PB88-127113
Investigation of Cancer Risk Assessment Methods
Volume 1. Introduction and Epidemiology
Clement Associates, Inc., Ruston, LA
Prepared for
Environmental Protection Agency, Washington, DC
Sep 87
L
-------
EFA/600/6-87/007b
Seotember 1987
Investigation of Cancer
Risk Assessment Methods:
Volume 1. Introduction and Epidemiology
Prepared by
Bruce C. Allen
Annette M. Shipp
Kenny S. Crump
Bryan Kilian
Mary Lee Hogg
Joe Tudor
Barbara Keller
Clement Associates, Inc.
1201 Gaines Street
Ruston, Louisiana 71270
Prepared for
U. S, Environmental Protection Agency
Contract (C68-01-6807
Research Triangle Institute, Prime Contractor
OFFICE OF HEALTH AND ENVIRONMENTAL ASSESSMENT
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, DC 20460
-------
TECHNICAL REPORT DATA
(Please read Inunctions on the reverse before completing!
1 REPORT NO. |2.
EPA/600/6-87/007b
4. TITLE AND SUBTITLE
Investigation of Cancer Risk Assessment Methods:
Volume 1. Introduction and Epidemiology
7. AUTHORS Bruce C- Allen, Annette M. Shipp, Kenny S.
Crump, Bryan Kilian, Mary Lee Hogg, Joe Tudor,
Barbara Keller
9 PERFORMING ORGANIZATION NAME AND ADDRESS
Clement Associates, Inc.
1201 Gaines Street
Ruston, LA 71270
12, SPONSORING AGENCY NAME AND ADDRESS
Office of Health and Environmental Assessment
Carcinogen Assessment Group (RD-689)
U.S. Environmental Protection Agency
Washington, DC 20460
3. RECIPIENT'S ACCESSION NO.
PB88S 1 2? 1 13/A*
5. REPORT DATE
September 1987
6. PERFORMING ORGANIZATION CODE
8 PERFORMING ORGANIZATION REPORT NO
10. PROGRAM ELEMENT NO.
I1. CONTRACT/GRANT NO.
68-01-6807
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
EPA/600/21
is SUPPLEMENTARY NOTES £pA Project officer: Chao Chen, Carcinogen Assessment Group
Office of Health and Environmental Assessment, Washington, DC (382-5719)
16 ABSTRACT yne maj0r focus of this study is upon making quantitative comparisons of
carcinogenic potency in animals and humans for 23 chemicals for which suitable
animal and human data exists. These comparisons are based upon estimates of risk
related doses (RRDs) obtained from both animal and human data. An RRD represents
the average daily dose per body weight of a chemical that would result in an extra
cancer risk of 25%. Animal data on these and 21 other chemicals of interest to the
EPA and the DOD are coded into an animal data base that permits evaluation by
computer of many risk assessment approaches.
This report is the result of a two-year study to examine the assumptions,
other than those involving low dose extrapolation, used in quantitative cancer risk
assessment. The study was funded by the Department of Defense [through an inter-
agency transfer of funds to the Environmental Protection Agency (EPA)j, the EPA,
the Electric Power Research Institute and, in its latter stages, by the Risk Science
Institute.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Held.'Group
2 OiSTHlBuT ON STATEMENT
Distribute to public
19. SECURITY CLASS lTh,s Krporl,
Unclassified
?1 NO. OF PAGES
334,
20 SECURITY CLASS iThnnafei
-------
DISCLAIMER
This document has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
publication. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use. The information in this document has
been funded by the U.S. Environmental Protection Agency, the Department of
Defense (through Interagency Agreement Number RW97075101), the Electric
Power Research Institute, and the Risk Science Institute.
-------
ACKNOWLEDGMENTS
We would like to acknowledge the guidance received from the Steering
Committee for this project. The members of this committee are Dr. Roy
Albert, Lt. Col. Dennis Naugle, Dr. Roger McClellan, Dr. Werner North,
Dr. Marvin Schneiderman, Dr. Abraham Silvers, and Dr. John Van Ryzin.
The support and interest of Dr. Chao Chen, the EPA task manager, and the
EPA Carcinogen Assessment Group is appreciated.
Further acknowledgment and thanks are due the computer support personnel
who created, edited, and analyzed the data base of animal bioassays.
These individuals include Cynthia Van Landingham, Rodger Harris, Tim
Martin, Karen Wright, Greg Couch, and Robin Nicholson. Last, but
certainly not least, the tireless patience of Lynn Williams, who typed
all of the many versions of this report, is greatly appreciated.
111
-------
CONTENTS
Section Poqe
1 INTRODUCTION
Background 1-1
Study Goals 1-2
Identification of the Chemical Data Base 1-7
2 EPIDEMIOLOGY
Introduction 2-1
Methods 2-2
Uncertainty in Exposure Estimates 2-3
Dose-Response Models 2-8
Calculation and Selection of RRD Estimates 2-12
Results 2-14
Aflatoxin 2-16
Arsenic 2-27
Asbestos 2-41
Benzene 2-49
Benzidine 2-65
Cadmium 2-73
Chlorambucil ' 2-82
Chromium 2-87
Cigarette Smoke 2-103
Diethylstilbestrol 2-110
Epichlorohydrin 2-117
Estrogen 2-126
Ethylene Oxide 2-147
Isoniazid 2-155
Melphalan 2-165
Methylene Chloride 2-171
Nickel 2-178
Polychlorinated Biphenyls 2-188
Phenacetin 2-201
Reserpine 2-211
Saccharin 2-218
Trichloroethylene 2-224
Vinyl Chloride 2-2o7
Summary of Results 2-263
Discussion 2-264
IV
-------
ILLUSTRATION
igure Poge
2-1 Representation of RRD Estimates Obtained for All 2-269
Chemicals and Each Putative Site of Action
-------
TABLES
Toble Page
1-1 Components of Risk Assessment: Choices to be Mode 1-3
1-2 Approaches to Risk Assessment Components 1-5
1-3 Chemicals for Which Minimal Human and Animal Data 1-8
Exist for Quantitative Risk Estimation
1-4 Other Chemicals for Which Data Has Been Collected 1-11
2-1 Distribution of Filippino Cases and Controls 2-22
With Respect to Daily Aflatoxin Intake
2-2 Data from Cross-Sectional Studies of Aflatoxin 2-23
Intone and Primary Liver Cancer Incidence
2-3 Dose and Response Data Estimation of a and 0 2-24
2-4 Population Statistics Used for Calculation of F 2-25
2-5 RRD Estimates for Lifetime Exposure to Aflatoxin 2-26
2-6 RRD Estimates for Aflatoxin 2-26
2-7 Dose and Response Data For the Cohort of Workers 2-36
Exposed to Arsenic at the Tocoma, Washington Smelter
2-8 Observed and Expected Deaths from Respiratory Cancer, 2-36
By Maximum Exposure to Arsenic and Length of
Employment, Anaconda Employees
2-9 Dose and Response Data for Anaconda Employees, 2-37
from the Lee-Foldstein Categorization
2-10 Dose and Response D^ta for the Welch et al. 2-38
Cohort of Anaconda Workers
2-11 Respiratory Cancer Potency Parameter Estimates for Arsenic 2-39
2-12 RRD Estimates for Arsenic 2-40
2-13 Values of KL and KM Obtained in the Analysis 2-47
of Eleven Studies of Asbestos Workers
2-14 HRD Estimates for Asbestos (Total Fibers) 2-48
2-15 RRD Estimates for Asbestos (mg/day) 2-48
2-16 Classification of Job Titles in the Ott et ol. 2-56
by Exposure to Benzene
-------
TABLES
Toble Poqe
2-17 Observed and Expected Numbers of Deaths in the 2-57
Ott et al. Cohort, by Cumulative Dose of Benzene
2-18 Benzene Exposure (ppm) by Operation Code and Year 2-58
for Location 1, Rinsky et al. Cohort
2-19 Benzene Exposure (ppm) by Operation Code and Year 2-59
for Location 2, Rinsky et al. Cohort
2-20 Observed and Expected Numbers of Deaths in the 2-60
Rinsky et al. Cohort, by Cumulative Dose of Benzene
2-21 Observed and Expected Numbers of Deaths in the 2-61
Wong Cohort, by Cumulative Dose of Benzene
2-22 Benzene Potency Parameter Estimates 2-62
2-23 RRD Estimates for Benzene 2-64
2-24 Concentrations of Benzidine in Atmosphere at 2-71
Different Locations of Benzidine Manufacturing Plant
2-25 Bladder Cancer Potency Parameter Estimates for 2-71
Benzidine, From Data in Zavon et al.
2-26 RRD Estimates for Benzidine 2-72
2-27 Estimates of Cadmium Inhalation Exposure, 2-79
by Plant Department and Time Period
2-28 Dose and Response Data for the Cadmium-Exposed 2-80
Cohort Studied by Thun et al.
2-29 Cadmium Lung Cancer Potency Estimates 2-80
((mg-yrs/m3)"1), for Thun et al. Cohort
2-30 RRD Estimates for Cadmium 2-81
2-31 Dose and Acute Leukemia Response Data for Chlorambucil 2-85
3-32 Leukemia Potency Parameter Estimates for Chlorambucil, 2-85
Based on the Study by Berk et al.
2-33 RRD Estimates for Chlorambucil 2-86
2-34 Chromate Exposure in Different Work Operations; 2-97
Langard and Norseth
-------
TABLES
Toble Poge
2-35 Chromate Exposure in Different Work Departments; 2-97
Langard et al.
2-36 Dose and Response Data from Chromate-Exposed Cohort 2-98
of Langard et al.
2-37 Estimates Concentrations of Chromium by Working Site; 2-98
Axelsson et al.
2-38 Dose and Response Data from Chromate-Exposed Cohort 2-99
of Axelsson et al.
2-39 Potency Parameter Estimates for Chromium 2-100
2-40 RRD Estimates for Chromium 2-102
2-41 Annual Death Rates per 100,000 among British Physicians 2-108
2-42 RRD Estimates for Cigarette Smoke 2-109
2-43 Dosage and Duration of Stilbestrol Therapy; 2-115
Herbst et al.
2-44 Dose and Response Data for Case-Control Study 2-115
of Herbst et al.
2-45 Vaginal Cancer Potency Parameter Estimates for DES, 2-116
From Data in Herbst et al.
2-46 RRD Estimates for DES 2-116
2-47 Duration of Exposure to Epichlorohydrin for the 2-122
Cohort Studies by Shellenberger et al.
2-48 Dose and Response Data for Epichlorohydrin-Exposed 2-122
Cohort of Shellenberger et al.
2-49 Duration of Exposure to Epichlorohydrin for the 2-123
Cohort Studies by Tassignon et al.
2-50 Dose and Response Data for Epichlorohydrin-Exposed 2-123
Cohort of Tassignon et al.
2-51 Epichlorohydrin Potency Parameter Estimates 2-124
2-52 RRD Estimates for Epichlorohydrin 2-125
2-53 Numbers of Breast Cancer Cases ana Controls by Total 2-136
Accumulated Dose of Conjugated Estrogen and Ovarian Status
-------
TABLES
Table Poge
2-54 Risk Ratios by Dose and Duration of Administration 2-137
of Conjugated Estrogen for Endometrial Carcinoma
Cases and Controls
2-55 Numbers of Endometrial Carcinoma Cases and Controls 2-138
by Drug-Free Days in Cycle and Wean Pill Size
2-56 Numbers of Endometrial Carcinoma Cases and Controls 2-139
by Total Dose
2-57 Distribution of Cases and Controls by Duration, Dose, 2-140
and Type of Administration of Conjugated Estrogens
2-5S Numbers of Endometrial Carcinoma Cases and Controls 2-141
in Olmsted County by Total Dose
2-59 Distribution of Endometrial Carcinoma Cases and Controls 2-141
by Duration of Use and Pill Strength of Conjugated Estrogen
2-60 Numbers of Endometrial Carcinoma Cases and Controls in a 2-142
Louisville, Kentucky Private Practice by Total Dose
2-61 Number of Endometrial Cancer Cases and Controls in 2-142
Baltimore-Area Hospitals by Daily Dose and Duration
of Use of Conjugated Estrogens
2-62 Distribution of Hypoestrogenic Patients by Total 2-143
Dose of Estrogen
2-63 Dose and Response Data for Hypoestrogenic Patients 2-143
2-64 Potency Parameter Estimates for Estrogens 2-144
2-65 RRD Estimates for Estrogens 2-146
2-66 Dose and Response Data for Ethylene Oxide-Exposed 2-152
Employees; Hogstedt et al.
2-67 Ethylene Oxide Leukemia Potency Parameter Estimates 2-153
2-68 RRD Estimates for Ethylene Oxide 2-154
2-69 Dose and Response Data for Tuberculosis Patients 2-160
Treated with Isoniazid
2-70 Cancer Deaths Among Household Members of Tubercular 2-161
Patients by Treatment Group and Year of Observation
-------
TABLES
Table Pogo
2-71 Career Deaths Among Mental Institution Tubercular 2-162
Patients by Treatment Group and Year of Observation
2-72 Potency Parameter Estimates for Isoninzid 2-163
2-73 RRD Estimates for Isoniazid 2-164
2-74 Observed and Expected Cases of Acute Nonlymphocytic 2-168
Leukemia by Trial and Initial Chemotherapy Dose
2-75 Dose and Response Data for Melphalan-Treated 2-169
and Control Patients
2 76 Leukemia Potency Parameter Estimates for Melpholan, 2-169
Based on the Study of Greene et aI.
2-77 RRD Estimates for Melpholan 2-170
2-78 Potency Parameter Estimates for Methylene Chloride 2-176
2-79 RRD Estimates for Methylene Chloride 2-177
2-80 Concentrations of Nickel from Personal Air Samplers 2-184
Worn by Welders and the Oak Ridge Gaseous
Diffusion Plant, Polednak
2-81 Dose and Response Data for Nickel-Exposed Workers 2-184
Studied by Polednak
2-82 Observed and Expected Deaths for 3 Groups of Male Nickel 2-185
Workers, 20 Years or More After First Exposure
2-83 Dose and Response Data for Nickel-Exposed Cohort 2-185
Studied by Enterline and Marsh
2-84 Nickel Respiratory Cancer Potency Parameter Estimates 2-186
2-85 RRD Estimates for Nickel 2-187
2-86 Mortality Experience of a PCB-Exposed Cohort of Workers 2-194
2-87 Concentrations of PCB at Plant 2 2-194
2-88 Duration of Employment Among Cohort Workers in PCB 2-19b
Exposure Jobs
2-89 Cancer Response Among Capacitor Manufacturers 2-195
-------
TABLES
Table Page
2-90 Observed and Expected Cancer Deaths by Length of Exposure 2-196
Among Capacitor Manufacturers
2-91 Dose and Response Data for Brown and Jones Cohort 2-197
of RGB-Exposed Workers
2-92 Potency Parameter Estimates for PCBs 2-198
2-93 RRD Estimates for PCBs 2-200
2-9
-------
TABLES
Toble Page
2-107 TCA Concentrations in Urine Samples from Workers 2-232
Exposed to Trichloroethylene at Various Concentrations
2-108 Dose and Response Data Derived form Tola et al. 2-233
Trichloroethylene-Exposed Cohort
2-109 Description of the Axelsson et al. Subcohort of 2-233
Trichloroethylene-Exposed Men With at Least
10 Years Latency
2-110 Dose and Response Data Derived from the Axelsson et al. 2-234
Cohort of Trichloroethylene-Exposed Workers
2-111 Potency Parameter Estimates for Trichloroethylene 2-235
2-112 RRD Estimates for Trichloroethylene 2-236
2-113 Duration of Exposure by Level of Exposure for Each 2-251
Exposure Grouping (Arsenic Workers Excluded),
Vinyl Chloride-Exposed Cohort
2-114 Observed and Expected Deaths by Exposure Intensity 2-252
and Duration of Exposure, 1942-1973, Ott et al.
Vinyl Chloride-Exposed Cohort (Arsenic Workers Excluded)
2-115 Dose and Response Information for the Vinyl Chloride- 2-253
Exposed Cohort of Ott et al. (Arsenic Workers Excluded)
2-116 Levels of Exposure and Lengths of Exposure for Men 2-254
in the Fox and Collier Vinyl Chloride-Exposed Cohort
2-117 Dose and Response Information for the Fox and Collier 2-255
Cohort of Vinyl Chloride-Exposed Workers
2-118 Estimated Vinyl Chloride Concentrations for Buffler et ol. 2-256
Cohort, by Time and Job Classification
2-119 Dose and Response Data for the Vinyl Chloride-Exposed 2-257
Cohort of Buffler et al.
2-120 Dose and Response Data for the Vinyl Chloride-Exposed 2-258
Cohort of Heldaas et al.
2-121 Potency Parameter Estimates for Vinyl Chloride 2-259
2-122 RRD Estimates for Vinyl Chloride 2-261
2-123 RRD Estimates Selected from the Epidemiologic Data 2-268
xi i
-------
Section 1
INTRODUCTION
BACKGROUND
In recent years there has been a growing awareness of the potential
health hazard from chemicals encountered in environmental or occupa-
tional settings. Paralleling this awareness has been an increasing
interest in the characterization of hu-nan health risks resulting from
exposures to potentially hazardous chemicals. Often, such characteri-
zations (risk assessments) must rely solely on experimental animal data
because of lack of relevant epidemiological studies. Lifetime animal
bioassays are frequently used to identify adverse health effects and to
estimate risk to humans. Risk assessments based on animal data involve
a series of assumptions concerning such issues as the dose response
model, the appropriateness of the animal data, and dose conversion
factors between animals and humans.
Carcinogenicity bioassays are usually designed as screening procedures
with the primary focus being identification of potential human hazards,
rather than human risk assessment. In these studies, a limited number
of animals may ba exposed to the maximum tolerated dose, which is a
level often several orders of magnitude higher than the doses encoun-
tered by man. To estimate human risk from such studies requires both
the extrapolation of results from high doses to low doses and from
animals to humans. The present study focuses i:pon extrapolation from
animals to humans, which has not L>een studied as extensively as
extrapolation from high to low dose.
Extrapolation from animals to humans involves a number of components,
each requiring a choice among several possible approaches. The
approaches to the components reflect the assumptions that are made with
respect to conditions likely to provide biological comparability of
1-1
-------
animals and humans. The choice of a particular approach can have a
considerable effect upon the resulting quantitative estimates. Although
these components are based on scientifically sound biological principles
insofar as possible, it is unfortunately the case that for most of the
components of risk analysis there is often no scientifically defensible
"correct" procedure. The NAS Committee on the Institutional Means for
Assessment of Risk to Public Health (i) listed the components of risk
assessment, and discussed possible approaches for each component. The
Committee also recommended that "detailed" but "flexible" guidelines be
developed for addressing each component of risk assessment.
STUDY GOALS
This study is designed to provide a detailed and systematic investiga-
tion of the components, other than low-dose extrapolation, that relate
to estimation of human carcinogenic risk from animal bioassay data.
Several components of quantitative risk assessment using animal models
are listed in Table 1-1. Those components relate to the choices that
must be made during the course of a risk assessment and involve such
things as experimental design, analysis of data, and choice of studies
to use when more than one is available. Components reflecting selection
of appropriate experimental protocols include the route of administra-
tion and the duration of dosing and observation. Analysis of an experi-
ment depends on the mathematical model relating dose and biological
response. Average dose levels can be calculated in several ways and
expressed in many units; the units chosen are those assumed to provide
human and animal equivalence with respect to carcinogenic response.
Response parameters depend on choice of site and type of tumor and
calculation of response rates. When more than one study is available,
one must determine which studies will be used and how they w.'.ll be
combined. Such choices may be based on selections of particular
species, studies, or sex of test animals. A multitude of analysis
procedures are possible, all of which may be more or less equally
supported by scientifically defensible arguments.
An empirical approach has been taken in the present study to investigate
the issue of human risk estimation based on animal carcinogenicity data.
1-2
-------
Table 1-1
COMPONENTS OF RISK ASSESSMENT: CHOICES TO BE MADE
A. Requirements for a Study
1. Length of the experiment
2. Length of dosing
3. Route of exposure
B. Analysis of a Study
1. Dose
a. Units assumed to give human-animal equivalence
b. Expression of dose level values
2. Response
a. Animals to use in analysis
b. Malignancy status to consider
c. Particular tumor type to use
3. Dose-Response model
C. Multiplicity of Sti'dies
1. Sex to use
2. Study to use
3. Species to use
1-3
-------
Having identified the steps in risk assessment at which crucial choices
must be made (Table 1-1), alternative approaches have been identified
for each of the steps (with the exception of the selection of a dose-
response mcdel, which has been studied extensively elsewhere in connec-
tion with the low-dose extrapolation problem). The approaches selected
(Table 1-2) either have been used in past assessments, seem potentially
useful, or seem particularly plausible because of biological considera-
tions. The goal is to derive estimates of risk from the bioassay data
by various combinations of the approaches selected and to compare those
estimates to risk estimates derived from epidemiologic data (cf.
Volume-3). Section 2 of this volume describes the manner in which the
human-based estimates are determined. These reprettsnt our best esti-
mates of risk to humans derived solely from epidemioloqical data; they
are the "targets" at which the bioassay-based estimates are aimed and
the standards against which they are Judged.
The detailed evaluation of the different methods of analysis of the
bioassay data allows one to address the question of uncertainty in
animal-to-human extrapolation. Examination of a wide range of methods
provides a range of risk estimates that are based on different but
scientifically acceptable assumptions, and so will aid in the develop-
ment of guidelines for presenting uncertainty. That examination also
allows identification of the set of assumptions, those relating to the
components of risk assessment as well as those relating to the quality
of the data (e.g. with respect to sample size), that produce the best
correlation of risk estimates between humans and animals. It is also
possible to study the uncertainty attributable to each component; by
including single-component variations of a standard protocol (i.e.
those that differ from the standard only in selection of an alternative
approach for one component, all other components remaining fixed) we
obtain ranges cT risk estimates that relate to the uncertainty
associated with each of the components. In all these respects, we
provide information about the extrapolation from animals to humans and
the uncertainty that is involved. Additional research may be warranted
in those areas that are less well characterized, i.e. those showing
considerable variability or uncertainty.
-------
Table 1-2
APPROACHES TO RISK ASSESSMENT COMPONENTS
1. Length of the experiment
a. Use data from any experiment but correct for short observation
periods.
b. Use data from experiments which last no less than 90* of the
standard experiment length of the test animal.
2. Length of dosing
a. Use data from any experiment, regardless of exposure duration.
b. Use data from experiments that expose animals to the test
chemical no less than 80* of the standard experiment
length.
3. Route of exposure
a. Use data from experiments for which route of exposure is most
similar to that encountered by humans.
b. Use data from any experiment, regardless of route of exposure.
c. Use data from experiments that exposed animals by gavage,
inhalation, any oral route, or by the route most similar to
that encountered by humans.
4. Units assumed to give human-animal equivalence
a. mg/kg body wt/day.
b. ppm in diet.
c. ppm in air.
d. mg/kg body wt/lifetime.
e. mg/m2 surface area/day.
5. Expression of dose-level values
a. Doses expressed as average dose up to termination of
experiment.
b. Doses expressed as average dose over the first 80* of the
experiment.
6. Animals to use in analysis
a. Use all animals examined for the particular tumor type.
b. Use animals surviving just prior to discovery of the first
tumor of the type chosen.
1-5
-------
Table 1-2 (continued)
APPROACHES TO RISK ASSESSMENT COMPONENTS
7. Malignancy status to consider
a. Consider malignant tumors only.
b. Consider both benign and malignant tumors.
8. Particular tumor type to use
a. Use combination of tumor types with significant
dose-response.
b. Use total tumor-bearing animals.
c. Use response that occurs in humans.
d. Use any individual response.
9. Sex to use
a. Use each sex within a study separately.
b. Combine the results of different sexes within a study.
10. Study to use
a. Consider every study within a species separately.
b. Combine the results of different studies within a species.
11. Species to use
a. Combine results from all available species.
b. Combine results from mice and rats.
c. Use data from a single, preselected species.
d. Consider all species separately.
1-6
-------
IDENTIFICATION OF THE CHEMICAL DATA BASE
To accomplish the goals of this investigation, a suitable collection of
chemicals is needed. Both human epidemiological and animal biassay
carcinogenicity data ure required for these chemicals. Moreover, the
data from those tests need to be available and adequate to support a
quantitative approach to risk assessment in both humans and animals.
Table 1-3 lists the chemicals for which were found at least minimal
human and animal data capable of supporting the comparative analysis
that is the goal of this study. This list has been compiled by review
of the literature, starting with the monographs produced by the
International Agency for Research on Cancer (IARC). Studies by the
National Academy of Science (2) and Crouch and Wilson (3) that made
similar comparisons of bioassay- and epidemiology-based risk estimates
were also reviewed. Primarily, however, the list has been developed by
perusal of the literature on specific cancer types, on chemical carci-
nogenesis in general, and on individual epidemiological investigations
in particular. It is almost always the case that the availability of
adequate, quantitative human data is a key factor for determining
whether a chemical could be included in the study, so thorough review of
the epidemiologic literature was essential. (Only one chemical,
treosulfan, had human data suitable for a quantitative approach but
lacked animal bioassay data. ) We are grateful to a number of persons
who suggested potentially useful substances, some of which were found to
be acceptable for our quantitative analysis.
For a chemical to be included in the comparative analysis, the animal
data or the human data, but not necessarily both, had to provide reason-
ably strong evidence of carcinogenicity. Note in this regard, that IARC
considers the evidence for the carcinogenicity of each chemical listed
in Table 1-3 (except methotrexate) to be either "limited" or "suffi-
cient" for either animals or humans. On the other hand, chemicals that
are carcinogenic in neither humans or animals provide no information on
the comparison of risk estimates and, hence, are not useful for this
investigation. For example, although methotrexate has been the subject
of both bioassay and epidemiologic investigations which provide the
necessary quantitative information for risk assessment, methotrrxate is
1-7
-------
Table 1-3
CHEMICALS FOR WHICH MINIMAL HUMAN AND ANIMAL
DATA EXIST FOR QUANTITATIVE RISK ESTIMATION
Evidence for Carcinogenicity
(IARC designation)
Chemical
In Humans
'In Animals
Benzene Sufficient Limited
Benzidine Sufficient Sufficient
Trichloroethylene Inadequate Limited
Vinyl Chloride Sufficient Sufficient
Aflatoxin Limited Sufficient
Arsenic Sufficient Insufficient
Nickel Limited Sufficient
Asbestos Sufficient Sufficient
Chlorombucil Sufficient Sufficient
Estrogens (conjugated) Sufficient Insufficient
Isoniazid (isonicotinic acid Inadequate Limited
hydrazide)
Melphalan ' Sufficient Sufficient
Methotrexate Inadequate Inadequate
Phenacetin (analgesics containing Sufficient Limited
phenacetin)
Reserpins Inadequate Limited
Tobacco smokea - - - -
Diethylstilbestrol (DES) Sufficient Sufficient
Ethylene oxide Inadequate Limited
Saccharin Inadequate Limited
Chromium Sufficient Sufficient
Polychlorinated Biphenyls Inadequate Sufficient
Methylene Chloride Inadequate Sufficient1*
Epichlorohydrin Inadequate Sufficient
Cadmium Limited Sufficient
°Not considered in IARC monographs, although acknowledged by IARC as a
known human carcinogen.
^Although classified as "Inadequate" by IARC (U), results of studies
completed since IARC evaluation indicate that the evidence for the
corcinogenicity of methylene chloride in animals is now "Sufficient"
(5).
1-8
-------
not used in this study because neither the animal nor human studies
provide evidence of carcinogenicity for methotrexate.
It is important to note that it is neither necessary nor sufficient that
a chemical be unequivocally carcinogenic in humans in order that that
chemical be included in the present investigation. Thus, a chemical
such as saccharin, which has been associated with cancer only in labora-
tory rodents, is included while bis(chloromethyl) ether is not included,
even though sufficient evidence apparently exists to establish that
bis(chloromethyl) ether is carcinogenic in humans (4). Of the 23 chemi-
cals or chemical groups that IARC considered in 1982 to have 'suffi-
cient* evidence of human carcinogenicity, 12 are included in this study
(including cigarette smoke). Eleven other chemicals have been included;
three were considered to provide "limited" evidence and eight to provide
•inadequate* evidence in support of human carcinogenic effects.
We feel that it i» essential to include in the study chemicals for which
carcinogenicity in humans is not yet established. One of the ultimate
gcals of the study is to compare the predictions of carcinogenic potency
of chemicals derived from animal data with the corresponding potency in
humans. Bias could result if such comparisons were restricted to
confirmed human carcinogens: the ability of the animal data to predict
human results might be overestimated. The same would be true if the
study were restricted to confirmed animal carcinogens. Although a
similar study by the National Academy of Sciences was restricted to
confirmed human carcinogens, the authors recognized the potential for
bias in this approach (2). Negative epidemiological studies con be used
in some cases to establish upper limits on the potential carcinogenic
potency of chemicals, and such limits can be evaluated for compatibility
with carcinogenic potencies estimated from animal data. These limits
are also useful for regulatory and other purposes. On the other hand,
we recognize the potential for misinterpretation of results derived from
data on chemicals not established to be carcinogenic in humans. We have
attempted, in the analyses and discussions which follow, to minimize the
possibilities for such misinterpretations.
In addition to the twenty-three chemicals that have appropriate epidem-
iological data, the contractor (the Environmental Protection Agency and
1-9
-------
the Department of Defense) supplied a list of 24 other substances that
are of interest or concern with respect to possible carcinogenicity
(Table 1-4). The available animal carcinogenicity bioassay data have
been collected for all the chemicals in Tables 1-3 and 1-4. When
available, IARC monographs, EPA criteria documents, EPA health assess-
ment documents, EPA Carcinogen Assessment Group (CAG) assessment docu-
ments, and National Cancer Institute (NCI) or National Toxicology
Program (NTP) technical reports have been consulted for references to
animal carcinogenicity bioassays. Computerized data bases were also
checked for suitable experimental studies. The data bases searched for
references include Medline, Chemical Exposure, Biosis, Embase, and
National Technical Information Service (NTIS). In addition, research
articles were checked for references to earlier bioassay data. The NCI
and NTP reports were particularly helpful in terms of supplying rodent
carcinogenicity bioassay data for many chemicals, as was the data base
published by Gold et al. (6). All bioassays listed by Gold and her
associates for the chemicals of interest have been collected.
Ths chemical data base is described in detail in Volumes 1 and 2.
Section 2 of this volume presents the epidemiological data that deter-
mine the direct estimates of cancer risk. Volume 2 devotes attention to
the content of the animal data base. Finally, Volume 3 presents the
comparison of the two sets of risk estimates, human and animal, and
discusses the implications of the results, especially with respect to
uncertainty and productiveness of bioassay-based risk assessment.
1-10
-------
Table 1-4
OTHER CHEMICALS FOR WHICH DATA HAS BEEN COLLECTED0
Acrylonitrile
Allyl chloride
4-Aminobiphenyl
Benzo(a)pyene
Carbon tetrachloride
Chlordane
Chlornaphazine
Chloroform
3,3-Dichlorobenzidine
1,2-Dichloroethane
Diphenylhydrazine
Ethylene dibromide
Formaldehyde
Hexachlorobenzene
Hydrazine
Lead
Mustard gas
2-Naphthylamine
Nitrilotriacetic acid
2, *,6-Trichlorophenol
TCDD
Tetrachloroethylene
Toxaphene
Vinylidene chloride
aRequested by the Environmental Protection Agency and the Department of
Defense. .
1-11
-------
REFERENCES
1. Notional Research Council (1983). Risk Assessment in the Federal
Government: Managing the Process. National Academy Press.
Washington, D.C.
2. National Academy of Sciences Executive Committee (1375).
Contemporary Pest Control Practices and Prospects. Pest Control :
An Assessment of Present and Alternative Technologies. Vol. 1.
3. Crouch, E. and Wilson, R. (1979). Interspecies comparision of
carcinogenic potency. Journal of Toxicology and Environmental
Health 5:1095-1118.
-------
Section 2
EPIDEMIOLOGY
.on 1, the risk estimates derived from epidemiologi-
Lhe basis for comparison of the bioassay analyses.
at yield predictions of risk that are closest to the
rived directly from the epidemic-logical data are deemed
• analyses that do not match the direct estimates as
estimates of carcinogenic potency based on epidemiologic
only as guides for evaluating the accuracy of estimates
. mal bioassay data; they are of fundamental importance in
jht. A realistic program for the protection of humans from
jnic effects of chemicals should take into account the
jrmation on potency in humans provided by the epidemiologic
or those chemicals with human data suitable for quantitative
jtion of risk, we have provided detailed analyses of their carcino-
ic potency in humans.
The data reported in the epidemiologic literature varies greatly in
format and quolity. Three types of studies are represented in the data
we have analyzed: prospective cohort studies (including clinical
trials), case-control studies, and (in the case of aflatoxin) a cross-
sectional comparison of cancer rates and levels of exposure in different
populations. Even within one of these categories, the individual
studies often differ in their handling of dose groups, latency, and
expected numbers of cancers. Such variation within the epidemiologic
literature makes our goal of a common, balanced treatment of all the
chemicals difficult to achieve. It has been necessary to tailor the
2-1
-------
analyses to some extent, so as to be able to take advantage of the
particular data available. Because of the wide variations in epidemio-
logicil studies, a systematic, standardized method of recording the
human data (like that developed for the bioassay data base) is not
considered feasible. The epidemiologic data for each chemical is
considered as a whole and risk estimates have been developed using
general guidelines. Those guidelines, described in this section, have
been developed so that, to the extent possible, the methodology 1) can
be employed with -a minimal amount of data, 2) makes best use of the
data, and 3) ensures that risk estimates made from data of differing
types and quality are comparable. While these requirements can be at
odds with one another, the approaches we describe have been able to
accommodate without modification all the chemicals listed in Table 1-3
with the exceptions of asbestos, cigarette smoke, and aflatoxin. The
modifications necessary for these exceptions are described in the
sections pertaining specifically to these chemicals.
METHODS
An epidemiological study provides adequate data for the- analyses
performed in the course of this investigation if dose can be estimated
quantitatively, if the observed numbers of responses (cancers) for each
dose group is known, and if a measure of the expected numbers of
responses for each dose group is available. The measure of expected
response varies according to the format of the study; for a prospective
study, expected numbers will be based on the response ratn in a refer-
ence population and for a case-control study the control series provides
the appropriate comparison groups. Regardless of the study format,
however, the major factor limiting the number of studies that are
suitable for a quantitative analysis is the sparsity of exposure data.
With very few exceptions, the amount of substance to which study parti-
2-2
-------
cipants have been exposed is incompletely known. The single major
uncertainty affecting the estimates of risk from an epidemiological
study is often this uncertainty in the dose variable. We have attempted
to quantitatively incorporate that uncertainty, as described below.
Also described below are the details of the methods used to estimate
risk from prospective and case-control studies and a description of the
approach used to determine the final risk related dose (RRD) estimates
for each chemical.
