table 18. For the Snellings et al.
3 (1984) data sets, the upper confidence bound on the sum of risks is used in the summary Table 18.
4 The results of the sum-of-risks calculations on the NTP bioassay time-to-tumor data are included
5 in the third row of Table 16. The estimate for the NTP females is 4.55 x 10~5 per |-ig/m3, which is
6 higher than the other two measures of total tumor risk in that bioassay. This value is preferable to
7 the other measures because it utilizes the individual animal data available for that bioassay.
8 Summary of results. The summary of unit risks from the five data sets is shown in Table
9 18. The data set giving the highest risk (4.55 x 10'5 per |_ig/m3) is the NTP (1987) data on
10 combined tumors in females. The other values are within about a factor of 2 of the highest value.
11
12 4.3. INHALATION UNIT RISK ESTIMATES—CONCLUSIONS
13 For both humans and laboratory animals, tumors occur at multiple sites. In humans, there
14 was a combination of tumors having lymphohematopoietic origins in males and breast cancer in
15 females, and in rodents both lymphohematopoietic tumors and tumors of other sites were
16 observed. From human data, an extra cancer unit risk estimate of 9.0 x 10"4 per |_ig/m3 (1.64 per
17 ppm) was calculated for lymphohematopoietic cancer incidence in males, and a unit risk estimate
18 of 5.0 x 10"4 per [ig/m3 (0.91 per ppm) was calculated for breast cancer incidence in females.
19 [The male lymphohematopoietic cancer incidence risk estimate was notably similar to the
20 estimate of 1.20 per ppm based on the subcategory of "lymphoid" cancers in males.] Unit risk
21 estimates derived from the three chronic rodent bioassays for EtO ranged from 2.2 x 10~5 per
22 |-ig/m3 to 4.6 x io~5 per |-ig/m3, about an order of magnitude lower than the estimates based on
23 human data.
24 The Agency takes the position that human data, if adequate data are available, provide a
25 more appropriate basis than do rodent data for estimating human risks (U.S. EPA, 2005a),
26 primarily because uncertainties in extrapolating quantitative risks from rodents to humans are
27 avoided. Although there is a sizeable difference between the rodent-based and the human-based
28 estimates, the similarity between the unit risk estimates based on the female breast cancer and
29 male lymphohematopoietic cancer results increases confidence in the use of the estimate based on
30 the male lymphohematopoietic cancers. Furthermore, the human data are from a large, high-
31 quality study, with EtO exposure estimates for the individual workers and little reported exposure
32 to chemicals other than EtO. Therefore, the extra cancer unit risk estimate of 1.64 per ppm (9.0 x
33 10"4 per |_ig/m3) calculated for lymphohematopoietic cancer incidence in males is the preferred
34 estimate for this assessment. This unit risk estimate is greater than the estimate of 0.91 per ppm
35 (5.0 x 10"4 per [ig/m3) calculated from the results of a breast cancer incidence study of the same
36 worker cohort, and is thus recommended as the potency estimate for Agency use. Although there
08/29/06 56 DRAFT—DO NOT CITE OR QUOTE
-------
1 was no clear exposure-response relationship for lymphohematopoietic cancer in females in the
2 Steenland et al. (2004) study, an increased risk in females cannot be ruled out. Nonetheless, the
3 Steenland et al. results suggest that if such a risk exists for females, it is likely to be lower than the
4 risk estimated for males. Thus, it is expected that the risk estimated based on
5 lymphohematopoietic cancer in males would be protective for females even if females have an
6 increased risk for both breast cancer and lymphohematopoietic cancer.
7 Because a mutagenic mode of action for EtO carcinogenicity (see Section 3.3.2) is
8 "sufficiently supported in [laboratory] animals" and "relevant to humans", and as there are no
9 chemical-specific data to evaluate the differences between adults and children, increased early-life
10 susceptibility should be assumed and, if there is early-life exposure, the age-dependent adjustment
11 factors (ADAFs) should be applied, as appropriate, in accordance with EPA's Supplemental
12 Guidance (U.S. EPA, 2005b). Consequently, applying the ADAFs to obtain a full lifetime unit
13 risk estimate yields
14
15 1.64/ppm x ((10 x 2 years/70 years) + (3 x 14/70) + (1 x 54/70))
16 = 2.72/ppm = 1.5 x 10-3/(ng/m3).
17
18 Using the above full lifetime unit risk estimate of 2.72 per ppm, the lifetime chronic exposure
19 level of EtO corresponding to an increased cancer risk of 10"6 can be estimated as follows:
20
21 (10~6)/(2.72/ppm) = 4 x 1Q'7 ppm = 0.0004 ppb = 0.0007 ng/m3.
22
23 [Note that for less-than-lifetime exposures scenarios (or for exposures that vary with age), the
24 adult-based estimate of 1.64 per ppm should be used, but, if there is early-life exposure, the
25 ADAFs should be applied in accordance with EPA's Supplemental Guidance. With respect to the
26 breast cancer estimates, evidence suggests that puberty/early adulthood is a particularly
27 susceptible lifestage for breast cancer (U.S. EPA, 2005b; Russo and Russo, 1999); however, EPA
28 has not, at this time, developed alternate ADAFs to reflect such a pattern of increased early-life
29 susceptibility, and there is currently no EPA guidance on an alternate approach for adjusting for
30 early-life susceptibility to potential breast carcinogens.]
31 If the linear regression model of the categorical cumulative exposure data used to derive
32 the POD is suitable, the inhalation unit risk estimate presented above, which is calculated based
33 on a linear extrapolation from the POD (LEC01), is expected to provide an upper bound on the risk
34 of cancer incidence. However, for certain applications, such as benefit-cost analyses, estimates of
35 "central tendency" for the risk below the POD are desired. Because a linear regression model was
36 used in the observable range of the human data and the POD was within the low-dose linear range
08/29/06 57 DRAFT—DO NOT CITE OR QUOTE
-------
1 for extra risk as a function of exposure, linear extrapolation below the LEC01 is a straight
2 continuation of the 95% UCL on the linear model used above the LEC01. This is illustrated in
3 Tables 6 and 7, where the 95% UCL on extra risk for lymphohematopoietic cancer mortality in
4 males is about 0.92 per ppm for exposures at or below about 0.01 ppm (Table 6), which is
5 equivalent to the mortality unit risk estimate of 0.92 per ppm derived from the LEC01 in Table 7
6 (rounded to two figures). The same holds for the central tendency (weighted least squares)
7 estimate of the mortality extra risk from the model (0.42 per ppm from Table 6 and 0.01 extra risk
8 divided by the EC01 of 0.0238 in Table 7 yields 0.42 per ppm, or 2.3 x 10'4 per |ig/m3).
9 The same also holds for the incidence estimates (i.e., the same model is used above and
10 below the POD); thus, one can calculate a central tendency estimate of the extra risk of
11 lymphohematopoietic cancer incidence in males for the linear model below the EC01 using the
12 results in Table 7 by calculating 0.01/EC01, or 0.017(0.013 ppm), which yields 0.77 per ppm (4.2 x
13 10"4 per |ig/m3). Similarly, because the same methodology was used for the breast cancer
14 calculations, one can calculate central tendency estimates for the linear model using the EC01
15 results in Tables 12 and 14, yielding a central tendency extra risk estimate of 0.26 per ppm (1.4 x
16 10"5 per |ig/m3) for breast cancer mortality and 0.42 per ppm (2.3 x 10"4 per |ig/m3) for breast
17 cancer incidence (invasive and in situ, based on results from subcohort with interviews). These
18 central tendency estimates are dependent on the suitability of the linear regression models to
19 reflect the lower end of the observable range as well as on the applicability of the linear models
20 below the observable range. The assumption of low-dose linearity is supported by the
21 mutagenicity of EtO (see Section 3.4). [If these central tendency estimates were to be used for
22 cost-benefit analyses or some other purpose, ADAFs should be applied, as appropriate, in
23 accordance with EPA's Supplemental Guidance, as discussed above.]
24
25 4.4. COMPARISON WITH OTHER ASSESSMENTS
26 The unit risk values derived in this document are compared with those of other
27 assessments in Table 19. One assessment is based on human data (EOIC, 2001; Kirman et al,
28 2004), and four assessments are based on laboratory animal data (California EPA, 1999; Health
29 Canada, 2001; EOIC, 2001/Kirman et al., 2004; and a study by Granath et al., 1999). The
30 comparisons are discussed in the sections below.
31
32 4.4.1. Assessments Based on Human Studies
33 The EOIC (2001) document uses human leukemia data only and pools data from both the
34 Stayner et al. (1993) and the UCC studies (Teta et al., 1993, 1999). Based on the assumption that
35 leukemias are due to chromosome translocations and that translocations require two independent
08/29/06 58 DRAFT—DO NOT CITE OR QUOTE
-------
1 events (chromosome breaks), the EOIC used a dose-squared model and derived a unit risk value
2 of 1.8 x 10~7 (jig/m3)"1 as the most appropriate value, with a range of 1.8 x 10~8 (jig/m3)"1 to
3 5.3 x 10~7 (jig/m3)"1 representing the uncertainty of this estimate. The EOIC document was
4 published as Kirman et al. (2004) with the unit risk estimate recalculated as 4.5 x 10~8 (jig/m3)"1,
5 with a range of values of 1.4 x 1Q'8 (jig/m3)'1 to 1.4 x 1Q'7 (jig/m3)-1.
6 The EOIC/Kirman et al. values are different from those in the current document because of
7 the different assumptions inherent in the EOIC's approach and because the study used
8 unpublished data from the two cohorts, which was necessary in order to combine the two data
9 sets. A key difference is that EPA uses a linear model rather than a quadratic (dose-squared)
10 model in the range of observation. Then, EPA uses a higher POD (1% extra risk), whereas
11 Kirman et al. used a POD of 10"5 for their best estimate and a risk range of 10"4 to 10"6 for their
12 range of values. The POD is not critical with a linear model, but with the quadratic model used
13 by EOIC/Kirman et al., the lower the POD, the greater the impact of the quadratic model and the
14 lower the resulting unit risk estimates.
15 In addition, EPA (1) includes all lymphohematopoietic tumor types likely to be caused by
16 EtO and does not restrict the data to leukemia alone, (2) includes ages up to 85 years in the
17 lifetable analysis rather than stopping at 70 years, (3) calculates unit risk estimates for cancer
18 incidence as well as mortality, (4) uses a lower bound as the POD rather than the maximum
19 likelihood estimate, and (5) uses the results of lagged analyses rather than unlagged analyses.
20 Another key difference is that the EOIC and Kirman et al. relied on earlier NIOSH results
21 (Stayner et al., 1993), whereas EPA uses the results of NIOSH's more recent follow-up of the
22 cohort (Steenland et al., 2004). Kirman et al. (2004) claim that a quadratic dose-response model
23 provided the best fit to the data in the observable range and that this provides support for their
24 assumed mode of action. However, the 2004 NIOSH data for lymphohematopoietic cancer in
25 males suggest a supralinear exposure-response relationship (see Section 4.1.1.2 and Figure 4),
26 which is inconsistent with a dose-squared model. Furthermore, EPA's review of the mode of
27 action evidence does not support the mode of action assumed by EOIC/Kirman et al. (see Section
28 3.4).
29
30 4.4.2. Assessments Based on Laboratory Animal Studies
31 The California EPA summarized cancer potency factors for EtO as well as 118 other
32 carcinogenic substances in a 1999 report (CalEPA, 1999). The EtO value is based on the
33 incidence of mononuclear cell leukemia in female rats in the Snellings et al. (1984) rat bioassay
34 because it was the most sensitive data set of all the sex, site, and species data sets available. The
35 95% upper bound value from a linear extrapolation procedure using these data gave a unit risk
36 value of 8.8 x 10'5 (ng/m3)'1.
08/29/06 59 DRAFT—DO NOT CITE OR QUOTE
-------
1 Health Canada (2001) based its carcinogenic hazard index value on the rodent data set
2 with the lowest ED05 value of the data sets in Tables 1 through 3, which had a lower confidence
3 limit of 1,500 |ig/m3. Using linear extrapolation, this is equivalent to an upper-bound risk of
4 0.05/1,500 = 3.3 x 10"5 (jig/m3)"1. The human data then available were not used because Health
5 Canada felt that the number of cases was relatively small and the uncertainty in the estimate was
6 relatively large.
7 The EOIC (2001) and Kirman et al. (2004) also used linear and dose-squared extrapolation
8 models to derive unit risk estimates based on the rat mononuclear cell leukemia data and the
9 mouse lymphoma data. First, they used the multistage model to calculate the LEC10 (LEC01 for
10 the male mouse lymphoma data) for the POD from the observable range. Then, using these PODs
11 for linear extrapolation, Kirman et al. obtained a unit risk range of 3.9 x 10"6 (jig/m3)"1 to 1.5 x
12 10 5 (jig/m3)"1. Alternatively, Kirman et al. used a quadratic extrapolation model below the
13 observable range to estimate secondary points of departure (LEC01-LEC000001; LEC001-LEC000001
14 for the male mouse) for final linear low-dose extrapolation, yielding unit risks ranging from 2.6 x
15 10~8 (jig/m3)"1 to 4.9 x 1Q~6 (jig/m3)"1. These values are all smaller than the unit risks derived
16 from the rodent data in this document and in the Health Canada document.
17 Another estimate was derived by Granath et al. (1999). They showed that the same rodent
18 data that we have used here fit a multiplicative risk model but do not fit an additive risk model.
19 They concluded from their multiplicative risk model that the doubling dose (the air concentration
20 necessary to double the background incidence of tumors, a measure that is assumed to be the same
21 for humans and rodents) is 22 ppm (95% confidence limits of 14-32 ppm) from the rat data and
22 19 (8-46) ppm from the mouse data. This is equivalent to a relative risk of approximately 2 per
23 20 ppm = 5.6 x 10"5 (jig/m3)"1. Using this relative risk and a background hematopoietic cancer
24 lifetime risk of 1.8 x 10"2, one obtains a unit risk of 1 x 1Q"6 (iig/m3)"1. This is 38 times smaller
25 than the total cancer unit risk estimate derived from the laboratory animal data in this document
26 and greater than 2 orders of magnitude smaller than the human estimate.
27 The multiplicative risk model has not been used in Agency carcinogen risk estimates that
28 have been based on data from laboratory animals, because there is seldom a direct tumor site
29 concordance between humans and rodents, and because the specific human tumor type must be
30 specified for this method. Kuo et al. (2002) examined the difference between additive risk models
31 and multiplicative risk models in predicting the results between mouse and rat data sets in the
32 NTP database and found that in most cases the differences between the two models were smaller
33 than the differences between chemical classes. Therefore, there is no compelling reason to use the
34 multiplicative risk model with laboratory animal data in general or for the case of EtO
35 specifically.
36
08/29/06 60 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3 4.5. RISK ESTIMATES FOR OCCUPATIONAL EXPOSURES
4 The unit risk estimates derived in the preceding sections were developed for
5 environmental exposure levels, where maximum modeled levels are on the order of 1-2 |ig/m3 (e-
6 mail dated October 3, 2005, from Mark Morris, U.S. EPA, to Jennifer Jinot, U.S. EPA), and are
7 not applicable to occupational exposures. However, occupational exposure levels are of concern
8 to EPA when EtO is used as a pesticide (e.g., fumigant for spices). Therefore, extra risk estimates
9 were calculated for a number of occupational exposure scenarios. For these occupational
10 scenarios, exposure-response models from the NIOSH cohort were used in conjunction with the
11 life-table program, as previously discussed in Section 4.1. A 35-year exposure occurring between
12 ages 20 and 55 years was assumed, and exposure levels ranging from 0.1 to 1 ppm 8-hour TWA
13 were examined (i.e., ranging from about 1,300 to 13,000 ppm x days).
14 For lymphohematopoietic cancer mortality in males, the best-fitting (natural) log
15 cumulative exposure Cox regression model (Steenland et al., 2004; see also Section 4.1.1.2),
16 lagged 15 years, was used. For lymphohematopoietic cancer incidence in males, the exposure-
17 response relationship was assumed to be the same as for mortality (see Section 4.1.1.3). The extra
18 risk results for lymphohematopoietic cancer mortality and incidence in males are presented in
19 Table 20. As can be seen in Table 20, the extra risks for these occupational exposure levels are in
20 the "plateau" region of the exposure-response relationships and increase less than proportionately
21 with exposure. [For occupational exposures less than about 1,000 ppm x days, or about 0.08 ppm
22 8-hour TWA for 35 years, risk estimates are no longer in the plateau region (see Figure 4) but
23 rather in a region of greater uncertainty for the log cumulative exposure model, and one might
24 want to use the linear regression of the categorical results that was used for environmental
25 exposures (see Section 4.1.1.2). Furthermore, if one is using the linear model in this range and
26 also estimating risks for exposure levels in the range between about 0.08 and 0.5 ppm (or where
27 the linear and log cumulative exposure Cox regression models meet) 8-hour TWA, one might
28 want to use the linear model for the entire range up to 0.5 ppm 8-hour TWA to avoid a
29 discontinuity between the two models; thus, results for the linear model for exposure levels up to
30 0.5 ppm 8-hour TWA are also presented in Table 20. While the best-fitting model would
31 generally be preferred in the 0.08 and 0.5 ppm 8-hour TWA exposure range, there is model
32 uncertainty, so the use of either model could be justified. For exposures higher than where the
33 linear and log cumulative exposure Cox regression models meet, the log cumulative exposure
34 model exclusively is recommended.]
35 For breast cancer mortality, the best-fitting (natural) log cumulative exposure Cox
36 regression model (Steenland et al., 2004; see also Section 4.1.2.2), lagged 20 years, was used.
08/29/06 61 DRAFT—DO NOT CITE OR QUOTE
-------
1 The breast cancer mortality risk estimates are presented merely for comparison; the breast cancer
2 incidence risk estimates are preferred because incidence estimates are the objective, and because
3 the incidence risk estimates are based on more cases and the incidence data (for the subcohort
4 with interviews) are adjusted for a number of breast cancer risk factors (see Section 4.1.2.3). In
5 terms of the incidence data, the subcohort data are preferred to the full cohort data because the
6 subcohort data are adjusted for these potential confounders and also because the full cohort data
7 suffer from incomplete ascertainment of breast cancer cases. For breast cancer incidence in the
8 subcohort with interviews, a number of Cox regression exposure-response models fit almost
9 equally well (Steenland et al., 2003; see also Section 4.1.2.3 and Table 13). These include a log
10 cumulative exposure model and a cumulative exposure model, both with a 15-year lag, and a log
11 cumulative exposure model with no lag. Steenland et al. (2003) also provide a duration of
12 exposure model with a marginally better fit; however, models using duration of exposure are less
13 useful for estimating exposure-related risks, and duration of exposure and cumulative exposure
14 are correlated, thus, only the cumulative exposure models are considered here.
15 The extra risk results for breast cancer incidence in females from the cumulative exposure
16 models listed above are presented in Table 21. Of these cumulative exposure models, the lagged
17 models are preferred because the inclusion of a 15-year lag for the development of breast cancer
18 seems more biologically realistic than not including a lag. As can be seen in Table 21, the extra
19 risk estimates for the lagged log cumulative and cumulative exposure models differ substantially.
20 Furthermore, the categorical results for breast cancer incidence in the subcohort with interviews
21 suggest that, for the lowest four exposure quintiles, the log cumulative exposure model
22 overestimates the RR, while the cumulative exposure model generally underestimates the RR,
23 with the categorical results largely falling between the RR estimates of those two models (see
24 Figure 8). [The lowest four exposure quintiles represent individual worker exposures ranging
25 from 0 to about 15,000 ppm x days, which covers the range of cumulative exposures for the
26 occupational exposure scenarios of interest in this assessment.] Therefore, the linear regression of
27 the categorical results for the lowest four exposure quintiles in the subcohort with interviews was
28 again used (see Section 4.1.2.3). Extra risk estimates using this linear regression model are also
29 presented in Table 21 and are the preferred estimates because, in the absence of a clearer "best-
30 fitting" model for the continuous data, the linear regression best represents the categorical RR
31 results for exposures below about 15,000 ppm x days. [For occupational exposures above this
32 level, the Steenland et al. log cumulative exposure Cox regression model (with a 15-year lag),
33 which additionally reflects the highest exposure quintile, would be preferred; however, for
34 estimating risks from future exposures, such high cumulative exposures are unlikely to occur
35 because the current Occupational Safety and Health Administration Permissible Exposure Limit is
36 1 ppm (8-hour TWA).]
08/29/06 62 DRAFT—DO NOT CITE OR QUOTE
-------
1 Extra risk estimates for a 45-year exposure to the same exposure levels were nearly
2 identical to those from the 35-year exposure for both lymphohematopoietic cancer in males and
3 breast cancer in females (results not shown). With the 15-year lag, the assumption of an
4 additional 10 years of exposure only negligibly affects the risks above age 70 and has little impact
5 on lifetime risk. For exposure scenarios of 35-45 years but with 8-hour TWAs falling between
6 those presented in the Tables, one can estimate the extra risk by interpolation. For exposure
7 scenarios with durations of exposure less than 30-35 years, one could roughly estimate extra risk
8 by calculating the cumulative exposure and finding the extra risk for a similar cumulative
9 exposure in Table 20 (or 21). For a more precise estimation, or for exposure scenarios of much
10 shorter duration or for specific age groups, one should do the calculation using the lifetable
11 presented in Appendix C.
08/29/06 63 DRAFT—DO NOT CITE OR QUOTE
-------
Table 1. Tumor incidence data in National Toxicology Program Study of
B6C3F! mice (NTP, 1987)a
Gender/tumor type
EtO concentration
(time-weighted average)b
0 ppm
50 ppm
(16.3 mg/m3)
100 ppm
(32.7 mg/m3)
EC10
(LEClor,
[mg/m3]
Unit risk
(0.1/LEC10)
[per mg/m3]
Males
Lung adenomas plus
carcinomas
11/49
19/49***
26/49***
6.94
(4.51)
2.22 x ID'2
Females
Lung adenomas plus
carcinomas
Malignant
lymphoma
Uterine
carcinoma
Mammary
carcinoma*
2/44
9/44
0/44
1/44
5/44
6/44
1/44
8/44*
22/49***
22/49*
5/49*
6/49
14.8
(9.12)
21.1
(13.9)
32.8
(23.1)
9.69
(5.35)
1.1 x IQ-2
7.18 x lO'3
4.33 x lO'3
1.87x 10'2
a Incidence data were adjusted by eliminating the animals that died prior to the occurrence of the first tumor or prior
to 52 weeks, whichever was earlier.
b Adjusted to continuous exposure from experimental exposure conditions of 6 hours/day, 5 days/week; 1 ppm =
1.83 mg/m3.
c Calculated using Tox_Risk program.
d Highest dose was deleted while fitting the dose-response data.
*p<0.05 (pairwise Fisher's exact test).
**p<0.01 (pairwise Fisher's exact test).
***p<0.001 (pairwise Fisher's exact test).
08/29/06
64
DRAFT—DO NOT CITE OR QUOTE
-------
Table 2. Tumor incidence data in Lynch et al. (1982,1984) study of male F344 rats
Tumor type
Splenic
mononuclear
cell leukemia0
Testicular
peritoneal
mesothelioma
Brain mixed-
cell glioma
Concentration (time-weighted average)"
0 ppm
24/77
3/78
0/76
50 ppm
(19.1 mg/m3)
38/79*
9/79
2/77
100 ppm
(38.1 mg/m3)
30/76
21/79**
5/79**
EC10
(LEC10)b,
[mg/m3]
7.11
(3.94)
16.7
(11.8)
65.7
(37.4)
Unit risk
(0.1/LEC10)
[per mg/m3]
2.54 x IQ-2
8.5 x ID'3
2.68 x ID'3
a Adjusted to continuous exposure from experimental exposure conditions of 7 hours/day, 5 days/week; 1 ppm =
1.83 mg/m3.
b Calculated using Tox_Risk program.
c Highest dose deleted while fitting the dose-response data.
*p<0.05 (pairwise Fisher's exact test).
**p<0.01 (pairwise Fisher's exact test).
08/29/06
65
DRAFT—DO NOT CITE OR QUOTE
-------
Table 3. Tumor incidence data in Snellings et al. (1984) and Garman et al. (1985) reports on F344 rats3
Gender/tumor type
Concentration (time-weighted average)6
Oppm"
10 ppm
(3.27 mg/m3)
33 ppm
(10.8 mg/m3)
100 ppm
(32.7 mg/m3)
EC10
(LEC10)<
[mg/m3]
Unit risk (0.1/LEC10)
[per mg/m3]
Males
Splenic mononuclear cell
leukemia
Testicular peritoneal
mesothelioma
Primary brain tumors
13/97
(13%)e
2/97
(21%)
1/181
(0.55%)
9/51
(18%)
2/51
(3.9%)
1/92
(1.1%)
12/39*
(32%)
4/39
(10%)
5/85*
(5.9%)
9/30*
(30%)
4/30*
(13%)
7/87**
(8.1%)
12.3
(6.43)
22.3
(11.6)
36.1
(22.3)
1.56 x lO'2
8.66 x 10'3
4.5 x ID'3
Females
Splenic mononuclear cell
leukemia
Primary brain tumors
11/116
(9.5%)
1/188
(0.53%)
11/54*
(21%)
1/94
(1.1%)
14/48**
(30%)
3/92
(3.3%)
15/26***
(58%)
4/80*
(5%)
4.46
(3.1)
63.8
(32.6)
3.23 x ID'2
3.07 x lO'3
Oi
Oi
O
O
O
HH
s
O
&
O
c
O
H
W
a Denominators refer to the number of animals for which histopathological diagnosis was performed. For brain tumors Garman et al. (1985) included animals
in the 18-month and the 24-month sacrifice and found dead or euthanized moribund of those alive at the time of the first brain tumor, whereas for the other sites
Snellings et al. (1984) included animals only at the 24-month sacrifice.
b Adjusted to continuous exposure from experimental exposure conditions of 6 hours/day, 5 days/week; 1 ppm = 1.83 mg/m3.