Uncertainty in Exposure Estimates
There are many sources of uncertainty in exposures in the epidemiologi-
cally studied populations. For example, exposures in occupational
cohorts are often measured infrequently and those measurements that are
made are sometimes of uncertain relevance to exposures of specific
workers. It was considered to be important for this study to quantify
these uncertainties, even though such quantification is difficult. The
approach adopted is to estimate uncertainty factors a and 7 (to be
described in detail below) that represent our impression of the uncer-
tainty of the dose estimates for any given study, and which are
developed from subfactors associated with different sources of uncei—
tainty. Some degree of subjectivity is unavoidable when determining
these factors. To promote uniformity, the subfactors discussed below
and the intervals from which they are chosen were formulated a priori.
A single investigator (B.A.) developed the bounds for each chemical for
each study. As additional studies were analyzed, the uncertainty bounds
derived earlier were reviewed and occasionally revised. All of the
analyses of the epidemiological data were performed independently of the
analyses of the animal data.
A set of three dose measures is estimated for each dose group in each
epidemioligical study used in this investigation. Let d^ represent the
2-3
-------
best estimate of cumulative dose for group i determined from the data
presented in a specific study. The two other dose estimates associated
with group i are upper and lower bounds on dj, labeled dj.u and d^ \_,
respectively, and are defined by
di,U
di,L
where
72 •*••••+ 18
The factors, a and 7, are uncertainty factors determined by the eight
subf actors as shown. Those subf actors correspond to eight sources of
potential uncertainty that may be present in a study. The following
describes the considerations that determine the magnitude of the sub-
factors, each of which is estimated independently for each dose group.
1. Length of exposure: a-) and T\ . It is often the case that there
is some uncertainty about length of exposure to the substance
in question. This might arise in a case-control study because
recollection by the patient (or his relatives or friends) may
be uncertain. Or, in a prospective study, length of exposure
may be completely undocumented, in which case we assume a
default value of 7 years. Most often, however, length of
exposure will be categorized but average duration values for
the categories will not be available. If duration of exposure
is completely unknown, then a-] and T\ are set equal to 1.5.
Otherwise, a-\ and T\ are assigned values from the interval [0,
0.3]. The specific values selected depend on the width of the
intervals defining the duration categories (the wider the
-------
intervals, the larger are a-j and TI ) and on additional informa-
tion pertinent to the estimation of average length of exposure.
2. Measurement of exposure early in the exposure period: «2 an(*
72- In prospective, occupational studies it is often the case
that concentrations of the chemical under investigation that
prevailed in the more distant past are not well documented.
Similarly, recall problems may entail greater uncertainty about
early exposure experiences in a case-control setting. To account
for such possibilities, 02 and ~*2 are selected from the interval
[0, 0.8]. The estimation of these subfactors depends on descrip-
tions of process changes affecting exposure, on the length of
time without adequate exposure documentation, and on the method
used to compensate for the lack of early measurements. If, for
instance, early concentrations are known to have been higher
than more recent concentrations but no extrapolation that
estimate* those higher concentrations is performed, then cxj,
the subfactor contributing to the lower bound, would be set
equal to zero (the uncertainty here does not affect the lower
bound, since exposures are underestimated) but 72 *°uld D*
positive.
3. Completeness of measurements (aside from the early exposure
period; cf. 2 above): 03 and 73. These subfactorm pertain to
the extent to which exposures have been measured and documen-
ted. Considerations such as the number of samples taken, the
amount of variability seen in the samples, and the completeness
of the sampling with respect to different areas or departments
of a facility (in an occupational setting), and with respect to
different periods of time are relevant to the determination of
013 and 73. These subfactors are chosen from the interval [0,
0.5].
k. Categorization of exposure: a^ and 74. As with duration of
exposure, intensity of exposure (or cumulative exposure) can be
2-5
-------
categorized into groups ar.d presented without average values
for those groups. Both a^ and 7^ are selected from [0, 0.3].
In this case also, the specific values depend on the presence
or absence of data pertinent to the estimation of those average
values.
5. Recording biases: 05 and 75. The presence of oiases in the
format of the presentation of results is a major contributor to
uncertainty. Common sources of bias include classification of
study participants by their maximum exposure and, more subtly,
duration and intensity of exposure being reported separately,
rather than cross-classified. In the former instance (classi-
fication by maximum exposure) 75 would equal zero but 015 would
be positive. Both 05 and 75 may be positive in the second case
— we may not know if those with less intense exposure were
exposed for longer or shorter periods, on average, than those
with more intense exposure. The subfactors, 05 and 75, are
chosen from th<» interval [0, 1.0], based on the estimated
degree of bias.
6. Applicability of reported exposures: ag and 75. Unless every
study participant had personal sampling performed during the
period of exposure, there is bound to be some uncertainty with
respect to the applicability of the reported exposures. That
uncertainty may be fairly minor, as when area as opposed to
personal concentration measurements determine an individual's
degree of exposure. In other cases, the uncertainty may be
rm -o substantial, as when respirator use is not considered or
concentrations from some different, though similar, facility
are used to estimate exposure. The subfactors ag and 73 are
chosen from the interval [0, 0.3].
7. Conversion of units: ay and 77. In several instances it has
been necessary to convert from one set of units to another.
This most often involves conversion of concentrations of a
2-6
-------
chemical found in the urine of exposed individuals to atmos-
pheric concentrations of the substance under investigation.
Such conversions involve both biological and statistical
variability. Whatever is known about that variability is used
to select ay and -77 from the interval [0, 1.0].
8. Expected numbers of cancers: 1x3 and 73. Uncertainty with
respect to the expected numbers of cancers is included here
even though it does not relate directly to bounds on the dose
values. Because epidemiology studies are observational in
nature rather than being carefully controlled experiments, it
is always possible that exposures to other substances or other
confounding effects may have been partly responsible for the
quantitative findings of a study. Whenever possible, we
restrict our analyses to cohorts or subcohorts with minimal
opportunity for exposure to multiple chemicals. Otherwise, we
can do little in the secondary analyses we conduct to directly
mitigate this problem, aside from noting those exposures that
may influence the results and to include them in the uncer-
tainty calculations, via the subfactors
-------
The bounds on the estimates of dose that result from the application of
the uncertainty factors are employed, as are the best estimates of dose,
in the dose-response models fit to the epidemiological data. In this
way, uncertainty with respect to exposure is incorporated in the deriva-
tion of risk estimates consistent with the data.
Dose-Response Models
A dose response model is used to relate the observed response to a
measure of cumulative dose for each group in a specific study. Both
prospective and case-control studies have been analyzed in such a way as
to provide a consistent approach to dose response and the subsequent
risk estimation.
Prospective Studies. The minimum amount of information required for an
analysis of i prospective study consists of a single group with known
cumulative dose (expressed in ppm-years, for example), an observed
number of cancers, and an expected number of cancers. Additional infor-
mation on observed and expected responses categorized by exposure group
is accommodated by the same approach and may provide better estimates of
carcinogenic potency.
The basic analytical treatment is as follows. Suppose we have data
divided into n groups (rot). For each group, i, we know the observed
number of cancers (0^), the expected number of cancers (E^) and a
measure of cumulative exposure (d^). A potency parameter, ft, is esti-
mated by maximum likelihood methods, assuming 0^ has a Poisson
distribution with mean
2-8
-------
Thus, o linear relationship is assumed between dose and relative risk.
The background cancer risk (i.e., risk assuming no exposure to the
chemical), Po, is calculated by a life table method from vital statis-
tics data for the cancer of interest.
The first method, to be known as Basic Method 1, for calculating a
risk related dose (RRD) corresponding to a given lifetime extra risk,
r, involves assuming that lifetime risk is a function of the total cumu-
lative dose received during life but is independent of the timing of
exposure. If that is the case, then the lifetime risk of cancer from
dose d is given by
P(d) - P0(1+/Jd) (2-1)
and the extra risk from dose d is defined as
(2-2)
The dose corresponding to extra risk w (RRD) is then
RRD - >(P«-t-1) . (2-3)
?
The assumption of risk being independent of timing of exposure is
obviously an oversimplification. Basic Method 2 does not make this
assumption but rather assumes only that Eq. 2-1 holds for each 5 year
age group with the dose used being the cumulative dose up to the 5-year
age interval in question. Since the estimated risk depends upon the
timing of exposure, risk is estimated for a specific exposure pattern
thought to be typical of occupational exposures: exposure for 2kO days
per year for i»5 years starting at age 20. Hence, the dose variable up
to age 20 will have the value zero; for the next nine five-year age
groups it will increase linearly, at which point it will level off. The
work day exposure that corresponds to a given extra risk, w, is our
2-9
-------
estimate of RRD.
The cancer endpoints used in the analyses depend on what is known about
the chemical in question. In every case possible, analyses are
performed using all malignant neoplasms. Any types of cancer Known to
be or suspected of being related to exposure (e.g., leukemia and benzene
exposure) are analyzed also.
Case-Control Studies. In a case-control study, cases of disease are
located and then suitable controls, often matched to the cases, are
found. Level of exposure to the substance under investigation is
subsequently determined for each case and control. Data from such a
study can be expressed as follows:
Average Dose No. of Coses No. of Controls
1l x-, V1
d2 x2 Y2
dg Xg Vg
To determine an appropriate analytic approach, consider the following
2x2 table:
Coses Controls
dT P11 P12
d2 P21 P22
where the PIJ'S represent the proportions of the population in the
various categories. In a prospective study persons are selected random-
ly from the tiose groups and then checked for disease status (the disease
status is the random variable). In that case, the relative risk of
those in dose group 2 compared to those in dose group 1 is given by
[P21/(P21 4 P22)3/[P11/
the approximation being valid whenever, as is usually the case, cases
2-10
-------
ore much rarer than controls. The right side of this expression is
called the odds ratio. This approximate relative risk can be estimated
from a retrospective study, also. In fact, xgyi/Cxiyj) i* on estimate
of
(tP21/(P21 * P11)][P12/3> • P21P12/(P11P22>-
This and related considerations lead to the conclusion that it is art
adequate approximation to analyze a retrospective study as if the data
had been collected prospectively (1_).
Moreover, if the population sampled is composed of persons all the same
age, then the p's refer to rates of occurrences given that that age is
attained — that is, the relative risk being estimated is the incidence
of disease in the exposed group at that age divided by the corresponding
incidence in the unexposed group. Thus, this relative risk has exactly
the same interpretation as the relative risk calculated from prospective
studies in the manner described earlier.
The basic assumption to be used is that the risk to dosed individuals
relative to unexposed individuals is the same for a given cumulative
dose independent of age. This same assumption is made in the recommen-
ded analysis of prospective studies. If age is confounded with dose
then an analysis of a retrospective study should be stratified on age.
Otherwise biases may occur even if the assumption of the same relative
risk for all ages holds Q)- Since we will generally hove access only
to the published data, we can use a stratified approach only if the
published analysis took such an approach. As this generally will not be
the case, we will present only an unstratified method here. However, a
stratified approach is preferred and should be used whenever possible.
2-11
-------
As stated above, assuming the data were collected prospectively is an
adequate approximation. Therefore, the log-likelihood to be used is the
prospective one,
L - S [xilog^di)) + yilog(1-P(di))].
where P(a) is the conditional probability of disease during the age
interval given survival to that age. With this notation the Adds ratio
is given by
In order to be consistent with the approach taken with prospective
studies we wish to have the odds ratio (the approximate relative risk)
given by
For that to be case, P(d) must be expressed as
P(d) - «(1 + 0d)/(1 + a + o«0d).
The potency parameter, 0, is estimated by maximum likelihood techniques
using this expression and the log-likelihood given above. The para-
meters so estimated are applied with Basic "tothods 1 and 2 to derive RRD
estimates, as described above for prospective studies.
Calculation and Selection of RRD Estimates
The methods discussed above in the context of prospective and case-
control studies estimate 90Jf upper confidence bounds and 904 lower
confidence bounds as well as the maximum likelihood estimate of p. As a
means of incorporating this statistical variability and the exposure
2-12
-------
uncertainty into the analysis, a set of three potency parameters is
estimated for each carcinogenic response in each study. A lower bound
for the potency, /?L, is estimated by the lower statistical confidence
limit obtained when using the upper bounds on dose, the dj u's.
Similarly the upper bound for potency, /Jy, is estimated by the upper
confidence limit derived using the lower bounds on dose, the d^^'s.
These potencies, /3|_, /3|j, and 0, the MLE estimate of potency usii.q ..he
best estimates of dose, are applied in Basic Methods 1 and 2 to derive
RRD(j, RRD|_. and a maximum likelihood estimate of RRD, respectively.
Thus, at this point, on interval (RRD(_ to RRDu) with a "midpoint" (the
MLE estimate of RRO) is defined for each carcinogenic response in each
study analyzed. These intervals are not statistical confidence inter-
vals; they combine statistical uncertainty with exposure uncertainty and
hence represent what might be called "reasonable limits" on the doses
corresponding to specific levels of extra risk. The triples of RRO
estimates have been calculated for extra risks of 10~6 and 0.25.
RRD't. derived from the epidemiological literature are converted to a
mg/kg/day equivalent, assuming 2
-------
Converting to 25°C,
1 pptn - f( 1000/22.4) x W x 10001 x 273 - 0.0*1 x W mg/m3.
IiQ6j 593
This volue is converted to mg/kg/doy by ossuming 10 m3 of air breathed
during an eight-hour work shift and a 70 kg body weight.
In many cases, more than one RRD interval for a chemical is available
from the epidemiologic literature either because of more than one study
or more than one carcinogenic response analyzed. Rather than combining
results for different responses or from different studies, a single
triple of RRD estimates is selected to represent the potency of a given
chemical. The triple that is selected is one that corresponds best with
the consensus of opinion about the carcinogenic effect of the chemical
determined from all the literature that was reviewed. However, the
results froiii a study or particular response in a study are not used if
the dose-response model provided a poor fit to the data or if the study
is deemed to be markedly inferior to other studies ; have analyzed a
particular resprnse. In the case of vinyl chloride, for example, a
liver cancer resprnse is chosen since angiosarcoma of the liver is
considered to be undeniably linked to vinyl chloride exposure whereas
respiratory cancer, another endpoint analyzed, is not so clearly linked.
Another example is provided by isoniazid. Overall, the literature on
isoniazid does not conclusively demonstrate its carcinogenicity in
humans let alone indicate any particular site of action. Hence, the
response selected is all malignant neoplasms, and, moreover, the triple
chosen is one that has an infinite upper bound (consistent with no
carcinogenic effect), since one meeting that criterion is available from
our quantitative estimation.
2-14
-------
RESULTS
Following below ore the descriptions of the epidemiologicol data
relating human exposure and carcinogenic response for the chemicals
listed in Table 1-3. Each analysis o* a particular chemical includes a
brief review of background information and a description of the human
carcinogenicity data as a whole. The latter of these is important in
light of the method, described above, of selecting the specific response
to represent the epidemiological risk estimation; i.e., we select the
response most representative of the entirety of the human cancer data
base for each substance. Derivation* of the RRD estimates and
discussions of the uncertainties relevant to those derivations conclude
each analysis.
The epidemiological assessments are intended to be independent, self-
contained examinations of the individual chemicals. Consequently, RRO
estimates are presented in the same units used by the authors of the
published reports, except that exposures reported in terms of total dose
(e.g. in total milligrams consumed) have been converted to cumulative
doses (e.g. milligram-years). Following the individual chemical write-
ups is a summary of the results for all the chemicals, specifying the
responses chosen and the conversions to the standard units of dose,
mg/kg/day.
2-15
-------
Aflotoxin
The aflatoxins are naturally occurring contaminants of food stuffs
produced by species of the fungus Aspergillus. According to the
International Agency for Research on Cancer (IARC) (2) aflatoxin B-| is a
potent mutagen and forms the same ONA adducts in human cells as in
rodent cells. Afiatoxin ingestion has been linked to primary liver
cancer so, in accordance with the analysis guidelines described above,
risk estimates will be based on liver tumors.
Primary liver cancer appears to be most prevalent in tropical areas,
those areas in whicn aflatoxin contamination of food is a problem but
also the areas where hepatitis B virus is common (3, 4). However, in
13 of 15 cases of liver cancer studied in CzeckoslovaKic (5) some
trace of aflatoxin B-| was found in or around the tumor. Aflatoxin
intake in the southeastern United States from 1910 to 1960 was estimated
to be about 100 times greater than in the north and west of the United
States (6). Stoloff concluded that the excess primary liver cancer rate
in the southeast (10)1), given the much greater aflatoxin contamination,
was not as large as expected from studies in Africa and Asia and that
the excess may not have been attributable to aflatoxin. Perhaps factors
present in Africa and Asia (hepatitis B virus, for example) may interact
with aflatoxin to enhance the rate of primary liver cancer in those
areca. The study by Wang et al. (f») suggests that this may be the
cas*.
Several studies (7-9) have reported conditions most conducive to
development of aflatoxin contamination of food and the types of foods
commonly found to be contaminated. Traditional harvesting and storage
techniques appear to be conducive to Aspergillus growth and a variety of
vegetable foodstuffs contain detectable levels of aflatoxin. Groundnuts
(peanuts) appear to be particularly susceptible; all peanut samples were
2-16
-------
contaminated in Pakistan (9) and samples from Mozambique averaged 1936
M9 of aflatoxin per kilogram of peanuts (7).
Studies in Swaziland (10) and China (^) report good correlation between
the pattern of liver cancer and the percentage of food samples found to
be contaminated with aflatoxin. In an occupational setting in the
Netherlands, 71 workers exposed to aflatoxin at a plant extracting oil
from peanuts were estimated to have respiratory exposures ranging from
0.0
-------
a,-e difficult to determine in cross-sectional investigations.
Nevertheless, all these reports document a positive co' -'elation between
estimated aflatoxin intake and primary liver cancer incidence. Table
2-2 presents intake and incidence data for those studies that used
ready-to-eat food samples to determine aflotoxin dose. Linsell and
Peers (13) consider the reported aflatoxin intake values to represent
minimal ingestions. The yearly and life-time cancer rates are the crude
rates.
These data are not in any of the forms assumed in our standard
approaches to quantify risk; a special analytic method is required.
Suppose that a yearly crude rate of liver cancer, cj, can be expressed
as a linear function of aflatoxin intake (ng/kg/day), dj/, i.e.,
Then, for any of the studies described in Table 2-2 for which the popu-
lation size is available, one would expect to observe c^-N^-y^ liver
cancers, where N^ is the population size and y^ is the number of years
of observation. If we assume that the observed liver cancers are
distributed as a Poisson distribution with the indicated expected value.
likelihood methods applied to the data in Table 2-2 allow estimation of
a and 0 and calculation of related confidence bounds (Table 2-3). Table
2-3 alio displays the bounds on the dose values that are used to inves-
tigate the sensitivity of the analysis to selection of the dose levels.
The upper bounds are arbitrarily set at 3 times the best estimate; the
lower bounds are 3 times smaller than the best estimate. The bound is
wider above the best estimates because, as suggested by Linsell and
Peers (12), the values given in Table 2-2 may reflect only minimal
aflatoxin intake levels.
2-18
-------
Given an annual crude rate of liver cancer, c^ - a + /Jdj, a lifetime
probability of liver cancer can be calculated as
P(di)
for some factor F. In this case, extra risk attributable to aflatoxin
is
1 - cxF
The choice of the factor F is open to some question. Carlborg (1*Q uses
a value of F equal to the life expectancy of the study population. This
may be too small, however, if we consider that in any given year a large
fraction of the population is under the age of 15, especially in the
third-world populations included in these calculations. This segment of
the population is essentially at no risk from liver cancer so that the
NJ values used to estimate ex and 0 are too large. To compensate, the
following procedure is suggested.
Let Oj be the proportion of the population in age group j , which con-
sists of all individuals with ages in an interval (j) years in length.
Suppose also that individuals in age group j experience an all-cause
yearly mortality rate of b j . The values a j , bj , and jj are available
from international demographic sources. Let Jj be the liver cancer rate
in group j without exposure to aflatoxin. This is not, in general,
known but we will assume that
where lus.j i» tne liver cancer rate in the U.S. in group j. This
assumption is made despite the fact that liver cancer in the U.S.
appears to have a substantially different age pattern than in the
2-19
-------
populations included in this study. The different pattern may very well
be attributable to the oflotoxin intake and hence not pertinent to
evaluation of the assumption. In any case, the background liver cancer
risk, a, estimated above, can be expressed as
« • Ejajaj • Ejaj1us,j'f.
which will serve to define f.
By lifetable iiethods, the lifetime probability of liver cancer in the
presence of an aflatoxin dose d is
P
-------
RRD estimates based on the derivation above are shown in Table 2-5.
These apply to a scenario corresponding to life long exposure to ofla-
toxin. To crudely approximate the RRD& associated with our standard
exposure scenario (45 years exposure starting at age 20) we have multi-
plied the estimates in Table 2-5 by (75/<»5), where 75 is the assumed
life span over which exposure occurs. These results are displayed in
Table 2-6.
2-21
-------
Table 2-1
DISTRIBUTION OF FILIPPINO CASES AND CONTROLS
WITH RESPECT TO DAILY AFLATOXIN INTAKE0
Daily Primary
Aflatoxin Liver Cancer Matched
Intoke(gg) Coses Controls
0-3 20 74
(1.5)"
4-6 15 it
(5)
7+ 55 12
(8.5)
°From Bulatao-Jayme «t al. (12).
''Assumed average for the group.
2-22
-------
Table 2-2
DATA FROM CROSS-SECTIONAL STUDIES OF AFLATOXIN
INTAKE AND PRIMARY LIVER CANCER INCIDENCE
Region
Mean Dose
of Aflatoxin
(ng/kg/day)
Popu-
lation
Size
Year* of
Obser-
vation
Yearly
Rick
(x1Q5)
Life
Expectancy0
(vr)
Lifetime
Risk
(x1C>5 )
Thailand15
Songkhla
Ratburi
Kenya0
High
Middle
Low
Swaziland*1
5.0
45.0
3.5
5.9
10.0
97867
99537
46279
187514
174525
7
7
2.0
6.0
1.2
2.5
4.0
54
54
47
47
47
108.0
324.0
56.4
117.5
188.0
Highveld 5.1
Middleveld 8.9
Lowveld 43.1
Lebombo 15.4
Mozambique**
Inhambane 222.4
100719
151430
91471
18747
5
5
5
5
2.2
3.8
9.2
4.3
13.0
41
41
41
41
41
90.2
155.8
377.2
176.3
533.0
°From Corlborg (1i»).
bFrom Shank et al. (1j>).
cFrom Peers and Linsell (V7).
dFrom Peers at al. (.18).
•Cited in Linsell and Peers (13).
2-23
-------
Table 2-3
DOSE AND RESPONSE DATA ESTIMATION OF a AND
Dose
Lower
(ng/kg/day)
Best
Bounds Estimotes
1.2
1.67
1 .7
2.0
3.0
3.3
5.1
15.0
U.%
Porometer Estimates:
3.5
5.0
5.1
5.9
8.9
10.0
15.*
1.5.0
M.I
Upper
Bounds
17.5
25.0
25.5
29.5
<»
-------
Table 2-4
POPULATION STATISTICS USED FOR CALCULATION OF F
Age Group
<1
1-4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
30-84
85+
Proportion of
Population"
.023
.138
.152
.118
.095
.092
.079
.067
.052
.043
.037
.031
.025
.018
.012
.008
.004
.002
.002
All-Cause
Death Rate0
. 07683
.01092
.00331
.00192
.00211
.00296
.00337
.0041
-------
Table 2-5
RRD ESTIMATES FOR LIFETIME EXPOSURE TO AFLATOXIN (ng/kg/day)
Level of Extra Risk
1Q-60.25
Method RRDi MLE RRDU RRDi MLE RRDU
Using F-47 3.13E-3 1.19E-2 7.78E-2 7.82E+2 2.98E+3 1.97E+4
(life
expectancy)
Using 6.81E-4 2.59E-3 1.71E-2 1.70E+2 6.V7E+2 4.28E+3
F-215.1*
(life table method)
Table 2-6
RRD ESTIMATES FOR AFLATOXIN (ng/kg/day)°
Level of Extra Risk
10-60.25
Method RRDi MLE RRDU RRDi MLE RRDU
Using F-V7 5.22E-3 1.98E-2 1.30E-1 1.30E+3 4.97E+3 3.28E+1*
(life
expectancy)
Using 1.14E-3 4.32E-3 2.85E-2 2.83E+2 1.08E-I-3 7.13E+3
F-215.*
(life table method)
°Converted so as to apply to the standard exposure scenario: 45
years of exposure starting at age 20.
2-26
-------
Arsenic
The metal arsenic and its compounds have had nany uses. Elemental
arsenic is used as an alloying additive and in electronic devices.
Various compounds containing arsenic have been used or medicines, as
defoliants, as intermediates in several manufacturing processes, as
pesticides, and as pigments (20). Human exposures have been associated
with ingestions of arsenic-containing medicines and drinking water
contaminated with arsenic, with insecticide operations (especially in
sheep dipping), and with certain smelting operations (20, 21).
The association between arsenic exposure and cancer in humans has been
known for some time. A preponderance of skin cancers was noted by
Neubauer (21) to be related to use of arsenic-containing medicines.
More recently, case reports of hepatic angiosarcoma attributed to use of
Fowler's solution (potassium arsenite) have appeared (22, 25). Skin
cancers hava been reported to be in excess in areas in which the drink-
ing water contains high levels of arsenic. Neubauer (21 ) reports
anecdotal information on increased frequency of skin cancer in a German
town whose drinking water contained 1.22 mg percent arsenic due to gold
smelting operations that release* arsenical fumes. Many reports (21)
document typical arsenic-associated skin cancers in an area of Argentina
whose wells had 0.28 - O.WJ mg percent arsenic. Tseng et al. (2Jt) and
Tseng (25) present similar findings for an area in Taiwan that had
arsenic in well water averaging 0.5 mg/!.
The mechanism by which arsenic causes cancers in humans is not entirely
clear. Arsenic does not appear to be mutagenic in the systems so far
tested (2). Reports of tests in human cells, however, document
arsenic's effect on chromosomal damage. Paton and Allison (26) describe
the results of in vitro tests in which chromosome damage in leucocytes
and fibroblasts was enhanced in the presence of arsenic salts.
2-27
-------
Chromatid breaks were frequent. In a case-control study, Petres et al.
(27) found more chromosomal aoerrations in individuals who hod been
exposed to arsenic (via psoriasis treatment or pesticides) than in
unexposed controls. These authors also found arsenic to impair nuclear
division, to reduce incorporation of thymidine in DNA (although the
effect on RNA uridine incorporation was not as severely affected), and
to decrease hyperdiploidism in PHA-stimulated lymphocytes.
The observations reported above have lead to the suspicion that arsenic
may be inhibiting DNA repair mechanisms. This may be accomplished by
interfering with the enzymes which are responsible for that operation,
especially those with sulfhydryl groups (27. 28). Another possibility
is that arsenic reduces, or competitively inhibits, the uptake of
phosphorus into the DNA chain. The arsenic would then cause weak spots
in the chromosomes that could lead to more frequent breakage (26, 27).
An argument against this mechanism is the fact that, while tne trivalent
forms of arsenic appear to have the highest carcinogenic potency, the
pentavalent form would be more likely to substitute for phosphorus in
DNA (29). No definitive conclusion about the mechanism of action of
arsenic in carcinogenesis has been reached.
Numerous investigations of occupationally exposed workers have impli-
cated arsenic as a human carcinogen. In an early proportional mortality
study, Hill and Faning (30) found an excess proportion of cancer deaths
among employees of a factory handling arsenicals. There was a sugges-
tion that lung nnd skin cancers were over represented. A case-control
study in Japan (31 ) strongly suggested that employment as a copper
smelter, which entails arsenic exposure, was closely linked to lung
cancer. A retrospective cohort study of those copper smelters (52)
confirmed the excess lung cancer risk and demonstrated a clear dose-
response relationship between lung cancer and a qualitative measure of
exposure to arsenic.
2-28
-------
Ott et al . (33) inv3»tigated the proportional mortality experience of
industrial workers manufacturing insecticides containing arsenic,
primarily lead and calcium arsenate. Nearly 2000 white males who died
between 1940 and 1972 were included in the study. Relative risks for
highly exposed workers were as high as 7.0. A retrospective cohort
study by the same authors confirmed a strong relationship between
arsenic-containing insecticide exposure and respiratory cancer
mortality.
Two cohorts of workers have been used to quontitate the relationship
between arsenic exposure and cancer: workers employed at the ASARCO
smelting plant in Tacoma, Washington, and copper smelters in Montana
working for the Anaconda Company. These cohorts form the basis of the
analysis reported below.
The ASARCO copper smelter in Tacoma, Washington is one of the nation's
largest, producing, at times, 10* of the refined copper in the U. S.
The copper it smelt* has been traditionally high in arsenic (3_4).
Several studies of workers and retirees from the plant have been
conducted. Pinto and Bennett (35) studied the proportional mortality of
229 decedents who had worked in the refinery. Excesses in the
proportions of lung cancers and of breast and genito-urinary cancers
were found. Pinto «t al . (36) studied 527 retirees who had had some
degree of arsenic expos-ire. A dose-response trend was evident for
respiratory cancer as related to an estimated cumulative exposure based
on urinary arsenic levels. Thes* authors also reported that atmospheric
concentrations could be related to urinary concentrations: air levels
(in /jg/m3) were 0.304 times the urine levels (in ng/9).
Enterline and Marsh ( 34, 37) defined a cohort of ASARCO workers who were
employed a year or more during the period 1940 through 1964 and studied
2-29
-------
its mortality experience through 1976. A total of 2802 men were
included in the cohort. Exposure information was based on the urinary
arsenic concentrations measured in 1948-1952 and 1973-1975. The average
concentrations for a department were assumed to apply to each worker
stationed in that department. Levels measured in 1948-1952 were assumed
to reflect earlier exposures and linear interpolation was used to
estimate the values for the years 1953 through 1972. Cumulative dose
was expressed in terms of /ig-years/liter of arsenic.
Table 2-7 displays the dose and response data for this cohort. The
cumulative dose values have been converted to cumulative atmospheric
exposures using the factor 0.304 determined by Pinto 0t al. (3J>). As
indicated in Table 2-7, reasonable bounds on dose (used to investigate
the sensitivity of the analysis to uncertainty in the estimates) are
derived from the best estimates of exposure and uncertainty factors
a - 2.35 and i - 2.75. The major uncertainties influencing those bounds
include the following:
1. Although the smelter has been in operation since 1913. the
first measurements of urinary arsenic occurred ir 1948.
Exposures prior to 1948 are estimated to be equal to those
measured in the period 1948-1952, even though the authors
believe this to underestimate those exposures. Consequently,
12 ha» been assigned a valui of 0.4. whereas 02 • 0.
2. Even after 1948, few measurements of urinary ursenic were
available, none between 1952 and 1972. Since linear interpo-
lation was used to estimate the exposures between 1952 and
1972, uncertainty is contributed to both upper and lower
bounds. Both 013 and 73 have been set equal to 0.4.
3. The levels of urinary arsenic assumed to be appropriate to the
different departments were determined from samples of workers
in those departments. Thus, the applicability of the values
2-30
-------
used to estimate cumulative exposure is uncertain. Both ag and
7g are given a value of 0.3.
4. Conversion of urinary to atmospheric concentrations of arsenic
is accomplished by one fixed conversion factor. This factor
undoubtedly varies from individual to inJividual and, perhaps,
with concentration level. A factor of 0.6 is assumed for 07
and 77.
5. Finally, expected values were based on U.S. national rates. A
nominal value of 0.05 is assumed for 09 and ~IQ to account for
uncertainty in the expected numbers.
The second cohort of workers that can supply quantitative risk estimates
is that employed by the Anaconda Company in Montana. The first report
on this cohort was that by Lee and Traumani (3_8). A total of 8,047
white males who worked at least one year before 1957 were observed from
1938 through 1963. In this report, only relative exposures to arsenic
("heavy", "medium", and "low") were discussed. An excess of respiratory
cancer was discovered that was related to relative exposure to arsenic
trioxide and also to SOj.
Lubin et al. (39), reporting on that portion of the Lee and Fraumani
(38) cohort alive in 1964, state that the mean ar**r:ic levels in the
•heavy", "medium", and "low" exposure areas were 11.3, 0.58, and 0.29
mg/mg3, respectively. Respirators were used, intermittently, especially
in the heavy exposure areas; the authors estimate this effect by
reducing the pertinent concentration in the heavy exposure jobs by a
factor of ten.
The follow-up of the 5,403 survivors through 1977 yielded 64,315 person-
years of observation. Multivariof* analyses revealed an association
between arsenic concentrations and respiratory cancer. SO2 did not
appear to be related to respiratory cancer although its interaction with
2-31
-------
arsenic could not be ruled out.
A pilot study (frO) and a larger sampling (VI) of the Lee and Fraumani
(38) cohort investigated the reported association between arsenic and
respiratory cancer, especially as it relates to cigarette smoking.
Although the cohort members were more likely to be smokers than men in
the general U. S. population, smoking could not be considered to
confound the relationship between arsenic and lung cancer. Dose-
response trends were observed among smokers as well as nonsmokers; no
interaction between smoking and arsenic exposure was apparent.
Finally, Lee-Feldstein (42) published an update of the original Lee and
Fraumani (38) study. Follow-up was extended through most of 1977.
Classification of a cohort member, as in the original study, was based
on the maximum exposure category (heavy, medium, light) experienced by
the worker. This tends to overestimate the actual exposures encountered
and complicates the calculation of cumulative dose. Table 2-8 displays
the distribution of respiratory cancer deaths cross-classified according
to this classification and length of employment. Using the estimates of
average airborne arsenic concentrations for the qualitative exposure
categories given by Lubin et ol. (59) and average lengths of exposure
for the groups defined in Table 2-8 yields cumulative dose and response
data as displayed in Table 2-9. It has been assumed that the category
into which the men were placed reflects their average exposure.
Cumulative exposures for those in the medium and heavy groups are apt to
be overestimated in this case. However, estimated exposures for those
in the low category may be representative because a person in this
category had all of his work experience in low exposure areas.
The uncertainties pertinent to the Lee-Feldstein (»2) study are
summarized as follows:
2-32
-------
1. Length of exposure is grouped into three categories. The
categories are presented without average values and so we have
used midpoints of the intervals defining the groups. This
contributes some uncertainty; o.-\ and T\ are both given values
of 0.2.
2. The arsenic concentrations that have been used to quantify the
exposure groups were measured between 1943 and 1958 (39).
There is some potential for underestimation of exposure, since
the plant has been in operation since before 1925 and some
members of the Lee-Feldstein cohort worked during this early
period. Consequently, «2 • 0 and 12 " 0.6.
3. The completeness of the concentration measurements is not
documented. Both 03 and 73 are set equal to 0.3.
-------
The resulting uncertainty factors are dose-group dependent, as follows:
•heavy" exposure: a - 2.5, -7 - 2.5
"medium" exposure: a - 2.i», 7-2.3
"low* exposure: a • 1.7, 7 • 2.3
Table 2-9 displays the bounds on cumulative dose that result when these
factors are applied.
Welch et al. (41i) consider a subset of the original Lee and Fraumani
cohort, a total of 1800 white males. Follow-up was continued to 1978.
Welch and her colleagues calculated and presented cumulative exposure
estimates based on measurements of arsenic concentrations prevalent
between 1943 and 1965 (Table 2-10). Data presented in Welch et al.
(41) indicates an average cumulative exposure for their cohort of about
26,500 pg-yrs/m3. We assume that the midpoints of the intervals
displayed in Table 2-10 represent average exposures in the first three
groups (i.e. 250, 1250, and 7000 jtg-yrs/m3, respectively). Employing
the fact that the entire cohort averaged 26,500 ^g-Vs/m3 and weighting
the exposure groups on the basis of their expected numbers of respira-
tory cancers (in lieu of actual numbers of men), we estimate the average
for the highest cumulative dose group to be 85,000 pg-yrs/m3. It should
be noted that Welch et ol. hove not considered the effect of respirator
use on reduction of exposure.
The following features of the Welch ft al. study affect uncertainty
estimation:
1 The same lack of measurements of concentrations before 1943 as
was seen in the Lee-Feldstein (42) study contributes a value of
0.6 to 72- Again 02" 0.
2. Between 1943 and 1965, 818 samples were taken in 18 of the 35
smelter departments. Estimates for departments with no
2-34
-------
meosurements were made by analogy to departments with Known
concentration. Both aj and 73 have been set equal to 0.3.
3. Average values for the cumulative exposure groups had to be
estimated. The knowledge of the average for the entire cohort
aided in estimation of the average for the highest dose group,
so a/t and 7^ both equal 0.2.
<». Welch et al. do not consider respirator use when accumulating
dose. As they soy. "Interview* with men who had worked in the
plant berore 1964 ... indicated that respirator usage was
sporadic at best during that time. It does not seem that
respirators would have made on important difference in the
concentrations to which the vast majority of men were exposed."
Nevertheless, the numbers presented by these authors are
probably overestimates of exposure; 05 • 0.3 and 75 - 0.
5. Although substantial discussion of smoking status and its
effect on respiratory cancer is provided by Welch et al., this
analysis did not consider that variable. Both 03 and 73 are
assumed to be 0.1, as they were for the Lee-Feldstein study.
The uncertainty factors, at • 1.9 and 7 - 2.2, determine the bounds on
cumulative dose, as seen in Table 2-10.
The respiratory cancer potency estimates for the three studies are giv*n
in Table 2-11. Unfortunately, the relative risk model does not. fit any
of these data sets. Crump and Ng (f*5) analyzed occupational arsenic
exposure data and found that an absolute risk model fit the data much
better than a relative risk model but that the risk estimates did not
differ greatly from one model to the other. We have continued to use
the relative risk model to compute PRO estimates (Table 2-12), knowing
that some error is thereby introduced.
2-35
-------
Table 2-7
DOSE AND RESPONSE DATA FOR THE COHORT OF WORKERS
EXPOSED TO ARSENIC AT THE TACOMA, WASHINGTON SMELTER
Dose (jig-yrs/m'}
Lower
Bound
39.1
112
281
588
1741
Best
Estimate
91.8
263
6S1
1381
4091
Upper
Bound
252
723
1818
5000
11250
Respiratory Cancer Deaths
Observed
8
18
21
26
31
Expected
4.0
11.4
10.3
14.1
12.7
°From data presented in Enterline and Marsh (34)
Table 2-8
OBSERVED AND EXPECTED DEATHS FROM RESPIRATORY
CANCER. BY MAXIMUM EXPOSURE TO ARSENIC AND
LENGTH OF EMPLOYMENT, ANACONDA EMPLOYEES0
Maximum Exposure to Arsenic (12 or more months)
Years of Heavy Medium Light
Exposure Observed Expected Observed Expected Observed Expected
25+ 13 2.7 49 7.2 51 16.2
10 - 24 12 1.9 23 6.7 25 13.0
1 - } 8 1.8 21 6.8 60 29.6
°From Lee-Feldstein (42).
2-36
-------
Toble 2-9
DOSE AND RESPONSE DATA FOR ANACONDA EMPLOYEES,
FROM THE LEE-FELDSTEIN (42) CATEGORIZATION
Maximum
Years of Exposure
Exposure Category
25+° Heavyb
Medium
Light
10 - 24d Heavy
Medium
Light
1 - 9» Heavy
Medium
Light
Cumulative Observed
Dose Respiratory
(MQ-yrs/m^) Cancer Deaths
39550
(15820. 98875)°
20300
(8458, 46690}
10150
(3759, 23345)
19436
(7774, 48590)
9976
(4157, 22944)
4988
(1847, 11472)
4656
(1862, 11640)
2390
(996, 5497)
1195
(443. 2748)
13
49
51
12
23
25
8
21
60
Expected
Respiratory
Cancer Deaths
2.7
7.2
16.2
1.9
6.7
13.0
1.8
6.8
29.6
aThe overage length of exposure for this group is assumed to be
years, the average of 25 and an assumed maximum of 45 years.
bThe airborne concentration for the "heavy" category has been reduced to
1.13 mg/m3 to account for possible respirator use.
cln parentheses are the lower bounds and upper bounds on dose for each
dose group.
dThe average length of exposure for this group, 17.2, is based on the
1138 men with 15-24 years employment (20 years average) and 678 men
with 10-14 years employment (12.5 years average).
•*The averogo length of exposure for this group, 4.12, is based on the
1032 men with 5-9 years employment (7.5 years average) and 3248 men
with 1-4 years employment (3 years average).
2-37
-------
Table 2-10
DOSE AND RESPONSE DATA FOR THE WELCH £T AL.
(in) COHORT OF ANACONDA WORKERS
Cumulative Arsenic
Exposure (>*g-Yrs7m3 )
< 500
(132, 250, 550)a
500 - 2000
(658. 1250. 2750)
2000 - 12000
(3684, 7000, 15400)
12000+
(44737, 85000, 187000)
Respiratory Cfincer Deaths
Observed Expected
4 5.8
9 5.7
27 6.8
40 7.3
aln parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for cumulative exposure in each dose group.
2-38
-------
Table 2-11
RESPIRATORY CANCER POTENCY PARAMETER ESTIMATES FOR ARSENIC
Dose Potencies ((^g-vs/m^)"1)
Study Measure Lower Limit0 MLE Upper Limit0
Enterline Upper 1. 1<»E-«»" 1.58E-* 2.07E-*
and Marsh Bounds
(34)
(chi-squared Best 3.30E-* 4.59E-4* 6.04E-4
(«O > 10.8) Estimates
Lower 7.757E-4 1.08E-3 1.42E-3"
Bounds
Lee-Feldstein Upper 8.80E-5* 1.00E-* 1.13E-4
(»2) Bounds
(chi-squared
(8) > 37.0) Best 2.08E-* 2.37E-*' 2.67E-*
Estimates
Lower 5.24E-* 5.97E-4 6.75E-<»"
Bounds
Welch et al. Upper 2.25E-5" 2.87E-5 3.57E-5
(chi-squared
(3) > 32.9) Best *.95E-5 6.31E-5" 7.85E-5
Estimates
Lower 9.<»OE-5 1.20E-<» 1.49E-4*
Bounds
°90< confidence limits ore shown.
"An asterisk marks the parameters used to derive RRO estimates.
2-39
-------
Table 2-12
RRD ESTIMATES0 FOR ARSENIC (/3/m3)
Level of Extra Risk
Study
Estimation 10~6
Method SRDi MLE
0.25
RRDU
RRDt
MLE
RRD,
Enterline
and Marsh
(5ft)
Lee-Feldstein 1
2
Welch et al. 1
(il)
2
1.92E-4 5.92E-4 2.38E-3 4.79E+1 1.48E+2 5.96E + 2
2.21E-4 6.84E-4 2.75E-3 6.81E+1 2.11E+2 8.47E+2
4.03E-4 1.15E-3 3.09E-3 1.01E+2 2.87E+2 7.72t
4.65E-4 1.33E-3 3.56E-3 1.43E+2 4.09E+2 1.10E+2
1.82E-3 A.31E-3 1.21E-2
2.10E-3 <».97E-3 1.«*OE-2
4.56E+2 1.08E+3 3.03E+3
1.53E+3 4.30E+3
°Based on the risk of respiratory cancer.
2-40
-------
Asbestos
With the exception of cigarette smoke and radiation, asbestos has
probably been more unequivocally linked with human carcinogenesis than
any other risk factor. Exposure to asbestos in the workplace has been
clearly shown in a number of studies to place workers at increased risk
of lung cancer, particularly among smokers, of plural and peritoneal
mesotrtelioma, and possibly of other cancers of the pulmonary and
gastrointestinal tracts. Several of these studies have data suitable
for quantitative risk assessment, and a number of risk assessments for
asbestos have been carried out by various individuals and governmental
agencies (frfr-»9).
Despite the fact thct the carcinogenicity of asbestos in human popula-
tions has been thoroughly documented and studied, there are some unique
difficulties associated with estimating risk to humans from exposure to
asbestos. First of all, the carcinogenic risk from asbestos depends
heavily upon fiber type and dimension. Risk of cancer from exposure to
asbestos in mines and mills is less than that from most other occupa-
tional exposures, including exposures in manufacturing plants, textile
mills, and in insulation applications. The potency of chrysotile
asbestos for causing lung cancer in Canadian miners, for example, is
1/100 of the corresponding risk in workers exposed to amosite asbestos
in a manufacturing plant during World War II (V*v). Risk of mesothelioma
appears to be much less after exposure to chrysotile than to a compara-
ble level of amosite or crocidolite (50).
Thus, a single carcinogenic potency for asbestos may not be appropriate
for all applicotons; different fiber clouds con entail different risks.
For purposes of setting regulatory policy, reQulatory agencies hove
calculated an average potency over a range of exposure conditions. We
shall take this approach here also.
2-41
-------
Asbestos fiber clouds used in animal studies of carcinogenicity are
generally created by hammer milling or ball milling raw asbestos. These
processes create fibers which may not be representative of those to
which humans have been exposed. This means that discrepancies in carci-
nogenic potencies obtained for asbestos from animal and human data may
be due to differences in the types and dimensions of fibers to which
animals and humans are exposed rather than to fundamental differences
between animals and humans in their in susceptibility to the carcino-
genic effects of asbestos.
Because of these limitations of the asbestos data, no detailed review of
the epidemiological literature will be carried out here and a detailed
risk assessment will be not be conducted for asbestos. Instead, risk
estimates will be adapted from risk assessments already available in the
literature. Several U.S. governmental aguncies have conducted risk
assessments for asbestos; these include the Consumer Product Safety
Commission (44), the Occupational Safety and Health Administration (45),
and the Environmental Protection Agency (46). The risk assessment
conducted by the CPSC is specifically relied upon here, although the
differences among these risk assessments are minor.
Asbestos has been shown unequivocally to cause lung cancer and mesothe-
lioma (both plurcl and peritoneal). There is also evidence that
asbestos can cause other types of cancer, including various cancers of
the gastrointestinal tract, although evidence for such a cause and
effect relationship are more limited." Following our guidelines for
selection of ccncer types for quantitative estimates, separate estimates
The CPSC Panel, for example, could not agree on the interpretation of
the evidence on these other cancers. "Some members [of the Panel]
thought it possible that these excesses could conceivably be due to a
combination of misdiagnosis of peritoneal mesothelioma and the use of
inappropriate expected numbers. Others members thought that after
allowance is made for these possible errors that some of the observed
excess must be attributable to asbestos (43).•
2-42
-------
\,
of riSK will be made for lung cancer and mesothelioma. Estimates will
also be made for all cancer, including lung cancer, mesothelioma and
other types of cancer that may be related to asbestos.
Due to the synergistic relationship that apparently exists between
asbestos and cigarette smoke in causing lung cancer (5J.), the risk of
lung cancer from asbestos is appreciably greater in smokers that in
nonsmokers. To account for this difference in susceptibilty, separate
estimates of risk for lung cancer and total cancer will be made for
smokers and nonsmokers.
Table 2-13 shows potency parameters for lung cancer (K|_'s) estimated by
C°SC (M») from 11 occupational studies, and corresponding potency para-
meters for nesotheliomo (K^'s) obtained from
-------
than the geometric average of I.0x10~?. CPSC used a range of values of
KM • 0.3x10~8 to 3.0x10-8 to estimate risk of mesothelioma. OSHA
justified using the middle of this range, 1.0x10"**, because the ratios
KM/KL for the four studies for which K^'s are available are closely
distributed about 1.0x10~6 and an value of 1.0x10~2 was used for KL on
the basis of eight studies. However, using data from three studies in
which exposure was primarily to chrysotile, Crump (55) estimated K^'s
that were much lower than 1.0x10~8; consequently the approach used by
CPSC and OSHA may overestimate the risk of exposure to chrysotile.
Nevertheless, we will follow the CPSC and OSHA approaches in the present
analyses and use KL - 1.0x10~2 and KM - 1.0x10~8 (or, equivalently, use
the middle of the range of risks estimated by CPSC).
CPSC (ftft) estimated risks for lung cancer and mosothelioma in nonsmoking
and smoking males and females separately for continuous exposure to 0.01
f/cc beginning at ages 0, 10, 20, 30 and 50, and lasting for 1,'5, 10,
and 20 years. These results are contained in CPSC's Table J-8A. In the
current analysis we are interested is estimating risk from occupational
exposure beginning at age 20 and lasting for
-------
due to differences in longevity between moles and female* and, more
importantly, between smokers and nonsmokers.
The continuous exposure assumed by CPSC is equivalent to an occupational
exposure (3 hours per day, 240 days per year) of
(0.01 f/cm3)(24 h/S h)(365 d/240 d) - 0.046 f/cm3.
Assuming a daily (8 hour) occupational breathing rate of 10 m3 per day,
the total exposure is
(0.046 f/cm3)(1.0x106 cm3/m3)(10 m3/d)(240 d/yr)(45 yr)
- S.OxlO9 fibers.
The corresponding RRD for a risk of 1x10~*> is therefore
(5.0x109/5.92x10-*)(1.Ox10~6) - 8.39x106 fibers.
The same approach will be used to estimate RRDs corresponding to a risk
of 0.25. Although this method does not fully account for attenuation of
risks at such a high risk level due to asbestos-related deaths, such an
adjustment would be very minor compared to the other sources of
uncertainty in these estimates.
This approach is also used to estimate lung cancer RRDs, the only
difference being that risks are estimated separately for smokers and
nonsmokers. Similar methods are also used to estimate RRDs for total
cancer. Following OSHA (45), cancer risks other than lung cancer and
mesothelioma are assumed to be 104 of the risk of lung cancer in
smokers.
Table 2-14 contains the resulting RRD estimates for asbestos in units of
total fibers. The number of fibers counted depends upon the equipment
2-45
-------
used and the method of counting. The estimates in Table 2-1* refer to
fibers longer than 5 microns as measured by an optical microscope.
It is necessary to convert RRD estimates based on numbers of fibers to
those based on a weight measure of asbestos exposure. The Ontario Royal
Commission (»7) estimated that there ore 30 fibers longer than 5
microns (measured by optical micorscope) per nanogram of asbestos
(measured with a transmission electron microscope). Experiments they
performed suggested conversion factors ranging from 9.1 to 770 fibers
per nanogram (conversions from samples in buildings ranged from 20 to
102 fibers per nanogram), so variability is considerable. Table 2-15
displays RRDs converted to units of milligrams per day.
Also shown in Table 2-15 are the bounds on the RRD estimates. Uncei—
tainty in this asbestos analysis has been handled differently from that
for other chemicals. We have arbitrarily chosen a factor of 6 to
specify the bounds, hoping to account for such uncertainties as the
conversion between number of fibers and weight of asbestos, the
different types of and setting for asbestos exposure, the range of
potency estimates obtained by CPSC, and possible differences in action
of the various forms of asbestos.
2-1*6
-------
Table 2-13
VALUES OF KL AND KM OBTAINED IN THE ANALYSIS
OF ELEVEN STUDIES OF ASBESTOS WORKERS (44)
Mortality Study
°In f-yr/mJ.
blncrease in SMR per f-yr/mJ/100.
McDonald et al. (52) 6x10~*
Henderson and Enterline (56) 3.3-5.0x10-3
Weill et al. (57) 3.1x10~3
Dement et ol. (58) 2.3-4.4x10-2
Rubino et ol. (54) 1.7x10~3
Berry and Newhouse (59) 6x10"*
Peto (60) 1.0x10-2 0.7x10~8
Finkelstein et ol. (61^, 62) 4.8x10~2 12.0x10-8
Nicholson et pi. (53) 1.2x10"3
Seidman et ol. (63) 6.8x10-2 5.7x10-8
Selikoff et ol. (64) 1.0x10-2 1.5x10-8
2-47
-------
Toble 2-14
RRD ESTIMATES FOR ASBESTOS (TOTAL FIBERS)
Neoclosm
All malignant
Lung
Mesothelioma
Population
Non smokers
Smokers
Nonsmokers
Smokers
All
Level of
10-6
6.5x106
3.6x106
5.5x107
6.9x106
8.4x106
Extra Risk
.25
1.6x1012
9.0X1011
1.4x1013
1.7x1012
2.1x1012
Table 2-15
RRD ESTIMATES FOR ASBESTOS (mg/day)
Level of Extro Risk
Neoplasm
All
Malignant
Lung
Population
Nonsmokers
Smokers
Nonsmokers
Smokers
Mesothelioma All
RRDi
3.3E-6
1.8E-6
2.8E-5
3.5E-6
4.3E-6
10-6
MLE
2.0E-5
1.1E-5
1.7E-4
2.1E-5
2.6E-5
RROu
1.2E-4
6.6E-5
1.0E-3
1 .3E-4
1.6E-4
RROi
8.3E-1
«f.7E-1
7.2
8.7E-1
1.1
.25
MLE
<». 9
2.8
5.2
6.5
RRD,j
2.9E-H
1 .7E+1
2.6E+2
3.1E-f1
3.9E-H
2-48
-------
Benzene
The aromatic hydrocarbon benzene (CAS No. 71-43-2) has been produced
from coal or petroleum for over one hundred years. At one time benzene
was blended with gasoline, but now it is used primarily as a chemical
intermediate (as ethylbenzene) in plastic manufacture and as a solvent
in the paint and rubber industries (65). Although benzene has been
tested in a number of systems, it does not appear to be mutagenic. It
has, however, induced chromosome anomalies in some rodent species and in
occupationally exposed people (2).
The literature describing the human benzene experience is extensive.
Several studies (65-68) document the metabolic fate of inhaled benzene.
Apparently, in addition to short-term excretion of benzene in expired
air and in urine (as phenol, primarily), some of the chemical accumu-
lates in the body, probably in the fat, and ifc only slowly eliminated.
Studies by McMichael et al. (70), Arp et al. (71) and Checkoway et al.
(72) review the health of workers in the rubber industry. These workers
are exposed to a variety of solvents, among which is benzene. Leukemia
is the cancer most commonly associated with benzene exposure. The first
two studies, along with the review article by Infante and White (73),
examine the distribution of leukemia types. Checkoway et al. (72) claim
a stronger association exists between lymphocytic leukemia and solvents
other than benzene (notably carbon disulfide and carbon tetrachloride)
than between lymphocytic leukemia and benzene. In the case reports
provided by Vigliani and Saita (7J») and Vigliani (75), none of the
leukemia deaths associated with benzene exposure were lymphocytic. Data
from France (75) reveal only 8 of <»<» benzene-related leukemia deaths of
the lymphocytic type.
2-49
-------
Th« eerie* of article* from Turkey (76-83) corroborate* these observa-
tions: only 9* of the benzene-exposed population with leukemia had
lymphocytic leukemia, as opposed to 264 of nonexposed leukemia patients
(§!)• These articles also suggest a relationship of benzene exposure to
malignant lymphoma, myeloid metaplasia, multiple myeloma, Hodgkins
disease, and possibly lung cancer. Aksoy and his associate* also
discuss the possibility that genetic factors may influence the predispo-
sition to, or expression of, benzene-induced leukemia. Unfortunately,
in none of these reports is a cohort of exposed individuals defined or
followed-up, so no risk analysis was performed on these studies.
Two articles (8», 85) document the occupational standards for and expo-
sures to benzene in recent years. It is concluded that occupational
exposures par so have been substantially reduced. However, background
levels of benzene as on atmospheric pollutant are present, especially in
urban areas. These studies apply primarily to Scandinavia and the
United States.
The three cohort studies for which exposure estimates have been derived
(86-88) will serve as the basis of our risk assessment. As leukemia is
the only generally accepted carcinogenic response associated with
benzene exposure, that outcome and all malignant neoplasms will be the
endpoints analyzed, in accordance with our guidelines. Similar risk
estimates have recently been developed from these data sets by two of us
(89). Thu exposure estimates contained in Appendix B of that report
were used in thic document without change.
A cohort of 59
-------
estimates of benzene concentrot.ii.oiis in the thr\»« production areas. Most
of the measurements of at.Tiosphisrie bensren* ware obtained subsequent to
1964, with none before 19<»A. Exposures before 1S44 were assumed to
parallel those in the earliest periods for which industrial hygiene
measurement* were available. If ore accept* that levels may hcve been
somewhat higher during World Wor II and earlier, exposure estimates may
tend to be sotrswhot low for peopls employ ad dur;rg th' earlier years.
Job categories were clotsifled by exposure level, as depicted in Table
2-16. Through the courtesy of Mr. Ott, we had access to a computer
listing of the data for this cohort, including work histories and the
exposure classification, so that cumulative dose (in ppm-years) ond
expected deaths were co;culabl% (Table 2-17). Age- and calendar-year-
specific mortality rates for United States white males were employed in
the calculation of expected numbers of deaths.
Uncertainty considerations for this study are summarized as follows:
1. Measurements of atmospheric benzene concentrations did not
occur before 1!»44, and in some departments not before 1952.
The authors claim, however, that few process changes occurred
in the earlier time periods so that the concentrations measured
in 1944 are probably representative of earlier exposures. We
have assigned a value of 0 to aj and 0.4 to 72-
2. To cover the uncertainties associated with use of area concen-
tration samples and with use of notional death rates to esti-
mate expected numbers of deaths, ag • 75 • 0.1 and
78 • 78 ' 0.05.
The resulting factors,
-------
application of these factors to cumulative dose estimates.
The cohort identified by Rinsky et al. (87) consists of white, male
employees of three facilities producing rubber hydrochloric!* at two
locations in Ohio. Through the courtesy of Mr. Rinsky, a computerized
listing of the cohort was obtained. It included detailed job histories,
mortality information, and follow-up through 1978. A total of 1713
white males employed between 1940 and closure of the plants (1965 in one
location, 1976 in the other) were examined.
%
Mr. Rinsky was also kind enough to supply us with the data necessary to
relate job codes in the cohort data tape to work areas, for which
exposure estimates were available. The occupational hygiene measure-
ments given in the original paper could then be used to estimate expo-
sures encountered in each job over the years (Tables 22 and 23). Note
that past exposures to high concentrations have been documented; the
industrial hygiene data for these plants was generally good and reason-
ably complete for many years, including most of the follow-up period.
No exposures to other potentially carcinogenic chemicals have been
reported.
As with the Ott et ol. cohort, detailed dose calculations were possible
owing to the availability of the individual job histories. The rela-
tionship between cumulative dose (in ppm-years) and occurrence of
leukemia and all malignant neoplasms is displayed in Table 2-20. A
total of 8 leukemias were observed; all of them were nonlymphocytic.
The expected numbers of deaths were calculated from age- and calendar-
year-specific rates.
The uncertainties in this investigation are much like those reported
above for the Ott et al. (86) study. The contributing features are as
follows:
2-52
-------
1. Benzene measurements were obtained only after 1945. However,
the plants supplying cohort members began operations in 1937
and 1939, so relatively little time of exposure is not docu-
mented. In this case, aj • 0 and 12 " 0.3.
2. Fewer concentration measurements were performed at the plants
in the second location. Rinsky et al. state that "there is no
evidenca to suggest that exposures between the two locations
differed widely either in type or severity." The concentra-
tions presented in their paper corroborate that claim; aj and
13 have been given a value of 0.1.
3. Once again, otg • 75 • 0.1 and ag • IQ - 0.05 to reflect uncer-
tainty in the applicability of area samples and of U.S.
national mortality rates.
The resulting bounds, shown in Table 2-20,.are derived from the uncer-
tainty factors
-------
state how far in the past measurements were taken.
Workers in the continuously-exposed group (so assigned if they worked at
one continuously exposed job, regardless of duration) were further
divided by cumulative exposure. This dose estimate was calculated using
all jobs held, those involving no exposure to benzene as well as those
associated with benzene exposure. Unfortunately, average exposures in
each group were not reported; the values we have estimated are presented
in Table 2-21.
Table 2-21 also lists observed and expected numbers of deaths for
leukemia and all cancers. Mote that the comparison group (no exposure)
experienced substantially fewer leukemias than were expected. A total
of six leukemias were observed in the group of men who at one time were
continuously exposed to benzene. Expected mortality was based on age-
and calendar-year-specific rates.
As shown in Table 2-21, the bounds on dose for this investigation are
wider than those in the two previously-reported benzene studies. The
following features contribute to the outcome:
1. Very little is said about the occupational hygiene measurements
that defined the exposure estimate associated with each job
title. We assume that some measurements in at least some of
the facilities formed the basis of those exposure estimates,
but beyond that we lack documentation. To account for possible
underestimation of early exposures, TJ - 0.5, and to cover the
possibility of not very complete measurement even in more
recent times (a very real possibility, given the fact that seven
different plants were included) aj and -73 have been assigned a
value of 0.5.
2-5<»
-------
2. The cumulative exposure groups were presented without average
values. Both a^ and ~i^ equal 0.2 to address uncertainty in the
estimation of the average.
3. As in the previous two studies, we assumed a value of 0.1 for
ag and 75 and a value of 0.05 for 013 and ~IQ in consideration of
the use of area samples and national mortality rates.
The overall uncertainty factors ore oc • 1.85 and 7 • 2.35.
The potency parameter estimates from the three studies discussed are
displayed in Table 2-22. These have been calculated for leukemia and
for oil malignant neoplasms. The corresponding RRDs are shown in Table
2-23. Despite the fact that benzene is fairly specifically linked to
leukemia, the lower bounds on RRO are smaller when studying all
neoplasms than when studying leukemia alone, within each study. Never-
theless, two of the studies (the weakest two — Ott et al. and Wong) are
consistent with a hypothesis of no effect of benzene on either leukemo-
genesis or carcinogenesis in general. This is indicative of the
variability and uncertainty associated with epidamiologic studies of
cancer.
2-55
-------
Table 2-16
CLASSIFICATION OF JOB TITLES IN THE OTT ET AL.
(86) COHORT, BY EXPOSURE TO BENZENE
Estimated
Exposure Category Range (ppm) Average (ppm)
Very Low <2 1
Low 2-9 5
Moderate 10-2^ 17
High >25 30
2-56
-------
Table 2-17
OBSERVED AND EXPECTED NUMBERS OF HEATHS IN THE
OTT £T AL. (86) COHORT, BY CUMULATIVE DOSE OF BENZENE
Cumulative Person-yeans
Doc* of
(ppm-yeors) Observation
0-5 3533
(1.29, 1.«*8.
2.29}°
5-20 2961
(9.<»8, 10.9
16.9)
20-80 3758
(39.0. <*<».a,
69. <»)
80-200 1790
109.5, 125.9.
195.1)
200-f 1229
(306.9. 352.9.
547.0)
Cause of Death
All Malignant Neoplasms Leukemia
Observed Expected Observed Expected
2 3.5 1 0.18
9
-------
Toble 2-i8
BENZENE EXPOSURE (ppm) BY OPERATION CODE
AND YEAR FOR LOCATION 1, RINSKY ET AL. COHORT
Operation Code
11
03,07,17
12
06,33
04
27
01,08,31
09,
18.
22,
26,
05
15
02.13
10, 14. 16,
19.20.21,
23,24.25,
28.29.30,34
32,35
-'46
62
60
56
51
111
31
259
259
259
76
5
--
'47
62
60
56
51
111
31
114
118
108
54
1
--
Year
'48 '49-'57
30 22
30 21
28 20
26 18
56 39
16 11
59 42
61 44
54 38
27 19
1 1
__
'58- '63
10
2
10
18
34
8
31
32
22
n
i
--
'64-'69
10
2
6
11
18
8
31
32
22
9
1
--
'70+
2
2
2
2
11
4
10
11
6
3
1
—
2-5B
-------
Table 2-19
BENZENE EXPOSURE (ppm) BY OPERATION CODE
AND YEAR FOR LOCATION 2, RINSKY £T AL. COHORT
Operation Code
18,56
03,16,46.47
10
07
04,
15,
25,
36.
43,
53,
05
21,40,52
12
22
34
, 14,35.54
.08.27.28,
44,48.57
29,30,55
01 .02
06.09.11.13.
19,20,23,^4,
26,31,32,33.
37,38,39,41.
45,i»9,50,51
59,60,61 .62,63
-'46
63
60
56
51
43
111
31
259
259
259
76
5
'47
63
60
56
51
43
111
31
240
249
240
70
5
Year
•48 '49- '57
31 22
30 2
28 20
26 18
21 15
56 39
16 11
120 42
129 44
120 36
36 19
5 1
'58- '63
12
2
10
18
15
34
8
31
32
20
14
1
'64- '69
18
2
6
11
3
18
8
31
32
20
9
1
2-59
-------
Tcble 2-20
OBSERVED AND EXPECTED NUMBERS OF DEATHS IN THE
RINSKY ET AL. COHORT, BY CUMULATIVE DOSE OF BENZENE
Cumulative Person-years Cause of Death
Dose of All Malignant Neoplasms
Leukemia
(ppm-yeors) Observation Observed Expected Observed Expected
0-5 19239
(0.96, 1.20,
1.86)a
5-20 8098
(8.8, 11.0,
17.0)
20-80 7003
(33.8, 42.2,
65.1*)
80-200 3746
(103.2. 12S,
200.0}
200-1000 3363
(336.4, 420.5,
651.8)
10004- 457
(1186, 1482.
2297)
28
15
15
26
13
13
7.9
8.8
1.4
1.2
0.55
0.52
0.32
0.34
0.051
°In parentheses are the lower bounds, best estimates, and upper bounds
for cumulative exposure in each dose group.
2-60
-------
Table 2-21
OBSERVED AND EXPECTED NUMBERS OF DEATHS IN THE
WONG COHORT, BY CUMULATIVE DOSE OF BENZENE
Cumulative
Dose
(ppm-years)
0
0-15
Cause of Death
All Malignant Neoplasms
Observed
53
56
Leukemia
Expected Observed Expected
82.5
54.6
0
2
3.40
2.07
(4.05, 7.5, 17.6)°
15-60 45 34.8 1 1.28
(20.3, 37.5, 88.1)
60+ 22 28.3 3 1.09
(42.8, 79.2, 186)
aln parentheses are the lower bounds, best estimates, and upper bounds
for cumulative exposure in each dose group.
2 61
-------
Table 2-22
BENZENE POTENCY PARAMETER ESTIMATES
Potencies ((ppm-yrs)""1)
Dose Lower Upper
Study Response Measure Limit0 MLE Limit0
Ott All Upper -5.35E-4* 7.61E-4 2.53E-3
et al. Malignant Bounds
(86) Neoplasms
(chi-squared Best -8.38E-4 1.18E-3* 3.93E-3
(
-------
Table 2-22 (continued)
BENZENE POTENCY PARAMETER ESTIMATES
Potencies ((ppm-yrs)"1)
Dose Lower Upper
Study Response Measure Limit0 MLE Limit0
Wong All Upper -1.33E-2" 0.00 1.05E-3
(88) Malignant Bounds
Neoplasms
(chi-squared Best -3.11E-3 0.00" 2.46E-3
(3) - 14.98) Estimates
Lower -5.75E-3 0.00 4.56E-3"
Bounds
Leukemia Upper -1.74E-4" 6.05E-3 1.56E-2
(chi-squared Bounds
(3) - <».11)
Best -4.08E-<» 1.42E-1" 3.67E-2
Estimates
Lower -7.62E-4 2.63E-2 6.78E-2"
Bounds
°900 confidence limits shown.
"An asterisk marks the parameters (f)\_, 0, 0U) used to derive RRDs for
each study and response.
2-63
-------
Table 2-23
RRD ESTIMATES FOR BENZENE (ppm)
ro
GB
Estimation
Study
Ott
et al.
(fig)
Rinsky
et al.
(02)
Wong
(28)
Response Method
All
Malignant
t 9oplasms
Leukemia
All
Malignant
Neoplasmr,
Leukemia
All
Malignant
r.'eoplasms
Leukemia
1
2
1
2
1
2
1
2
1
2
1
2
RRDL
1 74E-5
2.29E-5
1 . 33E-4
1.50E-4
3.53E-5
4.65E-5
6.95E-5
7.85E-5
1.73E-5
2.27E-5
4.06E-5
2.59E-5
Level of
ID'S
MLE RROL
6.67E-5 »
8.78E-5 »
CO 03
CO CO
8.66E-5 5.6*E-4
1..UE-* 7.42E-4
1.46E-4 4.39E-4
1.65E-4 4.97E-4
09 O>
CO CO
1 .94E-4 co
2.19E-4 »
Extra Risk
0.25
RRDL MLE
4.36 1.67E+1
7.34 2.81E+1
3.32E+1 <*>
5.03E+1 co
8.83 2.17E+1
1.49E+1 3.65E+1
1.74E+1 3.66E+1
2.63E+1 5.54E+1
4.32
7.28 co
1.02E+1 4.85E+1
1.D3E+1 7.35E+1
RRDU
CO
CO
CO
CO
1.41E+2
3.27E+2
1.10E+2
1.66E+2
CO
CD
CO
00
-------
Benzidine
The aromatic amine. benzidine, is an intermediary in the dyestuff
industry anr! has been used as a hardener in the rubber industry.
According to the International Agency for Research on Cancer (2),
benzidine is mutogenic when metabolically activated. Moreover,
occupational exposure to benzidine has been causally associated with
bladder cancer in humans. Risk estimation, in accordance with the
analysis guidelines described above, is based on bladder tumors.
Several studies (90-95) report occupational exposures to benzidine and
other aromatic amines, principally 0-naphthylamine. The incidence of
bladder tumors is increased in those workers exposed to benzidine.
BladJer cancer appears to be particularly aggravated by exposure to
0-naphth rlamine in addition to benzidine. There is a suggestion of
increased occurrence of some second primary cancers following joint
exposure (9J2). Moreover, it appears that exposure to azo-dyes made from
benzidine carries a carcinogenic risk to the bladder (93, 9J»). Walker
and Gerber (96) review data relating to that association and the reac-
tion of OSHA to the potential hazard. Unfortunately, none of these
occupational studies can document berzidine exposure levels.
Two reports (97, 98) relate physiological characteristics to aromatic
amine bladder carcinogenesis. Horton and Bingham (97), studying the
cohort defined by Zavon et al. (99) (see below), found that bladder
tumor occurrence was related to the serum properdin levels in exposed
men. Lower at al. (97) suggest that the genetically-determined ability
to acetylate arylamines influences aromatic amine-induced bladder
carcinogenesis; slow acetylators may be at increased risk.
Only the occupational study by Zavon et al. (99) gives and response data
essential for quantitative treatment of risk. This paper describes a
2-65
-------
cohort of workers engaged in the manufacture of benzidine. The number
of workers was quite small, 25, but 11 of them developed bladder carci-
nomas and 2 had benign bladder tumors. Some of these men were exposed
to other chemicals. In fact, 3 men were exposed to 0-naphthalamine
(including the 2 with benign tumors).