0 Using Tox_Risk program.
d Results for both control groups combined.
e Numbers in parentheses indicate percent incidence values.
^H
O * /?<0.05 (pairwise Fisher's exact test).
** p<0.0l (pairwise Fisher's exact test).
***£><0.001 (pairwise Fisher's exact test).
-------
Table 4. Cytogenetic effects in humans
Number exposed
(number of controls)
33(0)
Site I: 13
Site II: 22
Site III: 25-26
(171 total)
12 (12)
14 (14)
Factory I: 18
Factory II: 10
(20 total)
15 smokers (7)
10 nonsmokers (15)
10 (10)
Low dose: 9 (48)
High dose: 27 (10)
34 (23)
1 1 smokers
14 nonsmokers
(10 total)
75 (22)
56 (141)
22 (22)
19 (19)
10 (10)
Exposure time
(years)
Range
1-14
0.5-8
0.5-8
0.5-10
0.5-10
3-14
1-10
0.6^
1.5-15
Mean
3.2
1.7
5.7
4.5
3
4
15
8e
7
o
5
6.8
Ethylene oxide level in air
(ppm)a
Range
±0.05-8
0.5C
5-10c
5-20c
±36
<0.07-4.3C
20-123
20-123
60-69C
2.7-10.9
2.7-82
<0.1-2.4C
0.5^17f
0.5-208f
2-5c
1^0C
0.2-0.5C
3.7-20c
0-9.3C
Mean (TWA)
±0.01b
<1
<1
2.7
5.5
<0.3
0.35
10.7
1.84
Cytogenetic observations
CA
(+)
+
+
+
+
+
+
-
+
+
(+)
+
SCE
+
+
+
-
-
+
+
+
+
+
-
+
+
+
+
MN
+d
-
+
Reference
Clare etal. (1985)
Stolleyetal. (1984);
Galloway etal. (1986)
Garry etal. (1979)
Hansen etal. (1984)
Hogstedt etal. (1983)
Laurent etal. (1984)
Lerda and Rizzi ( 1992)
Major etal. (1996)
Mayer etal. (1991)
Popp etal. (1994)
Ribeiro etal. (1994)
Richmond et al. (1985)
Sarto etal. (1984)
Sarto etal. (1987)
a\
O
O
2
O
H
O
O
&
O
c
O
H
W
-------
Table 4. Cytogenetic effects in humans (continued)
Number exposed
(number of controls)
9
3
(27 total)
5
5
(10 total)
32
11
(8 total)
9 hospital workers (8)
15 factory workers
(15)
7
7
7
(7 total)
Low exposure: 9
High exposure: 5
(13 total)
19
17
(35 total)
Exposure time
(years)
Range
0.5-12
0.1-1
4-12
2-6
3-27
Accidental
<5
>15
1-5
6-14
Mean
5
2
8.6
5.1
9.5
4
12
Ethylene oxide level in air
(ppm)a
Range
0.025-0.38°
>0.38B
<1-4A
0-0.3°
0.1 3-0.3c
20-25
17-33
28^129°
0.005-0.02
O.005-0.01
O.05-8
0.05-8
Mean (TWA)
0.025
0.38
0.04
0.16
13J
50 1J
O.05
0.05
Cytogenetic observations
CA
|
-
-
SCE
-
+
+
+
|
-
-
-
+
MN
_
+h
_ i
-1
-
-
+
-
-
-
Reference
Sartoetal. (1990)
Sartoetal. (1991)
Schulteetal. (1992)
Tatesetal. (1991)
Tatesetal. (1995)
Yager etal. (1983)
van Sittert et al. 1985
oo
o
o
2
o
H
O
O
&
O
c
o
H
W
-------
O
O
2
O
H
O
O
?d
O
c
O
H
W
Table 4. Cytogenetic effects in humans (continued)
a 1 ppm =1.83 mg ethylene oxide/m3.
b Calculated by linear extrapolation.
c TWA (8-hour).
d Positive for erythroblasts and polychromatic erythrocytes (negative for lymphocytes).
e Maximum years exposed.
f Peak concentrations.
g Exposed acutely from sterilizer leakage in addition to chronic exposure.
h Nasal mucosa.
1 Buccal cells.
J Average 6-month cumulative exposure (mg).
CA = chromosomal aberrations
MN = micronucleus
SCE = sister chromatid exchange
TWA = time-weighted average
-------
Table 5. Cox regression results for all lymphohematopoietic cancer mortality in males8
Exposure variable15
Cumulative exposure
Log cumulative
exposure, 15-year lag
Categorical cumulative
exposure, 15-year lag
p value
0.12
0.02
0.15
Coefficient (SE)
0.0000040
(0.0000022)
0.119(0.052)
ORs by category0 (95% CI)
1.00, 1.23 (0.32-4.73), 2.52
(0.69-9.22), 3.13 (0.95-10.37),
3.42(1.09-10.73)
a Based on 37 cases.
b Cumulative exposure is in ppm x days.
c Exposure categories are 0, >0-1,199, 1,200-3,679,3,680-13,499, > 13,500 ppm x days.
Source: Steenland et al. (2004), Table 6.
08/29/06
70
DRAFT—DO NOT CITE OR QUOTE
-------
Table 6. Extra risk estimates for all lymphohematopoietic cancer mortality in
males from various levels of lifetime exposure to ethylene oxide
Exposure
concentration
(ppm)
0.0001
0.001
0.01
0.1
1
10
Extra risk
Continuous log
cumulative
exposure
model"
4.70 x 1Q-3
1.24 x 1Q-2
2.25 x 1Q-2
3.55 x 1Q-2
5.22 x 1Q-2
7.36 x 1Q-2
Continuous
cumulative
exposure model
6.22 x 1Q-7
6.22 x 1Q-6
6.23 x 1Q-5
6.32 x 1Q-4
7.28 x 1Q-3
3.34 x 1Q-1
95% UCL on
extra risk
Categorical cumulative
exposure modelb
4.22 x 1Q-5
4.22 x 1Q-4
4.21 x 1Q-3
—
—
—
9.25 x 1Q-5
9.25 x 1Q-4
9.19x 1Q-3
—
—
—
a With 15-year lag.
b From linear regression of categorical results as described in text. The linear regression model is intended for low
exposures; therefore, results for 0.1 ppm lifetime exposure and greater are not presented.
08/29/06
71
DRAFT—DO NOT CITE OR QUOTE
-------
Table 7. EC01, LEC01, and unit risk estimates for lymphohematopoietic
cancer in males"
Model"
Cumulative
exposure,
Olag
Log
cumulative
exposure,
15 -year lag
Categorical0;
cumulative
exposure,
15 -year lag
Incidence
EC01
(ppm)
0.784
0.000130
0.0133
LEC01
(ppm)
0.412
0.0000570
0.00608
Unit risk
(per ppm)
2.43 x 1Q-2
175.4
1.64
Mortality
EC01
(ppm)
1.31
0.000517
0.0238
LEC01
(ppm)
0.686
0.000124
0.0109
unit risk
(per ppm)
1.46x 1Q-2
80.6
0.917
a From lifetime continuous exposure. Unit risk = 0.01/LEC01.
b From Steenland et al. (2004), Table 6, Cox regression models.
0 Regression coefficient derived from linear regression of categorical results, dropping the highest exposure group, as
described in Section 4.1.1.2.
08/29/06
72
DRAFT—DO NOT CITE OR QUOTE
-------
Table 8. Cox regression results for "lymphoid" cancer mortality in males"
Exposure variable15
Cumulative exposure
Log cumulative
exposure, 15-year lag
Categorical cumulative
exposure, 15-year lag
p value
0.06
0.02
0.13
Coefficient (SE)
0.0000050
(0.0000022)
0.138(0.061)
ORs by category0 (95% CI)
1.00, 0.90(0.16-5.24), 2.89
(0.65-12.86), 2.74 (0.65-11.55),
3.76(1.03-13.64)
a Based on 27 cases of NHL, myeloma, and lymphocytic leukemia.
b Cumulative exposure is in ppm x days.
c Exposure categories are 0, >0-1,199, 1,200-3,679, 3,680-13,499, > 13,500 ppm x days.
Source: Steenland et al. (2004), Table 7.
08/29/06
73
DRAFT—DO NOT CITE OR QUOTE
-------
Table 9. EC01, LEC01, and unit risk estimates for "lymphoid" cancer in malesa'b
Model0
Cumulative
exposure,
Olag
Log
cumulative
exposure,
15 -year lag
Categoricald;
cumulative
exposure,
15 -year lag
Incidence
EC01
(ppm)
0.796
0.000154
0.0216
LEC01
(ppm)
0.462
0.0000620
0.00836
Unit risk
(per ppm)
2.16x 1Q-2
161.3
1.20
Mortality
EC01
(ppm)
1.39
0.000885
0.0427
LEC01
(ppm)
0.809
0.000166
0.0165
unit risk
(per ppm)
1.24x 1Q-2
60.2
0.606
a From lifetime continuous exposure. Unit risk = 0.01/LEC01.
b "Lymphoid" cancers include NHL, myeloma, and lymphocytic leukemia.
0 From Steenland et al. (2004), Table 7, Cox regression models.
d Regression coefficient derived from linear regression of categorical results, dropping the highest exposure group, as
described in Section 4.1.1.2.
08/29/06
74
DRAFT—DO NOT CITE OR QUOTE
-------
Table 10. Cox regression results for breast cancer mortality in females"
Exposure variable15
Cumulative exposure
Log cumulative
exposure, 20-year lag
Categorical cumulative
exposure, 20-year lag
p value
0.34
0.01
0.07
Coefficient (SE)
0.0000049
(0.0000048)
0.084(0.035)
ORs by category0 (95% CI)
1.00, 1.76(0.91-3.43), 1.77
(0.88-3.56), 1.97 (0.94-4.06), 3.13
(1.42-6.92)
a Based on 103 cases of breast cancer (ICD-9 174,175).
b Cumulative exposure is in ppm x days.
c Exposure categories are 0, >0-646, 647-2,779, 2,780-12,321, > 12,322 ppm x days.
Source: Steenland et al. (2004), Table 8.
08/29/06
75
DRAFT—DO NOT CITE OR QUOTE
-------
Table 11. Extra risk estimates for breast cancer mortality in females from
various levels of lifetime exposure to ethylene oxide
Exposure
concentration (ppm)
0.0001
0.001
0.01
0.1
1
10
Extra risk
Continuous log
cumulative exposure
model3
3.60 x 1Q-3
9.66 x 1Q-3
1.70x 1Q-2
2.57 x 1Q-2
3.63 x 1Q-2
4.89 x 1Q-2
95% UCL on
extra risk
Categorical cumulative exposure model" b
2.60 x 1Q-5
2.60 x 1Q-4
2.60 x 1Q-3
—
—
—
5.15 x 1Q-5
5.15 x 1Q-4
5.14 x 1Q-3
—
—
—
a With 20-year lag.
b From linear regression of categorical results as described in text. The linear regression model is intended for low
exposures; therefore, results for 0.1 ppm lifetime exposure and greater are not presented.
08/29/06
76
DRAFT—DO NOT CITE OR QUOTE
-------
Table 12. EC01, LEC01, and unit risk estimates for breast cancer mortality in
femalesa
Model"
Log cumulative
exposure, 20-year lag
Categorical0; cumulative
exposure, 20-year lag
EC01
(ppm)
0.00112
0.0387
LEC01
(ppm)
0.000219
0.0195
Unit risk
(per ppm)
45.7
0.513
a From lifetime continuous exposure. Unit risk = 0.01/LEC01.
b From Steenland et al. (2004), Table 8, Cox regression models.
0 Regression coefficient derived from linear regression of categorical results, as described in Section 4.1.2.2.
08/29/06
77
DRAFT—DO NOT CITE OR QUOTE
-------
Table 13. Cox regression results for breast cancer incidence in females
a,b
Cohort
Full incidence
study cohort
n = 7,576
319 cases
Sub cohort with
interviews
n = 5,139
233 cases
Exposure variable0
Cumulative exposure,
15-year lag
Log cumulative
exposure, 15-year lag
Categorical cumulative
exposure, 15-year lag
Cumulative exposure,
15-year lag
Log cumulative
exposure, 15-year lag
Categorical cumulative
exposure, 15-year lag
Coefficient
(SE),p value
0.0000054
(0.0000035),
p=O.U
0.037 (0.019),
p= 0.05
0.0000095
(0.0000041),
p= 0.02
0.050 (0.023),
/?=0.03
ORs by category" (95% CI)
1.00, 1.07(0.72-1.59), 1.00
(0.67-1.50), 1.24(0.85-1.90),
1.17(0.78-1.78), 1.74
(1.16-2.65)
1.00, 1.06(0.66-1.71), 0.99
(0.61-1.60), 1.24(0.76-2.00),
1.42(0.88-2.29), 1.87
(1.12-3.10)
a Invasive breast cancer (ICD-9 174) and carcinoma in situ (ICD-9 233.0).
b Cases and controls matched on age and race (white/nonwhite). Full cohort models include cumulative exposure
and categorical variable for year of birth (quartiles). Subcohort models include cumulative exposure, categorical
variables for year of birth (quartiles), breast cancer in first-degree relative, and parity.
0 Cumulative exposure is in ppm x days.
d Exposure categories are 0, >0-647, 647-2,026, 2,026-4,919, 4,919-14,620, >14,620 ppm x days.
Source: Steenland et al. (2003), Tables 4 and 5.
08/29/06
78
DRAFT—DO NOT CITE OR QUOTE
-------
Table 14. ECn,, LECn,, and unit risk estimates for breast cancer incidence in females - invasive and in situa
-'01'
-01'
Model"
Cumulative exposure,
15 -year lag
Log cumulative
exposure, 15 -year lag
Categorical0;
cumulative exposure,
15 -year lag
With interviews
EC01
(ppm)
0.125
0.0000693
0.0238
LEC01
(ppm)
0.0732
0.0000399
0.0110
Unit risk
(per ppm)
0.137
251
0.909
Full cohort
EC01
(ppm)
0.220
0.000108
0.0466
LEC01
(ppm)
0.107
0.0000492
0.0174
Unit risk
(per ppm)
9.35 x 1Q-2
203
0.575
a All-cause mortality adjusted (to dying of something other than breast cancer or developing breast cancer). Unit risk = 0.01/LEC01.
b From Steenland et al. (2003), Tables 4 and 5, Cox regression models.
0 Regression coefficient derived from linear regression of categorical results, as described in Section 4.1.2.3.
O
O
2
O
H
O
O
&
O
c
O
H
W
-------
Table 15. EC01, LEC01, and unit risk estimates for breast cancer incidence in females - invasive onlya
Model"
Cumulative exposure,
15 -year lag
Log cumulative
exposure, 15 -year lag
Categorical0;
cumulative exposure,
15 -year lag
With interviews
EC01
(ppm)
0.150
0.0000894
0.0287
LEC01
(ppm)
0.0875
0.0000460
0.0132
Unit risk
(per ppm)
0.114
217
0.758
Full cohort
EC01
(ppm)
0.263
0.000153
0.0561
LEC01
(ppm)
0.127
0.0000593
0.0210
Unit risk
(per ppm)
7.87 x IQ-2
169
0.476
a All-cause mortality adjusted (to dying of something other than breast cancer or developing breast cancer). Unit risk = 0.01/LEC01.
b From Steenland et al. (2003), Tables 4 and 5, Cox regression models.
0 Regression coefficient derived from linear regression of categorical results, as described in Section 4.1.2.3.
oo
o
O
O
2
O
H
O
O
&
O
c
o
H
W
-------
Table 16. Upper-bound unit risks (per jig/m3) obtained by combining tumor
sites
Combination method"
U.c.b. on sum of risks0
Sum of unit risksd
Time-to-tumor analysis
and u.c.b on sum of risks0
NTP (1987)
female mouse
2.71 x 1Q-5
4.12x ID'5
4.55 x 1Q-5
Lynch et al.
(1982, 1984)
male rat
4.17 x ID'5
3.66 x ID'5
—
Snellings et al. (1984)b
Male rat
2.19 x ID'5
2.88 x 1Q-5
—
Female rat
3.37 x ID'5
3.54 x ID'5
—
a Unit risk in these methods is the slope of the straight line extrapolation from a point of departure at the dose
corresponding to a value of 0.1 for the 95% upper confidence bound on total extra risk.
b Includes data on brain tumors from the analysis by Garman et al. (1985). See Table 3.
0 U.c.b. = 95% upper confidence bound. At a given dose, the maximum likelihood estimate (MLE) of the
combined extra risk was determined by summing the MLE of risk due to each tumor type. The variance
associated with this value was determined by summing over the variances due to each tumor type.
d Sum of values in last column of Tables 1 through 3.
08/29/06
81
DRAFT—DO NOT CITE OR QUOTE
-------
Table 17. Unit risk values from multistage WeibulP time-to-tumor modeling
of mouse tumor incidence in the NTP (1987) study
Tumor type
Unit risk,
0.1/LEC10
(per |J,g/m3)
from time to
tumor analysis
Unit risk,
0.1/LEC10
(per jig/m3)
(Table l)b
Ratio of unit
risks
time-to-tumor/
grouped data
Males
Lung: alveolar/bronchiolar
adenoma and carcinoma
3.01 x 1Q-5
2.22 xlO'5
1.4
Females
Lung: alveolar/bronchiolar
adenoma and carcinoma
Malignant lymphoma
Uterine carcinoma
Mammary carcinoma
2.40 x 1Q-5
1.43 x 1Q-5
6.69 x 1Q-6
8.69 x 1Q-6
l.lOx 1Q-5
7.18 x 1Q-6
4.33 x 1Q-6
1.87 x 1Q-5
2.2
2.0
1.5
0.5
a P(d,t) = 1 - exp[-(q0 + Qid + q2d2 + ... + c^d1")*^ - to)2], where d is inhaled ethylene oxide concentration in ppm,
t is weeks until death with tumor. In all cases, k = 1 provided the optimal model.
b Incidence data modeled using multistage model without taking time to tumor into account.
08/29/06
82
DRAFT—DO NOT CITE OR QUOTE
-------
Table 18. Summary of unit risk estimates (per ng/m3) in animal bioassays
Assay
NTP (1987), B6C3FJ mice
Lynch et al. (1982, 1984), F344 rats
Snellings et al. (1984), F344 rats
Males
3.01 x lQ-5a
3.66 x 10-5c
2.19 x lQ-5d
Females
4.55 x lQ-5b
-
3.37x lQ-5d
a From time-to-tumor analysis of lung adenomas and carcinomas, Table 17.
b Upper bound on sum of risks from the time-to-tumor analysis of the NTP data, Table 16.
0 Sum of (upper bound) unit risks (see text for explanation), Table 16.
d Upper bound on sum of risks, Table 16.
08/29/06
83
DRAFT—DO NOT CITE OR QUOTE
-------
Table 19. Comparison of unit risk estimates
Assessments
Data source
Inhalation unit risk estimate"
Based on human data
U.S. EPA
(this document)
Ethylene Oxide Industry
Council (Kirman et al.,
2004)
Lymphohematopoietic cancer
incidence in sterilizer workers
(NIOSH)b
Breast cancer incidence in
sterilizer workers (NIOSH)C
Leukemia mortality in
combined NIOSH and UCC
cohorts
9.0 x lO^dig/m3)"1
5.0 x lO^dig/m3)"1
4.5 x 1Q-8 (ng/m3)-1
Range of 1.4 x 10'8
to 1.4 x 10"7 (|ig/m3
(Hg/m3)-1
yl d
Based on rodent data
U.S. EPA
(this document)
California EPA
(CalEPA, 1999)
Health Canada
(Health Canada, 200 l)e
Ethylene Oxide Industry
Council (Kirman et al.,
2004)
Granathetal. (1999)
Female mouse tumors
Mononuclear cell leukemia in
female rats
Mononuclear cell leukemia in
female rats
Mononuclear cell leukemia in
rats and lymphomas in mice
Pooled data from all tumor sites
using multiplicative model; used
doubling concentration in rodents
and human background risk of all
hematopoietic cancers
4.6 x 1Q-5 (ng/m3)-1
8.8 x 1Q-5 (ng/m3)-1
3.3 x lO^dig/m3)'1
Range of 2. 6 x 10'8
to 1.5 x 10-5(ng/m3
(Hg/m3)-1
)-lf
'""""
Because the weight of evidence supports a mutagenic mode of action for EtO carcinogenicity, EPA believes increased early-life
susceptibility should be assumed in the absence of chemical-specific data. If estimating a unit risk for a constant lifetime
exposure for ages 0-70 years, EPA would apply ADAFs to the potency estimate of 9.0 x 104 dig/m3)"1 to obtain a full lifetime
unit risk estimate of 1.5 x 10'3 (ng/m3)'1, in accordance with EPA's Supplemental Guidance (U.S. EPA, 2005b). Other EPA
estimates in this table and the footnotes are not the final estimates recommended in this assessment and are left unadjusted for
early-life susceptibility. The non-EPA estimates in the table are shown as reported and do not account for potential increased
early-life susceptibility for lifetime exposures that include childhood.
Estimate based on lymphohematopoietic cancer mortality is 5.0 x 104 (iig/m3)"1.
Estimate based on breast cancer mortality is 2.8 x 104 (iig/m3)"1.
Estimates based on linear extrapolation from EC0001 - EC000001 obtained from the quadratic model.
WHO (2003) presents the same quantitative risk estimates for cancer as Health Canada (2001), Health Canada having provided
the first draft of WHO's assessment.
Estimates based on quadratic extrapolation model below the observable range of the data (i.e., below the LEG 10 or LEC01
obtained using multistage model) with various points of departure (LEC01 - LEC000001) for final linear extrapolation (see
Section 4.4.2).
08/29/06
84
DRAFT—DO NOT CITE OR QUOTE
-------
Table 20. Extra risk estimates for lymphohematopoietic cancer in males for various occupational exposure levels"
8-hour
TWA
(ppm)
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Lymphohematopoietic cancer mortality
log cumulative exposure
Cox regression modelb
MLE
0.025
0.029
0.031
0.033
0.034
0.035
0.036
0.037
0.038
0.039
95% UCL
0.060
0.071
0.078
0.084
0.088
0.092
0.095
0.098
0.10
0.10
linear modelc
MLE
0.0077
0.015
0.022
0.030
0.038
95% UCL
0.017
0.033
0.049
0.064
0.079
0.11
Lymphohematopoietic cancer incidenced
log cumulative exposure
Cox regression modelb
MLE
0.046
0.052
0.056
0.060
0.062
0.064
0.066
0.067
0.069
0.070
95% UCL
0.11
0.13
0.14
0.15
0.16
0.16
0.17
0.17
0.18
0.18
linear modelc
MLE
0.013
0.027
0.040
0.052
0.065
95% UCL
0.029
0.057
0.084
0.11
0.13
0.16
oo
fe
H
o
H
O
HH
H
W
O
&
O
c
o
H
W
a Assuming a 35-year exposure between ages 20 and 55 years (see Section 4.6).
b From the best-fitting (natural) log cumulative exposure Cox regression model for lymphohematopoietic cancer mortality in males;
15-year lag (Steenland et al., 2004; see also Section 4.1.1.2).
c Linear regression of categorical results (see Section 4.1.1.2); extra risk estimates from the linear model are provided only up to the
exposure level where the linear model meets the log cumulative exposure Cox regression model
d Assumes same exposure-response relationship as for lymphohematopoietic cancer mortality.
-------
Table 21. Extra risk estimates for breast cancer in females for various occupational exposure levelsa
8-hour
TWA
(ppm)
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Breast cancer
mortality
log cumulative
exposure modelb
(20-year lag)
MLE
0.018
0.021
0.022
0.024
0.024
0.025
0.026
0.026
0.027
0.027
95%
UCL
0.038
0.044
0.048
0.051
0.053
0.055
0.057
0.059
0.060
0.061
Breast cancer incidence0
log cumulative
exposure model
(15-year lag)
MLE
0.059
0.066
0.070
0.073
0.075
0.077
0.079
0.080
0.082
0.083
95%
UCL
0.11
0.13
0.14
0.15
0.15
0.15
0.16
0.16
0.16
0.17
log cumulative
exposure model
(no lag)
MLE
0.13
0.15
0.16
0.16
0.17
0.17
0.18
0.18
0.18
0.19
95%
UCL
0.27
0.31
0.33
0.35
0.36
0.37
0.38
0.39
0.39
0.40
cumulative
exposure model
(15-year lag)
MLE
0.0014
0.0028
0.0042
0.0056
0.0070
0.0085
0.0099
0.011
0.013
0.014
95%
UCL
0.0024
0.0048
0.0072
0.0097
0.012
0.015
0.017
0.020
0.023
0.025
linear regression
model"
(15-year lag)
MLE
0.0075
0.015
0.022
0.029
0.037
0.044
0.051
0.058
0.065
0.072
95%
UCL
0.016
0.032
0.048
0.063
0.078
0.092
0.11
0.12
0.13
0.15
oo
O
O
2
O
H
O
O
&
O
c
O
H
W
a Assuming a 35-year exposure between ages 20 and 55 years.
b Best-fitting (natural) log cumulative exposure Cox regression model for breast cancer mortality in females (Steenland et al., 2004).
c From incidence data for subcohort with interviews; invasive and in situ tumors (Steenland et al., 2003).
d Regression coefficient derived from linear regression of categorical results, excluding the highest exposure group, as described in
Section 4.1.2.3.
-------
CH2 — CH2
Ethylene oxide
Glutathione / \ Epoxide
transfer ase \ hydrolase?