The analysis presented below is based on the Zavon et al . (99) cohort.
All 25 men are included; exclusion of 3 men exposed to /J-naphthalamine,
for example, would not exclude any observed carcinoma responses but
would decrease the expected number and hence elevate the relative risk.
Only one of the cases was terminal (1 man died from metastases to the
abdomen); expected morbidity was determined from the age-specific inci-
dence rates of malignant bladder tumors presented in the SEER report
(100), Table 11E. These rates apply to the period 1973-1977; they would
tend to slightly overestimate the risks from the period 1958 to 1970,
judging by the trend in urinary system cancer mortality. The cohort was
followed up for a maximum of 13 years, for a total of nearly 200
person-years. Follow-up was terminated when death or bladder cancer
occurred. Roughly 0.05 bladder cancers would have been expected. With
11 cancers observed, the relative risk is 220.
Concentrations of benzidine measured at the manufacturing plant are
presented in Table 2-2*. If we assume that workers, rotated through each
work station and that only one employee at a time worked at the
shoveling location while the other 24 were spread evenly at the other
jobs, then a weighted aritr netic average concentration that these
employees encountered was 0.80
Zavon et al . also present data on urinary benzidine concentrations in
the workers. They found an average of about 0.04 mg/J after the shift,
0.01 mg/P before the shift, and 0.003 mg/J on Monday morning. Assuming
a linear increase in urine concentration during work gives an average
2-66
-------
concentration of (0.01 + 0.04)/2 - 0.025 mg/J during the 9 hours spent
at work, assuming an exponential decay during non-work hours gives an
average concentration of -0.04(1 - 0.25)/ln(.25) • 0.022 mg/P. This
results in an average concentration in urine during work days of about
0.023 mg/J. Assuming that 100* of the benzidine inhaled by humans is
taken up by the body, that humans excrete 1.4* of the inhaled benzidine
in urine (rhesus monkeys excrete this percentage of ingested benzidine
in urine, (101)), and that humans breathe 10 m3 per 8-hour work day, the
resulting estimate of benzidine exposure is
(0.023 mg/P)(1.5 P/day )(1/0.Oi5)/(10 m3/day) - 0.23 mg/m3.
By comparison, the National Academy of Sciences report (102) estimates
of exposure in this cohort correspond to a level of 0.38 mg/m3.
However, NAS assumes an average urine benzidine level of 0.05 mg/day as
opposed to the value
(0.023 mg/J)-(1.5 f/day) - 0.034 mg/day
that we have estimated. IARC (2), using the measurements of atmospheric
benzidine presented in Table 2-24, calculates an upper limit on concen-
tration of 0.5 mg/m3. The value 0.23 mg/m3 appears reasonable; it was
used in the calculations reported.
The men were exposed for on average of 11.24 years. Considered os a
whole, then, the cohort was exposed to an overage cumulative dose of
(0.23 mg/m3)-(11.24 years) - 2.59 mg-yrs/m3 .
2-67
-------
Uncertainty in this study stems from the following sources:
1. Relatively few urine benzidine measurements were obtained, and
these occurred lote in the history of the plant. There is a
possibility that higher exposures further in the past were not
documented although the authors state that they were assured
that the conditions they describe were typical of those that
prevailed over the past few years. We assigned a value of 0.2
to 12 Ca2 " 0) and a value of 0.3 to 013 and 73 to account for
the lack of extensive exposure determinations.
2. The largest uncertainty is associated with the conversion from
urine concentration to atmospheric concentration. First of
all, the percentage of benzidine oxcreted (1.5%) was based on a
study of govagdd monkeys. Secondly, that study measured
benzidine and mono-acetyl benzidine excretion; it is not known
if the mono-acetyl moiety was included in the Zavon et al.
measurements of urine concentration. In this case 07 • 1 and
77 • 1.5, the latter factor being larger to reflect the last-
mentioned concern, which would tend to make the atmospheric
concentration estimates larger if mono-acetyl benzidine was not
measured.
3. Expected numbers of bladder cancers were based on data from
1973 to 1977, after the follow-up period ended. A factor of
0.2 is assigned to ag and T8-
The uncertainty factors, a • 2.5 and i • 3.2, determine the lower and
upper bounds on cumulative exposure, 1.0<* ard 8.29 mg-yrs/m^, respec-
tively. Potency parameters based on these values, and the best estimate
of dose, are given in Table 2-25.
Since the expected number of bladder cancers is so small, we considered
an alternative to our standard analysis approach. Suppose the lifetime
2-68
-------
probability of those cancers, when exposed to tne indicated cumulative
doses is 11/25 • 0.44 as suggested by the Zavon et al. study. The
binomial 90X bounds on this probability are O.S1 and 0.57. The dose-
dependent probability of cancer is expressed as
P(d) - 1-exp(-oc-/?2d) (2-4)
The parameter a is calculated from the expected number of bladder
cancers, 0.05, if there had been no exposure. That is,
P(0) - 0.05/2-J
and
P(0) . 1-exp(-a).
Hence a • 0.002. Use of the best estimates of dose and lifetime
probability, 2.59 mg-yrs/m^ and 0,44, respectively, yields a best
estimate, 0%, of
- ln(1-0.44) + 0.002 - 0.22
2.59
The corresponding reasonable upper and lower bounds on potency, /?2u and
02\. < respectively are
- ln(1-0.57) + 0.002 - 0.81
and
- ln(1-0 31) + 0.002 • 0.045
5725
These potency parameters arise from a model different from the standard
model and have different interpretations. They should not be compared
2-69
-------
directly to the parameters in Table 2-25, but only by reference to the
corresponding RRD estimates (Table 2-26). Basic Methods 1 and 2 use the
parameters shown in Table 2-25 and calculate virtually safe doses based
on the methods outlined in earlier in this setion. The binomial method
is based on equation 2-4. If any more of the 25 men have developed
bladder cancer since Zavon et ol. studied that cohort, the binomial
method gives underestimates of risk and, consequently, overestimates of
RRDs.
2-70
-------
Table 2-24
CONCENTRATIONS OF BENZIDINE IN ATMOSPHERE AT
DIFFERENT LOCATIONS OF BENZIDINE MANUFACTURING PLANT13
Benzidine
Sampling Location Concentration,
Reducers <0.007
Conversion tubs <0.007
Clarification tub 0.005
Filter press 0.072 - 0.415
Salting-out tub 0.152
Centrifuge <0.005
Location for shoveling 17.600
benzidine into drums
QFrom Zavon et al. (99).
Table 2-25
BLADDER CANCER POTENCY PARAMETER ESTIMATES FOR
BENZIDINE, FROM DATA IN ZAVON ET AL. (99)°
Dose Potencies ((mq-yrs/m^)~b)
Measure Lower Limit" MLE Upper Limit"
Upper 1.55E+1" 1.93E+1 4.15E+1
Bounds
Best 4.96E+1 6.19E+1" 1.33E+2
Estimates
Lower 1.24E+2 1.54E+2 3.31E+2"
Bounds
QZero degrees of freedom are available to assess goodness-of-fit.
b905f confidence limits are shown.
*An asterisk marks the parameters used to derive RRD estimates.
2-71
-------
Table 2-26
RRD ESTIMATES0 FOR BENZIDINE (mg/m3)
Level of Extra Risk
Estimation IP"6 0.25
Method RRDi MLE RRDU RRDi MLE RRDU
1 2.35E-9 1.26E-8 5.02E-8 5.89E-* 3.15E-3 1.26E-2
2 2.60E-9 1.39E-8 5.54E-8 8.42E-4 4.50E-3 1.80E-2
Binomial" 8.24E-8 3.03E-7 1.48E-6 7.89E-3 2.91E-2 1.42E-1
QBased on the risk of bladder cancer and the data in Zavon et al.
(99).
bln this method, the model for lifetime probability of bladder cancer is
P(d) » 1-exp(-a-0d). Forty-five years of exposure is assumed.
2-72
-------
Cadmium
The metal, cadmium, is a relatively rare element found in the earth's
crust. It is produced for use mainly by sintering flue dusts and
roasting zinc ores. Cadmium and its compounds are used in metal
plating, in plastics stabilizers, in pigments, in pesticides, and in the
manufacture of batteries (103).
Short-term tests of cadmium have been equivocal. Although it was not
mutagenic in one test on Drosophila, it produced chromosomal anomalies
in human and mammalian cells in vitro. There are conflicting reports
about the production of chromosomal aberrations in exposed people
(2). The Kidney is the critical organ with respect to noncancer,
systemic effects (1(H).
The epidemiologic literature contains several descriptions of the human
cadmium experience. Pharmacokinetic studies (10<». 105) indicate that
ingested or inhaled cadmium is rapidly transported to the liver. It is
slowly released from the liver and appears in the kidneys where it again
resides for extended periods. The half-life for cadmium in the human
body may be as long as 30 years. Another study (106) indicates that
iron deficiency may increase the absorption of dietary cadmium. This
study also estimates the half-time of cadmium to be 93 to 202 days,
significantly shorter than the 30 years mentioned above.
Various types of studies have suggested carcinogenic effects of cadmium.
A case-control study (107) collected data on three main sources of cad-
mium exposure — diet, cigarette smoking, and occupation — and deter-
mined that renal cancer was associated with such exposure. A geogra-
phical investigation in Alberta, Canada (108) found that areas with high
cadmium concentrations in water, soil, and grains tended to have higher
incidences of prostate cancer. In a geographical investigation of trace
2-73
-------
metals in the water supplies of various parts of the United States, Berg
and Burbank (109) found an association between cadmium levels and
several tumors including cancers of the esophagus, large intestine, and
lung.
Studies of occupational cadmium exposure are also represented in the
literature. Potts (110) described conditions in an alkaline battery
factory. The employees were exposed to cadmium oxide dust. Of seventy-
four men exposed for at least 10 years, three prostatic cancer deaths
were noted. Further evidence of a link between cadmium and cancer of
the prostate was provided by Kipling and Woterhouse (111). In their
survey of 248 workers exposed for at least one year to cadmium oxide, 4
prostate cancers were observed whereas 0.58 would have been expected.
Lemon et ol. (112) followed up retirees from a cadmium smelter. The 292
white males who worked at least two years between 1940 and 1970 experi-
enced cadmium concentrations that were, for the most part, on the order
of 1 to 1.5 mg/m3. Excursions into areas that had concentrations up to
31.3 mg/m3 were documented. Follow-up to the beginning of 1974 revealed
4 prostate cancers (1.15 expected) and, moreover, an excess of respira-
tory system cancers (12 observed vs. 5.11 expected).
Kjellstrom et al. (113) have presented preliminary results on two groups
of workers exposed to cadmium: 269 workers in a battery factory and 94
exposed employees at a copper-cadmium alloy smelter. All members of the
cohort were exposed for at least 5 years. The first group, followed
from 1959 to 1975, showed a significant excess of nasopharnyx cancer
cases and nonsignificant excesses of prostate, lung, and colorectal
cancers. The second group was followed for prostate cancer deaths only,
from 1940 to 1975. Four were observed; 2.69 were expected. Exposures
in these cohorts ranged from 1 mg/m^ to 50 jig/m^ and perhaps down to 5
in recent times in the battery factory. In a recent update
2-74
-------
(114) with an expanded cohort and additional follow-up, nosopharyngeal
cancer was no longer significantly overrepresented. The nonsignificant
excesses of prostate, lung and colorectal cancers were still apparent.
Similar excesses of pancreatic and bladder malignancies were found.
Armstrong and Kazantzis (115) identified a cohort of 6995 men from 17
English plants processing cadmium. These men were born before 1940 and
had at least one year of employment between 1942 and 1970. The cohort
was categorized by the maximum exposure category ("high", "medium", or
•low") in which they were employed for a year or more. Unlike other
studies, prostatic cancer deaths were underrepresented (23 observed and
23.3 expected). A slight, but nonsignificant excess of lung cancers
(199 observed, 185.6 expected) was said to be unrelated to the exposure
grouping.
Sorahan and Waterhouse (116) provide an update and expansion of the
investigations by Potts (110) and Kipling and Waterhouse (111). The
cohort they described includes 3025 employees of the nickel-cadmium
battery factory who began work between 1923 and 1975 and who were
employed for a least one month. Mortality was investigated for the
period January 1, 1946 to January 31, 1981. Although exposure to nickel
hydroxide was not separable from cadmium exposure, the authors tenta-
tively conclude that there exists some indication of a risk of respira-
tory system cancer due to cadmium exposure. Prostate cancer was
associated with cadmium exposure also, but no new evidence of that
association was found over and above that presented by Kipling and
Waterhouse (111).
Unfortunately, the occupational studies described above provide very
little information on magnitudes of exposures, certainly not enough to
define quantitative dose measures. The study by Thun at al. (117) does
include mortality analyses by cumulative exposure and, hence, will serve
2-75
-------
as the basis for a quantitative assessment of risk due to cadmium
exposure.
The cohort identified by Thun et al. is an extension of the cadmium
production workers cohort described by Lemen et al. (112). In the
update, ohe cohort included all hourly employees who worked a minimum of
6 months in a production area of the plant between 1940 and 1969. A
total of 602 white males were followed up through 1978.
Calendar-year-specific cadmium concentrations were determined for each
department in the plant (Table 2-27). These were based on historical
area monitoring data, which began in 1940, but these have been adjusted
in two ways. First a correction was applied to convert area samples to
personal samples, based on the ratios of those two values found in
measurements taken from 1973 to 1976. Second, in those departments
where respirators were used, personal exposures were divided by 3.9, the
mean respirator protection factor determined in a plant survey.
Detailed job histories allowed computation of cumulative exposures over
time for each cohort member.
The total of 41 malignant neoplasms observed differed only slightly from
the 36.46 expected. On the other hand, the 20 respiratory system can-
cers were significantly in excess of the 12.15 expected. Four of the 20
were hired before 1926, when the plant functioned as an arsenic smelter.
When the analysis is limited to the subcohort hired on or after January
1, 1926, and so presumably with little or no arsenic exposure, a dose-
response trend for lung cancer was observed (Table 2-28). Lung cancer
is the only carcinogenic response categorized by cumulative exposure, so
the risk estimates are limited to that endpoint.
Table 2-28 presents the averages assumed for each exposure group —
1.22, 7.30, and 18.16 — and the bounds derived for those averages. The
2-76
-------
bounds were obtained by applying the factors a and 7, both of which have
the value 1.65 relating to the following uncertainties:
1. Industrial hygiene measurements began in 1940, but exposures
began as early as 1926. It is not known how concentration
estimates were extrapolated to this early period. The value
0.3 is assigned to aj and T2-
2. Also not documented is the completeness of the measurements
used to estimate concentrations post-1940. The data in Table
2-27 indicates that is was possible to estimate exposures that
differed from deportment to department and over time. A
nominal value of 0.1 is assumed for aj and 73.
3. The cumulative exposure groups defined by Thun et al. (Table
2-28} were presented without average values. The groups are
fairly narrow (except for the last) so a relatively small value
of 0.1 is assigned to on,, and i^.
4. The authors did an excellent job constructing exposure esti-
mates pertinent to individual inhalation by taking account of
respirator use and the difference between area and personal
samples. Some small amount of residual uncertainty with
respect to that conversion entails a value of 0.05 for ocg and
76-
5. The endpoint of interest is lung cancer, one very sensitive to
smoking behavior. A retrospective assessment of smoking
patterns in the cohort demonstrated little difference between
the cohort and notional values with respect to percentage of
smokers. Some of the cohort members were exposed to arsenic, a
known lung carcinogen. It would have been ideal if expected
values of lung cancer deaths could have been derived with these
factors in mind, but the effect is probably small. Both ag and
-JQ have been given a value of 0.1.
2-77
-------
The averages and bounds have been used to estimate the parameters /?, /?u,
and 0|_. These estimates are shown in Table 2-29. When applied to tho
standard exposure scenario — forty-five years exposure starting at age
20 — RRD estimates can be derived. Those corresponding to the esti-
mated potency parameter and its bounds are given in Table 2-30.
2-78
-------
Table 2-27
ESTIMATES OF CADMIUM INHALATION EXPOSURE, BY PLANT DEPARTMENT AND TIME PERIOD0
10
1
Nl
-------
Table 2-28
DOSE AND RESPONSE DATA FOR THE CADMIUM-EXPOSED
COHORT STUDIED BY THUN ET AL .
Cumulative
Exposure Lung Cancer Deaths
(mq-yrs/m3)0 Observed Expectedb
<2.43 2 3.77
(0.74, 1.22, 2.01)c
2.43 - 12.17 7 4.61
(4.42, 7.30, 12.0**)
>12.17 7 2.50
(11.07, 18.£6, 30.13)
°The original authors presented cumulative exposures in units of mg-
days/rn3. These have been converted by assuming that 240 days per year
are spent on the job, i.e. by dividing by 240.
^Expected values have been calculated from observed numbers and SMRs.
cln parentheses are the lower bounds, best estimates, and upper bounds
for cumulative exposures, respectively. These were not presented by the
original authors; see the text.
Table 2-29
CADMIUM LUNG CANCER POTENCY ESTIMATES
<(mg-yrs/m3)-1), FOR THUN £r AL . (117) COHORT
_ Dose Measure __ lover Limit0 _ MLE _ Upper Limit0
Upper Bounds 2.09E-2* 5.14E-2 9.03E-2
(chi-squared (2) • 1.21)
Best Estimates .V46E-2 8.48E-2" 1.49E-1
(chi-squared (2) - 1.21)
Lower Bounds 5.70E-2 1.40E-1 2.46E-1"
(chi-squared (2) - 1.21)
O90< confidence limits shown.
"An asterisk indicates potencies used to calculated RRDs.
2-80
-------
Table 2-30
PRO ESTIMATES FOR CADMIUM (mg/m3)a
Level of Extra Risk
Estimation 10~60.25
Neoplasm Method RRD| MLE RRDU RRDt MLE RRDU
Lung Cancer 1 1.12E-6 3.<»5E-6 1.<»OE-5 2.97E-1 8.61E-1 3.^9
2 1.36E-6 3.94E-6 1 .60E-6 4.20E-1 1.22 it.Sit
°From Thun «t ol. (117) data.
2-81
-------
Chlorombueil
Chlorambucil (CAS No. 305-03-3) is a drug used as an immunosuppressive
agent in the treatment of certain nonmalignant diseases (rheumatoid
arthritis, psoriasis, etc.) and as an antineoplastic agent for combat-
tiri hematopoietic system neoplasms and carcinomas of the breast, lung,
and genital organs (118). Chlorambucil is an alkylating agent,
interacts with DNA in mammalian cells in vitro, and is mutagenic in
bacteria and fungi (2). It is not surprising then that Chlorambucil is
highly suspected of being a carcinogen.
Four papers discussing the human carcinogenicity of Chlorambucil have
been reviewed. Rieche (119) provides an overview of the care ogenicity
of antineoplastic agents in general. He cites a study in which Chloram-
bucil was given for the treatment of non-Hodgkin's lymphoma. An eight-
fold risk of skin cancer was observed. Other investigators have noted
leukemia rather than skin cancer after Chlorambucil treatment. In a
study of aklylating-agent therapy of ovarian cancer, Reimer et ol.
(120) found a relative risk for acute nonlymphocytic leukemia of 36 (13
cases observed vs 0.36 expected). Chlorambucil was not the o-My chemo-
theraputic agent used in this multi-center study and some of the
patients received radiation therapy as well. That ic also the case in a
report of five clinical trials given by Greene et al. (121). These
authors report that 2 acute non-lymphocytic leukemias were observed, a''
opposed to 0.009 expected, in patients receiving more than 2003 mg of
Chlorambucil (6 mg/day for two years initially prescribed). These
ovarian cancer patients also received pelvic irradiation, however.
The study described by Berk ot al. (122) provides the information neces-
sary to estimate RRDs for chlorombucil. In a study of treatment of
polycythemia vera, they conducted a romdomized t^ial with three treat-
ment arms: phlebotomy alone, phlebotomy plus chlorombucil, and phle-
2-82
-------
botoniy plus radioactive phosphorus. These authors concluded that
chlorambucil significantly increased the risk of acute leukemia.
Table 2-31 displays the information pertinent to calculation of RRDs.
This data is based on the phlebotomy-only (untreated) group and that
portion of the chlorambucil-treated group with complete dose informa-
tion. Since the data is obtained from a randomized clinical trial, we
opted to use the untreated group to calculate expected numbers of
leukemias. This seems appropriate because the authors nofi that there
are reports of an association between polycythemia vera and leukemia
irrespective of chlorambucil treatment. Two out of 134 untreated
patients developed acute leukemia, so the expected value for the low-
dose group, for example, is
2 60-0.90.
T35 '
To calculate cumulative doses, we had to assume an average daily dose
for the two groups (<<* mg/doy and >4 mg/day) and a duration of dosing.
The authors state that only one patient received more than 10 mg per
day, so we assumed an average of 7 mg per day for the high-dose group.
A value of 2 mg/day was used for the low-dose group. The authors do not
clearly state the average length of dosing; they do state that chloram-
bucil treatment was stopped, apparently close to the date the article
was written. At that time, the chlorambucil group was followed for an
average of 5.
-------
of exposure, probably not a terrible assumption given that
chlorambucil treatment has been terminated. Nevertheless,
«1 " 71 " 0.2.
2. No averages are provided for the dosage-defined groups, so
again these were estimated. Consequently, a^ and -7^ have also
been given a value of 0.2.
3. The expected numbers of deaths is based on a very small popula-
tion of 134 phlebotomized, polycythemia vera patients. This
undoubtedly leads to substantial potential variability in the
leukemia responses seen. That being the case, ag and 73 are
equal to 0.3.
As a result of these considerations, a - T - 1.7. The resulting bounds
on cumulative dose are displayed in Table 2-31.
Table 2-32 displays the results of fitting the standard models to the
dose and response data described above. Table 2-33 provides the
corresponding RRD estimates when the calculated potency parameters are
applied to the standard exposure scenario.
2-84
-------
Table 2-31
DOSE AND ACUTE LEUKEMIA RESPONSE
DATA FOR CHLORAMBUCIL0
Average
(6.4.
(22.2
Daily
Dose
(mg)
0
<4
10.8, 18.4)c
>4
. 37.8, 64.3)
Number
of
Patients
134
60
31
Number
of
Leukemias
2
5
7
Expected
Number of
Leukemiasb
0.90
0.46
°From Berk at al. (122).
bBased on experience of the phlebotomy-only group; see text.
cln parentheses are the lower bounds, best estimates, and upper bounds
for cumulative dose, expressed in mg-yrs.
Table 2-32
LEUKEMIA POTENCY PARAMETER ESTIMATES FOR CHLORAMBUCIL,
BASED ON THE STUDY BY BERK ET AL.a
Potencies ((mq-vrs)"1)
Dose
Measure
Upper
Bounds
Best
Estimates
Lower
Bounds
Lower
Limit2
1.45E-1*
2.47E-1
4.20E-1
Upper
MLE Limit2
2.30E-1 3.38E-1
3.91E-1" 5.75E-1
6.64E-1 9.76E-1"
°Fit of the model to the data is adequate; chi-squared (1) < 0.03.
b90£ confidence limits are shown.
2-85
-------
Table 2-33
RRO ESTIMATES0 FOR CHLORAMBUCIL (mg/day)
Level of Extra Risk
Estimation
Method
1
2
RRDi
1.93E-6
2.17E-6
10-6
MLE RROU
4.82E-6 1.30E-5
5.43E-6 1.46E-5
RRDi
4.83E-1
7.21E-1
0.25
MLE
1.21
1 .80
RRO,,
3.24
4.84
QBased on the data of Berk et ol. (122) and the risk of leukemia
morbidity.
2-86
-------
Chromium
The metal chromium and its compounds are mined and produced in large
quantities for use in the metallugical, refractory, and chemical indus-
tries: ferrochromium alloys are used as additives in the production of
stainless and other special steels, chromite refractory materials are
used in iron and steel processing, nonferrous alloy refining, glass-
making, and cement processing (123). In addition to the elemental
state, two oxidation states (+3 and +6) are commonly found. Trivalent
and hexavalent compounds are the only ones known to play a role in bio-
logic systems. Apparently, the hexavalent forms are easily reduced to
the trivalent state after exposure to organic matter but the oxidation
of trivalent to hexavalent forms is unlikely in a biological context.
It is the case that hexavalent compounds are readily absorbed from the
lung whereas trivalent compounds dissipate slowly. On the other hand it
is believed that the reduction of hexavalent chromium to the trivalent
state and the formation of complexes is the mechanism by which hexava-
lent chromium reacts with nucleic acid (12fr). Indeed, the hexavalent
form has caused DNA damage, mutations, and chromosomal aberrations in
species ranging from bacteria to mammals. Trivalent chromium,
conversely, shows no evidence of producing these effects (2).
The epidemiologic literature contains many reports of the health effects
of chromium exposure. In a report on chromium plating industry employ-
ees, Royle (125) found nonmalignant respiratory symptoms to be more
prevalent than among industrial workers without chromium exposure. Also
found was a risk of skin and intranasal ulceration that increased with
duration of chromic acid exposure. As early as 1948, a study of
chromate plant workers revealed an increased risk of respiratory cancer
(cf. 12». 126-128 for review of many epidemiologic studies). This has
been corroborated in numerous occupational studies (129-136). Several
of these and other studies have suggested a link between chromium expo-
2-87
-------
sure and gastrointestinal cancer and at least one case-control study
(157) postulates an effect of chromium on nasal and sinonasal cancer.
It is unfortunate that these studies lack the exposure data needed for a
quantitative approach to risk assessment.
Three cohorts have been identified that provide information on exposures
to chromium. A small Norwegian company producing zinc chromate has been
the subject of study by Langard and his associates (138-140). The plant
produced only lead chromate from 1948 to 1951, adding zinc chromate
production at the latter date. Lead chromate production ceased in 1956.
Only 133 men worked in the facilities between 1948 and 1972 and only 24,
those exposed for at least 3 years before January, 1973, were analyzed.
By the time of the last update (follow-up through 1980), 6 lung cancers
were found compared to 0.135 expected from national rates. Expected
gastrointestinal cancer is reported only through 1975; 0.47 cases were
expected but 3 were observed.
Chromium concentrations were measured in 1972 (Table 2-34). Earlier
measurements were not available but the authors state that interviews
indicate that concentrations in the past were of the same magnitude as
those indicated. Plant C was built in 1972, the end of the cohort-
defining period. Consequently, Plant C is not considered in estimation
of cumulative exposures. Apparently, the workers rotated from job to
job, so an average exposure is calculated for the group as a whole. The
arithmetic average of the concentrations in Table 2-34 for plants A and
B is 0.45 mg/m^; this is the assumed average exposure. In addition, the
cohort members worked for 4 to 19 years; more than half of them worked
for 6 years or less, including six who worked for four years only. This
information is used to estimate an average length of exposure of 8
years. Consequently, the average cumulative dose for this cohort is
estimated to be
2-88
-------
(0.45 mg/m3)-(8 years) • 3.6 mg-yrs/m3.
Bounds on that best estimate havo been set to 1.85 and 8.82 mg-yrs/m3
(ex - 1.95 and 7 • 2.45). These result from the following considera-
tions:
1. No specific average duration of employment is given. The
information discussed above has allowed estimation of 8 years
as the average, but the uncertainty remaining contributes a
value of 0.2 to ocj and i-\.
2. No concentration measurements were performed before 1972.
Langard and Norseth believe that these are at least indicative
of the magnitude of concentrations prevailing at earlier times.
Nevertheless, a value of 0.2 is assumed for aj, and 12 *s set
equal to 0.5.
3. Several exposure estimates are presented, but no estimate of an
average, plant-wide exposure is given. By averaging those
estimates given, some uncertainty is perpetuated. Consequent-
ly, a
-------
confounder. Nevertheless, a value of 0.3 for 013 and IQ is
considered appropriate to cover the uncertainty introduced.
Langard et al. (1M ) studied o cohort of ferrosilicon and ferrochromium
workers specifically to see if trivalent forms of chromium entailed
risks similar to those of hexavalent chromium. Cohort membership was
restricted to men employed for at least one year who started before I960
(and who were alive after January 1, 1953). Follow-up extended from
1953 to 1977. A total of 976 individuals constitute the cohort; they
were categorized by the area of the facility in which most of their
employment was spent. The dichotomy studied by the authors was ferro-
chromium workers vs. all others. The former group experienced 23
cancers (7 lung cancers) with 23.49 (3.10 lung cancers) expected; among
the other workers
-------
(0.025 mg/m3)-(10 years) • 0.25 mg-yrs/m3.
On the other hand, the nonferrochromium workers include the maintenance
workers, who averaged (following the same procedure as described above
for ferrochromium workers) 0.011 mg/m3 of chromate. The 127 maintenance
workers are combined with the remaining 52* workers to provide an esti-
mated average cumulative exposure for nonferrochromium workers of
(0.011 mg/m3)-{10 years) •(127/52, is set to zero, but the upper bound
factor, 12- i* set equal to 0.8.
3. The measurements of concentration, even when they were pei—
formed are not terribly extensive, though they apparently
covered the areas and jobs expected to entail chromium expo-
sure. A value of 0.2 is assigned to 03 and 73.
i». No average chromium concentrations for various departments are
given. Nor is there information on how many men worked at each
2-91
-------
job. Either piece of information would have aided in the
calculation of average cumulative exposure. In general, the
range of concentration values is not too great, so the uncer-
tainty is not so drastic here, perhaps. A value of 0.1 is
ossMmed for a^ and 74.
5. Some recording bias is introduced by classifying cohort members
by the department they worked in for the longest time. Some
nominally unexposed persons may have spent some time in the
ferrochromium department, and ferrochromium workers may have
spent a considerable amount of time in jobs entailing no
chromium exposure. The factors 1x5 and 75 have been set to 0.5.
6. Total chromium concentrations were converted to chromate
concentrations. The factor used to accomplish this, 22%. is
subject to some variability (the percentages encountered ranged
from 110 to 33*) and, moreover, the method may not be precise.
Chromate content was based on solubility. The solubility of
chromic (trivalent) as well as chromate compounds varies (12fr).
In any case, a factor of 0.7 is assigned to 1x7 and 77 to
account for this uncertainty.
7. Expected numbers of cancers ore based on national rates, known
to exceed local rates for lung cancer incidence. Again,
smoking status may play a role, although the cohort's smoking
behavior appears not to differ from the notional pattern. A
factor of 0.1 is appropriate for 09 and IB-
The resulting uncertainty factors,
-------
year between 1930 and 1575. Cancer incidence was determined from 1958
to 1975 and compared to incidence rates from the county in which the
factory was located. No significant differences with respect to total
or respiratory cancers were seen in the factory when compared to
expected values derived from those rates.
The workers at this facility were primarily exposed to metallic and
trivalent chromium, although exposures to hexavalent chromium did occur.
Table 2-37 displays the concentrations at four working sites as esti-
mated by recent measurements and interviews with retired workers and
foremen employed in the 1930's. It should be noted that asbestos-
containing materials (textiles, plates, asbestos-isolated tubes) were
used in the plant. From 1931 to 1940 2100 kg of asbestos per year were
used; from 1941-1950 650 kg/yr were used. Workers were classified
according to the working site at which they spent the majority of their
time. They were also classified by length of employment, but unfor-
tunately no cross-classification by length and working site is provided.
Hence, it is necessary to estimate an average duration of employment for
the entire cohort a* opposed to working site-specific durations. A
total of 768 men worked 1 to 5 years (assumed average, 3 years) 538
worked 5 to 15 years (assumed average, 10 years), and 517 worked more
than 15 years (assumed average, 25 years). The estimated average
duration of exposure is
(768-3) + (538-10) + (517-25) -11.3 years.
1823
The cumulative exposure estimates based on the hexavalent chromium
concentrations displayed in Table 2-37 (using 0.03 mg/m3 for the trans-
port, metal grinding, and sampling site) and a duration of 11.3 years
are shown in Table 2-38. Also shown are the total malignancy and respi-
ratory cancer responses observed and expected. Two of the respiratory
cancers among the maintenance workers and one among the arc-furnace
2-93
-------
workers were due to pleurol mesotheliomo and hence may be attributed to
asbestos exposure. This observation strengthens the conclusion of the
original authors that exposure to chromium was not related to
respiratory cancer at this factory.
Also displayed in Table 2-38 are the bounds on cumulative exposure to
hexavalent chromium. The uncertainty factors used to derive those
bounds are based on the following features of the study:
1. Although length of smployment intervals of reasonable size are
presented, these had to be.used for all four job categories,
even though employment histories may well differ for different
jobs. Both <*i and 71 have been set equal to 0.3.
2. No hygiene data were available from the period when the
observed cancers would have been induced. The original authors
state that conditions have improved considerably over the post
ten year* (since 1970} and they warn against using the exposure
data to derive dose-response relationships. We have done so
despite their warning, but reflect the uncertainty in this
regard by setting 012 equal to 0.2 and 12 •Q.ual to 0.8. Since
general approximations to exposures were presented, it is
possible that they were overestimated (bearing on aj) but more
likely that they were underestimated and, hence, the wider
bound above the best estimate relative to that below.
3. The number of measurements forming the basis of the exposure
estimates is not documented. In addition, it appears that
exposure to hexavalent chromium also occurred in a chromate
reduction process not accounted for in the job categories
given, hio exposure estimates for that process were available.
To account for both of these factors, aj • 0.3 and 73 • 0.5.
The factor 73 is larger because the lack of exposure estimates
for the reduction process, where some employees may have
2-94
-------
worked, means that potential exposures ore not accounted for.
4. Some recording bios is presont in the format of the categori-
zation. Exposure ot jobs other than that defining the category
is ignored. This could entail o*,.. or greater or lesser cumula-
tive exposures in the transport and maintenance categories,
only lesser cumulative exposure for arc-furnace workers, and
only greater cumulative exposure for office and storage area
workei s. For the latter group, an arbitrary upper bound has
been selected to reflect this feature, 0.01 mg-yrs/m^. For the
other categories, this uncertainty is reflected as follows:
Arc-furnace workers: 05 • 0.2, 15 • 0;
Transport, metal grinder, sampling: ag • 0.1, 15 • 0.2;
Maintenance: 015 • 0.2, 75 • 0.1.
5. The report by Axelsson ot al. appears to indicate that area
samples are the basis of exposure estimates. This uncertainty
with respect to the applicability of the estimate is consistent
with «s and 75 equal to 0.2.
6. Although local, county incidence rates were used to calculate
expected numbers of cancers, it was not possible to account for
other chemical exposures, the most important of which was
asbestos exposure, nor for smoking habits. The asbestos expo-
sure is crucial given the observation of 3 mesotheliomas. The
extent of the influence of asbestos exposure on the development
of lung cancer in the cohort members is not known. In accor-
dance with this uncertainty, and that introduced by unknown
smoking behavior of the cohort, ag and ~IQ have been assigned a
value of 0.5.
In this case, the uncertainty factors a and 7 are group dependent. The
values entailed by the discussion above are:
2-95
-------
Arc-furnaces: a « 2.7, -j - 3.3;
Transport, metal grinder, sampling: ex - 2.6, i - 3.5;
Maintenance: a • 2.7, 7 • 3.if.
The potency parameteres calculrted from the data in the three studies
are displayed in Table 2-39. The parameters pertain to several end-
points, ranging from respiratory morbidity to mortality from all malig-
nancies. The morbidity rosults (1j*3) shoulu be viewed with the greatest
skepticism since asbestos exposure may have played an important role in
the observed responses and, moreover, the model fails to fit the respi-
ratory cancer data. Nevertheless, the RRD estimates corresponding to
the potencies displayed have been calculated and are presented in Table
2-96
-------
Table 2-34
XPOSURE IN DIFFERENT WORK
LANGARD AND NORSETH (139)
Mean Chromate
ation Concentration (mg/m3)
.ling 0.43
. raw materials 0.35
an (all departments) 0.19
jck filling 1.35
ixinq raw materials 0.33
foreman 0.04
Sack filling 0.08
Mixing raw materials 0.01
Table 2-35
TOTAL CHROMIUM CONCENTRATIONS IN DIFFERENT
DEPARTMENTS; LANGARD ET AL. (141)°
Number Mean Chromium
•r,c/Operotion of Samples Concentration (mg/m3)
chromium department
f'otmen 20 0.04
Cleaner baler 5 0.09
Crane driver 10 0.04
Packing 10 0.29
Maintenance people
General maintenance 9 0.09
Transport men 9 0.01
Other -- 0
°0nly personal samples are reported. Departments other than maintenance
and ferrochromium are assumed to have no atmospheric chromium, as
indicated by the authors.
2-97
-------
Table 2-36
DOSE AND RESPONSE DATA FROM CHROMATE-EXPOSED
COHORT OF LANGARO £T AL. (140)
All Malignant
Neoplasms Lung Cancers
Exposure Group Observed Expected Observed Expected
Nonferrocnromium 41 55.95 2 6.35
workers
(0.0066, 0.027, 0.13)°
Ferrochromium 23 23.49 7 3.10
workers
(0.061, 0.25, 1.22)
°In parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for average chrornate exposure in each group, expressed as
mg-yrs/m^.
Table 2-37
ESTIMATED CONCENTRATIONS OF CHROMIUM
BY WORKING SITE; AXELSSON ET AL. (145)
Concentrations (rng/m^)
Working Site Cr°+Cr3+ Cf-6*
Arc-furnaces 2.5 0.25
Transport, metal grinder, 0.5 - 2.5 0.01 - 0.05
sampling
Maintenance 2.5 0.05
Office, storage area 0 0
2-98
-------
Table 2-38
DOSE AND RESPONSE DATA FOR CHROMIUM-EXPOSED
COHORT OF AXELSSON ET AL. (143)
Exposure Group
All Malignant
Neoplasms
Observed Expected
Respiratory Cancer
Observed Expected
Arc-furnaces 31 30.8
(1.0«f, 2.82. 9.31)a
Transport, metal 26 30.0
grinder, sampling
(0.13, 0.3*. 1.19)
Maintenance 19 15.6
(0.21, 0.56, 1.90)
Office, storage area . 11 9.5
(0, 0, 0.01)
2.1
2.1
1.0
0.7
aln parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for cumulative exposure to hexavalent chromium
S) in each group.
2-99
-------
Toble 2-39
POTENCY PARAMETER ESTIMATES FOR CHROMIUM
Study
Response
Potencies ((mg-yrs/m3)~1)
Dose Lower Upper
Measure Limit0 MLE Limita_
Langard Lung Cancer Upper
and Mortality Bounds
Vigander (0 degrees
of freedom)
2.73"
Best 6.68
Estimates
61 Cancer
Mortality
(0 degrees
of freedom)
Lower
Bounds
Upper
Bounds
Best 4.85E-1
Estimates
Lower
Bounds
4.93
1.50"
9.44E-1 2.91
8.04
1.21E+1" 1.97E+1
1.30E+1 2.35E+1 3.83E+1'
1.98E-1" 6.10E-1 1.29
3.15
6.13*
Langard
et al .
Lung Cancer Upper 1.28E-1*
Mortality Bounds
(chi-squared
(1) • 3.7) Best 6.16E-1
Estimates
Lower 2.52
Bounds
All Upper -2.58E-1"
Malignancies Bounds
Mortality
(chi-squared Best -1.26
(1) - 4.0) Estimates
Lower -5.17
Bounds
7.82E-1 1.68
3-80" 8.18
1.56E+1 3.35E+11
0.00
0.00
1.62E-1
0-00" 7.90E-1
3.23"
2-100
-------
Table 2-39 (continued)
POTENCY PARAMETER ESTIMATES FOR CHROMIUM
Potencies
Dose Lower Upper
Study _ Response _ Measure Limita _ MLE _ Limit0
Axelsson Respiratory Upper -S.
-------
Table 2-40
RRD ESTIMATES FOR CHROMIUM (mq/m5)
o
NJ
Estimation
Study
Longard
and
Vigander
(143)
Langard
et al.
(m)
Axelsson
et al.
(Htl)
Response Method
All
Malignant
Neoplasms
GI Cancer
Mortality
Lung Cancer
Mortality
All
Malignancies
Mortality
Respiratory
Cancer
Morbidity
All
Malignancies
Morbidity
1
2
1
2
1
2
1
2
1
2
1
2
7.
8.
5.
6.
8.
9.
2.
3.
1.
2.
1.
2.
RROt
63E-9
73E-9
95E-8
71E-8
73E-9
99E-9
43E-8
20E-8
78E-7
08E-7
52E-7
32E-7
10~B
MLE
2.42E-8
2.77E-8
2.44E-7
2.75E-7
7.69E-8
8.61E-8
CD
ao
5.20E-6
6.08E-6
8.57E-6
1.S1E-5
Level of
RROU
1.07E-7
1.23E-7
1.84E-6
2.08E-6
2.29E-6
2.62E-6
CO
a
OB
00
CO
m
Extra Risk
RROL
1.91E-3
2.70E-3
1.
2.
2.
3.
6.
1.
4.
6.
3.
7.
.49E-2
19E-2
18E-3
09E-3
09E-3
03E-2
44E-2
36E-2
80E-2
30E-2
0.25
MLE
6.06E-3
8.57E-3
6.09E-2
8.96E-2
1 . 92E-2
2.72E-2
00
CO
1.30
1.86
2.U
-------
Cigarette Smoke
Cigarette smoke is probably the most widely studied cause of cancer.
The smoke is a heterogeneous mixture of many chemicals several of which
are carcinogens themselves. The constituents include nitrosamines,
hydrocarbons, formaldehyde, acrylonitrile, benzene, naphthylamine, PAH's
and various other particulate substances (Ijjjv). Cigarette smoke is
firmly linked to lung cancer and there are very strong indications that
it is associated with cancer of the larnyx, oral cavity, kidney,
pancreas, and bladder (145).
The epidemiologic literature on cigarette smoke is extensive and will
not be reviewed here. The Surgeon General's reports (cf. 1_fjA) provide
extensive references to the studies that have been conducted. The
remainder of this section is devoted to the development of quantitative
risk estimates for cigarette smoke.
The estimates are developed in terms of whole cigarette smoke, gas and
particulate phases together, with no reference to particular subfrac-
tions. The 1979 Surgeon General's report (144) lists the components of
cigarette smoke and provides an overall estimate of the weight of those
components that are in the mainstream smoke. The total weight of smoke
from one cigarette is about 500 mg. Conversions to weight-based
measurements of consumption are based on this value.
Doll and Peto (1_ftj5, 1»7) have reported on a study involving 34,400
British doctors who responded in 1951 to a questionnarie about their
smoking habits. Most of these were subsequently followed for 20 years
and causes of death for 10,072 decendents were recorded. Changes in
smoki: j habits were ascertained in follow-up questionnaries collected
during the seventh and fifteenth years of the study.
2-103
-------
The authors concluded that cancers of the lung, esophagus, and other
respiratory sites (lip, tongue, mouth, pharnyx excluding nasopharynx,
larynx, and trachea) could be directly attributed to cigarette smoking.
Although cancers of the rectum and pancreas occurred at significantly
higher rates in smokers, the authors did not consider these to be
related to emoking. Also, even though cancer of the bladder was not
found to be increased in this study, because cigarettes have been
implicated as the cause of bladder cancer in other studies, the authors
considered bladder cancer "probably wholly or partly attributable to
smoking."
Since cigarette smoking is so pervasive, our standard approach of esti-
mating a potency for a chemical in terms of a relative risk and applying
this to standard mortality rates could lead to serious errors; t'te
"background" rates, as given in vital statistics, include smokers and
nonsmokers and so do not represent mortality among the unexposed.
Consequently, specialized methods must be applied to estimate RRDs for
cigarette smoke. Doll and Peto (147) estimated that among smokers who
started smoking at ages 16 to 25 and who smoked 40 or less cigarettes
per day the annual lung cancer incidence* in the age range 40-79 was
0.273x10-12-(cigarettes/day + 6)-(age - 22.5)*-5. (2-5)
This equation is intended to apply to smokers who, once they begin
smoking, smoke a constant amount daily throughout life. Assuming that
the period from diagnosis until death for lung cancer is roughly 2
years, this expression for incidence can be converted to one for
mortality by replacing the 22.5 by 24.5.
"Age of onset of lung cancer was estimated for the subcohort used in
this analysis, thereby allowing the estimation of incidence rates as
opposed to mortal!'-y rates.
2-104
-------
Table 2-
-------
motes the default pattern to which all risk estimates are adjusted C*5
years beginning at age 20), no adjustments to the estimates are
required; they represent our best estimates of RRDs.
Uncertainty considerations are based on the discussion in the methods
portion of this section, as described below:
1. As with other studies relying on questionnaries and recall of
habits, uncertainty arises with respect to length of exposure
and exposure in the less recent past. With cigarette smoking,
this may not be as significant a problem, but a-j • TJ - 0.3 and
«2 • 12 ' 0-2.
2. Some recording bias is introduced by not adjusting smoking
habit categorization based on data obtained in later question-
naires. This should not be a serious confounder, though, so 015
and 75 have been assigned a value of 0.1.
3. In converting from number of cigarettes to weight of cigarette
smoke, some uncertainty is introduced. The conversion factor,
500 mg smoke per cigarette, is based on standard cigarettes
smoked to standard butt lengths. Actual smoking habits
undoubtedly vary, so the factors 017 and 77 have been set equal
to 0.2.
<». In the place of expected numbers of deaths, w* have estimated
annual mortality rates for nonsmoker*. This is based on the
equation that Doll and Peto explicitly warn against using for
nonsmoker8. Consequently, ag and 73 have been set equal to 0.5.
Another uncertainty not fitting into the standard categories discussed
above involves the indirect estimation of increased risk for all malig-
nant neoplasms. A factor of 0.2 is added to the total of the uncertain-
ty subfactors mentioned above. Hence, a and 7 equal 2.3 for lung
cancer; they equal 2.5 for all malignant neoplasms.
2-106
-------
The RRO estimates and their bounds derived from the Doll and Peto study
are given in Table 2-42.
2-107
-------
Table 2-41
ANNUAL DEATH RATES PER 100,000 AMONG BRITISH PHYSICIANS
Nonsmokers
Smokers
Age
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
Total0
I
132
151
221
540
844
1431
2081
4236
7529
9071
Lung Concerb
II
0.101
0.438
1.32
3.20
6.70
12.6
22.0
36.1
56.5
84.7
Total6
III
196
326
574
1090
1679
2767
4573
6642
9636
14261
Lunq Cancer**
IV
1.90
8.22
24.8
60.0
125.7
237.2
413.7
678.7
1060.0
1590.5
°From Doll and Peto (146). Table 13.
"From Doll and Peto (147). using 24.5 in their equation for incidence
rather than 22.5 in order to convert from incidence to mortality
(assuming an average delay of 2 years between diagnosis of lung cancer
and death) and assuming no cigarettes smoked.
eFrom Doll and Peto (145). Table 13.
dFrom Doll and Peto (146). using 24.5 in their equation for incidence
rather than 22.5 in order to convert from incidence to mortality
(assuming an average delay of 2 years between diagnosis of lung cancer
and death) and assuming an average of 20 cigarettes smoked per day.
2-108
-------
Table 2-^2
RRO ESTIMATES0 FOR CIGARETTE SMOKE (mg/doy)
Level of Extra Risk
0.25
Response RRDi _ MLE _ RRDu _ RRDi _ MLE _ RRDti
Lung 5.12E-2 1 . 18E-1 2.71E-1 1 . 28E+4 2.95E+* 6.78E+4
Cancer
Mortality
All *.OOE-2 1.00E-1 2.50E-1 1.00E+4 2.50E+'v 6.25E+4
Malignancies
Mortality
°Based on the data in Doll and Peto ( Tt6) .
2-109
-------
DES
The synthetic estrogen diethylstilbestrol (DES) is one of several estro-
gens that were commonly prescribed, in the period 19<»0 to I960, to women
who were threatening to oaort, had a history of prior pregnancy loss, or
had other complications (149). In fact, in the mid-1940's, it was
suggested that increasing amounts of DES be administered to all women
during pregnancy to decrease hazards of late complications for both
mothers and babies. As early as 1953, however, some indications that
DES may not, in fact, have therapeutic value were published (150). More
recently, of course, reports of the health hazards of intrauterine DES
exposure have proliferated.
Short-term tests are negative or equivocal. DES did not elicit unsched-
uled DNA synthesis in mammalian cells, nor did it induce mutations in
bacterial systems; the remits of chromosomal anomaly tests are equivo-
cal for a variety of test species; but cell transformations and sperm
abnormalities have been noted (2).
Several nonneoplastic abnormalities of the genito-urinary tract in
humans are reportedly linked to intrauterine DES exposure (151 ). In
females these include vaginal adenosis, cervical erosions, and ridges
(152). Intraepithelial glandular abnormalities are also documented
(155). Male offspring of DES-treated mothers have increased incidences
of anatomical and functional abnormalities, including hypotrophir
testes, epididymal cysts, and pathologic semen (154. 155).
The cancer hazard of DES is more widely known. Some reports have indi-
cated a risk to those treated with DES themselves, both male (156) and
female (157), although these are not unambiguously linked to the DES
exposure (158, 159). By far the clearest evidence of a carcinogenic
effect of DES is provided by studies of genital cancer in female off-
2-110
-------
spring of DES-treated mothers. Graenwald (160) and Greenwald and Nasca
(161) report seven case* of vag-.njl adenocarcinoma from the New York
State Cancer Registry; mothers of six (and possibly (ill seven) of the
cases were treated with DES whereas no mother of any of the 12 controls
was so treated. Among 830 young women in the Philadelphia area who were
exposed to DES in utero, 8 adenocarcinomas were discovered (162).
Even among females exposed transplacentally to DES, for whom an excess
risk is not seriously questioned, the risk is apparently small. Lanier
et al. (1_4J}) estimated the incidence of cervical or vaginal adenocar-
cinoma associated with in utero DES exposure to be no more than 4 per
1000 and Herbst at al. (163), using the Registry of Clear Cell
Adenocarcina-na of the Genital Tract in Young Females (RCCAGTYF), suggest
that the cumulative risk of that type of cancer in DES-exposed females
is between 0.14 and 1.4 per thousand.
The previously cited reports have not been able to quantify doses of DES
administered to the mothers. We have attempted to do so, and to derive
RRD estimates on the basis of a series of articles by Herbst and his
associates (164-166). The original case-control study identified 8
cases of v ginal adenocarcinoma and 32 individually-matched controls. A
control was selected from the same hospital as her corresponding case;
the controls were born within five days of their respective cases and on
the some type of service. Seven of the 8 cases were exposed in utero to
DES; none of the controls were exposed to DES.
Unfortunately, dosage information is not presented in the case-control
study. Dosage is estimated from the other two reports. The study of
adenocarcinoma cases identified in the RCCAGTYF published in 1972
provides information on done and duration of DES exposure for 46 cases
whose mothers were treated with DES (Table 2-43). Using the 32 coses
with known duration of therapy, we estimate average length of exposure
2-111
-------
to be
(0.5x2) +J1.5x3) + (3.5x3) + (8x24) - 6.5 months,
32
where the numbers in parentheses are the products of the midpoints of
the duration categories and the number of cases in each category. We
have assumed an average "throughout pregnancy" value of 8 months.
Average daily dose can be similarly estimated if the "increasing" dosage
is known. Herbst et al. (166) describe a standard, increasing treatment
pattern used at the time of pregnancy of the cases' mothers in the area
in which the cases \yere born:
"During or before the sixth week 2.5 mg daily was given, increasing
to 5 mg in the seventh week. An additional 5 ing/week administered
every two weeks until the fifteenth week, when 25 mg per day was
prescribed. The daily dosage was then increased by 25 mg each
month reaching a maximum of 150 mg...[T]he drug was discontinued at
the end of the 35th week."
This "increasing" dose pattern yields an average daily dose of
[(2.5x2) -i- (5x2) + (10x2) + (15x2) + (20x2) + (25.4)
+ (50x4) + (75x4) + (100x4) + (125x4)
+ (150x1)] / 31 - 57 mg.
Here we assumed that treatment began in the 4th week of pregnancy, on
average. Consequently, the average daily dose for the RCCA6TYF cases is
estimated to be
[(5x4) -f (30x11) + (100x6) + (57x9)]/30 - 49 mg
where the values in parentheses are the products of the midpoints of the
dosage categories (Table 2-43, assuming a value of 100 mg/day for the
2-112
-------
">50 mg daily" group) and the number of cases in those groups. A cumu-
lative dose estimate to be used in the case-control study is
(6.5 months) x (49 mg/dav) • 318.5 mg-months - 26.5 mg-years.
The dose and response data are combined in Table 2-44.
The bounds on dose displayed in Table 2-44 are based on the following
considerations:
1. Length of exposure is very roughly estimated from groups
presented without average values. Consequently, ocj and -\-\ are
set equal to 0.2.
2. Similarly, dose groups are presented without averages and they
too had to be combined. Hence, a^ and 74 are assumed to be
0.3.
3. Some recording bias is introduced by categorizing dose and
duration separately rather than a cross-classification, for
example. In this case, 05 and 75 have been assigned a value of
0.2
4. Neither dose nor duration estimates came from the exposed
individuals in the case-control study. Instead, other exposed
females, largely from the same geographical area and exposed at
the same time, were used to estimate these parameters. The
question of the applicability of these values leads to
estimates of
-------
variability is displayed in the RRD estimates (Table 2-46). It should
be noted that the RRD estimate pertains to 45 years of DES use, much
longer than the actual period of exposure, nine months or less.
-------
Table 2-*»3
DOSAGF AND DURATION OF STILBESTROL THERAPY;
HERBST £7 AL. (165)
Dosage
< 10 mg daily
10 - 50 mg daily
> 50 mg daily
"increasing"
unknown
Number
of
Cases
*
11
6
9
16
Duration
< 1 month
1-2 months
2.1 - 5 months
"throughout
pregnancy"
unknown
Number
of
Cases
2
3
3
24
1*
Table 2-M»
DOSE AND RESPONSE DATA FOR CASE-CONTROL
STUDY OF HERBST ET AL.
Cumulative Exposure
(mg-years) Cases0
0 1
26.5 7
. (13.3, 53.0)b
Controls
32
0
°The cases are odenocorcinomas of the vagina.
bln parentheses are the lower bound and upper bound for dose estimated
for the DES-exposed individuals.
2-115
-------
Table 2-45
VAGINAL CANCER POTENCY PARAMETER ESTIMATES FOR DES,
FROM DATA IN HERBST ET AL. (164)
Dose
Measure
Upper
Bounds
Best
Estimates
Lc
-------
Epichlorohydrin
The olkylating agent epichlorohydrin (CAS No. 106-89-8) is used in the
manufacture of glycerine and epoxide resins and in several other manu-
facturing settings. Acute reactions to epichlorohydrin in humans have
included skin burns, eye and throat irritation, and EE6 changes (167).
Short-term testing, as with many alkylating agents, revealed positive
results with respect to DNA damage, mutagenicity, and chromosomal
anomalies in bacteria, plants, insects, and mammalian species.
Indications of chromosomal abnormalities in epichlorohydrin-exposed
workers have been noted (2).
The epidemiologic data are scanty and ore ambiguous with respect to the
carcinogenicity of epichlorohydrin in humans. Studies of three occupa-
tional cohorts are available. The first set of studies investiaged
workers at Shell Oil (cf. 168. 169). There seems to be some increase in
lung cancer risk in the cohort, but it appears to be among those exposed
to isopropyl alcohol in addition to epichlorohydrin. In any case, no
estimates of atmospheric concentrations of either chemical were
available.
Shellenberger et al. (170) evaluated the mortality experience of
epichlorohydrin-exposed workers for Dow Chemical. Members of this
cohort, which included 553 individuals, could have been exposed to
epichlorohydrin as early as 1957, when production began. Each member
had at least one month of exposure at some time between 1957 and 1976,
at which time follow-up ceased. Two cancer deaths — an adenocarcinoma
of the stomach and a metastatic malignant melanoma — are reported.
Exposure measurements for epichlorohydrin were available only since the
early 1970's. Other sources, however, have allowed Shellenberger ot al.
to estimate exposures for the glycerine and epoxy resin departments.
2-117
-------
deaths, four neoplasms.
Exposure estimates, once again, were available only late in the exposure
period, 1977-1978. The personal exposure measurement obtained at that
time averaged less than 1 ppm. It is known, however, that conditions
were worse during the earlier time periods, with epichlorohydrin concen-
trations reaching peaks of 10-25 ppm. Nevertheless, we have selected 1
ppm as an assumed average exposure level. It should be noted that many
cohort members were exposed to a number of other chemicals, particularly
ollyl chloride.
Duration of exposure is documented in Table 2-49. Length of exposure
for the entire cohort averaged 9.3 years. Containing categories to get
the "10 or fewer years of exposure" and "more than 10 years of exposure"
groups (the groups for which we have observed and expected cancer
mortality) yields average duration estimates of
(2.5 x 190) * (7.5 x 1»»2) . i».6 years
and
(12.5 x 141) + (17.5 x 110) + (25 x 23) - 15.6 years,
respectively. These estimates use the midpoints of each duration cote-
gory (assuming a maximum of 30 for the last duration group) and yield an
average for the entire cohort of 9.6, fairly (ood agreement with the
stated value of 9.3.
The cumulative exposure groups defined by the assumed average concentra-
tion of epichlorohydrin (1 ppm) and durations of exposure ace displayed
in Table 2-50 along with the numbers of observed and expected responses.
2-120
-------
The bound* on dose (cf. Table 2-50) ore derived on the ba«i« of these
considerations:
1. Although duration of exposure is categorized without averages
for the categories, the overall average of 9.3 years gives us
guidance on the accuracy of our approximations. Since the
latter yield an average duration of 9.6 years, the factor for
the lower bound ought to be somewhat larger than that for the
upper bound. Hence, a-j - 0.2 and T\ - 0.1.
2. The most serious uncertainty involves exposures early in the
exposure period. Since those were undoubtedly larger than the
assumed 1 ppm average, 12 " °'8 but °"2 " "•
3. The exposure estimates presented came from a survey of European
facilities, not limited to the four plants included in the
present study. Hence, pichlorohydrin is carcinogenic
in humans, so the MIC estimates of potency are zero. In other words,
the best estimate of RRD is infinite although finite lower limits are
obtained (Table 2-52).
2-121
-------
Table 2-i*7
DURATION OF EXPOSURE TO EPICHLOROHYDR7N FOR THE
COHORT STUDIES BY SHELLENBERGER £T AL.
Months Exposed
1
7
13
19
25
37
49
61
121
- 6
- 12
- 18
- 24
- 36
- 48
- 60
- 120
- 180
181 +
Percent of Employees
30.7
13.2
8.6
6.2
8.7
3.6
4.3
12.0
4.3
2.4
Table 2-48
DOSE AND RESPONSE DATA FOR EPICHLOROHYDRIN-EXPOSED
COHORT OF SHELLENBERGER ET AL .
Cumulative
Exposure (ppm-vrs)
0
(0.34
3
(2.02
14
(7.64
.66
. 1.35)a
.93
. 8.06}
.9
, 30.5)
All Malignant Neoplasms
Observed Expected
2 1.10
0 1.33
0 1 . 07
aln parentheses ore the lower bound and upper bound on dose for each
dose group.
2-122
-------
Table 2-1*9
DURATION OF EXPOSURE TO EPICHLOROHYORIN FOR
THE COHORT STUDIES BY TASSIGNON FT AL.
Years Exposed
< 5
5-10
10 - 15
15 - 20
20+
Number and Percent of Employees
190 (31*)
142 (24*)
141 (23*)
110 (18*)
23 (4*)
Table 2-50
DOSE AND RESPONSE DATA FOR EPICHLOROHYDRIN
EXPOSED COHORT OF TASSIGNON ET AL.
Cumulative All Malignant Neoplasms
Exposure (ppm-yrs) Observed Expected
4.6 1 2.2
(2.42. 12.0)a
15.6 3 2.8
(8.21. 40.6)
°In parentheses are the lower bounds and upper bounds on dose for
each group.
2-123
-------
Table 2-51
EPICHLOROHYORIN POTENCY PARAMETER ESTIMATES0
Dose Potencies ((mq-yrs/m^)"1)
Study Measure Lower Limitb MLE Upper Limitb
Shellenberger Upper -6.49E-1* 0.00 1.95E-2
et al. (170) Boundw
(chi-squared
(2) - 3.U) Best -1.33 0.00* 3.99E-2
Estimates
Lower -2.58 0.00 7.77E-2"
Bounds
Tossignon Upper -1.45E-2" 0.00 2.03E-2
et ol. (171) Bounds
(chi-squared
(1) • 0.67) Best -3.77E-2 0.00* 5.28E-2
Estimates
Lower -7.16E-2 0.00 1.00E-1"
Bounds
aBosed on all malignant neoplasms.
b90< confidence limits are shown.
"An asterisk marks the parameters used to derive RRD estimates.
-------
Toble 2-52
RRD ESTIMATES0 FOR EPICHLOROHYDRIN (ppir)
Estimation
Study Method
Shellen-
berger
et al.
(170)
Tassignon
et al.
(Ill)
1
2
1
2
10~6
RRDi MLE
1.01E-6 oa
1.33E-6 «
7.84E-7 oo
1.03E-6 on
Level of Extra Risk
0.25
RRDU RRDi MLE RRDU
oo 2.53E-1 oo oo
oo <*.27E-1 oo >o
oo 1.96E-1 oo oo
oo 3 . 30E-1 o° oo
°Based on the risk of all malignant neoplasms.
2-125
-------
Estrogens
Various forms c>f estrogenic substances have been used in humans to treat
reproductive disorders and as palliative therapy for disseminated carci-
noma (172). Conjugated, or natural, estrogens have been used recently
to treat menopausal symptoms. The conjugated estrogens have been
reported to be nonmutagenic in bacteria and did not induce chromosomal
aberrations in mammalian cells in vitro (2). It is suspected that
exogenous estrogens act by enhancing the effect or altering the balance
of endogenous hormones. There have been suggestions that exogenous
estrogens are linked to breast, enilometrial, and cervical cancer and
possibly to malignant melanoma (172). The studies used to calculate
RRDs for estrogens have examined breast and endometrial cancer end-
points.
Hoover et al. (175) reviewed the records of 1891 white women who were
treated with conjugated estrogens and who were seen at one private
practice since 1939. Their vital status was ascertained through 1972.
Forty-nine breast cancer cases were observed; 39.1 were expected.
Unfortunately, the published report does not moke available data
necessary for calculation of cumulative dose.
A case-control study of menopausai estrogen therapy and breast cancer
was conducted by Ross et al. (174). The cases (and the matched con-
trols) were from two retirement communities and were 50 to 7*» years of
age at Jiagnosis. Estrogen-usage histories were obtained from inter-
views, medical charts, and pharmacy records. The total accumulated
doses of conjugated estrogens for the cases and controls are displayed
in Table 2-53. A significant trend for risk ratio with total dose is
evident for women with ovaries intact but not for women whose ovaries
have been removed.
2-126
-------
Horwitz and Stewart (175) present an interesting case-control investi-
gation of estrogen and breast cancer. These authors conclude that
breast cancer risk is not increased by exogenous estrogen use. No
exposure measures are given however. Moreover, this study investigates
the dependence of the risk ratios on the methods employed to select
cases and controls. It is the contention of these authors that hetero-
geneity of the patient subgroups and "inattention to recording bias" in
exposure classification have led to inappropriate conclusions concerning
the association of estrogens and breast cancer. These concerns may not
be pertinent to the study conducted by Ross «t al. (174) since community
rather than hospital controls were used and interviews formed at least
part of the basis for exposure classification. The uncertainties that
do affect the Ross et ol. study entail uncertainty factors, a and 7,
equal to 1.65, as discussed below:
1. Recall difficulties contribute to uncertainty about the length
of estrogen use and about the dosage taken early in the period
of use. These factors are mitigated by the use of medical
charts and pharmacy records, so a-), T) , 012- ond T2 nave been
set equal to 0.05.
2. No average doses are given for the dose groups. Since we had
to approximate these by the midpoints of the group-defining
intervals, a^ and 7^ are assigned the value 0.9.
3. Some of the estrogenic substances used by the study partici-
pants were not conjugated estrogens. Consequently, there is
some question about the applicability of the reported doses and
05 and 76 are assumed to be 0.2.
4. The control series appears to be well matched to the cases, so
a small value, 0.05, is assigned to 03 and 73.
The possible link between exogenous estrogen use and endometrlal cancer
has been extensively studied. Smi'.h et al. (176) first suggested a
2-127
-------
causal association; they conclude that risk of endometrial cancer was
"on the order of 5" times greater among users of exogenous estrogens.
Dosages, durations of use, and types of estrogen were not considered and
the risk varied with other factors, notably obesity and hypertension.
Mack at al. (177) performed a case-control study of a Los Angeles
retirement community. Sixty-three endometrial cancer cases and 252
matched controls were included. An overall risk ratio of 5.6 (95*
confidence interval 2.8 to 11.1} was associated with conjugated estrogen
use. Moreover, a trend was observed of increasing relative risk with
increasing duration or dose of estrogen (Table 2-5
-------
determined from the various sources. Also, duration was
grouped, without average values. A value of 0.2 is assigned to
o<-| and 0.3 to T\ .
2. Recall of early estrogen use is augmented by the medical and
pharmacy records. Since some use may be ignored (see 1.
above) it is somewhat more likely that early estrogen exposure
is underestimated. We assume a value of 0.05 for «2 and a
value of 0.1 for 12-
3. No average values are given for the "pill size" categories.
Hence a^ and 74 are given a value of 0.2.
4. It has been necessary to use the information on drug-free
intervals to convert to daily dose. The uncertainty introduced
is consistent with a value of 0.2 for 07 and 77.
5. Once again, well-matched controls introduce little uncertainty
with respect to "expected numbers." Both 019 and IQ are assumed
to be 0.05.
The sum of these consideration! entails the uncertainty factors « • 1,7
and 7 • 1.85.
Another case-control study is described by McDonald «t ol. (178). Cases
were selected from residents of Olmsted County, Minnesota who had
pathologically-confirmed diagnoses of endometrial carcinoma between 19<»5
and 1974. A total of T»5 such cases were found. Four controls were
age-matched to each cose. The controls were also residents of Olmsted
County, had intact uteri at the time of diagnosis of the case, and had
duration of medical care approximating that of the case. Sixteen cases
and sixteen controls hod a history of exposure to conjugated estrogens
for 6 months or more (cf. Table 2-57). Note that duration of exposure
>*id not differ substantially among the cases and controls who used the
medication for longer than six months. For those 32 individuals, an
estimate of average length of use is 3.55 years (using the midpoints of
2-129
-------
the intervals in Table 2-57 and 6 years for the "3-f" group). Assuming
that this estimate can be applied to each daily dose group, estimates of
total dose can be calculated (Table 2-58). We have not included those
who used conjugated estrogens for less than 6 months. Note also that
the "daily dose* given may in fact be "pill size" (see discussion of the
mack et al. report) (177). though this is not certain. We have not been
able to correct for this explicitly since the definitions of "inter-
mittent" and "cyclic" are not given. The control group, however,
contained more individuals on less than continuous regimens, so the
effective doses in that group may have been smaller than indicated.
The uncertainty considerations are as follows:
I. Length of estrogen use and the dose pattern were determined
from records of the Mayo Clinic. Some patients did not have
complete records at that institution. In addition, duration
categories are presented without average values. Consequently,
ac|> 0.2, T| • 0.3, 02 • 0 and 72 " 0.2. The subfactors
contributing to the upper bound are larger because of the
possible underestimation of duration and early dosing for
patients with incomplete records.
2. A possible recording bias is present since duration and dose
are not cross-classified. Hence both 05 and -75 have been
assigned a value of 0.2.
3. Since it is not certain whether the dosages listed are pill
size or daily doses, ag and 7$. pertaining to the applicability
of the reported values, have been set equal to 0.3.
k. A well-matched control series entails little uncertainty; both
03 and 73 equal 0.05.
The resulting uncertainty factors ore a» 1.75 and 7 • 1.95.
2-130
-------
Quantitative estimation can be based on the description of a case-
control study conducted by Gray et al. (179). The 205 endometrial
carcinoma cases diagnosed at one private practice between 1947 and 1976
were matched to the same number of controls who had hysterectomy based
on age, year of diagnosis (operation), parity, and weight. Estrogen use
was recorded ignoring use "just prior to diagnosis to control abnormal
bleeding." The distribution of cases and controls by duration of use
end strength of tablet usually taken is given in Table 2-59. In this
case calculation of total dose and definition of dose groups is more
problematical. Durations of use are not similar among users in the case
and control groups and, moreover, the data presented describes strength
of estrogen usually taken. Nevertheless, if we assume an average
duration of 7.6 years of daily use for all users, then dose groups can
be defined as in Table 2-60. Again, it is likely that the number of
controls has been artificially elevated in the higher dose groups.
Uncertainty is slightly greater in this study than in the previously
discussed studies. In this case, a • 1.95 and 7 • 2.15; these factors
are derived by consideration of the following features:
1. Since the records of only one private practice were accessed,
it is possible that early use of estrogen is missed. Duration
of use groups are again presented without average values. We
assign a value of 0.2 to 01 and a value of 0.3 to T\.
2. It is also possible that early estrogen use is underestimated
since records from only one practice form the base of exposure
data. In this case 012-0 but 12 " 0.1.
3. Recording biases occur in two instances. .First, duration of
use and dosage are noc cross-classified. Moreover, the pattern
of use is not the some for the cases and the controls, although
they have been grouped to calculate an average duration of use.
A value of 0.<» is assumed for 05 and 75 to cover these possible
2-131
-------
biases.
4. Estrogen use categories ore based on the dose usually taken.
The degree to which this makes the recorded values inapplicable
is not known; a5 and 75 are set equal to 0.9.
5. Little uncertainty is introduced by the well-matched control
series. Both 013 and 73 are assumed to be 0.05.
Yet another case-control study is that reported by Antunes et al. (180).
The endometrial cancer cases were drawn from all patients with primary
cancer of the uterine body admitted to six Baltimore-area hospitals
between 1973 and February 28, 1977. The controls were female patients
matched to the cases on hospital, race, age, and date of admission and
who were not in the gynecology, obstetrics, or psychiatry departments.