HOCH2CH2OH
S-2-(hydroxyethyl-glutathione) 1 ;2-ethanediol
i
HOCH2CHO
hydroxyacetaldehyde
i
CYS-CH2CH2OH Glycohc acid
S-2-(hydroxyethyl) cysteine I
OHCCO2H
Glyoxylic acid
N-acethyl-S-(2-hydroxyethyl)
HCO2H CO2HCO2H
Formic acid Oxalic acid
CO,
Figure 1. Metabolism of ethylene oxide
08/29/06 87 DRAFT—DO NOT CITE OR QUOTE
-------
6
5
I4
o
E
i 3
2
1
20
40 60 80
EtO exposure concentration (ppm)
100
120
Figure 2. Simulated blood AUCs for EtO following a 6-hour exposure to EtO
from the rat, mouse, and human PBPK models of Fennell and Brown (2001);
based on data presented in Fennell and Brown (2001). (Ratl and rat2 results
use different values for pulmonary uptake.)
08/29/06
DRAFT—DO NOT CITE OR QUOTE
-------
LED
0.0001
0.001
0.01
0.1
1.0
10
100
1000
10000
10
100
1000
10000
100000
1000000
HID
ETHYLENE OXIDE
75-21-8 GENETIC ACTIVITY PROFILE
I
s
A H
0 S
_ S M
A —
5 _
T
S D
I N CM
E E F C T X
B
R
D I
Ci[CB P -
• ^ 1 cS c" L •
- -F^M ^ M
1-nE G R F . ( D.
Iff" ,- - J k 4 M-
I 1
NU
...
ss
•*- AA
B79
o
P
- " t
1
s
G
R
-
F A
N
G
,1
H
t
PROKARY | LOWEUK PLNT INS
9
0
t
C
C
T
7
S
T
c
/I
3
HH G
1
MAMM VITRO
-H
S
H
c
B
A
R ?.M -
D V \?M <
F" C
NCEA-01
2001
s c
L L
H H
D
V M
HU_u
D ,
L
R
M
V As#v_c DH
H
: T
HUMAN
F
jC
=—
i i
j_
S
L
P
• c
V=H-
. .~
j. .
--
*
V A
A V
A
MAMM VIVO
HU
IARC human carcinogen (group 1: human - limited, animal - sufficient)
Figure 3. Display of 203 data sets, including bacteria, fungi, plants, insects,
and mammals (in vitro and in vivo), measuring the full range of genotoxic
endpoints. (This is an updated version of the figure in IARC, 1994.)
08/29/06
89
DRAFT—DO NOT CITE OR QUOTE
-------
•exp(B*exposure)
-exp(B*log exposure)
-linear regression of
categorical results
categorical
4000 5000 6000
mean cumulative exposure (ppm*days)
Figure 4. RR estimate for lymphohematopoietic cancer in males vs. mean
exposure (from Steenland et al., 2004, Table 6, Cox regression results, except
for linear regression [see text]; log and categorical exposures with 15-year
lag), unadjusted for continuous exposure. (Highest categorical exposure
quartile not shown.)
08/29/06
90
DRAFT—DO NOT CITE OR QUOTE
-------
4.0
3.5
3.0
2.5
E
t;2.o-|
1.5
0.5
0.0
•exp(B"exposure)
- exp(B"log exposure)
- linear recession of
categorical results
categorical
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
mean cumulative exposure (pprrfdays)
Figure 5. RR estimate for "lymphoid" cancer in males vs. mean exposure (from Steenland
et al., 2004, Table 7, Cox regression results, except for linear regression [see text]; log and
categorical exposures with 15-year lag), unadjusted for continuous exposure. (Highest
categorical exposure quartile not shown.)
08/29/06
91
DRAFT—DO NOT CITE OR QUOTE
-------
exp(B*log exposure)
-linear regression of low
exposure categorical results
categorical
1000 2000 3000 4000 5000 6000
mean cumulative exposure (ppm*days)
7000
8000
Figure 6. RR estimate for breast cancer mortality in females vs. mean
exposure (from Steenland et al., 2004, Table 8, Cox regression results, except
for linear regression [see text]; with 20-year lag), unadjusted for continuous
exposure. (Highest categorical exposure quartile not shown.)
08/29/06
92
DRAFT—DO NOT CITE OR QUOTE
-------
-exp(B*exposure)
-exp(B*log exposure)
-linear regression of
categorical results
categorical
4000 5000 6000
mean cumulative exposure (ppm*days)
Figure 7. RR for breast cancer incidence in females (full cohort) vs. mean
exposure (from Steenland et al., 2003, Table 4, Cox regression results, except
for linear regression [see text]; 15-year lag), unadjusted for continuous
exposure. (Highest categorical exposure quintile not shown.)
08/29/06
93
DRAFT—DO NOT CITE OR QUOTE
-------
-exp(B*exposure)
-exp(B*log exposure)
-linear regression of
categorical results
categorical
4000 5000 6000 7000
mean cumulative exposure (ppm*days)
Figure 8. RR for breast cancer incidence in females (subcohort with
interviews) vs. mean exposure (from Steenland et al., 2003, Table 5, Cox
regression results, except for linear regression [see text]; 15-year lag),
unadjusted for continuous exposure. (Highest categorical exposure quintile not
shown.)
08/29/06
94
DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX A: CRITICAL REVIEW OF EPIDEMIOLOGIC EVIDENCE
2 A.l. EARLY EVALUATIONS
3 In 1985, the U.S. Environmental Protection Agency (EPA) published a health assessment
4 of the potential carcinogenicity of ethylene oxide (EtO) (U.S. EPA, 1985). The report concluded
5 that exposure to this chemical was "probably carcinogenic to humans," and it was classified as
6 Group Bl according to Guidelines for Car cinogenic Risk Assessment (U.S. EPA, 1986). The
7 epidemiological evidence was considered to be "limited," bordering on "inadequate." Evidence
8 from three studies of two cohorts (Hogstedt et al., 1979a, b, 1984), in which the risk of leukemia
9 was reported to be statistically significant, provided the basis for this determination. A fourth
1 0 study by Morgan et al. (1981) reported an increased mortality from pancreatic cancer and
11 Hodgkin's disease that was statistically significant (p<0.05). Critical reviews of these studies are
1 2 found in the 1985 health assessment document and are not reviewed in the current document.
1 3 Studies published since 1985 are reviewed in this Appendix.
14 A.2. BACKGROUND
1 5 On the basis of studies indicating that EtO was a strong mutagen and that exposure to EtO
1 6 produced increased chromosomal aberrations in human lymphocytes (Rapoport, 1948; Ehrenberg
1 7 and Gustafsson, 1959; Ehrenberg and Hallstrom, 1967), Hogstedt and colleagues studied three
1 8 small, independent cohorts of workers from Sweden. Reports on two of these cohorts (Hogstedt
1 9 et al., 1979a, b, 1984) were reviewed in the earlier health assessment document (U.S. EPA, 1985).
20 These two small cohorts plus a third group of EtO-exposed workers from a third independent
21 plant in Sweden were then combined and studied as one cohort (Hogstedt et al., 1986; Hogstedt,
22 1988). A review of this reconstituted cohort study and subsequent independent studies is
23 presented in Section A3.
24 Shortly after the third Hogstedt study was completed, another independent study of EtO-
25 exposed employees was completed (Gardner et al., 1989) on a cohort of workers from four
26 companies and eight hospitals in Great Britain, and it was followed by a third independent study
27 on a cohort of exposed workers in eight chemical plants from the Federal Republic of Germany
28 (Kiesselbach et al., 1990). A followup study of the Gardner et al. (1989) cohort was recently
29 conducted by Coggon et al. (2004).
30 Greenberg et al. (1990) was the first in a series of studies of workers exposed to EtO at
31 two production facilities in the Kanawha Valley (South Charleston, WV). The workers at these
32 two facilities were studied later by Teta et al. (1993, 1999) and Benson and Teta (1993) and
08/29/06 95 DRAFT—DO NOT CITE OR QUOTE
-------
1 became the basis for several important quantitative risk assessment analyses (Teta et al., 1999;
2 EOIC, 2001).
3 Another independent study of EtO-exposed workers in 14 sterilizing plants from across
4 the United States was completed by the National Institute for Occupational Safety and Health
5 (Steenland et al., 1991; Stayner et al., 1993). The Stayner et al. (1993) paper presents the
6 exposure-response analysis performed by the NIOSH investigators. These same workers were
7 studied again from a different perspective by Wong and Trent (1993). The NIOSH investigators
8 recently completed a followup of the mortality study (Steenland et al., 2004) and a breast cancer
9 incidence study based in the same cohort (Steenland et al., 2003). The results of the Steenland et
10 al. (2003, 2004) analyses are the basis for the quantitative assessment in this document, for
1 1 reasons explained in the review and summary sections of this Appendix.
1 2 Several additional studies of lesser importance have been done on EtO-exposed cohorts of
1 3 workers in Sweden (Hagmar et al., 1991, 1995), Italy (Bisanti et al., 1993), Belgium (Swaen et
14 al., 1996), and western New York State (Norman et al., 1995), and other parts of the United States
1 5 (Olsen et al., 1997). These studies are discussed in the following review, but they provide limited
1 6 information to the overall discussion of whether EtO induces cancer in humans.
1 7 The more important studies, which are discussed in detail in the summary, are those at two
1 8 facilities in the Kanawha Valley in West Virginia (Greenberg et al., 1990; Benson and Teta, 1993;
1 9 Teta et al., 1993, 1999) and at 14 sterilizing plants around the country (Stayner et al., 1993;
20 Steenland et al., 1991, 2003, 2004). These studies indicate that a great deal of effort and care was
21 expended to ensure that they were done well. They have sufficient followup to analyze latent
22 effects, attempts were made to develop dose-response relationships using reasonable assumptions
23 about early exposures to EtO, and the cohorts appear to be large enough to test for small
24 differences.
25 A.3. INDIVIDUAL STUDIES
26 A.3.1. HOGSTEDT ET AL. (1986), HOGSTEDT (1988)
27 Hogstedt et al. (1986) combined workers from several cohorts for a total of 733 workers,
28 including 378 workers from two separate and independent occupational cohort mortality studies
29 by Hogstedt et al. (1979a, b) and 355 employees from a third EtO production plant who had not
30 been previously examined. The combined cohort was followed until the end of 1982. The first
31 cohort comprised employees from a small technical factory in Sweden where hospital equipment
32 was sterilized with EtO. The second was from a production facility where EtO was produced by
33 the chlorohydrin method from 1940 to 1963. The third was from a production facility where EtO
34 was made by the direct oxidation method from 1963 to 1982.
08/29/06 96 DRAFT—DO NOT CITE OR QUOTE
-------
1 In the update of the 1986 occupational mortality report (Hogstedt, 1988), the cohort
2 inexplicably was reduced to 709 employees (539 men; 170 women). Followup for mortality was
3 extended to the end of 1985. The author reported that 33 deaths from cancer had occurred,
4 whereas only 20 were expected in the combined cohort. The excesses that are significant are due
5 mainly to an increased risk of stomach cancer at one plant and an excess of blood and lymphatic
6 malignancies at all three. Seven deaths from leukemia occurred, whereas only 0.8 were expected
7 (standard mortality ratio [SMR] = 9.2). Ten deaths due to stomach cancer occurred versus only
8 1.8 expected (SMR = 5.46). The results tend to agree with those from clastogenic and short-term
9 tests on EtO (Ehrenberg and Gustafsson, 1959). The authors believe that the large number of
10 positive cytogenetic studies demonstrating increased numbers of chromosomal aberrations and
11 sister chromatid exchanges at low-level exposure to EtO indicate that the lymphatic and
1 2 hematopoietic systems are particularly sensitive to the genotoxic effects of EtO. They concluded
1 3 that the induction of malignancies even at low-level and intermittent exposures to EtO should be
1 4 "seriously considered by industry and regulating authorities."
1 5 The average air EtO concentrations in the three plants were as follows: In Plant 1
1 6 (Hogstedt et al., 1979b) in 1977, levels ranged from 2 to 70 ppm in the storage hall. The average
1 7 8-hour time-weighted average (TWA) concentration in the breathing zone of the employees was
1 8 calculated as 20 ppm +/- 10 ppm. Measured concentrations were 150 ppm on the floor outside of
1 9 the sterilized boxes and 1,500 ppm inside.
20 In Plant 2 (Hogstedt et al., 1979a), EtO was produced through the chlorohydrin process.
21 Between 1941 and 1947, levels probably averaged about 14 ppm, with occasional exposures up to
22 715 ppm. Between 1948 and 1963, levels were in the range of 6 ppm to 28 ppm. After 1963,
23 when production of EtO came to an end, levels ranged from less than 1 ppm to as much as 6 ppm.
24 In Plant 3 (Hogstedt et al., 1986), the 355 employees were divided into subgroups.
25 Subgroup A had almost pure exposure to EtO. Subgroup B had principal exposure to EtO but
26 also exposure to propylene oxide, amines, sodium nitrate, formaldehyde, and 1,2-butene oxide.
27 Workers in the remaining subgroup C were maintenance and technical service personnel, who had
28 multiple exposures, including EtO. Concentration levels in Plant 3 are shown in Table A-l.
29 In the earlier studies (Hogstedt et al., 1979a, b) of two of the plants that contributed
30 workers to this cohort, the authors allude to the fact that there was exposure to benzene, ethylene
31 dichloride, ethylene chlorohydrin, ethylene, and small amounts of bis-(2-chloroethyl) ether, as
32 well as other chemicals in the respective plants. Although 170 women were present in the
33 workforce, no gender differences in risk were analyzed separately by the investigators. Of 16
34 patients with tumors in the two exposed cohorts, there were three cases of leukemia (0.2
expected), six cases of alimentary tract cancer, and four cases of urogenital cancer. Of the 11
08/29/06 97 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-l. Estimated 8-hour time-weighted average ethylene oxide exposure,
Plant 3
Group
A (n = 128)
B (n = 69)
C(n= 158)
1963-1976
5-8 ppm
3 ppm
1-3 ppm
1977-1982
1-2 ppm
1 ppm
0.4-1. 6 ppm
Source: Hogstedt et al, 1986
1 cancer cases in the full-time exposed cohort, 5.9 were expected (p<0.05). This study was
2 criticized by Divine and Amanollahi (1986) for several reasons. First, they believed that the
3 study's strongest evidence in support of a carcinogenic claim for EtO was only a "single case of
4 leukemia" in subgroup C of Plant 3, where the workers had multiple chemical exposures;
5 however, there were no cases in subgroups A or B of Plant 3. Hogstedt et al. (1986) countered
6 that the expectation of leukemia in these two subgroups were 0.04 and 0.02, respectively, and
7 that the appearance of a case could only happen if EtO had "outstanding carcinogenic properties
8 at low levels." Divine and Amanollahi also pointed out that a study (Morgan et al., 1981) of a
9 cohort similar to that of Plant 3 found no leukemia cases or evidence of excessive mortality.
10 Hogstedt et al. replied that Morgan et al. stated in their paper that the statistical power of their
11 study to detect an increased risk of leukemia was not strong.
12 Divine and Amanollahi (1986) also stated that the exposures to EtO were considerably
13 higher in plants 1 and 2 than in Plant 3; therefore, combinations would "normally preclude
14 comparisons between the plants for similar causes of adverse health." This potential problem
15 could be resolved by structuring exposure gradients to analyze risk. Furthermore, they noted,
16 Plant 1 was a nonproduction facility involved in sterilization of equipment. Plant 2 used the
17 chlorohydrin process for making EtO, and Plant 3 used the direct oxygenation process. Although
18 these conditions are obviously different, they "are grouped together as analogous." This criticism
19 would, in most instances, be valid only because the methods for producing EtO differ and there
20 were differing exposures to multiple chemicals.
21 However, these concerns are not supported by the evidence. In all three plants the
22 leukemia risk was elevated, even if only slightly in Plant 3. This suggests that there may have
23 been a common exposure, possibly to EtO, endemic to all three plants that was responsible for the
24 measured excesses. Noteworthy is the elevated risk of leukemia seen in Plant 1 (3 observed vs.
25 0.14 expected), where the exposures were almost exclusively to EtO in the sterilization of
26 equipment. The argument that Plant 1 leukemias form a "chance cluster," as Shore et al. (1993)
08/29/06
98
DRAFT—DO NOT CITE OR QUOTE
-------
1 claim, and as such should be excluded from any analysis does not preclude the possibility that
2 these cases are in reality the result of exposure to EtO. Hogstedt argues that earlier remarks by
3 Ehrenberg and Gustafsson (1959) that EtO "constituted a potential cancer hazard" on the basis of
4 a considerable amount of evidence other than epidemiologic should have served as a warning that
5 the increased risk seen in Plant 1 was not necessarily a "chance cluster," and because the
6 chlorohydrin process was not in use in Plant 1, it cannot be due to exposure to a chemical in the
7 chlorohydrin process.
8
9 A.3.2. GARDNER ET AL. (1989)
10 Gardner et al. (1989) completed a cohort study of 2,876 men and women who had
11 potential exposure to EtO. The cohort was identified from employment records at four companies
12 that had produced or used EtO since the 1950s and from eight hospitals that have had EtO clinical
13 sterilizing units since the 1960s, and it was followed to December 31, 1987. All but 1 of the
14 1,012 women and 394 of the men in the cohort worked at one of the hospitals. The remaining
15 woman and 1,470 men made up the portion of the cohort from the four companies. By the end of
16 the followup, 226 members (8% of the total cohort) had died versus 258.8 expected. Eighty-five
17 cancer deaths were observed versus 76.64 expected.
18 No clear excess risk of leukemia (3 observed vs. 2.09 expected), stomach cancer (5
19 observed vs. 5.95 expected), or breast cancer (4 observed vs. 5.91 expected) was present as of the
20 cut-off date. "Slight excesses" of deaths due to esophageal cancer (5 observed vs. 2.2 expected),
21 lung cancer (29 observed vs. 24.55 expected), bladder cancer (4 observed vs. 2.04 expected), and
22 NHL (4 observed vs. 1.63 expected) were noted, although an adjustment made to reflect local
23 "variations in mortality" reduced the overall cancer excess from 8 to only 3. According to the
24 authors' published tabulations, all three leukemias identified in this study fell into the longest
25 latent category (20 years or longer), where only 0.35 were expected. All three were in the
26 chemical plants. This finding initially would seem to be consistent with experimental animal
27 evidence demonstrating excess risks of hematopoietic cancer in animals exposed to EtO. But the
28 authors note that since other known leukemogens were present in the workplace, the excess could
29 have been due to a confounding effect.
30 The hospitals began using EtO during or after 1962, whereas all of the chemical
31 companies had handled EtO from or before 1960. In the hospitals there was occasional exposure
32 to formaldehyde and carbon tetrachloride but few other confounding agents. On the other hand,
33 the chemical workers were exposed to a wide range of compounds including chlorohydrin,
34 propylene oxide, styrene, and benzene. The earliest industrial hygiene surveys in 1977 indicated
35 that the TWA average exposures were less than 5 ppm in almost all jobs and less than 1 ppm in
36 many. No industrial hygiene data were available for any of the facilities prior to 1977, although it
08/29/06 99 DRAFT—DO NOT CITE OR QUOTE
-------
1 is stated that peaks of exposure up to several hundred ppm occurred as a result of operating
2 difficulties in the chemical plants and during loading and unloading of sterilizers in the hospitals.
3 An odor threshold of 700 ppm was reported by both manufacturers and hospitals, according to the
4 authors. The authors assumed that past exposures were somewhat higher without knowing
5 precisely what they were. An attempt was made to classify exposures into a finite number of
6 subjectively derived categories (definite, possible, continual, intermittent, and unknown). This
7 exercise produced no discernable trends in risk of exposure to EtO. However, the exposure status
8 classification scheme was so vague as to be useless for determining risk by gradient of exposure
9 to EtO.
10 It is of interest that all three of the leukemia deaths entailed exposure to EtO, with very
11 little or no exposure to benzene, according to the authors. The findings are not inconsistent with
12 those of Hogstedt et al. (1986) and Hogstedt (1988). The possibility of a confounding effect other
13 than benzene in these chemical workers cannot entirely be ruled out. Other cancers were slightly
14 in excess, but overall there was little increased mortality from cancer in this cohort. It is possible
15 that if very low levels of exposure to EtO had prevailed throughout the history of these hospitals
16 and plants, the periods of observation necessary to observe an effect may not have been long
17 enough.
18 A followup study of this cohort conducted by Coggon et al. (2004) is discussed below.
19
20 A.3.3. KIESSELBACH ET AL. (1990)
21 Kiesselbach et al. (1990) carried out an occupational cohort mortality study of 2,658 men
22 from eight chemical plants in the Federal Republic of Germany (FRG) that were involved in the
23 production of EtO. The method of production is not stated. At least some of the plants that were
24 part of an earlier study by Thiess et al. (1982) were included. Each subject had to have been
25 exposed to EtO for at least 1 year sometime between 1928 and 1981 before person-years at risk
26 could start to accumulate. Most exposures occurred after 1950. By December 31, 1982, the
27 closing date of the study, 268 men had died (about 10% of the total cohort), 68 from malignant
28 neoplasms. The overall SMR for all causes was 0.87, and for total cancer the SMR was 0.97,
29 based on FRG rates. The authors reported that this deficit in total mortality indicates a healthy-
30 worker effect.
31 The only remarkable findings here are slightly increased risks of death from stomach
32 cancer (14 observed vs. 10.15 expected, SMR = 1.4), cancer of the esophagus (3 observed vs. 1.5
33 expected, SMR = 2), and cancer of the lung (23 observed vs. 19.86 expected, SMR = 1.2).
34 Although the authors claimed that they looked at latency, only stomach cancer and total mortality
35 has a latency analysis included. This was accomplished by not counting the first 10 years of
36 followup in the parameter "years since first exposure." This study is limited by the lack of further
08/29/06 100 DRAFT—DO NOT CITE OR QUOTE
-------
1 latency analyses at other cancer sites. The risk of stomach cancer shows only a slight
2 nonsignificant trend upward with increasing latency. Only two leukemias were recorded versus
3 2.35 expected.
4 This is a largely unremarkable study, with few findings of any significance. No actual
5 exposure estimates are available. The categories of exposure that the authors constructed are
6 "weak," "medium," and "strong." It is not known whether any of these categories is based on
7 actual measurements. No explanation of how they were derived is provided except that the
8 authors claim that the information is available on 67.2% of the members of the cohort. If the
9 information was based on job categories, it should be kept in mind that exposures in jobs that
10 were classified the same from one plant to the next may have produced entirely different
11 exposures to EtO. The tabular data regarding these exposure categories shows that only 2.4% of
12 all members of the cohort were considered "strongly" exposed to EtO. Although 71.6% were
13 classified as "weak," the remaining 26% were considered as having "medium" exposure to EtO.
14 This is largely a study in progress, and further followup will be needed before any definite
15 trends or conclusions can be drawn. The authors reported that only a median 15.5 years of
16 followup had passed by the end of the cutoff date, whereas the median length of exposure was 9.6
17 years. Before any conclusions can be made from this study several additional years of followup
18 would be needed with better characterization of exposure.
19
20 A.3.4. GREENBERG ET AL. (1990)
21 Greenberg et al. (1990) retrospectively studied the mortality experience of 2,174 men
22 employed in a Union Carbide Corporation (UCC) chemical plant in West Virginia that used or
23 produced EtO. The referent comparison population was general U.S. population death rates.
24 Regional population death rates were found to be similar to those of the U.S. general population.
25 Followup began either on January 1, 1940, if exposure to EtO began sooner, or on the date when
26 exposure began if it occurred after January 1, 1940. Followup ended on December 31, 1978.
27 Total deaths equaled 297 versus 375.9 expected (SMR = 0.79). Only 60 total cancer deaths were
28 observed versus 74.6 expected (SMR = 0.81). These deficits in mortality are a manifestation of
29 the healthy-worker effect. In spite of this, nonsignificant elevated risks of cancer of the liver,
30 unspecified and primary, (3 observed vs. 1.8 expected, SMR = 1.67), pancreas (7 observed vs. 4.1
31 expected, SMR = 1.7), and leukemia and aleukemia (7 observed vs. 3 expected, SMR = 2.33)
32 were noted.
33 The authors also reported that in 1976, 3 years prior to the end of followup, an industrial
34 hygiene survey found that airborne levels of EtO were low, less than 1 ppm 8-hour TWA. In
35 maintenance workers, levels averaged between 1 and 5 ppm 8-hour TWA. These measurements
36 were the first large-scale monitoring of respirable EtO at this company, and the results indicated
08/29/06 101 DRAFT—DO NOT CITE OR QUOTE
-------
1 that some exposure levels were as high as 66 ppm 8-hour TWA. Because of the lack of
2 information about exposures before 1976, the authors developed a prior exposure scheme on the
3 basis of the 1976 survey and the potential for exposure by each department on the basis of
4 occurrence of dermatological or other medical problems. This scheme envisioned designating
5 every department according to one of three categories of exposure to EtO: low, intermediate, or
6 high.
7 Except for two cases of leukemia, all the victims of pancreatic cancer and leukemia began
8 their work—and hence exposure to EtO—many years prior to their deaths. Four of the seven
9 leukemia victims had been assigned to the chlorohydrin department; only 0.8 deaths (SMR = 5)
10 would have been expected in this department of only 278 workers. Six pancreatic cancer victims
11 were assigned to the chlorohydrin department, whereas only 0.98 deaths would have been
12 expected to occur (SMR =6.12).
13 All seven leukemia victims, including the four in the chlorohydrin department, were listed
14 by the authors as having only low exposure to EtO. In contrast, in the department where exposure
15 to EtO was probably the highest, no leukemia deaths and only one pancreatic cancer death
16 occurred. However, according to the authors, exposures to EtO in the EtO direct oxidation
17 production department at a similar plant in Texas (discussed in Joyner, 1964) were "probably in
18 the range of 10-20 ppm" and when the chlorohydrin process was used, the "levels of exposure to
19 EtO were probably somewhat higher" (in the West Virginia Kanawha Valley plants) because
20 "technology, construction materials and work practices were from an earlier time" and there was
21 also "no control room and production equipment was indoors." These observations would seem
22 to argue against the assumption that the four victims who were assigned to the chlorohydrin
23 department would fall into a low exposure category, based on estimated levels of exposure in a
24 similar operation at the Texas plant.