The distribution of cases and controls from this study is given in Table
2-61. Not* how different the estrogen histories are for the two groups.
In fact, they differ to such an extent that it is not possible to
accurately define dose groups and no quantitative risk estimation is
based on this study. Had cases and controls been simultaneously cross-
classified, by dose and duration of estrogen use, the appropriate data
would have been available.
Hammond et al. (181) provide a follow-up study of hypoestrogenic women
who were either given exogenous estrogers or not. Cumulative dose of
estrogen, in ing-months, is categorized for those 301 women prescribed
estrogens (Table 2-62). These women were treated for at least 5 yaars.
If a woman received the lowest daily dose reported, 0.625 mg, the
minimum dose would be 37.5 mg-months. Th» average cumulative floeo for
the treated women is
52.6 x 96 + 106.3 K 86 * 300 x 105 • 159.2 mg-months
j^
where 52.6, 106.3, and 300 mg-months or« the assumed averogs dose* for
2-132
-------
the three treated groups shown in Tnble 2-62. Combination of the groups
is necessary because expected values are given only for the no-estrogen
and estrogen-treated women as a whole. Table 2-63 displays the resul-
tant dose and response data.
The bounds on dose in the treated group shown in that table result from
the uncertainty factors, a and -7, both of which equal 1.45. This value
stems from the following considerations:
1. Average values for the cumulative-dose groups are not provided.
Both
-------
elevated to 4.2 or 3.7 (compared to hospital or community controls,
respectively) for those using conjugated estrogen* for more than 31/2
years. Unfortunately, the data presented in this report does not allow
calculation of cumulative or total doses for conjugated estrogen* alone;
nonconjugated estrogens were considered simultaneously when dose and
duration were investigated. Hence, no quantitative estimates were
C'xnputed for use in our investigation.
Shapiro et al. (183) conducted a case-control study of exogenous estro-
gen and entfometrial cancer specifically designed to avoid a particular
bios thought to be present in some other case-control studies. That is,
if estrogens cause symptoms, mainly postmenopausal bleeding, leading to
closer examination and hence diagnosis of uterine cancer that might
otherwise have been overlooked, than selection of cases would be to some
degree conditional on use of estrogens. Shapiro and his colleagues
found that, even in women who had not been taking estrogens for a year,
the risk of endometriol cancer was elevated compared to never-users when
use lasted at least one year. Overall, use of conjugated estrogen
entailed a significant relative risk of 3.9. No dosage information was
available.
Another case-control study specifically designed to examine potential
biases is that described by Spengler et ol. (1Si»). These authors found
that the risk of endometriol cancer was increased among estrogen users
and that the risk increased with duration of use and daily dose. The
potential sources of bios (comparability of cases and controls, selec-
tion of controls, postmenopausal bleeding, medical surveillance, recall
of estrogen use, extent of disease, and influence of prior hysterectomy
in the controls) were found not to unduly influence the risk estimates.
Quantitative risk estimation was not possible, again becausw daily dose
and duration of use were not cross-classified, and the two groups showed
substantial difference in the distribution of these factors (cf.
-------
discussion cf Antunes et al. (ISO)).
The literature reviewed reveals a consistent pattern of increased risk
of endometrial cancer among estrogen users. The data for breast cancer
is not as conclusive. Five studies present data sufficient for potency
estimation. The potency parameters so estimated are given in Table
2-6U. When theset parameters are applied in our standard exposure
scenario (<»5 years of use starting at age 20. an unrealistic scenario
for conjugated estrogen use) the resulting RRO estimates are shown in
Table 2-65.
-------
Table 2-53
NUMBERS OF BREAST CANCER CASES AND CONTROLS BY TOTAL
ACCUMULATED DOSE OF CONJUGATED ESTROGEN AND OVARIAN STATUS0
Ovarian Status
Ovaries Intact
Ovaries Removed
Allb
Dose (mg)
0
1 - 1*99
1500+
0
1 - 1499
1500+
0
1 - 1U99
(1.24, 2.05. 3.38)c
1500+
(3.73, 6.16. 10.2)c
Cases
50
21
28
13
6
7
64
28
37
Controls
103
56
23
29
15
21
134
73
48
Ross et ol. (174).
^includes persons with ovary status unknown.
cln parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for cumulative dose (in mg-yrs) in each group.
2-136
-------
Table 2-5*
RISK RATIOS BY DOSE AND DURATION OF ADMINISTRATION OF
CONJUGATED ESTROGEN FOR ENDOMETRIAL CARCINOMA CASES AND CONTROLS0
Duration
of
Use
(months)
1 - 11
12 - 59
60 - 95
96+
Total"
Mean Dose
1 0.625 mg
Cases/
Controls RRfc
5/9 6.6
8/28 3.<»
3/6 6.0
*/10 0.8
23/55 5.0
> 0.625
Cases/
Controls
0/8
5/2
2/2
8/5
15/19
mg
RRC
0.0
29.8
11.9
19.1
9.*
Totalb
Cases/
Controls
6/26
15/M)
7/9
17/2^
59/109
RRC
2.8
*.*
9.*
8.8
5.6
°From Mock at al. (177).
^Including those for which dose or duration (or both) was not Known.
cBased on 12 coses and 1<»3 controls who were nonusers of conjugated
estrogens.
2-137
-------
2-55
NUMBERS OF ENDOMETRIAL CARCINOMA CASES AND
CONTROLS BY DRUG-FREE DAYS IN CYCLE AND MEAN PILL SIZEa
Drug-Free
Interval*3 (Days)
0-1
2-3
*+
i 0,
Cases
5
7
2
Pill
. 6?5 mg
Controls
1*
19
13
Size
> 0.
Cases
2
6
3
625 mg
Controls
5
5
5
°From Mack et al. (177).
bA cycle is assumed to last for 30 days, hence these are the number of
drug-free days per month.
2-138
-------
Toble 2-56
NUMBERS OF ENDOMETRIAL CARCINOMA
CASES AND CONTROLS BY TOTAL DOSE°
Total Dose (mq-yrs) Number of Cases
0 12
0.140 5
(8.22E-2. 0.258)b
0.458 0
(0.269, 0.8i«6)
0 . 778 8
(0.1*58, 1.44)
1.68 3
(0.992. 3.12)
2.54 5
(1.49. 4.69)
3.11 4
(1.83. 5.76)
5.49 2
(3.23, 10.2)
10.1 8
(5.97. 18.6)
Number of Controls
143
9
8
28
6
2
10
2
5
°Derived from data in Mack «t al. (177).
bln porenthese* or* the lower and upper bounds on cumulative dose for
each group.
2-139
-------
Table 2-57
DISTRIBUTION OF CASES AND CONTROLS BY DURATION, DOSE,
AND TYPE OF ADMINISTRATION OF CONJUGATED ESTROGENS0
Exposure Number of Coses Number of Controls
Duration:
None
<6 mo.
6 mo. - 1 yr.
1 - 2 yrs.
2 - 3 yrs.
3+ yrs
Daily Dose:b
0.625 mg
1.25
2.5
12*»
5
4
0
3
9
2
10
-------
Table 2-58
NUMBERS OF ENDOMETRIAL CARCINOMA
CASES AND CONTROLS IN OLMSTED COUNTY BY TOTAL DOSEa
Total Dos« (mg-yrs)
(1
(2
(5
0
2.22
.27, 4.33)b
4.44
.54. 8.66)
8.89
.08. 17.3)
Number of Cases
124
2
10
4
Number of Controls
533
8
8
0
0Based on data in McDonald «t al. (178).
bln parentheses are the lower and upper bounds, respectively, for
cumulative dose in each group.
Table 2-59
DISTRIBUTION OF ENDOMETRIAL CARCINOMA CASES AND CONTROLS BY
DURATION OF USE AND PILL STRENGTH OF CONJUGATED ESTROGENa-b
Exposure
Number of Cases
Number of Controls
Duration (yr«):
Pill
0-4
5-9
10 +
Strength (mg):c
0.3
0.625
1.25
8
7
10
7
14
11
8
2
1
2
9
1
°From Gray «t al. (179).
^Unknowns excluded in both cases. 150 cases and 174 controls received
no hormone*.
cStrength of pill usually used.
2-141
-------
Table 2-60
NUMBERS OF ENDOMETRIAL CARCINOMA CASES AND CONTROLS
IN A LOUISVILLE. KENTUCKY PRIVATE PRACTICE BY TOTAL DOSE°
Total Dose (mq-yrs) Number of Cases Number of Controls
0 150 174
2.28 7 2
(1.17. 4.90)b
4.75 14 9
(2.44. 10.2)
9.50 11 1
(4.87. 20.4)
°Based on data presented in Gray et al. (179).
"in parentheses are the lower and upper bounds, respectively, for
cumulative dose in each group.
Table 2-61
NUMBER OF ENDOMETRIAL CANCER CASES AND CONTROLS IN
BALTIMORE-AREA HOSPITALS BY DAILY DOSE AND
DURATION OF USE OF CONJUGATED ESTROGENS0
Exposure
Number of Cases
Number of Controls
Duration of use (vr):
Daily
Non*
<1
1 - 5
5*
Dose (mg):
None
<1
1 - 2
2+
274
11
17
36
274
23
27
6
390
7
8
3
390
9
5
2
°From Antunes ft al. (180).
2-142
-------
Table 2-62
DISTRIBUTION OF HYPOESTROGENIC
PATIENTS BY TOTAL DOSE OF ESTROGEN0
Estrogen Equivalents
(mg-months)
0
<67.6
67.6 - H»5
> 1
-------
Table 2-64
POTENCY PARAMETER ESTIMATES FOR ESTROGENS
Potencies ((mg-vrs)""1 )
Dose Lower Upper
Response Study Measure Limit0 MLE Limit0
Breast Ross et ol. Upper 1.06E-2* 5.27E-2 1.09E-1
Cancer (174) Bounds
Morbidity (chi-squared
(1) - 2.5) Best 1.74E-2 8.70E-2" 1.80E-1
Estimates
Lower 2.88E-2 1.44E-1 2.98E-11
Bounds
Endome- Mack et ol. Upper 5.94E-1* 1.07 1.78
trial (177) bounds
Cancer (chi-squared
Morbidity (7) - 14.8) Best 1.10 1.98" 9.30
Estimates
Lower 1.87 3.35 5.60"
Bounds
McDonald Upper 2.46E-1" 4.65E-1 8.16E-1
et ol. Bounds
(178) (chi-
squared (2) Best 4.80E-1 9.06E-1* 1.59
• 3.8) Estimates
Lower 8.40E-1 1.58 2.78"
Bounds
Gray et al. Upper 8.92E-2" 1.93E-1 3.71E-1
(179) Bounds
(chi-squared
(2) • 3.2) Best 1.92E-1 4.14E-1* 7.96E-1
Estimates
Lower 3.73E-1 8.06E-1 1.55"
Bounds
-------
Table 2-&k (continued)
POTENCY PARAMETER ESTIMATES FOR ESTROGENS
Potencies ((mg-yrs)"1)
Response
Study
Dose
Measure
Lower
Limit0
MLE
Upper
Limit0
Upper
Bounds
Endome- Hammond
trial et al.
Cancer (181)
Morbidity (chi- Best 3.31E-1
squared (1) Estimates
• 0.12)
Lower 5.52E-1
Bounds
2.62E-1" «».23E-1 6.31E-1
6.14E-1* 9.16E-1
S.91E-1 1.33"
°90> confidence limits shown.
*An asterisk marks the parameters used to derive PRO estimates.
2-U5
-------
I
.4
CD
Table 2-65
RRO ESTIMATES FOR ESTROGENS (mg/day)
Estimation
Study Response Method
Breast
Cancer
Morbidity
Endome-
triol
Cancer
Morbidity
Ross
ot ol.
(Ilk)
Mack
et al.
(177)
McDonald
et al.
(178)
Gray
et al.
(179)
Hammond
et al.
(181)
1
2
1
2
1
2
1
2
1
2
RROL
6.46E-7 2.
8.13E-7 2.
1.12E-7 3.
1.32E-7 3.
2.25E-7 6.
2.66E-7 8.
4.03E-7 1.
4.77E-7 1.
4.71E-7 1.
5.57E-7 1.
10~6
MLE
.21E-6
78E-6
16E-7
74E-7
90E-7
17E-7
51E-6
79E-6
02E-6
21E-6
1
2
1
1
2
3
7
8
2
2
RRDU
.82E-5
.29E-5
.05E-6
.25E-6
.5fcE-6
.OOE-6
.01E-S
.30E-6
-38E-6
.82E-6
RROL
1.62E-1
2.
2.
3.
5.
7.
1.
1.
1.
1.
39E-1
79E-2
82E-2
62E-2
69E-2
01E-1
38E-1
18E-1
61E-1
5
8
7
1
1
2
3
5
2
3
0.25
MLE
.53E-1
.17E-1
.90E-2
.08E-1
.73E-1
.36E-1
.78E-1
.17E-1
.55E-1
.48E-1
RRDU
*.55
6.73
2.63E-1
3.60E-1
6.35E-1
8.68E-1
1.75
2.40
5.96E-1
8.15E-1
-------
Tha oxiron* ethylane oxide (CAS No. 75-03-2) i« v:e«d in tru ^r.clnufac^u^a
of moriy othar chemical* onci as a furmgarvt ana sterilizing g^s, especial-
ly in the hospital industry OB^S). A» or. alkylatlrig o^aot, it is highly
suspect with respect to earcinoge-'icity. Ird««id, it d#monstrat*8 muta--
gsnic activity '.n short-tarn. t»sts on cells of bac*.*ria, plants,
Insects, and irammols. DNA aan*3g* Cmu chromosomcl onomolie* huve b»an
noteQ in several specien, including humans (2).
NIOSH has conducted industrial hygiene studies in o number of hospitals.
Time-weighted average exposures for the relevant personnel (tnose who
work with or around sterilization equipment) ranged from nordetectable
to about 10 ppm. Korpela et ol. (18S) measured only small amounts of
ethylene oxide dispersed from gas sterilizers. Nevertheless, the
epidemiologic literature contains studies of hospital workers (and
health instrument manufacturing employees) with observable cytogenotic
effects presumably due to ethylene oxide (187-189).
Landrigan »t al. (189) state that OSHA believes that ethylene oxide
exposure may increase the risk of malignancies, particularly leukemia.
This view is based, at least in part, on the results of three occupa-
tional studies. In a study of ethylene oxide producers, Morgan at al.
(185) observed cancers of several sites in exposed workers. No
leukemias were observed and the total number of malignancies did not
differ significantly from that expected on the basis of U.S. vital
statistics. The authors claim that small sample size reduced the power
to detect an effect of ethylene oxide, especially with respect to
leukemia.
On the other hand, Hogstedt and his colleagues have found significant
increases in leukemia in two small cohorts. Hogstedt et ol. (190)
2-107
-------
found three leukemias among workers at a technical factory that
sterilized hospital equipment with 50< ethylene oxide and 504 methyl
formate. A total of 230 individuals worked in or around the area
entailing exposure. National cancer rates predicted 0.2 cases of
leukemia among these workers. They worked between 196S and 1977 for <*
to 10 years (assumed average, 7 years). Seventy of the employees were
exposed to TWA concentrations of 20 ppm. The other 160 individuals were
only occasionally in this high exposure area. It is assumed that their
TWA exposures averaged 10 ppm. As a whole, the estimate of overage
exposure for this cohort is
(70x20) » (160x10) • 13 ppm.
230
This average exposure combined with the assumed average exposure period
yields a cumulative dose estimate of
(13 ppm)-(7 yrs) • 91 ppm-yrs.
Uncertainty factors a and 7 have been estimated to be 1.8 upon consi-
deration of the following features of the study:
1. Length of exposure varied from k to 10 years, but no average
length of exposure is given. Uncertainty attendant with
estimation of overage duration is reflected in the choice of
0.2 for «i and T\ .
2. The only measurements of ethylene oxide concentrations occurred
in 1977. Exposure may have begun as early as 1968 for some
individuals. It does not appear that concentrations would have
changed much over those 10 years, however, because the process
does not seem to have changed over that period. A factor of
0.1 is assigned to both 03 and 73-
2-U8
-------
3. It is not known how complete the measurements taken in 1977
were. Certainly, they were detailed enough to ascertain the
breathing zone concentrations for individuals who worked in the
storage hall. Time-weighted average concentrations are not
given for the majority of employees who only occasionally
ventured into the hall. Uncertainty is associated with the
approximation of 10 ppm TWA for those workers and is mirrored
in the selection of a value of 0.<» for 03 and 73.
k. Finally, the use of notional mortality rates does not, perhaps,
give the best estimate of expected values. These workers were
exposed to methyl formate as much as to ethylene oxide. A
fairly minimal value of 0.1 is assigned to 019 and 73-
The lower bound, best estimate, and upper bound for dose, 51, 91, and
164 ppm-yeors, respectively, have been used to derive estimates of /?u,
ft, and 0\_.
The other study by Hogstedt »t al. (191) utilized employment data from a
company that has produced ethylene oxide since the beginning of the
1940's. Included in the follow-up study were men who had hematologicol
investigations performed in 1960-61 and who were employed for more than
1 year. Follow-up started in January, 1961 and ended in December 1977;
however, accumulation of person-years at risk for each individual did
not begin until 10 years after the beginning of exposure. The m*n in
this cohort were exposed to other chemicals besides ethylene oxide,
including ethylene dichloride, ethylene chlorohydrin, and ethylene
itself.
Estimation of exposure in this cohort is difficult. Subcohorts of full-
time exposed, intermittently exposed, and unexposed individuals were
identified, but the level of exposure or duration of exposure is never
fully documented. During the mid-19^0'8 (19<»1-19<*7) it is believed that
2-H9
-------
ethylene oxide TWA exposures were somewhat less than 25 mg/m3. From
1950 through 1963, increased ethylene oxide production entailed expo-
sures between 10 and 50 mg/m3. Production of ethylene oxide terminated
iti 1963. It continued to be used in manufacture; the concentrations
associated with this operation ranged from 1 to 10 mg/m3 in the 1970's.
Since the cohort members must have been employed in 1960-61, the values
of ethylene oxide concentration from 1950 to 1963 (10-50 mg/m3) are most
important in determining exposure. Periods before and after were
associated with less exposure. Consequently, a value of 20 mg/m3 is
assumed for an average TWA exposure among the full-time exposed; half
that value, 10 mg/m3, is assumed for the intermittently exposed group.
Hogstedt et ol. (191) state that the majority of exposed persons were
employed before 1950. Evidently, most had more than ten years of
employment. Nothing more is Known about duration of exposure to
ethylene oxide. An average length of employment of 10 years is assumed.
The resulting best estimate of cumulative exposure is
(20 mg/m3) x (10 years) • 200 mg-yrs/m3
for the full-time exposed group and 100 mg-yrs/m3 for the intermittently
exposed group. The dose and response data are displayed in Table 2-66.
The bounds on dose indicated in that table are based on a and i values
derived with reference to the following considerations:
1. The average length of employment is not known, nor is a range
of durations presented. The value assumed, ten years, is based
solely on the fact that most of the workers were employed
before 1950 and must have worked until 1960-61 to be members of
the cohort. That being the case, the lower bound ought to oe
relatively closer to the best estimate, based on this feature
2-150
-------
at least, than the upper bound, but both should be fairly wide.
The value for a-) is 0.5 and for 7-) is 0.8.
2. The measurements of exposure are rather incomplete. This has
made it difficult to estimate exposures appropriate to the
full-time exposed group. It is not entirely clear when cohort
members were exposed. Both aj and 73 have been set equal to
0.5, the maximum value allowed to reflect incompleteness of
measurements.
I. For both exposed groups there is uncertainty with respect to
the applicability of the reported levels of atmospheric
ethylene oxide because the exact timing of exposure is not
known and because area samples may not be the best method to
determine effective, personal exposures. In the intermittently
exposed group there is an additional uncertainty: it is not
known to what extent intermittently-exposed individuals are
exposed. Merely halving the full-time exposed estimate may not
be accurate or appropriate. For the full-time exposed group,
ag and 7g equal 0.1 whereas ag and 7g are 0.3 for the
intermittently exposed group.
k. This cohort was exposed to several other chemicals besides
ethylene oxide. Ideally, expected numbers of deaths would
reflect these exposures. This uncertainty is in addition to
that corresponding to use of national mortality rates.
Overall, 09 and 73 hove been set equal to 0.2.
The uncertainty factors resulting from these considerations depend on
the dose group. For the full-time exposed group, a - 2.3 and 7 - 2.6.
For the intermittently-exposed group, a - 2.5 and 7 - 2.8.
The potency parameter estimates from this cohort are presented in Table
2-67, along with those from the first study discussed. The RRD esti-
mates derived from those potencies are displayed in Table 2-68.
2-151
-------
Table 2-66
DOSE AND RESPONSE DATA FOR ETHYLENE OXIDE-
EXPOSED EMPLOYEES; HOGSTEDT ET AL. (191)
All Malignant Neoplasms Leukemia
Dose Group Observed Expected Observed Expected
Unexposed 1 2.0 0 0
(0, 0. 0)a
Intermittently 3 3.4 1 0.13
Exposed
(40. 100, 280)
Full-time 9 3.4 2 0.14
Exposed
(87, 200, 520)
°In parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for exposure (mg-yrs/m5) in the dose groups. These have
been estimated; see text.
2-152
-------
Table 2-67
EThYLENE OXIDE LEUKEMIA POTENCY PARAMETER ESTIMATES
Dose Potencies ((mq-yrs/m^. 1)
Measure Lower Limit0 MLE Upper Limit0
Study
Hogstedt
«t al.
(190)
(one dose
group)
Upper
Bound
Best
Estimate
Lower
Bound
1.8*>E-2
4.73E-2 9.<»5E-2
3.32E-2 8.52E-2" 1.70E-1
5.927E-2 1.5?r-1 3.C«fE-l'
Hogstedt
• £ al.
(191)
(Chi-squared
Upper
Bounds
Be»t
Estimates
Lower
Bounds
9.38E-3" 2.50E-2 5.06E-2
2.49E-2 6.66E-2" 1.35E-1
5.86E-2 1.57E-1 3.18E-1'
confidence limits or a shown.
An asterisk marks tne parameters (i.e.
derive RRDs for each study.
. 0, und 0U) used to
2-153
-------
Table 2-68
RRD ESTIMATES FOR ETHYLENE OXIDE (mg/m3)"
Level of Extra Risk
Estimation 10~6 0.25
Study _ Method RRDi _ niE _ RRDU _ RRD| MLE _
Hogstedt 1 9.07E-6 3.2<»E-5 1.50E-4 2.27 8.09 3.74E+1
et al.
(190) 2 1.03E-5 3.66E-5 1.69E-* 3.<»3 1.22E+1 5.67E+1
Hogstedt 1 8.6flE-6 <».UE-5 2.9'»E-<» 2.17 1.04E+1 7.35E+1
et al .
(191) 2 9.81F-6 4.68E-5 3.32E-4 3.28 1.57E-f1 1.11E+2
QBosed on the risk of leukemia.
-------
Isoniozid
Isonicotinic ocid hydrozide or isoniazid (INH; CAS No. 54-85-3) was
introduced in 1952 as a chemotherapeutic agent to combat tuberculosis
(192). It is now the most widely used antituberculosis drug, given
alone in preventive therapy and in combination with other drugs for
treatment of the active disease (193).
Isoniazid has not been clearly linked to cancer in humans. INH has
induced DNA repair in bacteria. In mice treated early in embryogenesis,
isoniazid caused specific-locus mutations and in host-mediated assays in
rodents, it caused bacterial mutation, due partly to the formation of
hydrazine (2). Apparently, human metabolism of INH does not produce
hydrozine, an agent strongly suspected of being a carcinogen (194).
There was no evidence of DNA damage in lymphocytes of patients receiving
isoniozid (2).
Nevertheless, several investigations of a possible relationship between
isoniazid and cancer have been reported. Hammond et al. (195) consi-
dered this question a mere 14 years after isoniazid's introduction. Of
particular interest is their report on 311 tuberculosis patients treated
with isoniazid for various lengths of time [55 for less than 1 year; 97
for 1 to 2 years; 87 for 2 to 3 years; 29 for 3 to 4 years; 11 for 4 to
5 years; 9 for 5 to 6 years; 20 for 6 years or longer; 3 uncertain].
Using midpoints of those intervals (assuming 10 years average treatment
in the six-years-or-longer group), average duration of INH therapy was
2.57 years. The standard daily dose was 4 mg INH per kg body weight.
The average total dose for these patients is estimated to be 3752.2
mg/kg, equivalent to a cumulative dose of 10.3 mg-yrs/kg.
Possibly as many as 10 cancer cases were seen in these patients. Five
(and a probable sixth) were cancers of the respiratory tract, and there
2-155
-------
was one each of bladder cancer, chronic myelogenous leukemia, liver
cancer, and a death with a "hilar density suggesting tumour". Even
accepting the 10 as true cancer cases, the observed number is not
significantly greater than the 6.3 expected from local rates. Neverthe-
less, these values can contribute to the estimation of the quantitative
relationship between isoniazid use and cancer development, assuming one
exists.
Uncertainty associated with this study, leading to factors a and 7 both
equal to 1.3, is caused by two factors:
1. Length of exposure is categorized by one-year groups, but those
groups ore presented without average values. A small factor,
0.1, is assigned to of and 71.
2. It has been suggested (192) that there may exist an association
between tuberculosis itself and lung cancer. Hammond »t al.
failed to detect such a link in the tuberculosis patients
included in their prospective study. Nevertheless, this adds
some uncertainty to the expected number of deaths that Hammond
et al. present, so
-------
derive quantitative estimates and bound* on risk, even though the
evidence for carcinogenicity is negative. Arithmetic averages c.re
assumed to represent the overage dose of each group (cf. Table 2-69) and
these values can be used in the dose-response analysis. The uncertainty
in the dose values stems from two sources.
1. The dosage groups are presented without average values. The
factors ag/kg/doy. f°r ° total of 1825 mg/kg (5 mg-yr/kg). No particular
increase in all cancer or cancers of specific sites was noted among the
INH-treotment groups when follow-up was continued for 11 to 1
-------
(1/12595) x 12439 - 0.99.
Altogether, 107.6 cancer deaths ore expected during the fourteen years
for which results are available.
Uncertainty leads us to estimate bounds on cumulative dose of 3.8 and
6.25 mg-yrs/kg, based on the factors a • 1.3 and 7 - 1.25, respectively
derived as follows:
1. Medication was taken for at most 1 year. Some participants
stopped taking their medication, but we have no indication of
the overage duration of INH exposure. A factor of 0.05 is
assigned to oc|, but since the maximum duration is known, -7-1- 0.
2. It is stated that the daily dose of INH was between
-------
not linked to cancer but that tuberculosis was associated with lung
cancer, both of which are associated with cigarette smoking. Additional
negative evidence on the carcinogenicity of INH is provided by Howe
*>t ol. (201). These authors found no difference in cancer incidence or
j.iortality for a large group of Canadian tuberculosis patients. A year
later, Boice and Froumeni (202) again could discern no link between
isoniazid therapy and cancer. Finally, Costello and Snider (193).
reviewing a Public Health Service preventive therapy trial, could find
no evidence of an etiologic role of isoniazid in cancer development.
The Puerto Rican participants in the household contact trial (cf. Table
2-70) were followed for an average of 18 years.
Overall, the evidence does not support a conclusion that isoniazid is
carcinogenic in humans. The studies with quantitative data all provide
potency parameter estimaets that are positive (although the lower limits
for the Glassroth »t ol. study are negative), indicating some carcino-
genic potential for INH (Table 2-72). The RRD estimates derived from
those parameters, and assuming 45 years of exposure to isoniazid
starting at age 20, are displayed in Table 2-73.
2-159
-------
Table 2-69
>
DOSE AND RESPONSE DATA FOR TUBERCULOSIS
PATIENTS TREATED WITH ISONIAZID°
Total
Dose of
Isoniozid(q)
All Malignant
Neoplasms
Observed Expected
Malignant Neoplasm*
of Respiratory Tract
Observed Expected
<50
(17.9. 25. 35)b
50 - 99
(53.8. 75. 105)
100 - 199
(107, 150. 210)
200+
(357. 500, 700)
31
2
-------
Toble 2-70
CANCER DEATHS AMONG HOUSEHOLD MEMBERS OF
TUBERCULAR PATIENTS BY TREATMENT
CROUP AND YEAR OF OBSERVATION0
Plocebo
Year of
Observation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
TOTAL
°from Glassroth
Population
at Rick
12594
12568
12518
12484
12
-------
Table 2-71
CANCER DEATHS AMONG MENTAL INSTITUTION
TUBERCULAR PATIENTS BY TREATMENT
GROUP AND YEAR OF OBSERVATION0
Placebo
Year of
Observation
1
2
3
-------
Table 2-72
POTENCY PARAMETER ESTIMATES FOR ISONIAZID
Potencies (mg-yrs/kg)-'1
Dose Lower Upper
Study Response Measure Limit0 MLE Limit0
Hammond All Upper 2.08E-3" 4.38E-2 9.85E-3
et al. Malignant Bounds
(195) Neoplasms
(0 degrees Best 2.70E-3 5.70E-2" 1.28E-1
of freedom) Estimates
Lower 3.52E-3 7.*»3E-2 1.67E-1"
Bounds
Stott All Upper 1.<»2E-3" 1.37E-2 2.80E-2
et al. Malignant Bounds
(196) Neoplasms
(chi-squored Best 1.96E-3 1.92E-2* 3.92E-2
(3) - 7.9) Estimates
Lower 2.79E-3 2.69E-2 5.50E-2"
Bounds
Respiratory Upper 1.61E-3* 2.30E-2 5.00E-2
Cancer Bounds
(chi-squared
(3) - 5.9) Bast 2.22E-3 3.21E-2" 7.00E-2
Estimates
Lower 3.18E-3 4.50E-2 9.82E-J*
Bounds
Glass- All Upper -1.8<»E-2" 5.95E-<» 2.12E-2
roth Malignant Bounds
et al. Neoplasms
( 197) (0 degrees Best -2.30E-2 7.i»3E-*»" 2.65E-2
of freedom) Estimates
Lower -3.03E-2 9.78E-I* 3.49E-2'
Bounds
°901t confidunce limits are shown.
"An asterisk marks the parameters used to derive RRD estimotos.
2-163
-------
M
I
Table 2-73
RRD ESTIMATES FOR ISONIAZID (mg/kg/doy)
Estimation
Study
Hammond
et al.
(135)
Stott
et al.
(I5fi)
Gloss-
roth
et al.
(1SZ)
Response
All
malignant
neoplasms
All
malignant
neoplasms
Respiratory
cancer
All
malignant
neoplasms
Method
1
2
1
2
1
2
1
2
RRDL
4.71E-7
6.20E-7
1.(,IE-6
1.88E-6
2.77E-6
3.20E-6
2.26E-6
2.97E-6
Level of
10~6
MLE RRDU
1.38E-6 3.79E-5
1.82E-6 t.99E-5
V11E-6 5.55E-5
5.41E-6 7.31E-5
fa.«»8E-6 1.69E-4
9.78E-6 1.95E-4
1.06E-U a,
1.39E-4 «
Extra Risk
RRDL
1.18E-1
1 98E-1
3.58E-1
6.03E-1
6.93E-1
9.85E-1
5.64E-1
9.50E-1
0.25
MLE
3.45E-1
5.82E-1
1.03
1.73
2.12
3.01
2.65E+1
4.46E+1
RRDU
9.
-------
Melpholon
Melphalon (CAS No. 1^8-82-3) is a chemical used to treat various malig-
nant diseases, especially multiple myeloma, malignant melanoma, and
adenocarcinomas of the ovary (205). It is an alkylating agent and hence
highly suspect as a carcinogen. Melphalan is mutagenic in bacterial
tests, has induced chromosomal aberrations and sister chromatid
exchanges in mammalian cells, and has produced chromosome damage in the
lymphocytes of human patients treated therapeutically (2).
Numerous case reports have documented second malignancies in patients
whose first malignancy has been treated with melphalan. Gori «t al.
(20jf) describe a case of acute myeloblastic leukemia following treatment
of breast cancer with 300 tig of melphalan given over 72 weeks (12
courses of 5 mg/day for 5 days every 6 weeks). A cytoaenetic study of
this patient revealed chromosomal damage including the warn* tronsloca-
tion in 6 out of 8 karyotype analyses.
Law and Blom (205) noted 7 patients out of 57 with multiple myeloma who
had second neoplasms (<» with acute leukemia, 3 with solid-organ tumors).
All were treated with melphalan, although these authors speculate that
multiple myeloma itself may be a risk factor for other neoplasms.
Einhorn (206) described four cases of acute leukemia among i»7<» ovarian
cancer patients treated with melpholon. Other chemotherapy or radio-
therapy accompanied melphalon treai..-"<9nt in some cases.
Greene »t al. (121) review five clinical trials testing alkylating agent
therapy (with or without irradiation) of ovarian cancer. A total of
1399 women were included in these trials which tested 5-fluorouracil,
acinomycin-D, cyclophosphamide, and chlorambucil in addition to
melpholan. Eight different treatment regimens included melphalan
2-165
-------
therapy, sometimes with irradiation or other alkylating agents. Among
the 773 women in those treatment regimens, 10 acute nonlymphocytic
leukemios were observed whereas only 0.08 were expected.
Greene et al. report the leukemia pattern by total dose for two of the
trials that used melphalan alone (Table 2-74). We have selected the
groups that have not received any radiation therapy to investigate the
dose-response behavior of melpholan and we have converted total dose to
cumulative dose by dividing by 365 (Table 2-75).
Average doses assumed for the groups displayed in Table 2-75 are the
midpoint of the low dose group and on arbitrarily obtained value of 3.30
mg-years (1200 mg) for the high dose group. Some uncertainty is
associated with those estimates, as discussed below:
1. The dose groups presented are fairly wide and are presented
without averages. To account for this a^ and 7^ for the low
dose group are set equal to 0.2, whereas a^ and 7^ for the high
dose group are 0.3.
2. In this morbidity study, expected numbers of cases were deter-
mined from the Connecticut Tumor Registry. Some uncertainty as
to use of these numbers (as opposed to numbers based on
leukemia seen among those with ovarian cancer, for example)
leads us to select a value of C.1 for og and 73.
The group-specific uncertainty factors (a • -7 • 1.3 for the low dose
group, a - 7 • 1 . <» for the high dose group) determine the bounds on dose
given in Table 2-75. These ore used to determine sensitivity of the
analysis to uncertainty in the quantitative estimates.
Table 2-76 presents the potency parameters estimated from the ocutci
nonlymphocytic leukei.na response data in the melphalan-treated groups.
2-166
-------
The corresponding RHD estimotes, based on the exposure scenario of o
whit* mole being exposed for i»5 yeart starting at age 20, are given in
Tob'e 2-77
2-167
-------
Table 2-74
OBSERVED AND EXPECTED CASES OF ACUTE NONLYMPHOCYTIC
LEUKEMIA BY TRIAL AND INITIAL CHEMOTHERAPY DOSEa
Trial and
Treatment Regimen^
Number of
Patients
Observed
Expected
M.D. Anderson Hospital:
Pelvic/abdominal irradiation 89
< 700 mg melphalan 41
> 700 mg melphalan 42
0.019
0.015
0.011
Gynecologic Oncology Group
Trial 1 :
Observation after surgery
Pelvic irradiation
< 700 mg melphalan
> 700 mg melphalan
68
55
43
29
0
0
0
1
0.011d
0.008d
0.007
0.007
QFrom Greene et al. (121 ).
bThe treatment regimens with melphalan. are specified as follows: for the
M.D. Anderson Hospital trial, 1 mg/kg every 4 weeks for 12 cycles; for
the Gynecologic Oncology Group 1 trial, 1 mg/kg every 4 weeks for 18
cycles.
cThis patient was subsequently treated extensively with melphalan.
dThese expected values are approximated: the authors had combined the
expected values for the two non-melphalan regimens in this trial.
2-168
-------
Table 2-75
DOSE AND RESPONSE DATA FOR MELPHALAN-
TREATED AND CONTROL PATIENTS0
Cumulative Dose (mg-yrs)
Observed
Leukemia Cases
(0.74. 0.96, 1.25)b
Expected
Leukemia Cases
0
< 1.92
0
0
0.011
0.022
> 1.92
(2.36, 9.30, it.62)
0.018
°From Greene et ol. (121 ).
bln parentheses are the reasonable lower bounds, best estimates, and
reasonable upper bounds for average dose in each group.
Table 2-76
LEUKEMIA POTENCY PARAMETER ESTIMATES FOR MELPHALAN,
BASED ON THE STUDY OF GREENE ET AL.a
Dose
Measure Lower Limit*5
Potencies ((mg-yrs)"'1 )
MLE
Upper Limitb
Upper
Bounds
2.23E+1
3.47E+1 7.93E+1
Best
Estimates
3.09E+1
1.08E+2
Lower
Bounds
I».27E + 1
6.79E+1 1.i»8E>2'
°The fit of the model to the data is adequate; chi-squared (2) < 2.65.
b90£ confidence limits are shown.
"An asterisk marks the parameters used to derive RRD estimates.
2-169
-------
Toble 2-77
RRD ESTIMATES FOR MELPHALAN (mg/doy)°
Level of Extra Risk
Estimation _ 10~° _ _ 0.25 _
Method RRDi _ MLE _ RRDU _ RROi _ MLE _ RRDU
1 1.28E-8 3.88E-8 8.<»6E-8 3.19E-3 9.71E-3 2.11E-2
2 1.^E-8 «*.37E-8 9.52E-8 4.76E-3 Ht5E-2 3.16E-2
°Based on the risk of leukemia morbidity estimated from Greene et al .
(121).
2-170
-------
Methylene Chloride
The halogenoted hydrocarbon methylene chloride (dichloromethane, DCM) is
a nonflammable solvent used in a number of industrial capacities inclu-
ding degreasing; paint stripping; manufacture of photographic film,
textiles and plastics; and food additive extracting. Infrequent reports
of adverse effects in humans have appeared since 1936. Those effects
include headaches, incoordination, and irritability among others, but no
systematic investigation of toxic reactions is available (207). Pharma-
cokinetic analyses of DCM are available, however, including a report by
Riley et ol. (208) that postulates the involvement of only two tissues
(water and fat) in methylene chloride metabolism. More recently, it has
been found that two metabolic pathways exist, one involving production
of carboxyhemoglobin (209). DCM is mutagenic in prokaryote and insect
cells, but apparently not in mammalian cells (2).
Two recent occupational cohort studies provide information on the carci-
nogenicity of DCM. Both report negative findings: no increased risk of
cancer appears to be associated with methylene chloride exposure.
Nevertheless, risk estimates can be derived from these reports.
Friedlander et al. (207) describe a cohort of 751 hourly, male employees
of one department in a facility using DCM as the primary solvent since
the 1940's. Cohort members were those employed in 1964 and were
followed through 1980. An internal control group composed of hourly,
male employees at the same facility who were not exposed to methylene
chloride was used to calculate expected numbers of deaths. A total of
24 malignant neoplasms were observed in the cohort versus 28.64 expected
(210).
The authors of this study believe that the members of the cohort were
all exposed to similar levels of DCM. The fact that the employees in
2-171
-------
the DCM department all had similar tasks, that deployment in the depart-
ment was flexible, that methylene chloride is highly volatile, and that
the process did not change substantially for the 30 years prior to 1976
contribute to this belief. Area and personal DCM concentration measure-
ment, along with carboxyhemoglobin determinations, support estimated
time-weighted average exposures of between 30 and 125 ppm. An average
of 78 ppm is assumed for the following calculations. The members of
this cohort averaged about 26.5 years of exposure (Hearne, personal
communication). Consequently, the estimated average cumulative exposure
for the cohort is
(78 ppm) x (26.5 years) • 2067 ppm-years.
The uncertainty associated with this estimate leads to uncertainty
factors, a and 7, equal to 1.6 and 1.85, respectively, based on the
following considerations:
1. Although the process has remained relatively stable for many
years, some changes in the ventilation system did occur. The
authors suggest that exposures prior to 1959 may have been
somewhat higher than those used to derive our average exposure
estimate. Consequently, 73 ' 0-2 ond a2 " "•
2. The measurements of exposure since 1959 are fairly spotty,
occurring in 1959. 1966, 1973 and 197^ for a total of 307
samples. The exposures measured varied from 0 to 350 ppm,
indicating that the variability might have been better docu-
mented with more extensive measurement. A factor of 0.3 is
assumed for 013 and 73.
3. The overage exposure is not given, merely a range of 30 to 125
ppm as a reasonable interval for exposure. Since we had to
estimate an average, a factor of 0.2 is assigned to a^ and 74,
2-172
-------
4. Most of the measurements that determined the range of possible
exposures were area rather than personal samples. A nominal
value of 0.05 is assumed for a5 and 75 to account for this
feature. Similarly, some of the measurements were based on
carboxyhemoglobin determinations which were converted to esti-
mates of DCM exposure on the basis of experimental studies. In
this case 07 and 77 ore equated to 0.1 to cover the uncertainty
in that conversion.
The reasonable bounds on dose obtained from these considerations are
1253 and 382'* ppm-years. These are used to test the sensitivity of the
analysis to unknown features of this cohort.
The second occupational study of methylene chloride's health effects is
reported by Ott et al. (209, 211). These authors identified a plant
using DCM as' a solvent since 1954 in the production of cellulose triace-
tate fibers and a second plant that had similar production characteris-
tics but did not use DCM. The cohort studied consisted of employees who
worked at least three months in the preparation or extrussion areas of
either plant between 1954 and 1977. A total of 1271 DCM-exposed and 948
control workers were identified. Follow-up extended through June 1977.
Among the white male or female employees, 7 malignant neoplasms were
observed in the exposed group (11.5 expected on the basis of U.S.
national rates) and 7 in the control group (12.3 expected). Both
exposed and unexposed workers came into contact with acetone and had
other minor exposures.
Jobs were categorized by exposure to methylene chloride into three
groups: low (averaging 140 ppm, TWA), moderate (280 ppm), and high (475
ppm). Unfortunately, all we know about the distribution of the cohort
in these categories is that 25* of the workforce was assigned to the
high-exposure area and that most encountered only low DCM concentra-
2-173
-------
tions. If we assume that 60* had low exposures and 15* moderate
exposure, then the average exposure is
(.25 x 475) + (.15 x 280) + ( . 60 x 140) - 245 ppm.
Duration of exposure was similarly categorized, as follows:
less than 1 year (assumed average 0.5 yrs): 218 employees;
1 to 4 years (average 3 years): 496 employees;
5 to 9 years (average 7.5 years): 207 employees;
more than 10 years (average 17 years): 352 employees.
These numbers lead to an estimate of average length of exposure of
(218 x 0.5) + (496 x 3) +.. (207 x 7.5) + (552 x 17) - 7.2 years.
The average cumulative exposure estimate associated with this cohort is
245 ppm x 7.2 years - 1764 ppm-years.
The uncertainty associated with this estimate is reflected in the
factors a - 2.05 and i - 2.45, which lead to reasonable bounds on dose
of 860 and 4322 pprr-yeors. The following features of the study
contributed to the uncertainty of the estimate:
1 . Length of exposure was estimated for the entire cohort
combined, based on a categorization presented without average
values for the categories. Consequently, both a-) and TJ are
assumed to be 0.3.
2. Methylene chloride concentrations were measured extensively,
but only in 1977 and 1978. Nothing is recorded about concen-
trations that might have prevailed earlier in the employment
2-174
-------
history of the cohort. Nor is there any information about
process changes that might have affected exposures. To account
for this uncertainty, 02 • O.'f and 12 is assigned a value of
0.8.
3. The distribution of the cohort over three exposure classes is
not completely known, making the calculation of average expo-
sure for the entire cohort difficult. As a result, a^ and -74
are equal to 0.3.
-------
Table 2-78
POTENCY PARAMETER ESTIMATES FOR METHYLENE CHLORIDE0
Potencies ((ppm-yrs)~1 )
Study
Friedlander
et al. (207)
(0 degrees of
freedom)
Ott et al.
(209. 211 )
(chi-squared
(1 ) . ^.0*)
Dose
Measure
Upper
Bounds
Best
Estimates
Lower
Bounds
Upper
Bounds
Best
Estimates
Lower
Bounds
Lower
Limit" MLE
-1.05E-**" 0.00
-1.9i»E-4 0.00"
-3.19E-4 0.00
-1.70E-4" 0.00
-^.16E-*» 0.00"
-8.53E-<» 0.00
Upper
Limitb
3.53E-5
6.53E-5
1.08E-<»"
3.79E-5
9.28E-5
1.90E-1*"
°Based on the risk of all malignant neoplasms.
bgo£ confidence limits shown.
*An asterisk marks the parameters used to derive RRD estimates.
2-176
-------
Table 2-79
RRD ESTIMATES0 FOR METHYLENE CHLORIDE (ppm)
Level of Extra Risk
Estimation 10~6 0.25
Study Method RRDi MLE RRDU RRDi MLE RRDu
Friedlander 1 7.31E-4 » °° 1.83E+2 » =c
fit al .
(207) 2 9.62E-4 « oo 3.08E+2 » oc
Ott 1 ^.I'tE-'v uo oo 1.03E+2 °°
fit ol .
(209. 2 S.'f^E-'* o= o= 1.:
(211)
°Based on the risk of all malignant neoplasms.
2-177
-------
Nickel
Nickel is a silvery metal found in nature as the ores milerite (sulfide)
and garnierite (silicate), the latter being the most important commer-
cial" y (212). Nickel is liberated via conversion to the subsulfide,
Ni3S2, which is air-roasted to give nickel oxide, NiO, followed by
carbon reduction to the metal. Nickel is used chiefly in the production
of alloys, including stainless steels. It is also used in electropla-
ting, catalysts, coinage and pigments. Nickel alloys have been used in
jewelry, and in dental and surgical prostheses (213).
Various nickel compounds have been examined in short-term test Nickel
carbonyl has been found to inhibit DNA-dependont RNA polymerase activi-
ty, probably by binding to chromatin or DNA (214). In vitro ter,ts using
mammalian cells have shown nickel compounds to inhibit cellular uptake
of thymidine-^H and to induce chromosomal aberrations, somatic muta-
tions, and morphological transformation. Bacterial mutagenicity tests
have been uniformly negative.however (215).
Most of the epidemiological data on the carcinogenicity of nickel has
resulted from studies of occupationally exposed individuals, primarily
nickel refinery workers. Although a number of studies hove shown an
increased risk of respiratory cancer to be associated with work in
nickel refineries, there is no clear consensus as to which of the nickel
compounds is implicated. Increased risk of lung and nasal sinus cancer-
was first noted among nickel refinery workers in Clydach, South Wales
(216-218). Initially, the increased risk was thought to be associated
with nickel corbonyl. However, the risk of respiratory cancer dropped
dramatically after precautions were taken against exposure to du.it in
the refining process, although exposure to nickel carbonyl gas
continued. Norseth (219) claims that the slightly soluble nickel salts
ore the important occupational carcinogens.
2-178
-------
Studies of refinery worker* ot the Port Colborne ond Copper Cliff
sintering plants in Canada demonstrated increased risks of respiratory
cancer, particularly of the lung ond nasal sinuses (220-222). The
sintering operation produced dust containing nickel sulfides and oxides.
Pedersen »t ol. (223) found an increased risk of respiratory cancer,
including lung, nasal sinus and larynx, among workers at the
Falconbridge refinery in Norway. The highest risk was associated with
workers involved in roasting, smelting and electrolysis. A later study
by Kreyberg (22j*) confirmed that exposure to nickel dust and fumes was
associated with increased risk of lung cancer. Further confirmaticn was
supplied by Magnus »t al. (225). who also controlled for smoking habits
of workers. They concluded that the interaction of smoking and nickel
exposure with respect to lung cancer is closer to additive than to
multiplicative.
Lessard et al. (226) noted an increased risk of lung cancer for workers
in a nickel smelter in New Caledonia in the South Pacific, where the
refining of nicke-1 has been conducted for more than a century. They
concluded thot cigarette sroking, for which their analysis controlled.
was an important factor to consider, but that previous studies linking
nickel exposure to lung cancer hod rarely done so.
Several studies of occupational exposure to nickel other than in
'"fining operations have been carried out with mixed results.
.Iverstein et al. (227) found a significant excess of lung cancer among
••kors involved in die-casting and electroplating. However, Bernacki
,)J. (228) found no apparent increase in risk of lung cancer mortality
T workers in an aircraft engine factory, and Sodbold and Tompkins
found no evidence of increased risk of mortality due to respiro-
• -cor among nickel-exposed workers in a gaseous diffusion plant.
2-179
-------
A case-control study (230) implicated nickel exposure as a cause of
laryngeal cancer. In studying the mortality of men employed in a plant
manufacturing nickel alloys in England, Cox et al. (251) found no evi-
dence of an occupational hazard in men exposed regularly to atmospheres
containing on average between 0.04 and 0.84 mg/m' nickel and nickel
oxide. Although the study by Cox et al. included limited information on
exposure, this data could not be incorporated into a quantitative
analysis.
Norseth (219) noted the general lack of dose estimates for nickel expo-
sure both in the refining industry and in other industries where expo-
sure to nickel compounds occurs. He expressed the need for careful dose
registration before a quantitative cancer risk analysis can be
performed. In the present literature review, only two cohort studies
were found for which estimates of exposure to nickel could be derived.
The investigation of Polednak (232) involved welders at Oak Ridge
nuclear facilities who were employed between 1943 and 1973. Follow-up
ended on January 1, 1974. A total of 1059 white; male welders were
included, but our interest is focussed on 536 welders working at the
Gaseous Diffusion Plant who have some data on nickel oxide exposure
intensity and duration. Table 2-80 displays the TWA concentrations of
nickel associated with the welding operations conducted at the plant.
The operation specifically mentioned as a major part of the work at the
plant is welding of nickel-lined pipe. Consequently, we will assume the
average TWA exposure for the workers to be the level associated with
that procedure, 0.13 mg/m3, which is also the median of the values
displayed.
Length of exposure has been used to define two subgroups: those exposed
for less than 50 weeks and those exposed for 50 or more weeks. Polednak
states that those exposed for longer periods were employed for 104 to
378 weeks. We have assumed average durations of 25 and 241 weeks for
2-180
-------
the two groups. Coupled with the nickel concentration estimates
discussed above, these figures yield cumulative exposure groups as
displayed in Table 2-81.
Several uncertainties contribute to the factors a-2.0 and 7-2.8 that
determine the bounds on dose seen in Table 2-81. These are discussed
below:
1. Length of exposure is grouped very crudely and presented
without average values. Both a-| and TJ are assigned the value
0.3.
2. All the nickel concentration measurements were performed after
1975. Polednak states that these "provide a lower limit for
estimation of levels in earlier years". As a consequence
a2 • 0, but 12 ' 0.8.
3. It is not known how many personal samples were used to detei—
mine the exposure estimates. That being the case, ex3 and 73
both equal 0.3.
4. The range of possible exposures was 0.04 to 0.57 mg/m3. No
average value was presented. To cover the uncertainty of
estimating an average value, 04 and 7^ have been set equal to
0.3.
5. Respiratory cancer is the endpoint of interest. Hence, smoking
behavior undoubtedly influences calculation of actual expected
values. Polednak presents data suggesting that the proportion
of heavy smokers in the subcohort analyzed here was much the
same from 1950 to 1969 as that found in all U.S. white males,
the background population used to estimate expected values. To
account for other smoking differences and for use of national
rather than local rates of death,
-------
Enterline and Marsh (235) conducted a study of workers in a nickel
refinery in Huntington, West Virginia. A cohort of 185 men employed
between 1922 end 19
-------
state that none of the group? displayed significantly increased respira-
tory cancer mortality rotes. The overage exposures for the groups were
presented by Enterline and Marsh, but the bounds are based on the
following considerations:
1. Measurements of exposure early in the exposure period are
limited. Extrapolations backward from the plentiful, recent
measurements took into consideration changes in the process and
environmental controls. Nevertheless, some uncertainty persists
and is reflected in our choice of o<2 • 0.1 and 70 " 0.3.
2. Area samples, as opposed to personal samples, formed the basis
of the departmental exposure estimates. A fairly small value
of 0.1 is assigned to ag and 75 to account for possible differ-
ences between area and personal exposure.
3. Once again, smoking status could not be used to calculate
expected numbers of respiratory cancer deaths. Local rates of
death were used in those calculations, however. Both «Q and -73
have been assigned a value of 0.05.
The resulting uncertainty factors are a • 1.25 and 7 • 1.45.
Table 2-84 displays the potency parameter estimates obtained from the
Polednok (232) and Enterline and Marsh (235) studies. The best esti-
mates of and reasonable bounds on the potencies have been used to esti-
mate RRDs. Note that the negative lower bound on potency from the
Polednak study entails an infinite upper bound on the RRD estimate, i.e.
this study is consistent with the hypothesis that nickel is not carcino-
genic in humans.
2-183
-------
Table 2-80
CONCENTRATIONS OF NICKEL FROM PERSONAL AIR SAMPLERS
WORN BY WELDERS AT THE OAK RIDGE GASEOUS
DIFFUSION PLANT, POLEDNAK (252)°
TWA Average Air
Welding Procedure Concentrations(mg/m3)
MIG welding on Ni- 0.57
plated steel
TIG welding on Ni- 0.04.
plated steel
SMA welding on Ni- 0.13
lined pipe
MIG welding on carbon 0.25
steel
TIG welding on stainless 0.08
steel
aMeasuremerits are from 1975-1977.
Table 2-81
DOSE AND RESPONSE DATA FOR NICKEL-EXPOSED
WORKERS STUDIED BY POLEDNAK (232)
Cumulative Exposure Respiratory Cancer Deaths
(mq-vrs/m3) Observed Expected
0.0625 2 3.23
(0.0312, 0.175)a
0.602 5 2.86
(0.301, 1.69)
aln parentheses are the lower bounds and upper bounds for cumulative
exposure in each dose group.
2-184
-------
Table 2-82
OBSERVED AND EXPECTED DEATHS FOR 3 GROUPS OF MALE NICKEL
WORKERS, 20 YEARS OR MORE AFTER FIRST EXPOSURE0
Hired before 1947
Refinery Nonrefinery
Cancer Type 0 E 0 E
All Malignant
Neoplasms 27 19.22 133 131.24
Respiratory
Neoplasms 10 7.55 49 46.62
Hired after 1946
0 E
9 12.80
4 6.26
°From Enterline and Marsh (233).
Table 2-83
DOSE AND RESPONSE DATA FOR NICKEL-EXPOSED
COHORT STUDIED BY ENTERLINE AND MARSH (253)°
Cumulative Exposure
(mg-yrs/m')
< 0.83
(0.28. 0.35, 0.51)b
0.83 - 4.17
(1.83, 2.29, 3.32)
4.17 - 16.7
(6.63, 8.29, 12.0)
>16.7
(37.6, 47.0, 68.2)
Respiratory Cancer Deaths
Observed Expected
11 15.96
28 27 . 87
20 14.14
4 2.48
aCumulative exposure is calculated for the first twenty years of
employment and mortality is for the period starting at 20 years from
first employment.
bln parentheses are the lower bounds, best estimates, and upper bounds
for exposure in each dose group.
2-185
-------
Table 2-84
NICKEL RESPIRATORY CANCER POTENCY PARAMETER ESTIMATES
Dose Potencies ((mq-yrs/m^)"1 )
Study Measure Lower Limit0 MLE Upper Limit0
Polednak Upper -6.71E-2" 3.9
-------
Table 2-85
RRD ESTIMATES0 FOR NICKEL (mg/ffi3)
Level of Extra Risk
Estimation 10~60.25
Study Method RRDi MLE RRDU RRDi MLE RRDU
Polednak 1 4.54E-8 2.^6E-7 oo 1.14E-2 6.15E-2 °°
(232)
2 5.25E-8 2.84E-7 oo 1.61E-2 8.7«*E-2 «
Enterline 1 4.84E-6 1.34E-5 2.69E-4 1.21 3.35 6.71E+1
and Marsh
(253) 2 5.59E-6 1.55E-5 3.10E-4 1.71 ^.76 9.54E+1
QBased on respiratory cancer risk.
2-187
-------
PCB
Commercial preparations of polychlorinated biphenyls (PCBs) such as
Aroclors and Kanechlors are mixtures containing the biphenyl molecule
with a variable number of chlorines substituted at various locations on
the double biphenyl ring. These mixtures are thermally stable and
possess excellent dielectric properties. They were consequently used
extensively for heat transfer, as in electric transformers, and in
capacitors. PCBs have also been used in plasticizers, inks, and for
carbonless duplicating paper (25<»). Production and use of PCBs in the
U. S. has been severely limited, compared to earlier, large-scale use,
since 1979. Older transformers and capacitors still contain PCBs so
those maintaining them can be exposed (235). The stability of PCBs also
ensures their presence in the environment for extended periods.
Although PCBs have been shown to be carcinogenic to rodents when admin-
istered orally, the short-term tests have been largely negative. Only
two of many PCB mixtures tested have been shown to be mutagenic in the
Salmonella test (2).
Acute effects of PCB poisoning have been observed in two incidents of
accidental contamination of cooking oil. The long-term effects, such as
cancer, of low-level, chronic exposure to PCBs are not well documented.
This is so despite the fact that most individuals contain measurable
quantities of PCB in their tissues (236) due mostly to consumption of
contaminated food (257). By far the largest fraction of PCB that is
absorbed is stored in the fat (258), and the half-life for PCB in the
body is quite long.
Marino et al. (239, 240) have carried out an extensive occupational
hygiene and acute health effect study of capacitor workers. PCB blood
levels were determined for both current and post employees. The authors
2-188
-------
state that blood concentrations of trichlorobiphenyls may reflect the
current PCB exposure level more closely than that of pentachlorobi-
phenyls. Cancer follow-up was not available for this cohort.
Bahn at al. (241), in a preliminary report, document 2 malignant mela-
nomas in a group of men thought to be heavily exposed to PCBs. Since
standard rates would suggest that only 0.04 melanomas would be expected,
these authors suggest that close attention be given to investigating any
possible relationship. Melanomic response ties in with the other derma-
tological aspects of PCB exposure, chloracne and the contention of
Moroni et al. (242, 245) that dermal exposure is the primary route of
absorption. No further information is available on the cohort
experiencing the increased rate of melanoma.
Two cohorts have been followed for cancer mortality and also have some
data on PCB exposure. Bertazzi et al. (244. 245) discuss a cohort of
workers employed at a capacitor manufacturing plant. The cohort members
are men and women employed for at least one week between 1946 and 1978.
Mortality was monitored through 1982.
Measurement of atmospheric PCB concentrations was rare. Three determi-
nations in 1954 averaged 6100 pg/m^. The next reported samples were
done in 1977, by which time the concentrations were between 48 and 275
jig/m3. The only otner information relevant to estimation of exposure
pertains to the types of PCBs used: before 1964 the most commonly used
mixtures contained 545t chlorine, from 1965 to 1970 more 42£ chlorine
mixtures were used, and after 1970 only mixtures containing 42* chlorine
were used (until 1980, when PCB use ceased). For cumulative exposure
estimation, we assume an average concentration of 3000 /ig/m'.
The reports by Bertazzi et al. ore limited in several other respects.
The average duration of exposure is not given. We use the default value
2-189
-------
of 7 years. Moreover, mortality experience is broken down by sex, but
the same causes of death are not reported for each sex. The analyses
for this cohort are therefore limited to those cancers documented in
both males and females, i.e. all malignant neoplasms and hematologic
neoplasms (Table 2-86). No malanomas were observed, but digestive
system neoplasms were overrepresented in males.
The cumulative dose determined for this cohort is
(3000 /ig/m^)-(7 years) • 21000
The uncertainty associated with this estimate is considerable. The
uncertainty factors, a and 7, are both equal to 3.3, leading to bounds
on dose of 636^ and 69300 ^g-yrs/m3. The following summarizes the
features influencing uncertainty calculations:
1. Length of exposure is completely unknown. Consequently, a-) and
71 equal 1.5.
2. The plant was in operation, using PCBs, since 19^6. The first
PCB measurements were performed in 1954. This is a rather
short time of undocumented exposure. In this case, o<2 and 72
are set equal to 0.3.
3. PCB concentration measurements are very incomplete. The
average exposure we estimate is based on practically no
information. Both aj and 73 hove been set equal to 0.5.
U . A large uncertainty relates to the applicability of the
atmospheric PCB concentrations. It is strongly suspected that
skin absorption may be the route yielding the bulk of the PCB
dose. In this special circumstance, a value of 1.0 is assigned
to ag and 7g.
2-190
-------
Brown and Jones (255) identified a cohort of 2,567 employees engaged in
the manufacture of electrical capacitors at two plants. Both plants had
been using PCBs for more than 30 years. I rial hygiene measurements
(from 1977) at both plants revealed PCB levels, from personal air
samples, of between 2k and 393 ng/m~> in one plant and between 170 and
1260 ng/m^ in the other. The authors state that the first plant had
recently initiated new production techniques, so that the lower atmos-
pheric concentrations of PCBs were probably not representative of the
exposures experienced by the workers for the majority of their employ-
ment. We assumed that the values reported at the second plant (Table
2-87) are more representative and have used those values for workers at
both plants. The average exposure that is used in the subsequent
calculations is 631
The distribution of the cohort members according to duration of employ-
ment in PCB-exposed jobs is given in Table 2-88. The data in that table
suggest an average length of exposure of about
-------
cancers are not divided by duration of exposure. They are associated
with an overall estimate of cumulative exposure, namely
(631 ^9/m3)'C* years) « 252<* /ig-yrs/m3.
Uncertainty in this cohort study is described below:
1. Length of exposure is well documented, although the duration
categories are presented without average values. A value of
0.2 is assigned to
-------
the results are difficult to reconcile.
2-193
-------
Table 2-86
MORTALITY EXPERIENCE OF A PCB-EXPOSED COHORT OF WORKERS0
Cause of Death Observed Deaths
All neoplasms 26
Hematologic Neoplasms 7
Expected Deaths
12.9
2.2
°From Bertazzi et al. (245).
Table 2-87
CONCENTRATIONS OF PCB AT PUANT 2a
Number of
Job Title Samples
Degreaser
Solder
Tanker
Moveman
(soldering area)
Heat soak operator
Tester
Pump mechanic
Floorman
(pre-assembly )
• 1
3
9
3
3
3
1
6
Total Sampling
Time (min)
381
884
2120
752
872
917
377
1683
TWA
(M9/m3)
1260
1060
850
720
630
290
280
170
°From Brown and Jones (255); personal air samples are reported.
2-194
-------
Table 2-88
DURATION OF EMPLOYMENT AMONG COHORT MEMBERS
IN PCB EXPOSURE JOBS0
Duration
3 - 6 mo.
6 mo. - 1 yr .
1 - 2 yr.
2 - 3 yr.
3 - 10 yr.
10+ yr.
TOTAL
Plant 1
216
147
185
94
247
79
968
Plant 2
418
288
293
146
311
143
1599
Total
634
435
478
240
558
222
2567
aFrom Brown and Jones (255).
Table 2-89
CANCER RESPONSE AMONG CAPACITOR MANUFACTURERS0
Neoplasm Observed Deaths Expected Deaths
All maglignant 39 43.79
neoplasms
Digestive organs 13 9.85
and peritoneum
Respiratory system 7 7.98
Lymphatic and 2 4.34
hematopoietic
Other 17 21.62
°From Brown and Jones (255).
2-195
-------
Table 2-90
OBSERVED AND EXPECTED CANCER DEATHS BY
LENGTH OF EXPOSURE AMONG CAPACITOR MANUFACTURERS'3
Neoplasm
All malignant
neoplasms
Rectum
Liver
Length
Employment
0.25 -
5 -
10 -
15 -
20 +
0.25 -
5 -
10 -
15 -
20 +
0.25 -
5 -
10 -
15 -
20 +
of
(vrs)
5
9
1
-------
Table 2-91
DOSE AND RESPONSE DATA FOR BROWN AND JONES
COHORT OF PCS-EXPOSED WORKERS
Cumulative Dose All Malignant Neoplasms Liver Cancers
(^q-yrs/m^) Observed Expected Observed Expected
1578 31 30.99 3 0.7k
(619, 5286)°
4732 3 7.05 0 0.19
(1856, 15852)
7888 3 3.28 0 0.08
(3039, 26424)
11042 2 1.73 0 0.04
(4330, 36991)
15775 0 0.74 0 0.02
(6186, 52846)
Lymphatic and
Digestive Organ Cancers Hemotopoietic Cancers
Observed Expected Observed Expected
2524 13 9.85 2 4.34
(990, 8455)
aln parentheses are the lower bounds and upper bounds, respectively, for
cumulative dose in each group.
2-197
-------
Table 2-92
POTENCY PARAMETER ESTIMATES FOR PCBs
Potencies ((>ig-vrs/m^)~1 )
Dose Lower Upper
Study _ Response _ Measure Limit0 _ MLE _ Limit13
Bertazzi All Upper 7.9«fE-6* 1.^7E-5 2.26E-5
et al. Malignant Bounds
(2^*4, Neoplasms
(0 degrees Best 2.62E-5 ^.8*E-5" 7.U5E-5
of freedom) Estimates
Lower 8.65E-5 1.60E-<»
Bounds
Hematologic Upper 1.27E-5" 3.15E-5 5.75E-5
Neoplasms Bounds
(0 degrees
of freedom) Best 4.18E-5 1.04E-**" 1.90E-^
Estimates
Lower 1.38E-«f 3.«t3E-U 6.26E-4"
Bounds
Brown & All Upper -2.16E-5" 0.00 6.82E-6
Jones Malignant Bounds
(235) Neoplasms
(chi-squared Best -7.25E-5 0.00" 2.29E-5
CO • 3.13) Estimates
Lower -1.85E- < ^-S) Best -2.19E-1* 2.35E-*»" 1.0«>E-3
Estimates
Lower -5.57E-^ 6.06E-<» 2.68E-3*
Bounds
2-198
-------
Table 2-92 (continued)
POTENCY PARAMETER ESTIMATES FOR PCBs
Potencies ((/ig-yrs/m3)"1)
Dose Lower Upper
Study Response Measure Limit0 MLE Limit0
Brown & Digestive Upper -1.13E-5* 3.78E-5 LOOE-'f
Jones System Bounds
(235) Neoplasms
(0 degrees Best -3.78E-5 1.27E-<*" 3.35E-<*
of freedom) Estimates
Lower -9.64E-5 3.23E-^ 8.55E-4*
Bounds
Hematologic Upper -1.07E-4* 0.00 3.73E-5
Neoplasms Bounds
(0 degrees
of freedom) Best -3.60E-* 0.00* 1.25E-^
Estimates
Lower -9.17E-* 0.00 3.19E-4*
Bounds
°90£ confidence limits are shown.
*An asterisk marks the parameters usod to derive RRD estimates.
2-199
-------
Table 2-93
RRD ESTIMATES FOR PCBs
M
I
O
O
Estimation
Study
Bertazzi
et al.
(2M».
2 at>
00 00
5.99E-2 «
6.51E-2 »
2.88E-3 CD
3.25E-3 «
a.
00 CD
Extra Risk
0.25
RRDL MLE RRDU
8.00E-H *.07E+2 2.48E+3
1.35E+2 6.86E+2 ^.17E+3
2.12E+2 1.28E+3 1.05E+
5.67E+2 » oo
1.31E+3 1.50E+4 «
1.79E+3 2.04E+'* oo
1.07E+2 7.19E+2 «
1.57E+2 1.06E+3 «
^. 17E+2 °o •
6.37E+2 « »
-------
Pnonocetin
Phenacetin (CAS No. 62-44-2) is an aniline derivative that has been used
extensively in analgesic mixtures, usually in combination with phenazone
and caffeine. It has been produced in the U.S. for over 50 years (24J5).
Abuse of these analgesic mixtures has been associated with nephropathic
changes, notably renal papillary necrosis. Suggestions of carcinogeni-
city have come from animal studies and numerous human case reports (see
below). Short-term tests have been equivocal. Phenacetin was mutagenic
in the Salmonella test in the presence of hamster liver microsome prepa-
rations, but not in the presence of rat or mouse preparations. A minor
human metabolite of phenacetin was mutagenic. Some chromosome aberra-
tions due to phenacetin have been reported, but recessive lethal muta-
tions and micronuclei were absent in some test systems (2).
Case reports of renal pelvis carcinoma in association with renal papil-
lary necrosis have proliferated since the first report of such effects
by Hultengren et al. (247). In that report, 5 of the 6 necrosis-
carcinoma cases were in abusers of phenacetin-containing analgesics
(abuse was defined as intake of 1 g of phenacetin per day for at least
one year or a total consumption exceeding 1 Kg). Case descriptions that
followed came from Australia (248), the United States (249). and
particularly Sweden, where Bengtsson, Angervall, and their associates
have described a number of studies of urinary tract disorders. In one
early study (250) 242 patients with chronic non-obstructive pyelone-
phritis were classified according to analgesic abuse. Eight of the 104
abusers developed transitional cell tumors of the renal pelvis; none of
the nonabusers developed those tumors. The period of observation
averaged slightly over 5 years in the 79* of the patients who were
followed up. Angervall et al. (251) retrospectively studied the 15
cases of renal pelvis tumors seen in one hospital between 1960 and 1968.
Ten could with certainty be classified as abusers; two others may also
2-201
-------
have ingested large amounts of phenacetin. Further evidence of a link
between phenacetin-containing analgesic and renal pelvis cancer was
presented by Johansson at al. (252). Of 62 patients with such cancer,
38 had definite History of abuse. For the remaining patients "it could
only be stated that there had been heavy abuse ... over several years."
By 1978 (253) these authors were able to identify more than 100 cases of
uroepothelial renal pelvis tumors associated with abuse of phenacetin-
containing analgesics, mostly from Sweden.
Other aspects of the carcinogenicity of phenacetin have been noted.
Burnett «t al. (2fr9) describe a case of renal pelvis cancer that
developed in the remarkably short period of
-------
significantly greater mortality rate. The prospective study identified
98 patients with interstitial nephritis not associated with analgesic
abuse and 48 patients with interstitial nephritis associated with
analgesic abuse. These individuals were followed up for 3 to 5 years.