25 The authors hypothesized that leukemia and pancreatic cancers have been associated with
26 production of ethylene chlorohydrin or propylene chlorohydrin or both in the chlorohydrin
27 department. This reasoning is premised upon their exposure category construct, which places
28 most of the leukemia and pancreatic cancer victims in that department. Because the exposure
29 classification was not based on any actual individual TWA measurements prior to 1976 but only
30 on the potential for exposure, it is speculative to assume accuracy in the designation of
31 individuals to such categories.
32 If the exposure scheme is not considered—because it appears to be somewhat
33 subjective—a borderline significant excess risk of leukemia occurred in a group of workers
34 alleged to have had some exposure to EtO. The question is whether the slightly increased risk of
35 leukemia and pancreatic cancer was a consequence of exposure to EtO or to other chemicals such
08/29/06 102 DRAFT—DO NOT CITE OR QUOTE
-------
1 as ethylene dichloride, propylene chlorohydrin, bis-chloroethyl, and dichloroethane in the
2 chlorohydrin manufacturing process. This study does not resolve the issue.
3
4 A.3.5. STEENLAND ET AL. (1991)
5 In an industrywide analysis by the National Institute for Occupational Safety and Health,
6 Steenland et al. (1991) studied EtO exposure in 18,254 workers (55% female) identified from
7 personnel files of 14 plants that had used EtO for sterilization of medical equipment, treating
8 spices, or testing sterilizers. Each of the 14 plants (from 75 facilities surveyed) that were
9 considered eligible for inclusion in the study had at least 400 person-years at risk prior to 1978.
10 Within each eligible facility, at least 3 months of exposure to EtO qualified an employee for
11 inclusion in the cohort. Employees, including all salaried workers, who were "judged never to
12 have been exposed to EtO" on the basis of industrial hygiene surveys were excluded. Followup
13 ended December 31, 1987. The cohort averaged 16 years of latency. Approximately 86%
14 achieved the 9-year latent point, but only 8% reached the 20-year latency category. The average
15 year of first exposure was 1970, and the average length of exposure was 4.9 years. The workers'
16 average age at entry was not provided, nor was an age breakdown. Nearly 55% of the cohort
17 were women.
18 Some 1,137 workers (6.4%) were found to be deceased at the end of the study period,
19 upon which the underlying cause of death was determined for all but 450. If a member was
20 determined to be alive as of January 1, 1979, but not after and no death record was found in the
21 National Death Index through December 31, 1987, then that member was assumed to be alive for
22 the purposes of the life-table analysis and person-years were accumulated until the cut-off date.
23 Altogether, 4.5% of the cohort fell into this category. This procedure would tend to increase the
24 expected deaths and, as a consequence, potentially bias the risk ratio downward if a sizable
25 number of deaths to such persons during this period remained undiscovered to the researchers.
26 In the total cohort no significantly increased risks of death from any site-specific cancer
27 were noted. Analyses by job categories and by duration of exposure indicated no excess risks of
28 cancer when compared with the rate in the general population. However, there was an increased
29 trend in the risk of hematopoietic cancers, all sites, with increasing lengths of time since first
30 exposure. After 20 years latency, the SMR was 1.76, based on 13 cases. The test for trend was
31 significant at/>=0.03. For men (45%), without regard for latency, the SMR for hematopoietic
32 cancer was a significant 1.55 (p<0.05), based on 27 cases. Among men with long latency (greater
33 than 20 years) and the longest duration of exposure (greater than 7 years) the SMR for
34 hematopoietic cancers was 2.63, based on 7 deaths (p<0.05).
35 The authors pointed out that the SMR for leukemia among men was 3.45, based on 5
36 deaths (p<0.05), for deaths in the latter period of 1985 to 1987. For kidney cancer, the SMR was
08/29/06 103 DRAFT—DO NOT CITE OR QUOTE
-------
1 3.27, based on 6 deaths (p<0.05), after 20 years latency. The authors also reported on a significant
2 excess risk (p<0.05) of lymphosarcoma-reticulosarcoma in men (SMR = 2.6), based on 7 deaths.
3 Women had a lower nonsignificant rate. The risk of breast cancer was also nonsignificant (SMR
4 = 0.85 based on 42 cases). The authors hypothesized that men were more heavily exposed to EtO
5 than were women because "men have historically predominated in jobs with higher levels of
6 exposure." However, the lack of an association between EtO exposure and lymphohematopoietic
7 cancer in females was also observed in the exposure-response analyses of this cohort, including in
8 the highest exposure category, performed by Stayner et al. (1993) and discussed below.
9 Industrial hygiene surveys indicated that sterilizer operators were exposed to an average
10 personal 8-hour TWA EtO level of 4.3 ppm, whereas all other workers averaged only 2 ppm,
11 based on 8-hour samples during the period 1976 to 1985. These latter employees primarily
12 worked in production and maintenance, in the warehouse, and in the laboratory. This was during
13 a time when engineering controls were being installed to reduce worker's exposure to EtO; earlier
14 exposures may have been somewhat higher. The authors reported that no evidence of
15 confounding exposure to other occupational carcinogens was documented.
16 The authors concluded that there was a trend toward an increased risk of death from
17 hematopoietic cancer with increasing lengths of time since the first exposure to EtO. This trend
18 might have been enhanced if the authors had added additional potential deaths identified from the
19 820 (4.5%) "untraceable" members of the cohort from 1979 to 1987. The authors felt that their
20 results were not conclusive for the relatively rare cancers of a priori interest, based on the limited
21 number of cases and the short followup. The cohort averaged 16 years of latency and 86% had at
22 least 9 years but only 8% reached the 20-year latent category.
23 Exposure-response analyses were conducted by Stayner et al. (1993) and are discussed
24 below. More recently, a followup mortality study (Steenland et al., 2004) and a breast cancer
25 incidence study (Steenland et al., 2003) of this cohort were conducted; these are also discussed
26 below.
27
28 A.3.6. TETA ET AL. (1993)
29 In a reanalysis of the cohort of 2,174 male UCC workers studied by Greenberg et al.
30 (1990), Teta and her colleagues excluded the 278 workers in the chlorohydrin unit in which
31 Greenberg and colleagues found a high risk of leukemia and pancreatic cancer, thereby removing
32 the potential confounding of the chlorohydrin process. The 1,896 men remaining in the cohort
33 were followed for another 10 years, through all of 1988. It was determined from the Greenberg et
34 al. study that there were no elevated cancer risks in these remaining workers up to December 31,
35 1978; therefore it was not likely that any significant "risks" would be found without extending the
36 period of followup.
08/29/06 104 DRAFT—DO NOT CITE OR QUOTE
-------
1 Teta et al. (1993) reported that the average duration of exposure was more than 5 years and
2 the average followup was 27 years. The reanalysis demonstrated no increased risk of overall
3 cancer, nor of leukemia, NHL or brain, pancreatic, or stomach cancer. The SMR for total deaths,
4 based on comparison with mortality from the general population, was 0.8 (observed = 431). The
5 SMR for total cancer was 0.9 (observed = 110). No site-specific cancers were significantly
6 elevated. Although the authors concluded that this study did not indicate any significant trends of
7 site-specific cancer with increasing duration of potential exposure to EtO, there appeared to be a
8 nonsignificant increasing trend of leukemia and aleukemia as well as stomach cancer with
9 increasing duration of potential exposure (Table A-2).
Table A-2. Relative risk (RR) estimates and observed deaths (obs) for selected
causes by cumulative duration of assignments in departments in years using
or producing ethylene oxide
Cause of death
Stomach cancer
Leukemia and
aleukemia
Never assigned
RR (obs)
1(79)
1(75)
Less than 2 yrs
RR (obs)
0.64(1)
0(0)
2 to 9 years
RR (obs)
2.77 (5)
1.11(2)
10 or more
years RR (obs)
2.62 (2)
2.54 (3)
Source: Adapted from Teta et al., 1993
14 The average EtO levels reported by Greenberg et al. (1990) were less than 1 ppm in all
15 areas of the plant, based on the 1976 monitoring. Teta et al. estimated that in the 1960s, exposure
16 in the units producing EtO by direct oxidation ranged from 3 to 20 ppm 8-hour TWA, with
17 operator concentration estimated at between 5 and 10 ppm 8-hour TWA. These estimates were
18 based on an industrial hygiene survey conducted at another company facility in Texas that used
19 the same process as the two plants in West Virginia from which the Greenberg et al. cohort was
20 taken. Ethylene oxide was produced via the chlorohydrin process in a closed building during
21 these early years (1925 to 1957). Levels of exposure to EtO would have been higher than in the
22 direct oxidation production process because of start-up difficulties, fewer engineering controls,
1 less complex equipment, and the enclosed building. Nausea, dizziness, and vomiting were
2 documented in the employees who made numerous visits to the medical department in 1949.
3 These acute effects occur in humans at exposures of several hundred ppm, according to the
4 authors.
08/29/06
105
DRAFT—DO NOT CITE OR QUOTE
-------
1 During the time periods under investigation, the estimated exposure ranges were >14 ppm
2 from 1925 to 1939; 11-14 ppm from 1940 to 1956; 5-10 ppm from 1957 to 1973 ; and <1 ppm
3 from 1974 to 1988, with frequent peaks of several hundred ppm in the earliest period and some
4 peaks of similar intensity in the 1940s to mid-1950s. In the absence of monitoring data prior to
5 1976, these estimates cannot be confirmed. Furthermore, workers were eliminated from the
6 analysis if they had worked in the chlorohydrin unit because of the assumption that the increased
7 risks of leukemia and pancreatic cancer were possibly due to exposure to something in the
8 chlorohydrin process, as conjectured by Greenberg et al. (1990). It is still possible that exposure
9 to EtO in workers in this process might have produced the increased risks of leukemia seen in the
10 total cohort; however, even when the confounding influence of the chlorohydrin process is
11 removed there still remains the suggestion of a trend of an increasing risk of leukemia and
12 aleukemia with increasing duration of exposure to EtO in the remaining cohort members, based
13 on the numbers shown in Table A-2.
14 The authors indicated that their findings do not confirm the findings in experimental
15 animal studies and are not consistent with the earliest results reported among EtO workers. They
16 also noted that they did not observe any significant trend of increasing risks of stomach cancer (n
17 = 8) or pancreatic, brain, or nervous system cancer or leukemia (n = 5) with increasing duration of
18 exposure. No latency analysis was conducted in this study.
19 In a later analysis, Teta et al. (1999) updated their 1993 study of UCC workers and fitted
20 dose-response models to the revised UCC data and to the NIOSH data (Steenland et al., 1991).
21 They reported that latency and lagging of dose did not appreciably affect the fitted Poisson
22 regression models to these data, which they concluded were the best studies for evaluating dose-
23 response relationships. Because Teta et al. (1999) did not present aggregate risk ratios in the
24 categories used to model dose-response relationships, the only comparison that could be made
25 between the UCC and NIOSH data is based on the fitted models. These models are almost
26 identical for leukemia, but for the lymphoid category, the risk according to the fitted model for the
27 UCC data decreased as a function of dose, whereas the risk for the modeled NIOSH data
28 increased as a function of dose. It is possible that the difference is due to regional practices in
29 coding death certificates for specific types of leukemia.
30
31 A.3.7. BENSON AND TETA (1993)
32 In a companion mortality study (Benson and Teta, 1993), the remaining 278 employees
33 who were identified by Greenberg et al. (1990) as having worked at some time in the chlorohydrin
34 process and who were not included in the cohort of Teta et al. (1993) were followed to the end of
35 1988. Altogether, 40 cancer deaths occurred versus 30.8 expected (SMR = 1.3) in the subcohort
36 of chlorohydrin workers. In Greenberg et al., significant elevated risks of pancreatic cancer and
08/29/06 106 DRAFT—DO NOT CITE OR QUOTE
-------
1 leukemia and aleukemia occurred in only those workers assigned to the chlorohydrin process.
2 Benson and Teta noted a significantly increased risk of pancreatic cancer (SMR = 5, 8 observed
3 deaths, p<0.05) in the same group and a significantly increased risk of cancer in the enlarged
4 category of hematopoietic cancer (SMR = 3, 8 observed deaths,p<0.05), which included leukemia
5 and aleukemia, after an additional 10 years of followup.
6 The authors concluded that these cancers were likely work-related and some exposure in
7 the chlorohydrin process, possibly to the chemical ethylene dichloride, was probably the cause.
8 They pointed out that Greenberg et al. found that the chlorohydrin unit was likely to be a low-EtO
9 exposure area in the West Virginia facility. The other possibility was bis-chloroethyl ether, which
10 the authors pointed out is rated by the International Agency for Research on Cancer (IARC) as a
11 group 3 ("not classifiable as to its carcinogenicity to humans") chemical. Circumstantial evidence
12 seems to support the authors' contention that ethylene dichloride is the cause: IARC designated
13 ethylene dichloride as a group 2B chemical ("possibly carcinogenic to humans"), exposure was
14 likely heavier throughout the history of the facility, and plant medical records documented many
15 accidental overexposures occurring to the pancreatic cancer victims prior to diagnosis. However,
16 this conclusion is disputed by Olsen et al. (1997). Their analysis is discussed later.
17
18 A.3.8. STAYNER ET AL. (1993)
19 Stayner et al. (1993) provide an exposure-response analysis for the cohort study of EtO
20 workers described by Steenland et al. (1991). Nothing was modified concerning the followup,
21 cohort size, vital status, or cut-off date of the study. The exposure assessment and verification
22 procedures were presented in Greife et al. (1988) and Hornung et al. (1994). Briefly, a regression
23 model allows the estimation of exposure levels for time periods, facilities, and operations for
24 which industrial hygiene data were unavailable. The data consisted of 2,700 individual time-
25 weighted exposure values for workers' personal breathing zones, acquired from 18 facilities
26 between 1976 and 1985. Arithmetic mean exposure levels by facility, year, and exposure category
27 were calculated on the basis of grouping all sampled jobs into eight categories with similar
28 potential for EtO exposure. The data were divided into two sets, one for developing the
29 regression model and the second for testing it. Arithmetic means were logarithmically
30 transformed and weighted linear regression models were fitted. Seven out of 23 independent
31 variables tested for inclusion in the model were found to be significant predictors of EtO exposure
32 and were included in the final model. This model predicted 85% of the variation in average EtO
33 exposure levels.
34 Early historical exposures in jobs in the plants were estimated using this industrial
35 hygiene-based regression model. In the Stayner et al. study, cumulative exposure for each worker
36 was estimated by calculating the product of the average exposure in each job the worker held by
08/29/06 107 DRAFT—DO NOT CITE OR QUOTE
-------
1 the time spent in that job and then summing these over all the jobs held by that worker. This
2 value became the cumulative exposure index for that employee and reflected the working lifetime
3 total exposure to EtO. SMRs were generated based on standard life-table analysis. The three
4 categories of cumulative exposure were less than 1,200 ppm-days, 1,200 to 8,500 ppm-days, and
5 greater than 8,500 ppm-days. Additionally, the Cox proportional hazards model (SAS, 1986) was
6 used to model the exposure-response relationship between EtO and various cancer types, using
7 cumulative exposure as a continuous variable.
8 Stayner and colleagues noted a marginally significant increase in the risk of hematopoietic
9 cancers, with an increase in cumulative exposure by both the life-table analysis as well as the Cox
10 model, although the magnitude of the increased risk was not substantial. At the highest
11 level—greater than 8,500 ppm-days of exposure—the SMR was a nonsignificant 1.24, based on
12 13 cases. However, 12 of these cases were in males, whereas only 6.12 were expected. Thus, in
13 this highest-exposure category, a statistically significant (p<0.05) SMR of 1.96 in males was
14 produced. This dichotomy produced a deficit in females (1 observed vs. 4.5 expected, p<0.05).
15 The Cox analysis produced a significantly positive trend with respect to lymphoid cell
16 tumors (combination of lymphocyte leukemia and NHL) when EtO exposures were lagged 5
17 years. The authors stated that these data provide some support for the hypothesis that exposure to
18 EtO increases the risk of mortality from lymphatic and hematopoietic neoplasms. They pointed
19 out, however, that their data do not provide evidence for a positive association between exposure
20 to EtO and cancer of the stomach, brain, pancreas, or kidney or leukemia as a group. Breast
21 cancer was not analyzed in this report.
22 This cohort was not updated with vital status information on the "untraceables" (4.5%),
23 and cause of death information was not provided on deaths with unknown causes; thus, it lacks a
24 complete followup and, therefore, the risk estimates may be understated. Another potential
25 limiting factor is the information regarding industrial hygiene measurements of EtO that were
26 completed in the plants. According to the authors, the median length of exposure to EtO of the
27 cohort was 2.2 years and the median exposure was 3.2 ppm. It may be unreasonable to expect any
28 findings of increased significant risks because followup was too short to allow the accumulation
29 of mortality experience (average follow-up = 16 years; only 8% of cohort had > 20 years follow-
30 up).
31 The authors also remind us that there is a lack of evidence for an exposure-response
32 relationship among females or for a sex-specific carcinogenic effect of EtO in either laboratory
33 animals or humans. In fact, the mortality rate from hematopoietic cancers among the women in
34 this cohort was lower than that of the general U.S. population. Therefore the contrast seen here is
35 unusual.
08/29/06 108 DRAFT—DO NOT CITE OR QUOTE
-------
1 The positive findings are somewhat affected by the presence in the cohort of one heavily
2 exposed case (although the authors saw no reason to exclude it from the analysis), and there is a
3 lack of definite evidence for an effect on leukemia as a group. Despite these limitations, the
4 authors believe that their data provide support for the hypothesis that exposure to EtO increases
5 the risk of mortality from hematopoietic neoplasms.
6
7 A.3.9. WONG AND TRENT (1993)
8 This study is a reanalysis of the same cohort that was studied by Steenland et al. (1990)
9 and Stayner et al. (1993), with some differences. The cohort was incremented without
10 explanation by 474 to a total of 18,728 employees and followed one more year, to the end of
11 December 1988. This change in the cohort resulted in the addition of 176 observed deaths and
12 392.2 expected deaths. The finding of more than twice as many expected deaths as observed
13 deaths is baffling. A reduced total mortality of this magnitude suggests that many deaths may
14 have been overlooked. This resulted in a further reduction of the overall SMR to a significant
15 deficit of 0.73. Sixty additional cancer deaths were added versus 65.9 expected, for an SMR =
16 0.9, based on 403 total cancer deaths observed versus 446.2 expected.
17 The authors reported no significant increase in mortality at the cancer sites found to be of
18 most interest in previous studies, that is, stomach, leukemia, pancreas, brain and breast. They also
19 reported the lack of a dose-response relationship and correlation with duration of employment or
20 latency. They did report a statistically significant increased risk of NHL among men (SMR =
21 2.47; observed =16, expected = 6.47; /><0.05) that was not dose-related and a nonsignificant
22 deficit of NHL among women (SMR = 0.32; observed = 2, expected = 6.27). The authors
23 concluded that the increase in men was not related to exposure to EtO but could in fact have been
24 related to the presence of acquired immune deficiency syndrome (AIDS) in the male population.
25 When this explanation was offered in a letter to the editor (Wong, 1991) regarding the excess of
26 NHL reported in Steenland et al. (1991), it was dismissed by Steenland and Stayner (1993) as pure
27 speculation. Steenland and Stayner responded that most of the NHL deaths occurred prior to the
28 AIDS epidemic, which began in the early 1980s. They also indicated that there was no reason to
29 suspect that these working populations would be at a higher risk for AIDS than was the general
30 population, the comparison group.
31 Wong and Trent also reported a slightly increased risk of cancer in other lymphatic tissue
32 (14 observed vs. 11.39 expected). In men, the risk was nonsignificantly higher (11 observed vs.
33 5.78 expected). Forty-three lymphopoietic cancers were observed versus 42 expected. In men,
34 the risk was higher (32 observed vs. 22.22 expected). Fourteen leukemia deaths were noted
35 versus 16.2 expected. The authors did not derive individual exposure estimates for exposure-
08/29/06 109 DRAFT—DO NOT CITE OR QUOTE
-------
1 response analysis, such as in Stayner et al. (1993). Rather, they used duration of employment as a
2 surrogate for exposure.
3 This study has many of the same limitations as the Stayner et al. (1993) study. The
4 authors assumed that those individuals with an unknown vital status as of the cut-off date were
5 alive for the purposes of the analysis, and they were unable to obtain cause of death information
6 on 5% of the known deaths.
7 The differences between this cohort study and that of Stayner et al. (1993) are in the
8 methods of analysis. Stayner et al. used the 9th revision of the International Classification of
9 Diseases (ICD) to develop their site-specific cancer categories for comparison with expected
10 cancer mortality, whereas Wong and Trent used the 8th revision. This could account for some of
11 the differences in the observed numbers of site-specific cancers, because minor differences in the
12 coding of underlying cause of death could lead to a shifting of some unique causes from one site-
13 specific category to another. Furthermore, Wong and Trent did not analyze separately the
14 category "lymphoid" neoplasms, which includes lymphocytic leukemia and NHL, whereas
15 Stayner et al. did. Stayner et al. further developed cumulative exposure information using
16 exposure estimates, whereas Wong and Trent used length of employment as their surrogate for
17 exposure but did not code detailed employment histories.
18 Because Wong and Trent made no effort to quantify the exposures, as was the case in
19 Stayner et al., this study is less useful in determining a exposure-response relationship.
20 Furthermore, the assumption that a member of the cohort should be considered alive if a death
21 indication could not be found will potentially tend to bias risk ratios downward if, in fact, a large
22 portion of this group is deceased. In this study all untraceable persons were considered alive at
23 the end of the followup; therefore, the impact of the additional person-years of risk cannot be
24 gauged.
25
26 A.3.10. BISANTI ET AL. (1993)
27 These authors reported on a cohort mortality study of 1,971 male chemical workers
28 licensed to handle EtO by the Italian government, whom they followed retrospectively from 1940
29 to 1984. Altogether, 76 deaths had occurred in this group by the end of the study period, whereas
30 98.8 were expected. Of those, 43 were due to cancer versus 33 expected. The cause of one death
31 remained unknown, and 16 workers were lost to followup. A group of 637 individuals from this
32 cohort was licensed to handle only EtO; the remaining 1,334 had licenses valid for handling other
33 toxic gases as well. Date of licensing for handling EtO became the initiating point of exposure to
34 EtO, although it is likely that some of these workers had been exposed previously to EtO. The
35 regional population of Lombardia was used as the reference group from which comparison death
36 rates were obtained.
08/29/06 110 DRAFT—DO NOT CITE OR QUOTE
-------
1 Although there were excess risks from almost all cancers, one of the greatest SMRs was in
2 the category known as "all hematopoietic cancers," where 6 observed deaths occurred when only
3 2.4 were expected (SMR = 2.5). In the subgroup "lymphosarcoma, reticulosarcoma" there were 4
4 observed deaths whereas only 0.6 were expected (SMR = 6.7,/><0.05); the remaining 2 were
5 leukemias. The authors note that five hematopoietic cancers occurred in the subgroup of workers
6 who were licensed to handle only EtO but no other chemicals versus only 0.7 hematopoietic
7 cancers expected (SMR = 7.1,/?<0.05). These deaths occurred within 10 years from date of
8 licensing (latent period), which is consistent with the shorter latent period anticipated for this kind
9 of cancer. According to the authors, all workers began their employment in this industry when
10 the levels of EtO were high, although no actual measurements were available. The fact that this
11 subgroup of workers was licensed only for handling EtO reduces the likelihood of a confounding
12 chemical influence.
13 The authors concluded that the excess risk of cancer of the lymphatic and hematopoietic
14 tissues in these particular EtO cohort members support the suggested hypothesis of a higher risk
15 of cancer found in earlier studies, but they added that the lack of exposure information on the
16 other industrial chemicals in the group that had a license for handling other toxic chemicals made
17 their findings inconclusive.
18 This study was of a healthy young cohort, and most person-years were contributed in the
19 latter years of observation. Many years of followup may be necessary in order to fully verify any
20 trend of excess risks for the site-specific cancers of interest and to measure latent effects.
21 Furthermore, the unusual deficit of total deaths versus expected contrasted with an excess of
22 cancer deaths versus expected raises a question about the potential for selection bias when the
23 members of this cohort were chosen for inclusion. Also, one of the study's major limitations is
24 the lack of exposure data.
25
26 A.3.11. HAGMAR ET AL. (1991,1995)
27 Cancer incidence was studied in a cohort of 2,170 EtO-exposed workers from two plants
28 in Sweden that produced disposable medical equipment. To fit the definition for inclusion, the
29 subjects, 1,309 women and 861 men, had to have been employed for a minimum of 12 months
30 and some part of that employment had to have been during the period 1970-1985 in the case of
31 one plant and 1965-1985 in the case of the other. The risk ratios were not dichotomized by
32 gender. No records of anyone who left employment or died before January 1, 1972 in one plant
33 and January 1, 1975 in the other were included. Expected incidence rates were generated from
34 the Southern Swedish Regional Tumor Registries.
35 Because of a short followup period and the relative young age of the cohort, little
36 morbidity had occurred by the end of the cutoff date of December 31, 1990. Altogether, 40
08/29/06 111 DRAFT—DO NOT CITE OR QUOTE
-------
1 cancers occurred, compared with 46.3 expected. After 10 years latency, 22 cases of cancers were
2 diagnosed versus 22.6 expected. However, 6 lymphohematopoietic tumors were observed versus
3 3.37 expected, and when latency is considered, this figure falls to 3 versus 1.51 expected. The
4 authors pointed out that for leukemia the standard incidence ratio (SIR) is a nonsignificant 7.14,
5 based on 2 cases in 930 subjects having at least 0.14 ppm-years of cumulative exposure to EtO
6 and a minimum of 10 years latency. The authors believed that the results provided some minor
7 evidence to support an association between exposure to EtO and an increased risk of leukemia.
8 However, for breast cancer, no increase in the risk was apparent for the total cohort (SIR = 0.46,
9 OBS = 5). Even in the 10 years or more latency period, the risk was less than expected (SIR =
10 0.36, OBS = 2).
11 The authors made a reasonably good attempt to determine exposure levels during the
12 periods of employment in both plants for six job categories. Sterilizers in the years 1970-1972
13 were exposed to an average 40 ppm in both plants. These levels gradually dropped to 0.75 ppm
14 by 1985-1986. Packers and developmental engineers were the next highest exposed employees,
15 with levels in 1970-1972 of 20 to 35 ppm and by 1985-1986 of less than 0.2 ppm. During the
16 period 1964-1966 in the older plant, EtO levels averaged 75 ppm in sterilizers and 50 ppm in
17 packers. Peak exposures were estimated to have ranged from 500 to 1,000 ppm during the
18 unloading of autoclaves up to 1973. The levels gradually dropped to less than 0.2 ppm in both
19 plants by 1985-1986 in all job categories (developmental engineers, laboratory technicians, repair
20 men, store workers, controllers, foremen, and others) except sterilizers.
21 These exposure estimates were verified by measurement of hydroxy ethyl adducts to N-
22 terminal valine in hemoglobin in a sample of subjects from both plants. The adduct levels reflect
23 the average exposure during the few months prior to the measurement of EtO. The results of this
24 comparison were close except for sterilizers, whose air monitoring measurements were 2 to 3
25 times higher.
26 The authors pointed out two limitations in their study: a minority of subjects had a high
27 exposure to EtO, and the median followup (11.8 years) was insufficient to assess a biologically
28 relevant induction latency period. Although this study has good exposure information and the
29 authors used this information to develop an exposure index per employee, they did not evaluate
30 dose-response relationships that might have been present, nor did they follow the cohort long
31 enough to evaluate morbidity. The strength of this study is the development of the cumulative
32 exposure index as well as the absence of any potential confounding produced by the chlorohydrin
33 process, which was a problem in workers who produced and manufactured EtO in other similar
34 studies.
35
08/29/06 112 DRAFT—DO NOT CITE OR QUOTE
-------
1 A.3.12. NORMAN ET AL. (1995)
2 These authors conducted a mortality/incidence study in a cohort of 1,132 workers, mainly
3 women (82%), who were exposed to EtO at some time during the period July 1, 1974, through
4 September 30, 1980. Followup was until December 31, 1987. Ethylene oxide was used at the
5 study plant to sterilize medical equipment and supplies that were assembled and packaged there.
6 This plant was selected for the study because in an earlier small study at this plant (Stolley et al.,
7 1984) there was an indication that in a sample of workers the average number of sister chromatid
8 exchanges was elevated over that of a control group selected from the nearby community. Cancer
9 morbidity was measured by comparing cancers occurring in this cohort with those predicted from
10 the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program for
11 the period 1981-1985 and with average annual cancer incidence rates for western New York for
12 1979-1984. Observed cancers were compared to expected cancers using this method.
13 Only 28 cancer diagnoses were reported in the cohort; 12 were for breast cancers. Breast
14 cancer was the only cancer site in this study where the risk was significantly elevated, based on
15 the SEER rates (SIR = 2.55,/><0.05). No significant excesses were seen at other cancer sites of
16 interest: leukemia (1 observed, 0.54 expected), brain (0 observed, 0.49 expected), pancreas (2
17 observed, 0.51 expected) and stomach (0 observed, 0.42 expected). The authors offered no
18 explanation except chance as to why the risk of breast cancer was elevated in these workers.
19 In 1980, three 2-hour samples from the plant provided 8-hour TWA exposures to sterilizer
20 operators that ranged from 50 to 200 ppm. Corrective action reduced the levels to 5 to 20 ppm.
21 This study has little power to detect any significant risk of cancer at other sites because
22 morbidity was small, chiefly as a consequence of the short followup period. The mean number of
23 years from the beginning of followup to the end of the study was 11.4 years. In fact, the authors
24 stated that breast cancer was the only cancer site for which there was adequate power to detect an
25 increased relative risk. Additional weaknesses in this study include no historic exposure
26 information and too short a period of employment in some cases (<1 month) to result in breast
27 cancer. The authors maintained that their study was inconclusive.
28
29 A.3.13. SWAEN ET AL. (1996)
30 A significant cluster of 10 Hodgkin's disease cases in the active white male workforce of
31 an unidentified large chemical manufacturing plant in Belgium led to a nested case control study
32 by Swaen et al. (1996) to determine which, if any, chemical agents within the plant may have led
33 to the increase. By comparison with regional cancer incidence rates, the SIR for this disease was
34 4.97 (95% CI = 2.38-9.15) over a 23-year period, from 1966 to 1992. This suggested that an
35 occupational exposure may have produced the significant excess risk of Hodgkin's disease seen
36 in these workers.
08/29/06 113 DRAFT—DO NOT CITE OR QUOTE
-------
1 The investigators randomly selected 200 individuals from a computerized sampling frame
2 of all men ever employed at the facility. From this list of 200, workers who were actively
3 employed at the time of diagnosis of each case were chosen as controls. No age matching was
4 done because the authors stated that age-specific incidence rates for Hodgkin's lymphoma in the
5 United States were relatively flat for men between ages 18 and 65. The investigators felt that a
6 control could serve for more than one case.
7 Verification of the 10 cases revealed that 1 case was, in reality, a large-cell anaplastic
8 lymphoma. Two others could not be confirmed as Hodgkin's lymphoma due to the lack of tissue.
9 The remaining 7 were confirmed as Hodgkin's disease. In the ensuing case-control analysis,
10 significant odds ratios (ORs) for Hodgkin's disease were observed for five chemicals, ammonia
11 (6 cases, OR = 5.6), benzene (5 cases, OR = 11), EtO (3 cases, OR =8.5), NaOH (5 cases, OR =
12 8) and oleum (3 cases, OR = 6.9), based on the number of cases and controls known to be
13 exposed to the chemicals in question. This does not mean they were exposed only to the chemical
14 in question.
15 The availability of exposure information made it possible to calculate cumulative
16 exposure to the cases and controls of two chemicals, benzene and EtO. The cumulative exposure
17 for benzene-exposed cases was 397.4 ppm-months versus an expected 99.7 ppm-months for the
18 matched controls. The authors stated that one heavily exposed case was chiefly responsible for
19 the high cumulative total for all the benzene-exposed cases; however, it was not statistically
20 significant. Only a few studies have suggested that exposure to benzene could possibly be related
21 to an increase in the risk of Hodgkin's disease. The cumulative total exposure to EtO for the
22 cases was 500.2 ppm-months versus 60.2 for the matched controls, which was statistically
23 significant, the significance being due to one extreme case.
24 This study is limited because the authors enumerated only cases among active employees
25 of the workforce; therefore, the distinct possibility exists that they could have missed potential
26 cases in the inactive workers. It is possible that latent Hodgkin's disease cases could have been
27 identified in the controls after the controls left active employment. However, given that there
28 were many different possible exposures to the chemicals produced in the workplaces of these
29 employees, it is not likely that EtO or benzene could be considered solely responsible for the
30 excess risk of Hodgkin's disease in this working group.
31
32 A.3.14. OLSEN ET AL. (1997)
33 Olsen et al. (1997) studied 1,361 male employees of four plants in Texas, Michigan, and
34 Louisiana who were employed a minimum of 1 month sometime during the period 1940 through
35 1992 in the ethylene chlorohydrin and propylene chlorohydrin process areas. These areas were
08/29/06 114 DRAFT—DO NOT CITE OR QUOTE
-------
1 located within the EtO and propylene oxide production plants. Some 300 deaths had occurred by
2 December 31, 1992.
3 Plant A in Texas produced EtO beginning in 1941 and ceased production in 1967. Bis-
4 chloroethyl ether, a byproduct of EtO continued to be produced at this plant until 1973. The plant
5 was demolished in 1974. Plant B, which was nearby, manufactured EtO from 1951 to 1971 and
6 then again from 1975 until 1980. This plant continues to produce propylene oxide. The
7 Louisiana plant produced EtO and propylene oxide through the propylene chlorohydrin process
8 from 1959 until 1970, when it was converted to propylene oxide production. The Michigan plant
9 produced ethylene chlorohydrin and subsequently EtO beginning in 1936 and continuing into the
10 1950s. This plant produced propylene chlorohydrin and propylene oxide up to 1974.
11 The authors noted that exposure to EtO was likely at the plants studied in this report, but
12 exposure was unlikely in the 278 chlorohydrin unit workers who were excluded from the cohort
13 studied by Teta et al. (1993). Unfortunately, no actual airborne measurements were reported by
14 Olsen et al., and thus only length of employment could be used as a surrogate for exposure.
15 The SMR for all causes was 0.89 (300 observed). For total cancer the SMR was 0.94 (75
16 observed, 79.7 expected). There were 10 lymphohematopoietic cancers versus 7.7 expected
17 (SMR = 1.3). No significantly increased risks of any examined site-specific cancer (pancreatic,
18 lymphopoietic, hematopoietic, and leukemia) were noted even after a 25-year induction latency
19 period, although the SMR increased to 1.44 for lymphopoietic and hematopoietic cancer. When
20 only the ethylene chlorohydrin process was examined after 25 years latency, the SMR increased
21 to 1.94, based on six observed deaths. The data to support the latter observation by the authors
22 were not presented in tabular form.
23 The authors concluded that there was a weak, nonsignificant, positive association with
24 duration of employment for lymphopoietic and hematopoietic cancer with Poisson regression
25 modeling. They stated that the results of their study provide some assurance that their cohort has
26 not experienced a significant increased risk for pancreatic cancer and lymphopoietic and
27 hematopoietic cancer in ethylene chlorohydrin and propylene chlorohydrin workers. They
28 believed that this study contradicted the conclusions of Benson and Teta (1993) that ethylene
29 dichloride, in combination with chlorinated hydrocarbons, appeared to be the causal agent in the
30 increased risk of pancreatic cancer and hematopoietic cancer seen in their study.
31 Although the authors did not specifically state that ethylene dichloride was a byproduct of
32 the chlorohydrin process in the plants they studied, the clear implication was that it must have
33 been present. They pointed out that ethylene dichloride is readily metabolized and rapidly
34 eliminated from the body after gavage or inhalation administration; therefore, they questioned
35 whether experimental gavage studies (NCI, 1978) are appropriate for studying the effects of
36 ethylene dichloride in humans. One study (Maltoni et al., 1980) found no evidence of tumor
08/29/06 115 DRAFT—DO NOT CITE OR QUOTE
-------
1 production in rats and mice chronically exposed to ethylene dichloride vapor concentrations up to
2 150 ppm for 7 hours a day. Also, because this chemical is a precursor in the production of vinyl
3 chloride monomer, the authors wondered why an increase in these two site-specific cancers had
4 not shown up in studies of vinyl chloride workers. However, they believe that an additional 5 to
5 10 years of followup of this cohort would be necessary to confirm the lack of risk for the two
6 types of cancer described above.
7 Another major weakness of this study is the lack of any actual airborne measurements of
8 EtO and the chlorohydrin chemicals.
9
10 A.3.15. STEENLAND ET AL. (2004)
11 In an update of the earlier mortality studies of the same cohort of workers exposed to EtO
12 described by Steenland et al. (1991) and Stayner et al. (1993), an additional 11 years of followup
13 were added. This increased the number of deceased to 2,852. Work history data were originally
14 gathered in the mid-1980s. Approximately 25% of the cohort continued working into the 1990s.
15 Work histories on these individuals were extended to the last date employed. It was assumed that
16 these employees continued in the job they last held in the 1980s. Little difference was noted
17 when cumulative exposure was calculated with and without the extended work histories, chiefly
18 because the exposure levels after the mid-1980s were very low. Again overall, no excess risk of
19 hematopoietic cancer was noted based on external rates. However, as in the earlier paper,
20 exposure-response analyses reported positive trends for hematopoietic cancers limited to males
21 (p=0.02 for the log of cumulative exposure with a 15-year lag) using internal comparisons and
22 Cox regression analysis. (See Table A-3 for the categorical exposure results.)
Table A-3. Cox regression results for hematopoietic cancer mortality
(15-year lag) in males
Cumulative exposure (ppm-days)
>0-1,199
1,200-3,679
3,680-13,499
13,500+
Odds ratio (95% CI)
1
1.23(0.32-4.73)
2.52 (0.69-9.22)
3.13(0.95-10.37)
3.42(1.09-10.73)
08/29/06
116
DRAFT—DO NOT CITE OR QUOTE
-------
1 The excess of these tumors was chiefly lymphoid (NHL, myeloma, lymphocytic leukemia) (see
2 Table A-4), as in the earlier paper. A positive trend was also observed for Hodgkin's disease in
3 males, although this was based on small numbers.
Table A-4. Cox regression results for lymphoid cell line tumors (15-year lag)
in males
Cumulative exposure (ppm-days)
0
>0-1,199
1,200-3,679
3,680-13,499
13,500+
Odds ratio (95% CI)
1
0.9(0.16-5.24)
2.89(0.65-12.86)
2.74(0.65-11.55)
3.76(1.03-13.64)
7 The hematopoietic cancer trends were somewhat weaker in this analysis than were those reported
8 in the earlier studies of the same cohort. This is not unexpected because most of the cohort was
1 not exposed after the mid-1980s, and the workers who were exposed in more recent years were
2 exposed to much lower levels because EtO levels decreased substantially in the early 1980s. No
3 association was found in females, although average exposures were only twice as high in males
4 (37.8 ppm-years) as in females (18.2 ppm-years), and there was enough variability in female
5 exposure estimates to expect to be able to see a similar trend if it existed.
6 This study also reports a significant excess risk of breast cancer in the highest cumulative
7 exposure quartile, with a 20-year lag (SMR = 2.07, 95% CI 1.1-3.54, n = 13) in female
8 employees. The results using internal Cox regression analyses with a 20-year lag time produced
9 anOR=3.13 (95% CI 1.42-6.92) in the highest cumulative-exposure quartile. The log of
10 cumulative exposure with a 20-year lag was found to be the best model (p=0.01) for the analyses
11 of breast cancer. As for hematopoietic cancer in males, cumulative exposure untransformed
12 showed a weaker trend (p=0.16). A breast cancer incidence study of this cohort is discussed in
13 Steenland et al. (2003).
14
15 A.3.16. STEENLAND ET AL. (2003)
16 In a companion study on breast cancer incidence in women employees of the same cohort
17 discussed in Steenland et al. (2004), the authors elaborated on the breast cancer findings in a
18 subgroup of 7,576 women from the cohort (76% of the original cohort). They had to be employed
08/29/06
117
DRAFT—DO NOT CITE OR QUOTE
-------
1 at least 1 year and exposed while employed in commercial sterilization facilities. The average
2 length of exposure was 10.7 years. Breast cancer incidence analyses were based on 319 cases
3 identified via interview, death certificates, and cancer registries in the full cohort, including 20 in
4 situ carcinomas. Interviews on 5,139 women (68% of the study cohort) were obtained; 22% could
5 not be located. Using external referent rates (SEER), the SIR was 0.87 for the entire cohort based
6 on a 15-year lag time. When in situ cases were excluded, the overall SIR increased to 0.94. In the
7 top quintile of cumulative exposure, with a 15-year lag time, the SIR was 1.27 (95% CI
8 0.94-1.69, n = 48). A significant positive linear trend of increasing risk with increasing
9 cumulative exposure was noted (p=0.002) with a 15-year lag time. Breast cancer incidence was
10 believed to be underascertained owing to incomplete response and a lack of coverage by regional
11 cancer registries (68% were contacted directly and 50% worked in areas with cancer registries).
12 An internal nested case-control analysis, which is less subject to concerns about
13 underascertainment, produced a significant positive exposure-response with the log of cumulative
14 exposure and a 15-year lag time (p=0.05). The top quintile was significant with an OR of 1.74 (CI
15 1.16-2.65) based on all 319 cases (the entire cohort).
16 The authors also conducted separate analyses using the subcohort with interviews, for
17 which there was complete case ascertainment and additional information on potential
18 confounders. In the subcohort with interview data, the odds ratio for the top quintile equaled 1.87
19 (CI 1.12-3.1), based on 233 cases in the 5,139 women and controlled for with respect to parity
20 and breast cancer in a first-degree relative. Information on other risk factors was also
21 collected—e.g., body mass index, SES, diet, age at menopause, age at menarche, breast cancer in
22 a first-degree relative, and parity—but only parity and breast cancer in a first-degree relative were
23 significant in the model. Continuous cumulative exposure, as well as the log cumulative
24 exposure, lagged 15 years, produced/* values for the regression coefficient of 0.02 and 0.03,
25 respectively, for the Cox regression model, taking into account age, race, year of birth, parity, and
26 breast cancer in a first-degree relative.
27 The authors concluded that their data suggest that exposure to EtO is associated with
28 breast cancer, but because of inconsistencies in exposure-response trends and possible biases due
29 to nonresponse and incomplete cancer ascertainment, the case for breast cancer is not conclusive.
30 However, monotonically increasing trends in categorical exposure-response relationships are not
31 always the norm owing to lack of precision in the estimates of exposure. Furthermore, positive
32 trends were observed in both the full cohort and the subcohort with interviews, lessening concerns
33 about nonresponse bias and case underascertainment.
34
08/29/06 118 DRAFT—DO NOT CITE OR QUOTE
-------
1 A.3.17. KARDOS ET AL. (2003)
2 These authors reported on a study completed earlier by Muller and Bertok (1995) of cancer
3 among 299 female workers who were employed from 1976 to 1993 in a pediatric ward at the
4 county hospital in Eger, Hungary, where gas sterilizers were used. Their observation period for
5 cancer was begun in 1987 on the assumption that cancer deaths before 1987 were not due to EtO,
6 based on a paper by Lucas and Teta (1996). Information about the Muller and Bertok (1995)
7 study is unavailable because the paper is in Hungarian and no translated copy is available. Kardos
8 and his colleagues evaluated mortality among these women and found a statistically significant
9 excess of total cancer deaths in the period from 1987 to 1999 when compared with expected
10 deaths generated from three different comparison populations (Hungary, Heves County, and city
11 of Eger). Altogether, 11 deaths were observed compared with, respectively, 4.38, 4.03, and 4.28
12 expected deaths. The SMRs are all significant at the/? < 0.01 level. Site-specific rates were not
13 calculated. Among the 11 deaths were 3 breast cancer deaths and 1 lymphoid leukemia death.
14 The authors claim that their results confirm "predictions of an increased cancer risk for the Eger
15 hospital staff." They suggest an etiological role for EtO in the excess risk.
16
17 A.3.18. TOMPA ET AL. (1999)
18 The authors reported a cluster of 8 breast cancer cases and 8 other malignant tumor cases
19 that developed over a period of 12 years in 98 nurses who worked in a hospital in the city of Eger,
20 Hungary, and were exposed to EtO. These nurses were exposed for 5 to 15 years in a unit using
21 gas sterilizer equipment. The authors report that EtO concentrations were in the neighborhood of
22 5 to 150 mg/m3. The authors state that the high breast cancer incidence in the hospital in Eger
23 indicates a combined effect of exposure to EtO and naturally occurring radioactive tap water,
24 possibly due to the presence of radon. This case report study is discussed further in the
25 genotoxicity section.
26
27 A.3.19. COGGON ET AL. (2004)
28 Descriptive information about this cohort is available from the earlier study (Gardner et
29 al., 1989). This current update of the 1,864 men and 1,012 women described in the Gardner et al.
30 study were followed to December 31, 2000. This added 13 more years of followup resulting in
31 565 observed deaths versus 607.6 expected. For total cancer, the observed number of deaths
32 equaled 188 versus 184.2 expected. For NHL, 7 deaths were observed versus 4.8 expected. For
33 leukemia, 5 deaths were observed versus 4.6 expected. All 5 leukemia deaths fell into the subset
34 with definite or continual exposure to EtO, where only 2.6 were expected. In fact, the total
35 number of deaths classified to the lymphohematopoietic cancer category was 17 with 12.9
36 expected. This increased risk was not significant. When definite exposure was established, the
08/29/06 119 DRAFT—DO NOT CITE OR QUOTE
-------
1 authors found that the risk of lymphatic and hematopoietic cancer was increased with 9 observed
2 deaths versus 4.9 expected. Deaths from leukemia were also increased in chemical workers with
3 4 leukemia deaths versus 1.7 expected. No increase was seen in the risk of hematopoietic cancer
4 in the hospital sterilizing unit workers, who are mostly female. Another finding of little
5 significance was that of cancer of the breast. Only 11 deaths were recorded in this cohort up to
6 the cutoff date versus 13.1 expected. Since there were no female workers in the chemical
7 industry, the results on breast cancer reflect only work in hospital sterilizing units. The
8 researchers concluded that the risk of cancer must be low at the levels sustained by workers in
9 Great Britain over the last 10 or 20 years.
10
11 A.4. SUMMARY
12 The initial human studies by Hogstedt et al. (1979a, b, 1986) and Hogstedt (1988), in
13 which positive findings of leukemia and blood-related cancers suggested a causal effect, have
14 been followed by studies that either do not indicate any increased risks of cancer or else suggest a
15 dose-related increased risk of cancer at certain sites. These are chiefly cancers of the
16 hematopoietic system and include leukemia, lymphosarcoma, reticulosarcoma and NHL. More
17 recently, an association with breast cancer has also been suggested. However, the evidence is not
18 conclusive because of inadequacies and limitations in the epidemiological database. The main
19 effects and limitations are presented in Table A-5.
20 Exposure information, where available, indicates that levels of EtO probably were not high
21 in these study cohorts. If a causal relationship exists between exposure to EtO and cancer, the
22 reported EtO levels may have been too low to produce a significant finding. Exposures in the
23 earlier years (prior to 1970) in most of the companies, hospitals, and other facilities where EtO
24 was made or used are believed to have been in the range of 20 ppm, with excursions many times
25 higher, although few actual measurements are available during this period. (One exception is the
26 environmental study by Joyner (1964), who sampled airborne levels of EtO from 1960 to 1962 in
27 a Texas City facility owned by Union Carbide.)
28 Almost all actual measurements of EtO were taken in the 1970s and 1980s at most plants
29 and facilities in the U.S. and Europe, and levels have generally fallen to 5 ppm and below. Some
30 plants may have never sustained high levels of airborne EtO. Assuming that there is a true risk of
31 cancer associated with exposure to EtO, then the risk is not evident at the levels that existed in
32 these plants except under certain conditions, possibly due to a lack of sensitivity in the available
33 studies to detect associated cancers at low exposures.
34 The best evidence of a dose-response relationship comes from the NIOSH study of
35 sterilizer workers by Steenland et al. (2004, 1991) and Stayner et al. (1993). This study estimated
36 cumulative exposure (i.e., total lifetime occupational exposure to EtO) in every member of the
08/29/06 120 DRAFT—DO NOT CITE OR QUOTE
-------
1 cohort. The investigators quantified exposure from the best available data on airborne levels of
2 EtO throughout the history of the plants and used regression techniques to estimate individual EtO
3 exposures. Industrial hygienists from NIOSH met with plant personnel to identify workers who
4 were potentially exposed to EtO. An added advantage to this study, besides its diversity and size,
5 is the absence of other known confounding exposures in the plants, especially benzene.
6 In the recent followup of the NIOSH cohort, as in the earlier study, Steenland et al. (2004)
7 observed no overall excess of hematopoietic cancers (ICD-9 codes 200-208). In internal analyses,
8 however, they found a significant positive trend (p=0.02) for hematopoietic cancers for males
9 only, using log cumulative exposure and a 15-year lag, based on 37 male cases. In the Cox
10 regression analysis using categorical cumulative exposure and a 15-year lag, a positive trend was
11 observed and the OR in the highest exposure quintile was statistically significant (OR = 3.42;
12 95% CI 1.09-10.73). Similar results were obtained for the "lymphoid" category (lymphocytic
13 leukemia, NHL, and myeloma). No evidence of a relationship between EtO exposure and
14 hematopoietic cancers in females in this cohort was observed. The reasons for this discrepancy
15 are unknown.
16 In the analysis by Teta et al. (1999) of UCC workers, the authors discussed the
17 development of an age-dependent exposure history on each worker at the facility in West Virginia
18 (Greenberg et al., 1990; Benson and Teta, 1993; Teta et al., 1993), based on departmental
19 assignments and estimated exposure levels. Eight-hour TWA concentrations (ppm) were
20 estimated over four time periods (1925-1939, 1940-1956, 1957-1973, and 1974-1978) at the two
21 facilities in three exposure intensity categories (high, medium, low exposure departments) defined
22 in the earlier study of the same two plants by Greenberg et al. Average exposures in the latter
23 time period (1974-1978) were based on industrial hygiene monitoring conducted at the locations
24 where the study subjects worked. Estimates for the earlier time periods were inferred from data
25 on airborne exposure levels in "similar" manufacturing operations during the time periods of
26 interest. These estimates are from the EtO production facility at Texas City studied by Joyner
27 (1964) and the Swedish company described by Hogstedt et al. (1979b).
28 These inferred estimates of exposure formed the basis of the UCC dose-response
29 assessment of the UCC study described in Teta et al. (1999). The authors fit several different
30 models to the UCC exposure data and to the earlier NIOSH data (Stayner et al., 1993) and used
31 Poisson regression techniques to estimate site-specific cancer risks for leukemia and "lymphoid"
32 cancer. The results for leukemia were similar using either source of information (UCC or
33 NIOSH) but were very different for "lymphoid" tumors due to the fact that none of the leukemia
34 death certificates in the UCC study specified "lymphocytic" leukemia as the histologic type,
35 whereas most of the NIOSH death certificates listed "lymphocytic" leukemia on the death
36 certificate. The UCC data produced a "flat" relationship between ppm-years and the risk of
08/29/06 121 DRAFT—DO NOT CITE OR QUOTE
-------
1 "lymphoid" cancer. The analysis of the NIOSH data produced an increased risk of "lymphoid"
2 cancer with increasing exposure.
3 The many different analyses of the UCC data completed by Teta et al. (1999) are
4 weakened by the reliance on inferred exposure data from other plants, mainly Joyner (1964) and
5 Hogstedt et al. (1979a), which may not have been as similar as they assume to those for the
6 employees of the two UCC plants that provided the basis of their own risk assessment. Although
7 the workers at the Texas plant were all production workers, descriptions of the job processes at
8 the UCC facilities in West Virginia lead one to believe that only a minority of the workers there
9 were production workers, based on Teta et al. (1993, 1999). Most of the workers at the West
10 Virginia facilities appear to have been employed in the use of EtO but not in the making of it.
11 Greenberg et al. (1990), in their partitioning of the job categories at the West Virginia
12 facilities, considered production workers as "highly" exposed to EtO, whereas those who only
13 used it were classified in a "medium" or "low" exposure category. Because of this potential
14 difference in exposure between the two types of workers, it may not be appropriate to assume that
15 the workers at the West Virginia facilities had "similar" exposures to the workers at the Texas
16 plant prior to 1976. If, in fact, the workers at the West Virginia facilities had generally lower
17 exposures, then risk calculations based on extrapolating levels from the Texas plant to the West
18 Virginia plants could potentially underestimate the risk of cancer in those workers.
19 The NIOSH investigators developed estimates of exposure on the basis of knowledge and
20 information developed from plant personnel at each of the 14 chemical plants in the cohort and
21 meetings with NIOSH industrial hygienists. Furthermore, the NIOSH cohort was a much larger,
22 more diversified group of workers who have been exposed to fewer confounders that might have
23 skewed the findings. Hence, the Teta et al. (1999) exposure estimates are potentially less reliable
24 than those of Steenland et al., despite the extensive risk assessment analyses and meta-analysis in
25 Teta et al.
26 Although not explicitly citing any numbers, Teta et al. (1999) concluded that there was no
27 evidence that EtO causes brain, stomach, or pancreatic cancer. They also concluded that their
28 meta-analysis and tests of heterogeneity provide "compelling" evidence that the high risk of
29 leukemia seen in the Hogstedt studies was an incorrect inference. They stated that findings for
30 cancers of the lymphopoietic tissues (leukemia, NHL) were "inconclusive." Yet, in contrast, they
31 concluded that EtO is a "probable human carcinogen," based partially on "limited evidence in
32 humans."
33 One other study that provides cumulative exposure estimates is the incidence study by
34 Hagmar et al. (1991, 1995). The short followup period and relative youthfulness of the cohort
35 produced little morbidity by the end of the study, although some support for an excess risk of
36 leukemia and lymphohematopoietic cancer had appeared.
08/29/06 122 DRAFT—DO NOT CITE OR QUOTE
-------
1 In a separate analysis of the NIOSH cohort by Wong and Trent (1993), duration of
2 exposure to EtO was used as a surrogate for exposure. These authors did not find any positive
3 exposure-response relationships. They did observe an elevated significant risk of "NHL" in males
4 (SMR = 2.47,/?<0.05), based on 16 deaths, which was not dose- related or time-related.
5 However, a deficit in females remained.
6 Increases in the risk of hematopoietic cancers or Hodgkin's lymphoma are also suggested
7 in several other studies (Gardner et al., 1989; Coggon et al., 2004; Norman et al., 1995; Bisanti et
8 al., 1993; Swaen et al., 1996; Olsen et al., 1997). However, in all these studies the deaths were
9 few and the risk ratios were mostly nonsignificant except at higher estimated exposures or after
10 long observation periods. They were not robust and there were potentially confounding
11 influences, such as exposure to benzene and/or chlorohydrin derivatives.
12 In those plants where there were no detectable risks (Kiesselbach et al., 1990; Norman et
13 al., 1995), the cohorts were relatively youthful or had not been followed for a sufficient number of
14 years to observe any effects from exposure to EtO. In the study by Olsen et al. (1997), although a
15 slight increase in the risk of cancer of the lymphopoietic and hematopoietic system was evident,
16 the authors stated that their study provided some assurance that working in the chlorohydrin
17 process had not produced significantly increased risks for pancreatic cancer or lymphopoietic or
18 hematopoietic cancer, thus contradicting the findings of Benson and Teta (1993). This study lacks
19 any measurement of airborne exposure to any of the chemicals mentioned and the authors
20 indicated that an additional 5 to 10 years of followup would be needed to confirm the lack of a
21 risk for the cancers described in their study.
22 Although the strongest evidence of a cancer risk is with cancer of the hematopoietic
23 system, there are indications that the risk of stomach cancer may have been elevated in some
24 studies (Hogstedt et al., 1979a, 1986; Kiesselbach et al., 1990; Teta et al., 1993); however, it
25 attained significance only in the study by Hogstedt et al. (1979a), with 9 observed versus 1.27
26 expected. It was reported by Shore et al. (1993) that this excess may have been due to the fact that
27 early workers at this plant "tasted" the chemical reaction product to assess the result of the EtO
28 synthesis. This reaction mix would have contained ethylene dichloride and bis-chloroethyl ether.
29 Ethylene dichloride is a suspected carcinogen, whereas bis-chloroethyl ether is not. This
30 increased risk of stomach cancer was not supported by analyses of intensity or duration of
31 exposure in the remaining studies, except that Benson and Teta (1993) suggested that exposure to
32 this chemical increased the risk of pancreatic cancer and perhaps hematopoietic cancer but not
33 stomach cancer.
34 A significant risk of pancreatic cancer first reported by Morgan et al. (1981) was also
35 reported by Greenberg et al. (1990) in his cohort of chemical workers, but only in those workers
36 assigned to the ethylene chlorohydrin production process, where the authors reported that
08/29/06 123 DRAFT—DO NOT CITE OR QUOTE
-------
1 exposure to EtO was low. Benson and Teta (1993) attributed the increase in pancreatic cancer
2 seen in Greenberg et al. (1990) to exposure to ethylene dichloride in the chlorohydrin process.
3 However, Olson et al. (1997) refuted this finding in their study. The pancreatic cancers from the
4 study by Morgan et al. (1981) also occurred in workers in a chlorohydrin process of EtO
5 production. The possibility that exposure to a byproduct chemical such as ethylene dichloride
6 may have produced the elevated risks of pancreatic cancer seen in these workers cannot be ruled
7 out.
8 In addition to the cancer risks described above, some recent evidence indicates that
9 exposure to EtO may increase the risk of breast cancer. The study by Norman et al. (1995) of
10 women who sterilized medical equipment observed a significant twofold elevated risk of breast
11 cancer, based on 12 cases. A study by Tompa et al. (1999) reported on a cluster of breast cancers
12 occurring in Hungarian hospital workers exposed to EtO. In another Hungarian study of female
13 hospital workers by Kardos et al. (2003), 3 breast cancers were noted out of 11 deaths reported by
14 the authors. Although expected breast cancer deaths were not reported, the total expected deaths
15 calculated was just slightly more than 4, making this a significant finding for cancer in this small
16 cohort.
17 The most compelling evidence on breast cancer comes from the NIOSH cohort. In the
18 recent update of this cohort, no overall excess of breast cancer mortality was observed in the
19 female workers; however, a statistically significant SMR of 2.07 was observed in the highest
20 cumulative exposure quartile, with a 20-year lag. In internal Cox regression analyses, a positive
21 exposure-response (p=0.01) was observed for log cumulative exposure with a 20-year lag, based
22 on 103 cases. Similar evidence of an excess risk of breast cancer was reported in a breast cancer
23 incidence study of a subgroup of 7,576 female workers from the NIOSH cohort who were exposed
24 for 1 year or longer (Steenland et al., 2003). A significant (p=0.002) linear trend in SIR was
25 observed across cumulative exposure quintiles, with a 15-year lag. In internal Cox regression
26 analyses, there was a significant regression coefficient with log cumulative exposure and a 15-
27 year lag, based on 319 cases. Using categorical cumulative exposure, the OR of 1.74 was
28 statistically significant in the highest exposure quintile. In a subcohort of 5,139 women with
29 interviews, similar results were obtained based on 233 cases, and the models for this subcohort
30 were also able to take information on other potential risk factors for breast cancer into account.
31 Additionally, the coefficient for continuous cumulative exposure was also significant (p=0.02),
32 with a 15-year lag.
33 Several other studies with female employees in the defined cohorts reported no increased
34 risks of breast cancer due to exposure to EtO (Coggon et al., 2004; Hogstedt et al., 1986; Hagmar
35 et al., 1991, 1995). However, these studies have much lower statistical power than the NIOSH
36 studies, as evidenced by the much lower numbers of breast cancer cases that they report. The
08/29/06 124 DRAFT—DO NOT CITE OR QUOTE
-------
1 largest number of cases in any of these other studies is 11 cases in the Coggon et al. (2004) study.
2 Furthermore, none of these other studies conducted internal (or external) exposure-response
3 analyses, which are the analyses that provided the strongest evidence in the NIOSH studies.
4
5 A.5. CONCLUSIONS
6 Experimental evidence demonstrates that exposure to EtO in rodents produces
7 lymphohematopoietic cancers; therefore, an increase in the risk of lymphohematopoietic cancer in
8 humans should not be unexpected. An increase in mammary gland carcinomas was also observed
9 in mice. Although several human studies have indicated the possibility of a carcinogenic effect
10 from exposure to EtO, especially for lymphohematopoietic cancers, the total weight of the
11 epidemiologic evidence does not provide sufficient evidence to support a causative determination.
12 The causality factors of temporality, coherence, and biological plausibility are satisfied. There is
13 also evidence of consistency and specificity in the elevated risk of lymphohematopoietic cancer as
14 a single entity in the human studies. The earlier significant risk of leukemia seen in the Hogstedt
15 studies was supported in some studies and not in others. In fact, not all human studies of EtO
16 have suggested an elevated risk of cancer and in those that do, the marginally elevated risks vary
17 from one site to another within the lymphohematopoietic system. When combined under the
18 rubric "lymphohematopoietic cancers," this loosely defined combination of blood malignancies
19 produces a slightly elevated risk of cancer in some studies but not in all. There is evidence of a
20 biological gradient in the significant dose-response relationship seen in the large, high-quality
21 Steenland et al. (2004) study.
22 The best evidence of a carcinogenic effect produced by exposure to EtO is found in the
23 NIOSH cohort of workers exposed to EtO in 14 sterilizer plants around the country (Steenland et
24 al., 1991, 2004; Stayner et al., 1993). A positive trend in the risk of lymphohematopoietic and
25 "lymphoid" neoplasms with increasing log cumulative exposure to EtO with a 15-year lag is
26 evident. But there are some limitations to concluding that this is a causal relationship at this time.
27 For example, there was a lack of dose-response relationship in females.
28 An elevated risk of lymphohematopoietic cancers from exposure to EtO is also apparent in
29 several other studies. In some of these studies, confounding exposure to other chemicals
30 produced in the chlorohydrin process concurrent with EtO may have been partially responsible for
31 the excess risks. In other studies, where the chlorohydrin process was not present, there are no
32 known confounding influences that would produce a positive risk of lymphohematopoietic cancer.
33 Overall, the evidence on lymphohematopoietic cancers in humans is considered to be strong but
34 not sufficient to support a causal association, i.e., limited.
35 There also exists the possibility that exposure to EtO may increase the risk of breast
36 cancer, based chiefly on the Steenland et al. (2003, 2004) studies discussed earlier with some
08/29/06 125 DRAFT—DO NOT CITE OR QUOTE
-------
1 corroborating evidence from the Norman et al. (1995) study of breast cancer in women exposed to
2 EtO. The risk of breast cancer was narrowly analyzed in a few other studies (Hagmar et al., 1991;
3 Hogstedt, 1988; Hogstedt et al., 1986; Coggon et al., 2004), and no increase in the risk of breast
4 cancer was found. However, these studies had far fewer cases to analyze, did not have individual
5 exposure estimates, and relied on external comparisons. The Steenland et al. (2003, 2004)
6 studies, on the other hand, used the largest cohort of women potentially exposed to EtO and
7 clearly show significantly increased risks of breast cancer incidence and mortality, based on
8 internal exposure-response analyses. However, the authors suggest that the case is not conclusive
9 of a causal association "due to inconsistencies in exposure-response trends and possible biases
10 due to non-response and an incomplete cancer ascertainment." While these are not crippling
11 limitations—exposure-response relationships are often not strictly monotonically increasing
12 across finely dissected exposure categories, and the consistency of results between the full cohort
13 (less nonresponse bias) and the subcohort with interviews (full case ascertainment) alleviates
14 some of the concerns about those potential biases—we agree that the evidence for a causal
15 association between breast cancer and EtO exposure is less than conclusive at this time.
08/29/06 126 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer
Population/
industry
Sterilizers,
production
workers, Sweden
Hogstedtetal.,
(1986); Hogstedt
(1988)
Sterilizing
workers in 8
hospitals and
users in 4
companies,
Great Britain
Gardner et al.
(1989)
Number
of
subjects
709
(539
men,
170
women)
2,876
(1,864
men,
1,012
women)
Extent of exposure to
ethylene oxide
Plant 1: mean = 20 ppm
in sterilizer room
Plant 2: mean =14 ppm
in early years, less than 6
ppm later
Plant 3: less than 8 ppm
in early years, less than 2
ppm later
In early years, odor
threshold of 700 ppm
noted; in later years,
5 ppm or less was noted
Health outcomes
33 cancer deaths vs. 20
expected
7 leukemia deaths vs.
0.8 expected
10 stomach cancer
deaths vs. 1.8 expected
3 leukemia deaths vs.
0.35 expected
(after 20+ years latency)
5 esophageal cancer
deaths vs. 2.2 expected
4 bladder cancer deaths
vs. 2.04 expected
4 non-Hodgkin's
lymphoma deaths vs.
1.6 expected
29 lung cancer deaths
vs. 24.6 expected
Other chemicals to which subjects
were potentially exposed
Benzene, methyl formate, bis-(2-
chloroethyl) ether, ethylene, ethylene
chlorohydrin, ethylene dichloride,
ethylene glycol, propylene oxide,
amines, butylene oxide, formaldehyde,
propylene, sodium
Aliphatic and aromatic alcohols,
amines, anionic surfactants, asbestos,
butadiene, benzene, cadmium oxide,
dimethylmine, ethylene, ethylene
chlorohydrin, ethylene glycol,
formaldehyde, heavy fuel oils,
methanol, methylene chloride,
propylene, propylene oxide, styrene,
tars, white spirit, carbon tetrachloride
Limitations
No personal exposure
information from which
to estimate dose
No latency analysis
Mixed exposure to other
chemicals
Insufficient followup
Exposure classification
scheme vague, making it
difficult to develop dose-
response gradient
No exposure measure-
ments prior to 1977, so
individual exposure
estimates were not made
Mixed exposure to several
other chemicals
to
o
o
2
o
H
O
O
&
O
c
o
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Coggon et al.
(2004)
Update of
Gardner et al.
(1989)
Number
of
subjects
Same
cohort
followed
addi-
tional 13
years
Extent of exposure to
ethylene oxide
Ibid.
Health outcomes
Recent Findings
5 leukemia deaths vs.
2.6 expected (definite
or continual exposure)
7 non-Hodgkin's
lymphomavs. 4.8
expected
1 1 breast cancers vs.
13.1 expected
17 hematopoietic
cancers vs. 12.9
expected
9 lymphatic and/or
hematopoietic cancers
vs. 4.9 expected (definite
exposure)
Other chemicals to which subjects
were potentially exposed
Ibid.
Limitations
Ibid, and, in addition,
no latency evaluation
to
oo
fe
H
O
O
o
H
O
HH
H
W
O
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Production
workers
(methods
unspecified)
from 8 chemical
plants in West
Germany
Kiesselbach et al.
(1990)
Number
of
subjects
2,658
men
Extent of exposure to
ethylene oxide
No exposure information
available
Health outcomes
14 stomach cancer
deaths vs. lO.lexpected
3 esophageal cancer
deaths vs. 1.5 expected
23 lung cancer deaths
vs. 19.9 expected
Other chemicals to which subjects
were potentially exposed
Beta-naphthylamine, 4-amino-
diphenyl, benzene, ethylene
chlorohydrin, possibly alkylene oxide
(ethylene oxide/propylene oxide),
based on inclusion of plants that were
part of a cohort study by Thiess et al.
(1982)
Limitations
Insufficient followup;
few expected deaths in
cancer sites of
significance with which
to analyze mortality
Production methods not
stated; information vague
on what these plants do
Latency analysis given
only for total cancer and
stomach cancer mortality
Although categories of
exposure are given, they
are not based on actual
measurements
No actual measurement
data are given; dose-
response analysis is not
possible
to
VO
fe
H
O
O
O
H
O
HH
H
W
O
O
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
_1
>J
D
3
0
>
Tj
H
3
2
Population/
industry
Production
workers and
users at 2
chemical plants
in West Virginia
Greenberg et al.
(1990)
Number
of
subjects
2,174
men
Extent of exposure to
ethylene oxide
Exposure prior to 1976
not known (estimates
based on 1976
measurements and
occurrence of medical
conditions)
1^.3 ppm TWA in
later years
Health outcomes
7 leukemia and
aleukemia deaths vs.
3 expected;
SMR = 2.3
3 liver cancer deaths vs.
1.8 expected;
SMR =1.7
7 pancreatic cancer
deaths vs. 4.1 expected;
SMR= 1.7
Suggestion of increasing
risk of stomach cancer
and leukemia/aleukemia
with cumulative duration
of potential exposure
Other chemicals to which subjects
were potentially exposed
Acetaldehyde, acetonitrile, acrolein,
aldehydes, aliphatic and aromatic
alcohols, alkanolamines, allyl
chloride, amines, butadiene, benzene,
bis-(chloroethyl) ether, ethylene
dichloride, diethyl sulphate, dioxane,
epichlorhydrin, ethylene, ethylene
chlorohydrin, formaldehyde, glycol
ethers, methylene chloride, propylene
chlorohydrin, styrene, toluidine
Limitations
Exceptionally low
exposure levels to
ethylene oxide, less than
1 ppm (from a 1976
survey)
No actual measurements of
exposure to ethylene
oxide exist prior to 1976;
inferences of levels of
exposure to ethylene
oxide at this plant were
assumed to be similar to
exposure levels measured
at two other plants during
early years (prior to
1976)
Exposure occurred to
many other chemicals,
some of which may be
carcinogenic
Lack of quantitative
estimates of individual
exposure levels
-*,
J
-H
n
D
0
D
-H
3
H
n
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Same cohort as
Greenberg et al.
(1990) minus
all chlorohydrin-
exposed
employees,
followed an
additional 10
years
Tetaetal. (1993)
Only the
chlorohydrin-
exposed
employees from
Greenberg et al.
(1990) cohort,
followed an
additional 10
years
Benson and
Teta (1993)
Number
of
subjects
1,896
men
278
men
Extent of exposure to
ethylene oxide
Estimated exposure prior
to 1956: 14+ ppm; after
1956: less than 10 ppm
Prior to 1976 estimates
were based on
measurements taken at
"similar" facilities
Reported to be very low
exposure to ethylene
oxide in the chlorohydrin
process
Health outcomes
Trend of increasing risk
of leukemia and
aleukemia death with
increasing duration of
exposure
8 pancreatic cancer
deaths vs. 1.63 expected
(p<0.05)
8 hematopoietic cancer
deaths vs. 2.72 expected
(p<0.05) SMR = 2.9
Other chemicals to which subjects
were potentially exposed
Same (except for chemicals specific to
the chlorohydrin process)
Same
Limitations
Same
Same
O
O
2
O
H
O
HH
H
W
O
&
O
c
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Sterilizers of
medical
equipment and
spices; and
manufacturers
and testers of
medical
sterilization
equipment,
in 14 plants in
the United States
Steenland et al.
(1991);Stayneret
al. (1993)
Number
of
subjects
18,254
(45%
male,
55%
female)
Extent of exposure to
ethylene oxide
1938-1976 (estimated):
16 ppm for sterilizer
operators, 5 ppm for
remainder
1977-1985 (mean): 4.3
for sterilizers, 2 ppm
for remainder
Individual cumulative
exposure estimates
calculated for workers in
13 of the 14 facilities
Health outcomes
36
(lympho)hematopoietic
cancer deaths vs. 33.8
expected
8 lymphosarcoma and
reticulosarcoma deaths
vs. 5.3 expected
After 20+ years latency,
SMR= 1.76 for
hematopoietic cancer, a
significant trend with
increasing latency
(/?<0.03)
Significantly increasing
hematopoietic cancer
and "lymphoid" cancer
risks with cumulative
exposure
Other chemicals to which subjects
were potentially exposed
No identified exposures to other
chemicals
Limitations
Potential bias due to lack
of followup on
"untraceable" members
(4. 5%) of the cohort
Short duration of exposure
and low median exposure
levels
Individual exposures were
estimated prior to 1976
before first industrial
hygiene survey was
completed
Short followup for most
members of the cohort;
only 8% had attained 20
years latency
Little mortality (6.4%) had
occurred in this large group
of employees
No exposure-response
relationship among female
workers
to
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Same cohort as
Stayner et al.
(1993) and
Steenland et al.
(1991), plus 474
additional
members,
followed 1 more
year
Wong and Trent
(1993)
Number
of
subjects
18,728
(45%
male,
55%
female)
Extent of exposure to
ethylene oxide
Same as Steenland et al.
(1991) and Stayner etal.
(1993)
Health outcomes
16 non-Hodgkin's
lymphoma deaths in
men vs. 6.47 expected
43 lymphohematopoietic
cancer
deaths observed vs.
42 expected (in men
32 observed vs. 22.2
expected)
14 other lymphatic
cancer deaths vs. 11.4
expected (in men 1 1
observed vs. 5.8
expected)
14 leukemia deaths vs.
16.2 expected
Other chemicals to which subjects
were potentially exposed
No identifiable exposures to other
chemicals
Limitations
All of the limitations of
Steenland etal. (1991)
apply here
Although this group is the
same as Steenland et al.
(1991), an additional
unexplained 474
employees were added
It is questionable that one
additional year of followup
added 392.2 expected
deaths but only 176
observed deaths
No effort was made to
develop exposure-
response data such as in
Stayner etal. (1993) on
the basis of individual
cumulative exposure data
but only on duration of
employment
fe
H
O
O
o
H
O
HH
H
W
O
O
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Steenland et al.
(2004)
Update of
Steenland et al.
(1991), Stayneret
al. (1993)
Number
of
subjects
18,254
(45%
male,
55%
female)
Extent of exposure to
ethylene oxide
Same as Steenland et
al. (1991), with extension
of worker histories based
on job held at end of
initial exposure
assessment for those still
employed at end of 1991
study (25% of cohort)
Health outcomes
With 15 -year lag, in
internal Cox regression
analyses, OR = 3.42
(p<0.05) in highest
cumulative exposure
group for
(lympho)hematopoietic
cancer in males;
significant regression
coefficient for
continuous log
cumulative exposure
Similar results for
"lymphoid" cancers in
males
For females, with 20-
year lag, in internal Cox
regression analyses, OR
= 3. 13 (^0.05) for
breast cancer mortality
in highest cumulative
exposure group;
significant regression
coefficient for
continuous log
cumulative exposure
Other chemicals to which subjects
were potentially exposed
No identified exposures to other
chemicals
Limitations
Potential bias due to lack
of followup on
"untraceable" members
(4.5% of the cohort)
Individual exposures were
estimated prior to 1976
before first industrial
hygiene survey was
completed
No increase in
hematopoietic cancer risk
with increase in exposure
in women
fe
H
O
O
o
H
O
HH
H
W
O
O
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Women
employees from
Steenland et al.
(2004) employed
in commercial
sterilization
facilities for at
least 1 year
Steenland et al.
(2003)
Number
of
subjects
7,576
women
Extent of exposure to
ethylene oxide
Same as in Steenland et
al. (2004)
Minimum of 1 year
Health outcomes
SIR = 0.87
319 cases of breast
cancer
SIR = 0.94
20 in situ cases
excluded
A positive trend in SIRs
with 15 -year lag time
for cumulative exposure
(/?=0.002)
In internal nested case-
control analysis, a
positive exposure-
response log of
cumulative exposure
with 15 -year lag, top
quintile had OR = 1.74,
Similar results in
subcohort of 5,139
women with interviews
(233 cases)
Other chemicals to which subjects
were potentially exposed
Same as in Steenland et al. (2004),
Stayneretal. (1993)
Limitations
Interviews were available
for only 68% of the
women; thus, there is
underascertainment of
cancer cases. Also, there
are potential nonresponse
biases in the subcohort
with interviews
Exposure-response trends
not strictly monotonically
increasing
fe
H
O
O
o
H
O
HH
H
W
O
O
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Chemical workers
licensed to
handle ethylene
oxide and other
toxic chemicals,
Italy
Bisanti et al.
(1993)
Two plants that
produced
disposable
medical
equipment,
Sweden
Hagmar et al.
(1991, 1995)
Number
of
subjects
1,971
men
2,170
(861
men,
1,309
women)
Extent of exposure to
ethylene oxide
Levels were said to be
high at beginning of
employment; no
actual measurements
were available
637 workers were
licensed only to handle
ethylene oxide and no
other toxic chemicals
1964-1966, 75 ppm in
sterilizers, 50 ppm in
packers
1970-1972, 40 ppm in
sterilizers, 20-35 ppm in
packers and engineers
By 1985, levels had
dropped to 0.2 ppm in all
categories except
terilizers and to 0.75
ppm in sterilizers
Health outcomes
43 total cancer deaths
vs. 33 expected
6 hematopoietic cancer
deaths vs. 2.4 expected
4 lymphosarcoma and
reticulosarcoma deaths
vs. 0.6 expected
5 hematopoietic cancer
deaths vs. 0.7 expected
in group licensed to
handle only ethylene
oxide
6 lymphohematopoietic
cancer cases vs. 3.37
expected
Among subjects with at
least 0.14 ppm-years of
cumulative exposure
and 10 years latency,
the SIR for leukemia
was 7. 14, based on two
cases
Other chemicals to which subjects
were potentially exposed
Toxic gases, dimethyl sulphate,
methylene chloride, carbon
disulphide, phosgene, chlorine,
alkalic cyanides, sulfur dioxide,
anhydrous ammonia, hydrocyanic
acid
Fluorochlorocarbons, methyl formate
(1:1 mixture with ethylene oxide)
Limitations
Lack of exposure data
Insufficient followup for
this young cohort
Potential selection bias
Possible earlier exposure
than date of licensing
would indicate
Short followup period;
authors recommend
another 10 years of
followup
Youthful cohort — few
cases and fewer deaths;
unable to determine
significance or
relationships in categories
Only a minority of
subjects had high
exposure to ethylene
oxide
O
O
2
O
H
O
O
&
O
c
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Sterilizers of
medical
equipment and
supplies that
were assembled
at this plant,
New York
Norman et al.
(1995)
Nested case-
control study;
cases and
controls from a
large chemical
production
plant, Belgium
Swaen et al.
(1996)
Number
of
subjects
1,132
(204
men,
928
women)
10 cases
of Hodg-
kin's
disease
(7 cases
con-
firmed)
and 200
controls;
all male
Extent of exposure to
ethylene oxide
In 1980, levels were
50-200 ppm (8-hr TWA);
corrective action reduced
levels to less than 20 ppm
Cumulative exposure to
ethylene oxide in cases
was 500.2 ppm-months
vs. 60.2 ppm-months in
controls
Health outcomes
Only 28 cancers were
diagnosed
1 leukemia case vs. 0.54
expected
12 breast cancer cases
vs. 4.7 expected
(p<0.05)
2 pancreatic cancer
cases vs. 0.51 expected
3 cases indicated
exposure to ethylene
oxide, producing an
OR = 8.5 (p<0.05)
Other chemicals to which subjects
were potentially exposed
No other chemical exposures cited
Fertilizers, materials for synthetic
fiber production, PVC, polystyrene,
benzene, methane, acetone, ammonia,
ammonium, sulfate, aniline,
caprolactam, ethylene, Nah., oleum
Limitations
Little power to detect any
significant risk chiefly
because a short followup
period produced few
cancer cases
Insufficient latency
analysis
This was a hypothesis-
generating study; the
authors were not looking
for ethylene oxide
exposure alone but for
other chemical exposures
as well to explain the
excess risk
Only one disease —
Hodgkin's lymphoma —
was analyzed
O
O
2
O
H
O
HH
H
W
O
&
O
c
O
H
W
-------
Table A-5. Epidemiological studies of ethylene oxide and human cancer (continued)
Population/
industry
Four ethylene
oxide production
plants in 3 states
utilizing the
chlorohydrin
process (both
ethylene and
propylene)
Olsen et al.
(1997)
Female worker at
Markhot Fereng
Provincial
hospital and clinic
of Egerinthe
Pediatric
Department
Kardos et al.
(2003)
Number
of
subjects
1,361
men
299
female
em-
ployees
Extent of exposure to
ethylene oxide
No actual measurements
were taken
EtO sterilizing units with
unknown elevated
concentrations
Health outcomes
10 lymphohematopoietic
cancer deaths vs. 7.7
expected
After 24 years, the SMR
increased to 1.44,
based on 6 observed
deaths
No increase in pancreatic
cancer
1 1 cancer deaths
observed compared with
4.38, 4.03, or 4.28
expected (p<0.0l), based
on comparison
populations of Hungary,
Heves County, and city
of Eger, respectively
Other chemicals to which subjects
were potentially exposed
Bis-chloroethyl ether, propylene
oxide, ethylene chlorohydrin,
propylene chlorohydrin, ethylene
dichloride, chlorohydrin chemicals
No identifiable exposures to other
chemicals
Limitations
No actual airborne
measurements of
ethylene oxide or other
chemicals such as
ethylene dichloride were
reported; only length of
employment was used as
a surrogate
Increase in risk of
lymphocytic and
hematopoietic cancers
after a 25-year latency is
not shown in tabular form
An additional 5 to 10 years
of followup is needed to
confirm the presence or
lack of risk of pancreatic
cancer and lymphopoietic
and hematopoietic
cancers
Underlying cause of death
provided on all 1 1 cases
but no expected deaths
available by cause
Possible exposure to
natural radium, which
permeates the region
oo
o
O
2
O
H
O
HH
H
W
O
&
O
c
O
H
W
-------
APPENDIX B: REFERENCES FOR FIGURE 3
The references in this list correspond to the additional data that was added to Figure 3
since the IARC (1994) genetic toxicity profile was published. See the Figure 3 legend for details.
DeSerres, FJ; Brockman, HE. (1995) Ethylene oxide: induction of specific-locus mutations in the ad- 3 region of
heterokaryon 12 of Neurospora crassa and implications for genetic risk assessment of human exposure in the
workplace. MutatRes 328:31-47.
Hengstler, JG; Fuchs, J; Gebhard, S; et al. (1994) Glycolaldehyde causes DNA-protein crosslinks: a new aspect of
ethylene oxide genotoxicity. MutatRes 304:229-234.
Major, J; Jakab, MG; Tompa, A. (1996) Genotoxicological investigation of hospital nurses occupationally exposed to
ethylene-oxide: 1. chromosome aberrations, sister-chromatid exchanges, cell cycle kinetics, and UV-induced DNA
synthesis in peripheral blood lymphocytes. Environ Mol Mutagen 27:84-92.
Major, J; Jakab, MG; Tompa, A. (1999) The frequency of induced premature centromere division in human
populations occupationally exposed to genotoxic chemicals. MutatRes 445:241-249.
Nygren, J; Cedervall, B; Eriksson, S; et al. (1994) Induction of DNA strand breaks by ethylene oxide in human
diploid fibroblasts. Environ Mol Mutagen 24:161-167.
Oesch, F; Hengstler, JG; Arand, M; et al. (1995) Detection of primary DNA damage: applicability to biomonitoring
of genotoxic occupational exposure and in clinical therapy. Pharmacogenetics 5 SpecNo:S118-S122.
Ribeiro, LR; Salvadori, DMF; Rios, ACC; et al. (1994) Biological monitoring of workers occupationally exposed to
ethylene oxide. MutatRes 313:81-87.
Sisk, SC; Pluta, LJ; Meyer, KG; et al. (1997) Assessment of the in vivo mutagenicity of ethylene oxide in the tissues
of B6C3F1 lacl transgenic mice following inhalation exposure. Mutat Res 391:153-164.
Swenberg, JA; Ham, A; Koc, H; et al. (2000) DNA adducts: effects of low exposure to ethylene oxide, vinyl chloride
and butadiene. Mutat Res 464:77-86.
Tates, AD; vanDam, FJ; Natarajan, AT; et al. (1999) Measurement of HPRT mutations in splenic lymphocytes and
haemoglobin adducts in erythrocytes of Lewis rats exposed to ethylene oxide. Mutat Res 431:397-415.
vanSittert, NJ; Boogaard, PJ; Natarajan, AT; et al. (2000) Formation of DNA adducts and induction of mutagenic
effects in rats following 4 weeks inhalation exposure to ethylene oxide as a basis for cancer risk assessment. Mutat
Res 447:27-48.
Vogel, EW; Nivard, MJM. (1997) The response of germ cells to ethylene oxide, propylene oxide, propylene imine
and methyl methanesulfonate is a matter of cell stage-related DNA repair. Environ Mol Mutagen 29:124-135.
Vogel, EW; Nivard, MJM. (1998) Genotoxic effects of inhaled ethylene oxide, propylene oxide and burylene oxide
on germ cells: sensitivity of genetic endpoints in relation to dose and repair status. MutatRes 405:259-271.
Walker, VE; Sisk, SC; Upton, PB; et al. (1997) In vivo mutagenicity of ethylene oxide at the hprt locus in T-
lymphocytes of B6C3F1 lacl transgenic mice following inhalation exposure. Mutat Res 392:211-222.
Walker, VE; Wu, KY; Upton, PB; et al. (2000) Biomarkers of exposure and effect as indicators of potential
carcinogenic risk arising from in vivo metabolism of ethylene to ethylene oxide. Carcinogenesis 21:1661-1669.
08/29/06 139 DRAFT—DO NOT CITE OR QUOTE
-------
Appendix C. Extra risk calculation" for environmental exposure to 0.00608 ppm (the LEC01 for lymphohematopoietic cancer
incidence in males)b using a linear exposure-response model based on the categorical cumulative exposure results of Steenland
et al. (2004), with a 15-year lag, as described in Section 4.2.1.
G
H
K
M
O
interval
number
0)
1
2
3
4
5
6
7
8
^ 9
£ 10
O
11
12
13
0 14
^d is
fe 16
H 17
6 18
o
age all cause LH
interval mortality cancer
(x!05/yr) incidence
(xl05/yr)
<1 801.5 4.1
1-4 38.5 10.1
5-9 19.2 5.4
10-14 25.3 5.5
15-19 96.3 8.0
20-24 137.6 10.4
25-29 138.5 12.2
30-34 161.0 15.6
35-39 211.1 19.4
40-44 302.9 26.2
45-49 455.8 33.9
50-54 654.6 49.7
55-59 1026.1 71.3
60-64 1595.9 102.1
65-69 2479.5 154.5
70-74 3816.5 204.1
75-79 5719.8 250.7
80-84 9156.8 286.2
all cause
hazard
rate (h*)
0.0080
0.0015
0.0010
0.0013
0.0048
0.0069
0.0069
0.0081
0.0106
0.0151
0.0228
0.0327
0.0513
0.0798
0.1240
0.1908
0.2860
0.4578
prob of
surviving
interval
(q)
0.9920
0.9985
0.9990
0.9987
0.9952
0.9931
0.9931
0.9920
0.9895
0.9850
0.9775
0.9678
0.9500
0.9233
0.8834
0.8263
0.7513
0.6326
prob of
surviving
up to
interval
(S)
1
0.9920
0.9905
0.9895
0.9883
0.9835
0.9768
0.9701
0.9623
0.9522
0.9379
0.9167
0.8872
0.8428
0.7782
0.6875
0.5680
0.4267
LH cancer
hazard
rate (h)
0.00004
0.00040
0.00027
0.00028
0.00040
0.00052
0.00061
0.00078
0.00097
0.00131
0.00170
0.00249
0.00357
0.00511
0.00773
0.01021
0.01254
0.01431
Ro =
cond prob
ofLH
cancer
incidence
in interval
0.00004
0.00040
0.00027
0.00027
0.00039
0.00051
0.00059
0.00075
0.00093
0.00124
0.00157
0.00224
0.00308
0.00414
0.00565
0.00639
0.00619
0.00490
0.03956
exp dur
mid
interval
(xtime)
0
0
0
0
2.5
7.5
12.5
17.5
22.5
27.5
32.5
37.5
42.5
47.5
52.5
57.5
62.5
67.5
cum exp
mid
interval
(xdose)
0
0
0
0
16.88
50.63
84.38
118.13
151.88
185.63
219.38
253.13
286.88
320.63
354.38
388.13
421.88
455.63
exposed
LH cancer
hazard
rate (hx)
0.00004
0.00040
0.00027
0.00028
0.00041
0.00054
0.00065
0.00085
0.00108
0.00149
0.00198
0.00296
0.00434
0.00635
0.00981
0.01322
0.01655
0.01927
exposed
all cause
hazard
rate
(h*x)
0.0080
0.0015
0.0010
0.0013
0.0048
0.0069
0.0070
0.0081
0.0107
0.0153
0.0231
0.0332
0.0521
0.0810
0.1261
0.1938
0.2900
0.4628
exposed
prob of
surviving
interval
(qx)
0.9920
0.9985
0.9990
0.9987
0.9952
0.9931
0.9931
0.9919
0.9894
0.9848
0.9772
0.9673
0.9493
0.9222
0.8816
0.8238
0.7483
0.6295
exposed
prob of
surviving
up to
interval
(Sx)
1
0.9920
0.9905
0.9895
0.9883
0.9835
0.9768
0.9700
0.9622
0.9519
0.9375
0.9161
0.8862
0.8412
0.7757
0.6838
0.5633
0.4215
Rx =
exposed
cond
prob of
LH
cancer in
interval
0.00004
0.00040
0.00027
0.00027
0.00040
0.00053
0.00063
0.00082
0.00104
0.00141
0.00183
0.00267
0.00375
0.00513
0.00715
0.00822
0.00810
0.00650
0.04915
Q extra risk = (Rx-Ro)/(l-Ro) = 0.00998
O
M column A:
|-[-j column B:
interval index number (i)
5-year age interval (except <1 and
1-4) up to age 85
O
^w'
3
column C: all-cause mortality rate for interval i (x 105/year) (1999 data from NCHS; males)
column D: lymphohematopoietic cancer incidence rate for interval i (x lOVyear) (1996-2000 SEER data; males)0
column E: all-cause hazard rate for interval i (h*;) (= all-cause mortality rate x number of years in age interval)"
column F: probability of surviving interval i without being diagnosed with lymphohematopoietic cancer (q;) (= exp(-h*;))
column G: probability of surviving up to interval i without having been diagnosed with lymphohematopoietic cancer (S;) (Sj = !
qui, fori>l)
column H: lymphohematopoietic cancer incidence hazard rate for interval i (h;) (= lymphohematopoietic cancer incidence rate x number of years in interval)
-------
Appendix C. Extra risk calculationa for environmental exposure to 0.00608 ppm (the LEC01 for lymphohematopoietic cancer
incidence in males)b using a linear exposure-response model based on the categorical cumulative exposure results of Steenland
et al. (2004), with a 15-year lag, as described in Section 4.2.1. (continued)
column I: conditional probability of being diagnosed with lymphohematopoietic cancer in interval i (= (h/h*;) x S; x (l-q;)), i.e., conditional upon surviving up to interval i without having been
diagnosed with lymphohematopoietic cancer [Ro, the background lifetime probability of being diagnosed with lymphohematopoietic cancer = the sum of the conditional probabilities across
the intervals]
column J: exposure duration at mid-interval (taking into account 15-year lag) (xtime)
column K: cumulative exposure mid-interval (xdose) (= exposure level (i.e., 0.00608 ppm) x 365/240 x 20/10 x xtime x 365) [ 365/240 x 20/10 converts continuous environmental exposures to
corresponding occupational exposures; xtime x 365 converts exposure duration in years to exposure duration in days]
column L: lymphohematopoietic cancer incidence hazard rate in exposed people for interval i (hx;) (= h; x (1 + p x xdose), where p = 0.000347 + (1.645 x 0.000251) = 0.000760) [0.000347 per ppm x
day is the regression coefficient obtained from the weighted linear regression of the categorical results, dropping the highest exposure group (see Section 4.2.1.1). to estimate the LEC01, i.e.,
the 95% lower bound on the continuous exposure giving an extra risk of 1%, the 95% upper bound on the regression coefficient is used, i.e., MLE + 1.645 x SE]
column M: all-cause hazard rate in exposed people for interval i (h*x;) (= h*; + (hx; - h;))
column N: probability of surviving interval i without being diagnosed with lymphohmematopoietic cancer for exposed people (qx;) (= exp(-h*x;))
column O: probability of surviving up to interval i without having been diagnosed with lymphohematopoietic cancer for exposed people(Sx;) (Sxj = 1; Sx; = Sx^ x qx^, for i>l)
column P: conditional probability of being diagnosed with lymphohematopoietic cancer in interval i for exposed people (= (hx/l^X;) x Sx; x (l-qx;)) [Rx, the lifetime probability of being diagnosed with
lymphohematopoietic cancer for exposed people = the sum of the conditional probabilities across the intervals]
fe
H
6
O
O
H
O
HH
H
W
O
O
O
H
W
a using the methodology of BEIRIV (1988)
b the estimated 95% lower bound on the continuous exposure level that gives a 1% extra lifetime risk of lymphohematopoietic cancer incidence in males
0 background cancer incidence rates are used to estimate extra risks for cancer incidence under the assumption that the exposure-response relationship for cancer
incidence is the same as that for cancer mortality (see Section 4.1.1.3)
d for the cancer incidence calculation, the all-cause hazard rate for interval i should technically be the rate of either dying of any cause or being diagnosed with the
specific cancer during the interval, i.e., (the all-cause mortality rate for the interval + the cancer-specific incidence rate for the interval - the cancer-specific
mortality rate for the interval [so that a cancer case isn't counted twice, i.e., upon diagnosis and upon death]) x number of years in interval, for the
lymphohematopoietic cancer incidence calculations, this adjustment was ignored because the lymphohematopoietic cancer incidence rates are small when
compared with the all-cause mortality rates, for the breast cancer incidence calculations, on the other hand, this adjustment was made in the all-cause hazard rate
(see Section 4.1.2.3)
MLE: maximum likelihood estimate, SE: standard error
-------
APPENDIX D: MODEL PARAMETERS IN THE ANALYSIS OF ANIMAL TUMOR
INCIDENCE
Table D-l. Analysis of grouped data, NTP mice study (NTP, 1987)a; multistage
model parameters
Tumor
Multistageb
polynomial
degree
Qo
(mg/m3)1
(mg/m3)2
(mg/m3)2
p value
(chi-square
goodness of
fit)
Males
Lung adenomas
plus carcinomas
1
2.52 x 10'1
1.52 x 10'2
0.92
Females
Lung adenomas
plus carcinomas
Malignant
lymphoma
Uterine carcinoma
Mammary
carcinoma
2
3
2
ld
3.87 x 10'2
1.74 x lO'1
0.0
2.27 x 10'2
0.0
0.0
0.0
1.09 x 10'2
4.80 x 10'4
0.0
9.80 x 10'5
1.13 x 10'5
0.39
0.18
0.90
a The exposure concentrations were at 0, 50 ppm, and 100 ppm. These were adjusted to continuous exposure.
b P(d) = 1 - exp[-(q0 + qjd + q2d2 + ... + c^d1")], where d is inhaled ethylene oxide exposure concentration.
0 Even though q{ is zero in some cases, the upper bound of q{ is nonzero.
d The 100-ppm dose was deleted; the fit was perfect with only two points to fit.
08/29/06
142
DRAFT—DO NOT CITE OR QUOTE
-------
Table D-2. Analysis of grouped data, Lynch et al. (1982,1984) study of male F344
ratsa; multistage model parameters
Tumor
Splenic mononuclear
cell leukemia
Testicular peritoneal
mesothelioma
Brain mixed-cell
glioma
Multistage15
polynomial
degree
lc
1
1
Qn
3.12 x ID'1
3.54 x ID'2
0
Qi
(mg/m3)1
1.48 x ID'2
6.30 x ID'3
1.72 x ID'4
p value
(chi-square goodness
of fit)
_
0.34
0.96
a The exposure concentrations were at 0, 50 ppm, and 100 ppm. These were adjusted to continuous exposure.
b P(d) = 1 - exp[-(q0 + qjd + q2d2 + ... + c^d1")], where d is inhaled ethylene oxide exposure concentration.
0 The 100-ppm dose was deleted; the fit was perfect with only two points to fit.
08/29/06
143
DRAFT—DO NOT CITE OR QUOTE
-------
Table D-3. Analysis of grouped data, Snellings et al. (1984) and Garman et al.
(1985) reports on F344 ratsa; multistage model parameters
Tumor
Multistage15
polynomial
degree
Qn
Qi
(mg/m3)1
p value
(chi-
square
goodness
of fit)
Males
Splenic mononuclear cell
leukemia
Testicular peritoneal
mesothelioma
Primary brain tumors
1
1
1
1.63 x ID'1
2.38 x 1Q-2
5.88 x 1Q-3
8.56 x ID'3
4.74 x ID'3
2.92 x ID'3
0.34
0.68
0.46
Females
Splenic mononuclear cell
leukemia
Primary brain tumors
1
1
1.08 x 1Q-1
5.94 x ID'3
2.37 x ID'2
1.65 x 1Q-3
0.75
0.80
a The exposure concentrations were at 0, 10 ppm, 33 ppm, and 100 ppm. These were adjusted to continuous
exposure.
b P(d) = 1 - exp[-(q0 + qjd + q2d2 + ... + c^d1")], where d is inhaled ethylene oxide exposure concentration.
08/29/06
144
DRAFT—DO NOT CITE OR QUOTE
-------
Table D-4. Time-to-tumor analysis of individual animal data, NTP mice study
(NTP, 1987)a; multistage-Weibull model" parameters
Tumor
Multistage
polynomial
degree
Qn
Qi
(mg/m3)1
z
Males
Lung adenomas plus
carcinomas
1
3.44 x ID'1
2.03 x IQ-2
5.39
Females
Lung adenomas plus
carcinomas
Malignant lymphoma
Uterine carcinoma
Mammary carcinoma
1
1
1
1
5.35 x ID'2
1.91 x ID'1
0.0
3.78 x 10-2
1.76 x IQ-2
8.80 x IQ-3
3.81 x IQ-3
5.10 x 10-3
7.27
1.00
3.93
1.00
3 The exposure concentrations were at 0, 50 ppm, and 100 ppm. These were adjusted to continuous
exposure.
b P(d, t) = 1 - exp[-(q0 + q: d + q2d2 + .... + qkdk)*(t - t0)z], where d is inhaled ethylene oxide exposure
concentration.
The length of the study was 104 weeks. The times t and t0 as expressed in the above formula are
scaled so that the length of the study is 1.0. Then, q0 is dimensionless, and the coefficients qk are
expressed in units of (mg/m3)"k.
08/29/06
145
DRAFT—DO NOT CITE OR QUOTE
-------
REFERENCES
BEIR ([Committee on the] Biological Effects of Ionizing Radiation). (1988) Health risks of radon and other internally
deposited alpha emitters. BEIR IV. Washington, DC: National Academy Press.
Benson, LO; Teta, MJ. (1993) Mortality due to pancreatic and lymphopoietic cancers in chlorohydrin production
workers. Br J Ind Med 50:710-716.
Beranek, DT. (1990) Distribution of methyl and ethyl adducts following alkylation with monofunctional alkylating
agents. Mutat Res 231:11-30.
Bisanti, L; Maggini, M; Raschetti, R; et al. (1993) Cancer mortality in ethylene oxide workers. Br J Ind Med
50:317-324.
Bolt, HM. (1996) Quantification of endogenous carcinogens: the ethylene oxide paradox. Biochem Pharmacol 52:1-5.
Brown, CD; Wong, BA; Fennell, TR. (1996) In vivo and in vitro kinetics of ethylene oxide metabolism in rats and
mice. Toxicol Appl Pharmacol 136:8-19.
Brown, CD; Asgharian, B; Turner, MJ; et al. (1998) Ethylene oxide dosimetry in the mouse. Toxicol Appl Pharmacol
148:215-221.
CalEPA (California Environmental Protection Agency). (1999) Air toxics hotspots program risk assessment
guidelines. Part II: Technical support document for describing available cancer potency factors. Office of
Environmental Health Hazard Assessment, Secretary for Environmental Protection, California Environmental
Protection Agency. Sacramento, CA.
Chaganti, SR; Chen, WY; Parsa, N; et al. (1998) Involvement of BCL6 in chromosomal aberrations affecting band
3q27 inB-cell non-Hodgkin lymphoma. Genes Chromosomes Cancer 23:323-327.
Clare, MG; Dean, BJ; de Jong; G; et al. (1985) Chromosome analysis of lymphocytes from workers at an ethylene
oxide plant. Mutat Res 156:109-116.
Clewell, HJ; Teeguarden, J; McDonald, T; et al. (2002) Review and evaluation of the potential impact of age- and
gender-specific pharmacokinetic differences on tissue dosimetry. Clinical Reviews in Toxicology 32(5):329-389.
Coggon, D; Harris, EC; Poole, J; et al. (2004) Mortality of workers exposed to ethylene oxide: extended follow up of a
British cohort. Occup Environ Med 61(4):358-362.
Csanady, GA; Denk, B; Ptitz, C; et al. (2000) A physiological toxicokinetic model for exogenous and endogenous
ethylene and ethylene oxide in rat, mouse, and human: formation of 2-hydroxyethyl adducts with hemoglobin and
DNA. Toxicol Appl Pharmacol 165:1-26.
Dellarco, VL; Generoso, WM; Sega, GA; et al. (1990) Review of the mutagenicity of ethylene oxide. Environ Mol
Mutagen 16:85-103.
Divine, BJ; Amanollahi, KS. (1986) Ethylene oxide and cancer [letter]. JAMA 256:1726-1727.
Dunkelberg, H. (1982) Carcinogenicity of ethylene oxide and 1,2 propylene oxide upon intragastric administration to
rats. Br J Cancer 46:924-932.
Ehrenberg, L; Gustafsson, A. (1959) Chemical mutagens: their uses and hazards in medicine and technology. A report
of February 1959 to the National Board of Health (English translation published by Bloms Boletryckeri, Lund, 1970).
08/29/06 146 DRAFT—DO NOT CITE OR QUOTE
-------
Ehrenberg, L; Hallstrom, T. (1967) Hematologic studies on persons occupationally exposed to ethylene oxide. In:
Radiosterilization of medical products. Vienna, Austria: International Atomic Energy Agency; pp. 327-334.
Ehrenberg, L; Hussain, S. (1981) Genetic toxicology of some important epoxides. Mutat Res 86:1-113.
EOIC (Ethylene Oxide Industry Council). (2001) Toxicological review of ethylene oxide in support of summary
information on the integrated risk information system. Ethylene Oxide Industry Council, Arlington, VA.
Fennell, TR; Brown, CD. (2001) A physiologically based pharmacokinetic model for ethylene oxide in mouse, rat, and
human. Toxicol Appl Pharmacol 173:161-175.
Filser, JG; Denk, B; Tornqvist, M; et al. (1992) Pharmacokinetics of ethylene in man; body burden with ethylene
oxide and hydroxyethylation of hemoglobin due to endogenous and environmental ethylene. Arch Toxicol
66:157-163.
Galloway, SM; Berry, PK; Nichols, WW; et al. (1986) Chromosome aberrations in individuals occupationally exposed
to ethylene oxide and in a large control population. Mutat Res 170:55-74.
Gardner, MJ; Coggon, D; Pannett, B; et al. (1989) Workers exposed to ethylene oxide: a follow-up study. Br J Ind
Med 46:860-865.
Garman, RH; Snellings, WM; Maronpot, RR. (1985) Brain tumors in F344 rats associated with chronic inhalation
exposure to ethylene oxide. Neurotoxicol 6:117-138.
Garman, RH; Snellings, WM; Maronpot, RR. (1986) Frequency, size and location of brain tumours in F-344 rats
chronically exposed to ethylene oxide. Food Chem Toxicol 24:145-153.
Garry, VF; Hozier, J; Jacobs, D; et al. (1979) Ethylene oxide: evidence of human chromosomal effects. Environ
Mutagen 1:375-382.
Gelehrter, TD; Collins, FS. (1990) Principles of medical genetics. Baltimore, MD: Williams and Wilkins; p. 234.
Granath, F; Vaca, C; Ehrenberg, L; et al. (1999) Cancer risk estimation of genotoxic chemicals based on target dose
and a multiplicative model. Risk Anal 19:309-320.
Greenberg, HL; Ott, MG; Shore, RE. (1990) Men assigned to ethylene oxide production or other ethylene oxide
related chemical manufacturing: a mortality study. Br J Ind Med 47:221-230.
Greife, A; Hornung, R; Stayner, LG; et al. (1988) Development of a model for use in estimating exposure to ethylene
oxide in a retrospective cohort mortality study. Scand J Work Environ Health 14:29-30.
Hagmar, L; Welinder, H; Linden, K; et al. (1991) An epidemiological study of cancer risk among workers exposed to
ethylene oxide using hemoglobin adducts to validate environmental exposure assessments. Int Arch Occup Environ
Health 63:271-277.
Hagmar, L; Mikoczy, Z; Welinder, H. (1995) Cancer incidence in Swedish sterilant workers exposed to ethylene
oxide. Occup Environ Med 52:154-156.
Hansen, JP; Allen, J; Brock, K; et al. (1984) Normal sister chromatid exchange levels in hospital sterilization
employees exposed to ethylene oxide. J Occup Med 26:29-32.
Harada, H; Harada, Y; Tanaka, H; et al. (2003) Implications of somatic mutations in the AML1 gene in radiation-
associated and therapy-related myelodysplastic syndrome/acute myeloid leukemia. Blood 101:673-680.
08/29/06 147 DRAFT—DO NOT CITE OR QUOTE
-------
Health Canada. (2001) Priority substances list; assessment report ethylene oxide. Available at:
http://www.ec.gc.ca/substances/ese/eng/psap/final/ethyleneoxide.cfm.
Hill, AB. (1965) The environment and disease: association or causation? Proc R Soc Med 58:295-300.
Hogstedt, LC. (1988) Epidemiologic studies on ethylene oxide and cancer: an updating. In: Bartsch, H; Hemminki, K;
O'Neil, IK, eds. Methods for detecting DNA damaging agents in humans: applications in cancer epidemiology and
prevention. Lyon, France: IARC; pp. 265-270.
Hogstedt, C; Rohlen, O; Berndtsson, BS; et al. (1979a) A cohort study of mortality and cancer incidence in ethylene
oxide production workers. Br J Ind Med 36:276-280.
Hogstedt, C; Malmqvist, N; Wadman, B. (1979b) Leukemia in workers exposed to ethylene oxide. JAMA
241:1132-1133.
Hogstedt, B; Gullberg, B; Hedner, K; et al. (1983) Chromosome aberrations and micronuclei in bone marrow cells and
peripheral blood lymphocytes in humans exposed to ethylene oxide and propylene oxide. Hereditas 98:105-113.
Hogstedt, C; Aringer, L; Gustavasson, A. (1984) Ethylene oxide and cancer: a literature review and follow-up of two
epidemiologic studies. Arbete ochHalse (translation from Swedish of pre-publication manuscript).
Hogstedt, C; Aringer, L; Gustafson, A. (1986) Epidemiologic support for ethylene oxide as a cancer-causing agent.
JAMA 255:1575-1578.
Hornung, RW; Greife, AL; Stayner, LT; et al. (1994) Statistical model for prediction of retrospective exposure to
ethylene oxide in an occupational mortality study. Am J Ind Med 25:825-836.
IARC (International Agency for Research on Cancer). (1994) Ethylene oxide. In: IARC monographs on the evaluation
of carcinogenic risks to humans and their supplements. Vol. 60. Some industrial chemicals. Lyon, France: World
Health Organization; pp. 73-157.
Johanson, G; Filser, JG. (1992) Experimental data from closed chamber gas uptake studies in rodents suggest lower
uptake rate of chemical than calculated from literature values on alveolar ventilation. Arch Toxicol 66(4):291-295.
Joyner, RE. (1964) Chronic toxicity of ethylene oxide: a study of human responses to long-term low-level exposures.
Arch Environ Health 8:700-710.
Kardos, L; Szeles, G; Gombkoto, G; et al. (2003) Cancer deaths among hospital staff potentially exposed to ethylene
oxide: an epidemiological analysis. Environ Mol Mutagen 42:59-60.
Kiesselbach, N; Ulm, K; Lange, H-J; et al. (1990) A multicentre mortality study of workers exposed to ethylene oxide.
BrJ Ind Med 47:182-188.
Kirman, CR; Sweeney, LM; Teta, MJ; et al. (2004) Addressing nonlinearity in the exposure-response relationship for
agenotoxic carcinogen: cancer potency estimates for ethylene oxide. Risk Anal 24:1165-1183.
Krewski, D; Crump, KS; Farmer, J; et al. (1983) A comparison of statistical methods for low dose extrapolation
utilizing time-to-tumor data. Fundam Appl Toxicol 3:140-160.
Krishnan, K; Gargas, ML; Fennell, TR; et al. (1992) A physiologically based description of ethylene oxide dosimetry
in the rat. Toxicol Ind Health 8(3):121-140.
Kuo, J; Linkov, I; Rhomberg, L; et al. (2002) Absolute risk or relative risk? A study of intraspecies and interspecies
extrapolation of chemical-induced cancer risk. Risk Anal 22:141-157.
08/29/06 148 DRAFT—DO NOT CITE OR QUOTE
-------
Laurent, C; Frederic, J; Leonard, AY. (1984) Sister chromatid exchange frequency in workers exposed to high levels
of ethylene oxide in a hospital sterilization service. Int Arch Occup Environ Health 54:33-43.
Lerda, D; Rizzi, R. (1992) Cytogenetic study of persons occupationally exposed to ethylene oxide. Mutat Res
281:31-37.
Losses, IS; Levy, R. (2000) Mutation analysis of the 5 ' noncoding regulatory region of the BCL-6 gene in non-
Hodgkin lymphoma: evidence for recurrent mutations and intraclonal heterogeneity. Blood 95:1400-1405.
Lucas, LJ; Teta, MJ. (1996) Breast cancer and ethylene oxide exposure. Int J Epidemiol 25(3):685-686.
Lynch, DW; Lewis, TR; Moorman, WJ; et al. (1982) Effects on monkeys and rats of long-term inhalation exposure to
ethylene oxide: major findings of the NIOSH study. Toxicologist 2(1):11.
Lynch, DW; Lewis, TR; Moorman, WJ; et al. (1984) Carcinogenic and toxicological effects of inhaled ethylene oxide
and propylene oxide in F344 rats. Toxicol Appl Pharmacol 76:69-84.
Major, J; Jakab, MG; Tompa, A. (1996) Genotoxicological investigation of hospital nurses occupationally exposed to
ethylene-oxide. 1. chromosome aberrations, sister-chromatid exchanges, cell cycle kinetics, and UV-induced DNA
synthesis in peripheral blood lymphocytes. Environ Mol Mutagen 27:84-92.
Major, J; Jakab, MG; Tompa, A. (2001) Genotoxicological investigation of hospital nurses occupationally exposed to
ethylene-oxide. II Hprt mutation frequencies. Central European Journal of Occupational and Environmental Medicine
7:195-208.
Maltoni, C; Valgimigli, L; Scarnato, C. (1980) Long-term carcinogenic bioassays on ethylene dichloride administered
by inhalation to rats and mice. In: Ames, B; Infante, P; Reitz, R, eds. Ethylene dichloride: a potential health risk?
(Banbury Report 5). Cold Spring Harbor Laboratory, Cold Spring Harbor, NY; pp. 3-29.
Mayer, J; Warburton, D; Jeffrey, AM; et al. (1991) Biologic markers in ethylene oxide-exposed workers and controls.
Mutat Res 248:163-176.
Memisoglu, A; Samson, L. (2000) Base excision repair in yeast and mammals. Mutat Res Fundam Mol Mech Mut
451:39-51.
Morgan, RW; Claxton, KW; Divine, BJ; et al. (1981) Mortality among ethylene oxide workers. J Occup Med
23:767-770.
Muller, E; Bertok. A. (1995) Tumor cases in the personnel of Eger hospital: hygienic conditions and consequences.
(In Hungarian) Korhaz-es Orvostechnika 33:17-22.
Natarajan, AT; Preston, RJ; Dellarco, V; et al. (1995) Ethylene oxide: evaluation of genotoxicity data and an
exploratory assessment of genetic risk. Mutat Res 330:55-70.
NCHS (National Center for Health Statistics). (2002) United States life tables, 1999. National Vital Statistics Reports,
vol. 50, no. 6; March 21, 2002, Tables 2 and 3. National Center for Health Statistics, Hyattsville, MD.
NCI (National Cancer Institute). (1978) Bioassay of 1,2-dichloroethane for possible carcinogenicity. Publication
(NIH) 1978; pp. 78-1361. National Cancer Institute, Bethesda, MD.
NCI (National Cancer Institute). (2003) SEER Cancer Statistics Review, 1975-2000. National Cancer Institute,
Bethesda, MD. Available at: http://seer.cancer.gov/csr/1975_2000/.
08/29/06 149 DRAFT—DO NOT CITE OR QUOTE
-------
NCI (National Cancer Institute). (2004a) SEER Cancer Statistics Review, 1975-2001. National Cancer Institute,
Bethesda, MD. Available at: http://seer.cancer.gov/csr/1975_2001/.
NCI (National Cancer Institute). (2004b) Breast cancer (PDQR): screening. National Cancer Institute, Bethesda, MD.
Available at: http://www.cancer.gov/cancertopics/pdq/screening/breast/HealthProfessional/page 5.
NTP (National Toxicology Program). (1987) Toxicology and carcinogenesis studies of ethylene oxide (CAS No. 75-
21-8) in mice (Inhalation studies). Tech. report Series No. 326. National Toxicology Program, Research Triangle Park,
NC.
NTP (National Toxicology Program). (2000) Ninth report on carcinogens. Public Health Service, U.S. Department of
Health and Human Services, Washington, DC. Available at: http://ehp.niehs.nih.gov/roc/toc9.html.
Norman, SA; Berlin, JA; Soper, KA; et al. (1995) Cancer incidence in a group of workers potentially exposed to
ethylene oxide. Int J Epidemiol 24:276-284.
Olsen, GW; Lacy, SE; Bodner, KM; et al. (1997) Mortality from pancreatic and lymphopoietic cancer among workers
in ethylene and propylene chlorohyrdrin production. Occup Environ Med 54:592-598.
Paz-y-Mino, C; Perez, JC; Fiallo, BF; et al. (2002) A polymorphism in the hMSH2 gene (gIVS12-6T>C) associated
with non-Hodgkin lymphomas. Cancer Genet Cytogenet 133:29-33.
Pedersen-Bjergaard, J; Andersen, MK; Christiansen, DH; et al. (2002) Genetic pathways in therapy-related
myelodysplasia and acute myeloid leukemia. Blood 99:1909-1912.
Pfeiffer, P; Goedecke, W; Obe, G. (2000) Mechanisms of DNA double-strand break repair and their potential to
induce chromosomal aberrations. Mutagenesis 15:289-302.
Popp, W; Wahrenholz, C; Przygoda, H; et al. (1994) DNA-protein cross-links and sister chromatid exchange
frequencies in lymplocytes and hydroxyethyl mercapturic acid in urine of ethylene oxide-exposed hospital workers. Int
Arch Occup Environ Health 66:325-332.
Preston, RJ. (1999) Cytogenetic effects of ethylene oxide, with an emphasis on population monitoring. Crit Rev
Toxicol 29:263-282.
Preston, RJ; Fennell, TR; Leber, AP; et al. (1995) Reconsideration of the genetic risk assessment for ethylene oxide
exposures. Environ Mol Mutagen 26:189-202.
Preudhomme, C: Warot-Loze, D; Roumier, C; et al. (2000) High incidence of biallelic point mutations in the Runt
domain of the AML1/PEBP2 alpha B gene in Mo acute myeloid leukemia and in myeloid malignancies with acquired
trisomy 2. Blood 96:2862-2869.
Rapoport, IA. (1948) Action of ethylene oxide, glycidol and glycols on gene mutations: Dokl. Akad. Nauk. SSSR
60:469-472.
Recio, L; Abernathy, D; Donner, M; et al. (1999) Assessment of the in vivo mutagenicity of ethylene oxide in the bone
marrow of B6C3F1 lacl transgenic mice following a chronic inhalation exposure. Toxicologist 48:368.
Recio, L; Donner, M; Abernethy, D; et al. (2004) In vivo mutagenicity and mutation spectrum in the bone marrow and
testes of B6C3F1 lad transgenic mice following inhalation exposure to ethylene oxide. Mutagenesis 19:215-222.
Ribeiro, LR; Salvador!, DMF; Rios, ACC; et al. (1994) DNA-protein cross-links and sister chromatid exchange
frequencies in lymphocytes and hydroxyethyl mercapturic acid in urine of ethylene oxide-exposed hospital workers.
Int Arch Occup Environ Health 66:325-332.
08/29/06 150 DRAFT—DO NOT CITE OR QUOTE
-------
Richmond, GW; Abrahams, RH; Nemenzo, JH; et al. (1985) An evaluation of possible effects on health following
exposure to ethylene oxide. Arch Environ Health 40:20-25.
Rothman, KJ. (1986) Modern epidemiology. Worcester, MA: Little, Brown and Co.; pp. 343-344.
Russo, J.; Russo, IH. (1999) Cellular basis of breast cancer susceptibility. Oncol Res 11:169-178.
Sarto, F; Cominato, I; Pinton, AM; et al. (1984) Cytogenetic damage in workers exposed to ethylene oxide. Mutat Res
138:185-195.
Sarto, F; Clonfero, E; Bartolucci, GB; et al. (1987) Sister chromatid exchanges and DNA repair capability in sanitary
workers exposed to ethylene oxide: evaluation of the dose-effect relationship. Amer J Indust Med 12:625-637.
Sarto, F; Tomanin, R; Giacomelli, L; et al. (1990) The micronucleus assay in human exfoliated cells of the nose and
mouth: application to occupational exposures to chromic acid and ethylene oxide. Mutat Res 244:345-351.
Sarto, F; Tornqvist, MA; Tomanin, R; et al. (1991) Studies of biological and chemical monitoring of low-level
exposure to ethylene oxide. Scand J Work Environ Health 17:60-64.
SAS. (1986) SUGI supplemental library user's guide, version 5. SAS Institute, Carey, NC.
Schulte, PA; Boeniger, M; Walker, JT; et al. (1992) Biologic markers in hospital workers exposed to low levels of
ethylene oxide. Mutat Res 278:237-251.
Shore, RE; Gardner, MJ; Pannett, B. (1993) Ethylene oxide: an assessment of the epidemiological evidence on
carcinogenicity. Br J Ind Med 50:971-997.
Sisk, SC; Pluta, LJ; Meyer, KG; et al. (1997) Assessment of the in vivo mutagenicity of ethylene oxide in the tissues of
B6C3F1 lacl transgenic mice following inhalation exposure. MutatRes 391:153-164.
Snellings, WM; Weil, CS; Maronpot, RR. (1984) A two-year inhalation study of the carcinogenic potential of ethylene
oxide in Fischer 344 rats. Toxicol Appl Pharmacol 75:105-117.
Sorensen, KJ; Zetterberg, LD; Nelson, O; et al. (2000) The in vivo dose rate effect of chronic gamma radiation in
mice: translocation and micronucleus analyses. MutatRes 457:125-136.
Stayner, L; Steenland, K; Greife, A; et al. (1993) Exposure-response analysis of cancer mortality in a cohort of
workers exposed to ethylene oxide. Am J Epidemiol 138:787-798.
Steenland, K; Stayner, L. (1993) An epidemiological study of workers potentially exposed to ethylene oxide [letter].
Br JInd Med 50:1125-1126.
Steenland, K; Beaumont, J; Spaeth, S; et al. (1990) New developments in the Life Table Analysis System of the
National Institute for Occupational Safety and Health. J Occup Med 32:1091-1098.
Steenland, K; Stayner, L; Greife, A; et al. (1991) Mortality among workers exposed to ethylene oxide. N Engl J Med
324:1402-1407.
Steenland, K; Whelan, E; Deddens, J; et al. (2003) Ethylene oxide and breast cancer incidence in a cohort study of
7576 women (United States). Cancer Causes and Control 14:531-539.
Steenland, K; Stayner, L; Deddens, J. (2004) Mortality analyses in a cohort of 18,235 ethylene oxide exposed workers:
follow up extended from 1987 to 1998. Occup Environ Med 61:2-7.
08/29/06 151 DRAFT—DO NOT CITE OR QUOTE
-------
Stolley, PD; Soper, KA; Galloway, SM; et al. (1984) Sister-chromatid exchanges in association with occupational
exposure to ethylene oxide. MutatRes 129:89-102.
Swaen, GM; Slangen, JM; Ott, MG; et al. (1996) Investigation of a cluster often cases of Hodgkin's disease in an
occupational setting. Int Arch Occup Environ Health 68:224-228.
Swenberg, JA; Fedtke, N; Fennell, TR; et al. (1990) Relationships between carcinogen exposure, DNA adducts and
carcinogenesis. In: Clayson, DB; Munro, 1C; Shubik, P; et al., eds. Progress in predictive toxicology. Amsterdam:
Elsevier; pp.161-184.
Tates, AD; Grummt, T; Tornqvist, M; et al. (1991) Biological and chemical monitoring of occupational exposure to
ethylene oxide. Mutat Res 250:483-497.
Tates, AD; Boogaard, PJ; Darroudi, F; et al. (1995) Biological effect monitoring in industrial workers following
incidental exposure to high concentrations of ethylene oxide. Mutat Res 329:63-77.
Tates, AD; vanDam, FJ; Natarajan, AT; et al. (1999) Measurement of HPRT mutations in splenic lymphocytes and
haemoglobin adducts in erythrocytes of Lewis rats exposed to ethylene oxide. Mutat Res 431:397-415.
Teta, MJ; Benson, LO; Vitale, JN. (1993) Mortality study of ethylene oxide workers in chemical manufacturing: a 10
year update. Br J Ind Med 50:704-709.
Teta, MJ; Sielken, RL, Jr; Valdez-Flores, C. (1999) Ethylene oxide cancer risk assessment based on epidemiological
data: application of revised regulatory guidelines. Risk Anal 19:1135-1155.
Thier, R; Bolt, HM. (2000) Carcinogenicity and genotoxicity of ethylene oxide: new aspects and recent advances. Crit
Rev Toxicol 30:595-608.
Thiess, A; Frentzel-Beyme, R; Link, R; et al. (1982) Mortality study on employees exposed to alklylene oxides
(ethylene oxide/propylene oxide) and their derivatives. In: Prevention of occupational cancer. Geneva: International
Labour Office; pp. 249-259.
Tompa, A; Major, J; Jakab, MG. (1999) Is breast cancer cluster influenced by environmental and occupational factors
among hospital nurses in Hungary? Pathol Oncol Res 5:117-121.
Tucker, JD; Tawn EJ; Holdsworth D; et al. (1997) Biological dosimetry of radiation workers at the Sellafield nuclear
facility. Radiat Res 148:216-226.
U.S. EPA (Environmental Protection Agency). (1985) Health assessment document for ethylene oxide. National
Technical Information Service, Washington, DC. EPA/600/8-84/009F.
U.S. EPA (Environmental Protection Agency). (1986) Guidelines for carcinogen risk assessment. Office of Research
and Development, Washington, DC. EPA/630/R-00/004.
U.S. EPA (Environmental Protection Agency). (1994) Methods for derivation of inhalation reference concentrations
and application of inhalation dosimetry. Environmental Criteria and Assessment Office, Office of Health and
Environmental Assessment, Washington, Washington, DC; EPA/600/8-90/066F. Available from: National Technical
Information Service, Springfield, VA; PB2000-500023.
U.S. EPA (Environmental Protection Agency). (1997) Chemical and radiation leukemogenesis in humans and rodents
and the value of rodent models for assessing risks of lymphohematopoietic cancers. Office of Research and
Development, Washington, DC. EPA/600/R-97/090.
08/29/06 152 DRAFT—DO NOT CITE OR QUOTE
-------
U.S. EPA (Environmental Protection Agency). (1999) Guidelines for carcinogen risk assessment (proposed). NCEA-
F-0644. Available at: www.epa.gov/ncea/raf/cancer.htm.
U.S. EPA (Environmental Protection Agency). (2002) Health assessment of 1,3-butadiene. Office of Research and
Development, Washington, DC. EPA/600/P-98/001F.
U.S. EPA (Environmental Protection Agency). (2005a) Guidelines for carcinogen risk assessment. EPA/630/P-
03/001F. Available at: www.epa.gov/cancerguidelines.
U.S. EPA (Environmental Protection Agency). (2005b) Supplemental guidance for assessing susceptibility from early-
life exposure to carcinogens. EPA/630/R-03/003F. Available at: www.epa.gov/cancerguidelines.
Van Sittert, NJ; De Jong, G; Clare, MG; et al. (1985) Cytogenetic, immunological, and haematological effects in
workers in an ethylene oxide manufacturing plant. Brit J Indust Med 42:19-26.
Van Sittert, NJ; Beulink, GD, Van Vliet, EW; et al. (1993) Monitoring occupational exposure to ethylene oxide by the
determination of hemoglobin adducts. Environ Health Perspect 99:217-220.
Van Sittert, NJ; Boogaard, PJ; Natarajan, AT; et al. (2000) Formation of DNA adducts and induction of mutagenic
effects in rats following 4 weeks inhalation exposure to ethylene oxide as a basis for cancer risk assessment. Mutat Res
447:27-48.
Walker, VE; Skopek, TR. (1993) A mouse model for the study of in vivo mutational spectra: sequence specificity of
ethylene oxide at the hprt locus. Mutat Res 288:151-162.
Walker, VE; Fennell, TR; Boucheron, JA; et al. (1990) Macromolecular adducts of ethylene oxide: a literature review
and a time-course study on the formation of 7-(2-hydroxyethyl)guanine following exposures of rats by inhalation.
Mutat Res 233:151-164.
Walker, VE; MacNeela, JP; Swenberg, JA; et al. (1992a) Molecular dosimetry of ethylene oxide: formation and
persistence of N-(2-hydroxyethyl)valine in hemoglobin following repeated exposures of rats and mice. Cancer Res
52:4320-4327.
Walker, VE; Fennell,TR; Upton, PB; et al. (1992b) Molecular dosimetry of ethylene oxide: formation and persistence
of 7-(2-hydroxylethyl)guanine in DNA following repeated exposures of rats and mice. Cancer Res 52:4328-4334.
Walker, VE; Sisk, SC; Upton, PB; et al. (1997) In vivo mutagenicity of ethylene oxide at the hprt locus in T-
lymphocytes of B6C3F1 lacl transgenic mice following inhalation exposure. Mutat Res 392:211-222.
Walker, V; Meng, Q; Clement, N. (1999) Spectra of mutations in HPRT exon 3 of T-cells from F344 rats and LAC I
transgenic and nontransgenic B6C3F1 mice exposed by inhalation to ethylene oxide. Environ Mol Mutagen 29:54.
Wong, O. (1991) Mortality among workers exposed to ethylene oxide. N Engl J Med 325:1254.
Wong, O; Trent, L. (1993) An epidemiologic study of workers potentially exposed to ethylene oxide. Br J Ind Med
50:308.
World Health Organization (WHO). (2003) Concise International Chemical Assessment Document 54: ethylene oxide.
Availabe at: http://www.inchem,org/documents/cicads/cicad54.htm.
Yager, JW; Hines, CJ; Spear, RC. (1983) Exposure to ethylene oxide at work increases sister chromatid exchanges in
human peripheral lymphocytes. Science 219:1221-1223.
08/29/06 153 DRAFT—DO NOT CITE OR QUOTE
-------
Yong, LC; Schulte, PA; Wiencke, JK; et al. (2001) Hemoglobin adducts and sister chromatid exchanges in hospital
workers exposed to ethylene oxide: effects of glutathione S-transferase Tl and Ml genotypes. Cancer Epidemiol
Biomarkers Prev 10:539-550.
Zhao, CY; Kumar, R; Zahlsen, K; et al. (1997) Persistence of 7-(2-hydroxyethyl)guanine-DNA adducts in rats exposed
to ethene by inhalation. Biomarkers 2:355-359.
08/29/06 154 DRAFT—DO NOT CITE OR QUOTE
------- |