Four of the 48 and none of the 98 patients developed transitional-cell
carcinoma of the urinary tract, a significant difference.
A case-control study in the Netherlands investigating phenacetin use and
bladder cancer was reported by Fokkens (257). A total of 1084 bladder
cancer patients and 1094 control patients were interviewed about their
consumption of analgesic drugs. Eighteen cases and 16 controls admitted
to more than incidental intake of phenacetin-containing analgesics. The
majority of the preparations contained 250 mg phenacetin, 250 mg
ocetosal, and 50 mg of caffeine. This information has been used to
convert consumption to grams of phenacetin-contoining analgesics, so as
to be compatible with other studies (see below). That is, 1 g of
phenacetin is assumed to be equivalent to 2.2 g of phenacetin-containing
analgesic. The distribution of consumption in these units is displayed
in Table 2-94.
Uncertainty in this study stems from a variety of sources which combine
to yield uncertainty factors a and t both equal to 1.55. Specific items
are discussed below:
1. As in most case-control studies, the subjects were required to
recall both duration and dosage of analgesic use. This feature
has prompted us to assign a value of 0.1 to a-j and T| (for
uncertainty in duration) and to agand IQ (for uncertainty in
daily dose).
2. We converted from measurements of phenacetin to measurements of
analgesics. To the extent that consumption deviated from the
250 mg of phenacetin per ^50 mg analgesic (45< phenacetin),
2-203
-------
uncertainty is introduced. Fokkens merely states that the
majority of the analgesics were of that mixture; other authors
have reported mixtures containing <*3* phenacetin (255) and IARC
(2fr6) lists mixtures that have as little as 32< and as much as
86* phenacetin. In any case, this factor is not believed to
contribute greatly to uncertainty, so 07 and 77 have been set
equal to 0.05.
3. Finally, in lieu of expected numbers of cancers, the study
design calls for selection of an appropriate control series.
Since the controls were not individually matched to the cases
in this study, the value 0.3 has been chosen for 03 and 73.
The bounds on dosr> resulting from these uncertainty considerations are
displayed in Table 2-94.
McCredie »t al. (258) conducted o case-control study carried out in New
South Wales, Australia that investigated renal pelvis cancer rather than
bladder cancer. Forty women and 27 men were histologically confirmed to
have renal pelvis cancer. Two control series were selected. The first
(<*9 women and 35 men) were friends or relatives of cancer or renal
disease patients. The second (61 women and 35 men) were selected from
among individuals attending a walk-in clinic. Interviewers elicited
information on analgesic consumption and smoking.
The authors present results relating to phenocetin-containing for women
only, since so fsw men consumed analgesics. Ignoring analgesics
consumed le-is than 5 years before the diagnosis or interview, the
distribution of cases and controls by total phenacetin-containing
analgesic consumption is shown in Table 2-95. The average consumption
values shown in that table have been estimated; the value of 25 g in the
lowest consumption group is due to the fact that those with absolutely
no consumption have been merged with those who consumed some, but less
2-20<»
-------
than 100 g, of pnenacetin-containing analgesics.
Uncertainty considerations in this study are in many ways similar to
those pertaining to the Fokkens study, as discussed as follows:
1. Recall uncertainty contributes the same value of 0.1 to a.-\, T\,
- 1.8, result in the bounds on dose shown
in Table 2-95.
McCredie and her associates conducted another case-control study among
females in Australia, this time examining both bladder and renal pelvis
cancer cases and nonmatched controls drawn from the electoral rolls of
New South Wales (259). Mailed questionnaires elicited information on
analgesic consumption, smoking history, as well as demographic data.
One hundred fifty-four (15
-------
Those bounds come from the following features of the study, the some
ones affecting the earlier McCredie et al. study:
1. Recall uficertainty contributes a value of 0.1 to a-j, T\ , ag,
and 75, as in the previous two studies.
2. Again, no average values for the groups are presented. Both a^
and 74 equal 0.3
3. Nonmatched controls entail a value of 0.3 for 03 and ~IQ.
Overall uncertainty is reflected in the factors oc • 7 • 1.8.
All three of the case-control studies described above have been used to
derive potency estimates (Table 2-97). The RRD estimates obtained by
application of thosw potency parameters in the standard exposure
scenario are also available (Table 2-98).
2-206
-------
Table 2-94
NUMBER OF BLADDER CANCER CASES AND CONTROLS FROM FOKKENS (257).
BY TOTAL PHENACETIN-CONTAINING ANALGESIC CONSUMPTION
Total Phenacetin-
Containing Analgesic
Consumption (g)
Cases
Controls
0
<4400
1003
6
1017
13
(1296, 2009, 3114)°
4400 - 12100
14684. 7260, 11253)
>12100
(17268. 26765, 41485)
aln parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for total dose in each group.
Table 2-95
NUM8 R OF FEMALE RENAL PELVIS CANCER CASES AND
CONTROLS FROM MCCREDIE £T AL. (258). BY
CONSUMPTION OF PHENACETIN-CONTAINING ANALGESICS
Phenacetin-Containing
Analgesic Consumption (g)Q
Cases
Controls'3
< 100
(13.9, 25, 45)c
100 - 5000
(1417, 2550, 4590)
> 5000
(4157. 7500, 13500)
15
18
87
17
algnoring consumption less than 5 years prior to diagnosis of cancer or
interview.
''Both control series have been combined.
eln parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for total dose in each group.
2-207
-------
Table 2-96
NUMBER OF BLADDER CANCER AND RENAL PELVIS CANCER
CASES AND CONTROLS FROM MCCREDIE ET AL. (259)
BY CONSUMPTION OF PHENACETIN-CONTAINING ANALGESICS0
Phenacetin-Containing Bladder Renal Pelvis
Analgesic Consumption (g) Cancer Coses Concer Cases Controls
< 100 113 17 384
(14, 25, 45)b
100 - 1000 60 9
(305, 550, 990)
> 1000 35 14 47
(1111, 2000, 3600)
aThe numbers are estimated from percentages given in the original
article.
^In parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for .total dose in each group.
2-208
-------
Table 2-97
PHENACETIN POTENCY PARAMETER ESTIMATES
Potencies ((g-yrs)"1)
Dose Lower Upper
Study Response Measure Limit0 MLE Limit0
Fokkens Bladder Upper 7.34E-3" 3.16E-2 7.63E-2
(257) cancer Bounds
(chi-squared
(2) - 5.5) Best 1.fvE-2
-------
10
I
M
Table 2-98
RRD ESTIMATES FOR PHENACETIN (g/day)
_Level of Extro Risk
Estimation
Study
Fokkens
(252)
McCredie
et al.
(258)
McCredie
et al.
(259)
Response Method
Bladder
Cancer
Renal
Pelvis
Cancer
Bladder
Cancer
Renal
Pelvis
Cancer
1
2
1
2
1
2
1
2
RRDL
4.25E-6
4.69E-6
1.60E-6
1.83E-6
8.89E-7
9.81E-7
7.35E-7
8.42E-7
10~6
MLE
1.59E-5
1.75E-5
4.86E-6
5.57E-6
2.58E-6
2.84E-6
2.23E-6
2.55E-6
RRDU
1.06E-4
1.17E-4
1.55E-5
1.78E-5
8.37E-6
9.23E-6
7.32E-6
8.39E-6
RRD|_
1.06
1.52
3.99E-1
5.62E-1
2.22E-1
3.18E-1
1.8AE-1
2.58E-1
0.25
MLE
3.98
5.68
1.22
1.71
6.44E-1
9.20E-1
5.56E-1
7.83E-1
RRDU
2.65E+1
3.79E+1
3.88
5.46
2.09
2.99
1.83
2.58
-------
Reserpine
Reserpine, a medicinal drug extracted from the roots of the plant
Rauvolfia serpentina, is used to treat hypertension and has been used in
the past as a sedative (260). It has been suggested (261) that
reserpine may act as a "promoter" of breast cancer, that it affects
cancer incidence only while it is being used. Short-term tests for
mutations, chromosomal aberrations, or DNA synthesis have been uniformly
negative (2). In fact, the epidemiologic literature presents a mixed
picture of reserpine as a carcinogen. A portion of that literature is
reviewed below.
The carcinogenic response suggested by some studies to be associated
with reserpine use is breast cancer. Ross et al. (262) compared
prolactin levels in long-term reserpine users to those in non-users. It
is suspected that prolactin may play a role in breast carcinogenesis.
Although the mean prolactin level was 50$ higher among the reserpine
users, Ross et al. suggest that that increased level would most likely
cause only a small increase in breast cancer incidence.
Two case-control studies published in 1975 (263, 264) failed to find a
purported association between reserpine use and breast cancer. A
further study (265) found that a link between reserpine use and breast
cancer disappeared after correcting for other variables (in this case,
duration of hypertension, the condition treated by reserpine). Kewitz
et al. (266) corroborate the lack of a strong etiologic contribution
from reserpine. In one of two prospective studies, Labarthe and
O'Fallon (267) did not discern an effect due to reserpine use; among
users 11 cases were observed with 10.8 expected. On the other hand,
Danielson et al. (261) followed up women in a group health organization
and found a relative risk of 1.7 for recent users of reserpine with
respect to breast cancer. This result was not significant however; the
2-211
-------
90jf confidence limit on the relative risk included 1.
One case-control study ard one clinical trial report provide information
necessary for a quantitative estimation of reserpine risk. The case-
control study (268) selected 275 breast cancer patients from 5
Baltimore-area hospitals and two individually-matched control series,
one from the hospitals and one from the neighborhoods of the cases.
Overall, reserpine use was not associated with breast cancer, no matter
which control group was used. For analysis of response by total dose, a
significant portion of the cases and controls could not be used; they
had incomplete fata on reserpine use. The portions of the groups that
do have adequate data used reserpine as shown in Table 2-99. Displayed
in that table are the cases matched to the neighborhood controls and the
neighborhood controls themselves; this case-control pairing provided the
highest proportion of cases with adequate data.
The dose data presented in Table 2-99 results from the following consi-
derations. The best estimates are arithmetic averages of the group-
defining ranges. Uncertainty considerations are as follows:
1. Dose determination depends on recall of past behavior. In this
case, the cases and controls were asked who had treated them
for hypertension. Then, at that point, doctors' or hospitals'
records were checked for dosage and duration. That being the
case, uncertainty related to recall affects only the upper
bound: dose will not be smaller than determined by this method,
but if periods of treatment are not recalled, dose could be
larger. The factor related to upper bound calculation and
completeness of measurement, 73, is set equal to 0.2; 013 * 0.
2. The dose groups were presented without average values. The
limits of the groups are fairly small. A nominal value of 0.1
is assigned to a<, and -7^.
2-212
-------
3. The neighborhood controls were matched to the cases on the
basis of age (+5 years), race, and neighborhood of residence.
Other health and demographic characteristics did not differ
either (except for number of nulliparous or primiparous women:
k2% of the cases as opposed to 27* of the controls). On the
other hand, a large fraction, of the cases and controls did not
have dose information. It is not known if this has influenced
the comparability of the two groups. A value of 0.1 is assumed
for ag and IQ.
The resulting factors, a - 1.2 and 7 - 1.4, determine the bounds on
dose. These bounds and the best estimates are used in the estimation of
/?, /JL. and /3U (Table 2-100).
The Hypertension Detection and Follow-Up Program provides data on more
than 2500 women treated for hypertension. Curb et al. (269) have used
this program to investigate the reserpine-breast cancer association. A
total of 1036 women received reserpine at one time or another, for an
average of 1.97 years. The dosage of reserpine prescribed was between
0.1 and 0.25 mg/day (assumed average 0.175 mg/doy). The resulting
cumulative dose for those taking reserpine is estimated to be
(1.97 years) • (0.175 mg/day) - 0.345 mg-yrs.
Seven breast cancers were observed among the reserpine users. Based on
the experience of the 1493 women who never took reserpine during the
study, the relative risk for breast cancer among the reserpine users is
estimated to be 1.28, adjusting for age, race, and whether or not hyper-
tension treatment had begun before the study began. This corresponds to
an expected value of 5. 47 breast cancers. The observed and expected
responses (7 vs. 5.47) are used with the dose data to determine the
potency parameter estimates shown in Table 2-100. The best estimate of
2-213
-------
dose is presented obove, 0.345 mg-yrs, ond the bounds, 0.265 and 0.586
mg-yrs, are consequences of the uncertainty factors « and i, 1.3 and
1.7, respectively, which are determined as follows:
1. The medications prescribed prior to the study for those with
hypertension are not documented. Over a quarter of the women
in the study hod hod hyoertension for 10 years or longer. A
substantial portion of time, with potential reserpine use, is
not factored into the dose calculations. Once again, this
affects only the upper bound on average dose. The factor 72 is
set equal to 0.4 to account for this, whereas aj is set to
zero.
2. The range of dosage (0.1 to 0.25 mg/day) is presented but no
average dosage is presented. This fairly narrow range is
consistent with a^ and -r^ equal to 0.1.
3. The expected number of breast cancers is based on specially-
treated hypertensives who never received reserpine and is
adjusted for age, race, and prior antihypertension medication
use. Some variability in that calculation is presented by Curb
et al. Consequently, ag and 73 have been assigned a value of
0.2 to account for that variability and the uncertainty with
regard to selection of the model used to estimate relative
risk.
RRD estimates derived from these two sources, the case-control study of
Lilienfeld et al. and the prospective study by Curb et al. are based on
the potency parameters shown in Table 2-100. The RRDs are displayed in
Table 2-101. The prospective study (269) is not inconsistent with a
hypothesis of no carcinogenic effect of reserpine.
2-214
-------
Table 2-99
BREAST CANCER CASES AND MATCHED NEIGHBORHOOD CONTROLS WITH RESPECT
TO CUMULATIVE DOSE OF RESERPINE; DATA FROM LILIENFELD £T AL.
Cumulative
Dose (mq-yrs)° Coses Controls
0 121 119
0 - 0.068 3 4
(0.028. 0.03i», 0.0i»8)b
0.071-0.137 2 5
(0.086, 0.103. 0.1^'t)
0.140-0.822 5 3
(0.399, 0.(»79, 0.671)
0.822+ 3 0
(1.028, 1.233, 1.726)
aLilienfeld et al. present total doses in mg. Cumulative doses were
derived by dividing by 365. 1 mg-yr is equivalent to taking 1 mg of
reserpine per day for a year.
bln parentheses are the lower bounds, best estimates, and upper bounds,
respectively, for average dose in each group. These were not presented
by the original authors.
2-215
-------
Table 2-100
BREAST CANCER POTENCY PARAMETER ESTIMATES FOR RESERPINE
Dose Potencies ((mg-yrs)""1 )
Study Measure Lower Limit0 MLE Upper Limit0
Lilienfeld Upper 1.51E-1* 1.
-------
Table 2-101
RRD ESTIMATES0 FOR RESERPINE (mg/day)
Level'of Extra Risk
Estimation 10~6 0.25
Study Method RRD| MLE RRDU RRD) MLE RRDU
Lilien- 1 2.63E-8 9.53E-8 1.28E-6 6.58E-3 2.38E-2 3.19E-1
feld
et al. 2 3.31E-8 1.20E-7 1.61E-6 9.72E-3 3.52E-2
-------
Soccharin
The artificial sweetener saccharin (CAS No. 128-44-9) was discovered
accidentally in 1879 and has been the subject of debate with respect to
its safety for human consumption since about 1890. Not only is
saccharin a sweetener, it is used as a brightener in automobile bumpers,
as an intermediary in fungicide production, and previously as an
antiseptic and food preservative. The primary exposure for humans is
from low-sugar food and beverages and sugar substitutes (270).
Short-term testing of saccharin has produced mixed results. It is
mutagenlc in cells of some animal and plant species but not in others or
in bacteria. In mammalian cells, there are conflicting reports about
induction of chromosomal anomalies in vitro, about production of domi-
nant lethal mutations, and about specific-locu* somatic mutation test
results. No data on humans are available (2).
A pharmacokinetic study in women who us* saccharin daily showed that
saccharin is rapidly absorbed into the blood and i* retained with a
half-life of about 7.5 hours (271). The data from this stjdy also
indicate that one or more high-retention areas may exist. Animal
Species tested stored saccharin in high concentration in the kidney and
urinary bladder. The lower urinary tract is the area of greatest
concern for human saccharin carcinogenesis.
In fact, several investigations have examined the relationship between
soccharin consumption and cancer of the bladder. In 1974, Morgan and
Jain (272) conducted a case-control study that could find no association
between artificial sweetener use and bladder cancer, although the risk
was increased among smokers and, especially, male smokers who drank cola
or alcohol. On trie other hand, Howe et ol. (273) claim that a signifi-
cantly increased relative risk of 1.6 exists in males who ever used
2-218
-------
saccharin as opposed to those who never used saccharin. These authors
also claim a dose-response relationship was seen for both duration and
frequency of use.
Further investigations have failed to confirm this finding. Morrison
and Buring (27*»), in a case-control study of lower urinary tract cancer,
found no consistent pattern of relative risk with respect to frequency
or duration. Hoover and Strasser (275) performed a large, population-
based case-control investigation. Again, no overall relationship
between saccharin consumption and bladder cancer was noted, although
non-smoking women and heavy-smoking men who also consumed artificial
sweeteners, those groups suggested to be at greater risk on the basis of
animal studies, did have increased risk. Wynder and Stellman (276) also
conducted a case-control study, of bladder cancer in this case, and
found no dose-response relationship with respect to saccharin consumed,
measured by quantity, duration, or the combination of the two. These
authors detected no evidence that artificial sweeteners promote the
carcinogenic effect of tobacco smoke. Finally, a study from Denmark
(277) failed to detect any increased risk among regular artificial
sweetener users.
One other study has addressed this question and also provides quantita-
tive estimates of exposure. Armstrong »t al. (278) report the results
of a prospective study of cancer mortality among members of the British
Diabetic Association (BOA). The study included 5971 members, 3003 men
and 2968 women, most of whom registered between 1965 and 1968. Deaths
were monitored until July 1, 1973. A total of 16,2<»7 person-years of
observation were contributed by the males, 15,932 by the females. Four
bladder cancer deaths were recorded whereas 5.8 were expected.
Saccharin consumption was determined from a survey of a sample of all
BDA members polled in 1973. The males averaged 0.99 mg/kg body weight/
2-219
-------
day of saccharin and the females 0.65 mg/kg/doy, based on the reported
average number of saccharin tablets per day and assumed average body
weights like those of other diuretics. The duration of diabetes among
those surveyed is distributed as in Table 2-102. If we use the mid-
points of the intervals shown in that table (and use 35 years for the
longest duration group) we can calculate an average duration of diabetes
of
[(2.5 x 539) • (7.5 x 353) + (15 x 38
-------
and 12•
3. Substantial uncertainty is involved in the mere use of the
numbers derived from the BOA survey. Tho subset included in
the mortality study may have consumed saccharin ct a rate
different than the general membership. Consequently, we assign
a value of 0.3 to cxg and 75.
4. The authors converted consumptions measured in tablets of
saccharin por day to mg/kg/day. That conversion relied on an
average weight that was assumed to apply to the study partici-
pants, even though weights were not available for the partici-
pants themselves. The uncertainty in that conversion is
probably no. too great, so 1x7 and 77 have been assumed to be
0.1.
5. Expected numbers of deaths were determined from a 105f random
sample of deaths which occurred in England or Wales in 1972.
It seems that this may introduce more uncertainty than using
the entire population as the reference. We might also consider
that bladder cancer is related to smoking, a factor not
controlled for in the calculation of expected values. As a
result, we assume that
-------
and 72.
3. Substantial uncertainty is involved in the mere use of the
numbers derived from the BOA survey. The subset included in
the mortality study may hove consumed saccharin at a rate
different than the general membership. Consequently, we assign
a value of 0.3 to ag and 75.
<*. The authors converted consumptions measured in tablets of
saccharin per day to mg/kg/day. That conversion relied on an
average weight that was assumed to apply to the study partici-
pants, even though weights were not available for the partici-
pants themselves. The uncertainty in that conversion is
probably not too great, so ay and 77 have been assumed to be
0.1.
5. Expected numbers of deaths were determined from a 10jf random
sample of deaths which occurred in England or Wales in 1972.
It seems that this may introduce more uncertainty than using
the entire population as the reference. We might also consider
that bladder cancer is related to smoking, a factor not
controlled for in the calculation of expected values. As a
result, we assume that 013 and 73 equal O.U.
The bounds on cumulative dose resulting from these considerations are
-------
Toble 2-102
DURATION OF DIABETES AMONG BOA MEMBERS,
FROM DATA IN ARMSTRONG ET AL. (278)
Duration of
Diabetes (yrs)
0-4
5-9
10 - 19
20-29
30+
Number of
Men
539
353
384
172
99
Subjects
Women
475
306
408
156
77
Table 2-103
BLADDER CANCER POTENCY PARAMETERS FOR SACCHARIN,
FROM DATA IN ARMSTRONG ET AL.a
Dose
Measure
Upper
Bounds
Best
Estimates
Lower
Bounds
Potencies
Lower Limitb
-3.30E-2"
-7.60E-2
-1 .75E-1
((mq-yrs/kq) 1)
MLE Upper
0.00 1.
0.00" 3.
0.00 8.
Limitb
62E-2
72E-2
57E-2"
°No degrees of freedom allow evaluation of the fit of the model.
b90£ confidence limits are shown.
"An asterisk marks the parameters used to derive RRD estimates.
2-222
-------
Table 2-104
RRD ESTIMATES0 FOR SACCHARIN (mg/kg/day)
Level of Extra Risk
Estimation 10~» 0.25
Method RRDi MLE RRPU RRD| MLE RRDU
1 3.37E-5 8.
-------
Trichloroethylene
Trichloroethylene (CAS No. 79-01-6; abbreviated here TCI) is an asymme-
trical ethylene similar in structure to vinyl chloride. Trichloro-
ethylene is used primarily as a metal degreaser; it has been used as a
solvent in the textile industry and for adhesives. In addition,
trichloroethylene has served as a general anesthetic and for many years
was used as an extraction solvent (for decaffeinating coffee, for
example) (279).
Trichloroethylene is considered to be weakly mutagenic (and only when
activated by an enzyme system) and weakly active in cell transformation
(2, 280). In fact, it is believed that the short-lived, epoxide
intermediate formed during trichloroethylene metabolism is the ultimate
carcinogen which binds to cellular macromolecules.
Many articles have described human trichloroethylene pharmacokinetics
and metabolism (281-287). Several of these describe the time course of
trichloroethylene metabolites in blood, urine, and expired air under
experimental conditions. One of them (282) presents data from an
experiment in which volunteers were exposed to 200 ppm of trichloro-
ethylene for 7 hours each of 5 consecutive days. If equilibrium between
exposure to trichloroethylene and trichlorocetic acid (TCA) excretion in
urine occurs by the end of a working week (287), then the amount of TCA
in urine corresponding to an occupational exposure to 200 ppm is roughly
400 mg/24 hr. If we assume that 1.2 to 2.4 liters of urine ore voided
in a 24-hour period (284), then the corresponding concentration of TCA
is 167 to 333 mg/J.
Occupational investigations of this relationship are also represented in
the literature (284, 287). In a preliminary study. Smith (287 found
that the ratio of time-weighted, atmospheric trichloroethylene (in ppm)
2-224
-------
to urinary TCA concentration (in tng/f) was roughly 1:2. He does not
report the atmospheric concentrations studied. A more detailed study
(284) presents TCA concentrations for trichloroethylene exposures
ranging from 0 to 175 ppm. A nonlinear relationship was found;
progressively larger atmospheric levels produce smaller increases in TCA
excretion. That relationship is consistent with the data reported by
Stewart et ol. (282).
Several mortality studies have been reported (288-292). Only the
studies by Axelson et al. and Tola et al. provide estimates of exposure,
both of them in terms of TCA concentrations in urine. None of these
mortality studies establish an association of trichloroethylene with a
particular type of cancer. The study by Blair (290) of metal polishers
and plater* included workers exposed to TCI in degreasing operations.
Increased proportional mortality was noted for liver and esophageol
cancer but unfortunately no levels of TCI were recorded and exposure to
other putative carcinogens including nickel and chromium compounds was
also common. The brief report by Paddle (292) is intended merely to
document the lack of liver cancers among TCI manufacturers, and the
proportional mortality study by Blair et al. (290) of dry cleaning
employees records no trichloroethylene exposures, which would in any
case have been confounded by extensive contact with tetrachloroethylene
and carbon tetrachloride. Due to limitations in the other studies,
quantitative risk estimates will be derived from Tola at al. (291) and
Axelson et al. (288) only.
The cohort of workers identified by Tola et al. is known from files of a
Finnish laboratory that measured urine TCA concentrations of trichloro-
ethylene-exposed workers. The total cohort contains 2084 men and women
who contributed some 13933 person-years of follow-up. No single cancer
type predominated; cancers of the gall bladder, lung, breast, uterus,
and testis and multiple myeloma have been diagnosed.
2-225
-------
Table 2-105 presents the distribution of cancer deaths found by Tola
at al.—11 in all—by urine TCA concentrations. This table, and
analysis of this cohort, will be limited to the authors' original list
of workers who had their TCA levels measured (i.e. excluding the
poisoning cases). No expected values are presented in this format and,
unfortunately, the presentation of expected numbers in other places in
the paper includes the poisoning cases. Bearing this in mind, two
options are used to provide mortality data in the form needed:
1. Assume that the age distribution for each group in Table 2-105
is roughly the same and distribute the Tf.3 expected cancer
deaths over the groups according to the total number of workers
in each group (Table 2-106). Tola et al. state that the
expected number of cancers in their highest 3 categories
(highest 2 in Table 2-106) is about 2 as opposed to our
estimated 2.6, so our calculations could be considered to
underestimate the relative risk in the higher groups.
2. Combine the groups from Table 2-105 into on aggregated group
with one exposure estimate, 10 observed cancer deaths, and 13.8
expected cancer deaths.0
Both of these options require estimates of mean TCA concentrations,
either for the groups as defined in Table 2-106 or for the cohort as a
whole. For estimation of cumulative doses, this poses some problem
because, in the first place, the values recorded are highest TCA
measured with no accounting for lower levels and, secondly, no average
duration of exposure is provided for members of this cohort. The second
of these issues is dealt with by assuming a default value of 7 years
average length of employment.
"The13.3 expected deaths is based on inclusion of 200
-------
Urine TCA concentrotions were converted to otmospheric TCI levels by
using data from Ikeda ot ol. (28fr) (Table 2-107). We determined that a
simple linear regression of loge(TCA) on loge(TCI) provided an adequate
description of the data. The resulting equation,
loge(TCA) - 1.57 + 0.82-loge(TCI) ,
was inverted to obtain estimates of TCI dependent on our estimated
average TCA concentrations. The best estimates for each dose group.
expressed in ppm-yeors, derived from the equation above, an assumed
seven-year average exposure, and using midpoints of the intervals shown
in Table 2-106 (300 in the lost group) are given in Table 2-108.
Table 2-108 also presents reasonable upper and lower bound on those dose
estimates. The group-dependent factors a and ^ that determine the
bounds are derived on the basis of the following considerations:
1. Length of exposure is completely unknown. Hence a-j and T\ are
set equal to 1.5.
2. The number of urine TCA concentration measurement* for each
individual varied and it is not Known how frequently the
measurements were taken. It appears that there is great
uncertainty with respect to the completeness of the exposure
profile of each individual, and so we set. 03 and Tjequol to
0.5.
3. The groups displayed in Table 2-106 are given without average
values. As these had to be estimated, 04 and it, equal 0.2.
4. A serious recording bias exists due to the fact that the
exposed individuals were classified according to thsir maximum
TCA concentrations, as opposed to their average concentrations.
We have no way to determine the averopes, so 05 has bean set
equal to 0.8 for all but th« lowest grot,}, for which a^ » 0. i«.
2-227
-------
In all cases 75 • 0.
5. Some uncertainty about the applicability of the reported TCA
values for use in estimating TCI exposures is raised by Tola
et al. when they state "it is possible that some of the workers
with low urinary trichloracetic acid values were not actually
exposed to trichloroelhylene at all but to some other degreaser
such as perchloroethylene." To reflect this ag - 0.2 in the
lowest group, ag • 0 in the other groups, and 75 - 0 in all
groups.
6. Another important uncertainty concerns the conversion from
urine TCA concentrations to atmospheric TCI concentrations. We
judge that 07 and 77 should be equal to 1 to adequately cover
all the variability inherent in that conversion.
7. Expected values were determined from national mortality rates,
which is probably not too bad for a small country like Finland.
However, we had to estimate the expected numbers of cancer
deaths that were appropriate for each dose group. This adds
uncertainty, reflected in the choice of 0.2 for ag and 79.
The overall uncertainty factors are a • 5.0 in the lowest dose group,
a - 5.2 in all other dose groups, and 7 - '*.'».
Axelson et al. (288) have identified a cohort of workers employed at a
TCI producer in Sweden. Follow-up was successful for 518 men exposed to
TCI with known duration of exposure, although that duration is not
reported. Those men with a latency period of a least 10 years contri-
buted 36i»3 person-years with 9 cancers observed. Again, no single type
of cancer prodominated - one each of stomach, colon, pancreas, pulmo-
nary, prostate, kidney, and brain cancer and of melanoma and leukemia
art represented. Men witn ot Iftast 10 years latency were divided into
higt^ arnl low exposure groups bu^ud on avoroga reported urine TCA concen-
trations, with 100 mg/J as the cut point. This subcohort is the subject
2-?28
-------
of our analyses (Table 2-109).
The high exposure group corresponds in exposure to the highest group in
the Tola at al. cohort. Consequently, we assumed the average TCA
concentration for that group to be 300 mg/P. A value of 50 mg/f was
used for the low exposure group. The corresponding atmospheric
trichloroethylene concentrations, using the Ikeda et al. (284) data, are
154.7 and 17.4 ppm, re pectively. Since duration of exposure was not
reported, the default value of 7 years was used so that the best
estimates of dose are 121.8 and 1082.9 ppm-years for the low and high
exposure groups, respectively.
Uncertainty considerations are very similar for the Axelson et nl. data
and the Tola et al. data. The uncertainty factors corresponding to the
Axelson cohort have been derived as follows:
1. Length of exposure is, again, completely unknown so that a-) and
T) are set equal to 1.5.
2. The laboratory that determined TCA concentrations discarded
most records prior to 1967. Consequently the categorization is
based only on more recent determinations. We have assigned a
value of 0.2 to «2 and a value of 0.4 to 72 to try to account
for this uncertainty.
3. Once more, the completeness of the measurements is not docu-
mented, so that aj and 13 have been given a value of 0.5
4. No average values are provided for either exposure group. The
value 0.2 is assigned to a^ and 7^.
2-229
-------
5. The some uncertainties with respect to the applicability and
conversion of urine TCA values are present here as were present
in the Tola et al. study. Consequently ag » 0.2 in the low
dose group, 05 - 0 in the high dose group, 75 - 0 in both
group, and 07 • 77 • 1.0 in both groups.
The resulting uncertainty factors are dose group dependent: for the low
dose group a • 4.6 and 7 * <».6; for the high dose group a • i*.it and
7 - >».6. The bounds on dose are displayed in Table 2-110.
The potency parameter estimates from these two studies are presented in
Table 2-111 and the corresponding RRD estimates in Table 2-112. The
data are consistent with a hypothesis of no carcinogenic activity for
trichloroethylene (i.e. the upper bounds on RRD from both studies are
infinite).
2-iJO
-------
Table 2-105
DISTRIBUTION OF 2004 WORKERS EXPOSED TO
TRICHLOROETHYLENE WITH REGARD TO THE HIGHEST
MEASURED VALUE OF URINARY TRICHLORACETIC ACID (TCA)a
TCA in
Urine (mg/J)
<10.0
10.0-49.9
50.0-99.9
100.0-499.9
>500.0
Cancer Deaths
4
2
2
2
0
Others
883
736
192
164
19
Total
887
738
194
166
19
°From Tola «t ol. (291)
Table 2-106
OBSERVED AND EXPECTED NUMBERS OF
CANCER DEATHS BY HIGHEST MEASURED VALUE
OF URINARY TRICHLOROACETIC ACID.
TOLA £T AL. TRICHLOROETHYLENE-EXPOSED COHORT
TCA in
Urine (mg/Pj
<10.0
10.0-49.9
50.0-99.9
>100.0
Total Observed
Number Cancer Deaths
887 4
738 2
194 2
185 2
Expected
Cancer Deaths0
6.1
5.1
1.3
1 .3
°Expected values calculated bv distributing the 14.3 deaths expected for
the 2084 people by the number in each group.
2-231
-------
Table 2-107
TCA CONCENTRATIONS IN URINE SAMPLES FROM WORKERS
EXPOSED TO TRICHLOROETHYLENE AT VARIOUS CONCENTRATIONS0
Trichloroethylene
Concentrations No. of
in the Air (ppm) Measurements
3
5
10
25
40
45
50
60
120
175
9
5
6
4
it
5
5
5
4
4
TCA
Concentration13
(mq/f)
12.7
20.2
17.6
77.2
90.6
138.4
146.6
155.4
230. 1
235.8
(8.8-18.2)
(10.0-40.8)
(10.3-30.0)
(51 .6-115.6)
(50.2-163.8)
(83.2-216.5)
(76.3-281 .7)
(104.3-231 .4)
(199.0-267.4)
(187.2-296.9)
°From Ikedo at al. (284).
''Geometric means together with SO ranges in parentheses.
2-232
-------
Table 2-108
DOSE AND RESPONSE DATA DERIVED FROM TOLA FT AL.
TRICHLOROETMLSNE-EXPOSED COHORT
Dose (ppm-years)
Lower
Group
1
2
3
4
Entire
Cohort
Bound
1 .
12.
38.
208.
22.
47
6
4
2
4
Best
Estimate
7.34
65.3
199. b
1082.6
116.5
Upper
Bound
32.
287.
87fl.
4763,
512,
3
3
2
.4
.6
Observed
Cancer
Deaths
4
2
2
2
10
Expected
Cancer
Deaths
f>. 1
5.1
1 .3
1 .3
13.8
Table 2-109
DESCRIPTION OF THE AXELSON ET AL. SUBCOHORT OF
TRICHLOROETHYLENE-EXPOSED MEN WITH AT LEAST 10 YEARS LATENCY0
Group
Observed Expected
Person-years Number of Number of
of Observation Cancer Deaths Cancer Deaths
1:Low Exposure 3095
(<100 mg/P TCA)
2:High Exposure 548
(>100 mg/P TCA)
7.7
1.8
°From Axelson et al. (288).
2-233
-------
Table 2-110
DOSE AND RESPONSE DATA DERIVED FROM THE
AXELSON ET AL. COHORT OF TRICHLOROETHYLENE-EXPOSED WORKERS
Group
Dose (ppm-yeors)
Lower Best Upper
Bound Estimate Bound
Observed Expected
Cancer Cancer
Deaths Deaths
26.5
246.1
121 .8
1082.9
560.3
4981.3
6
3
7.7
1.8
2-234
-------
Table 2-111
POTENCY PARAMETER ESTIMATES FOR TRICHLOROETHYLENEa
Study
Tola et ol .
(291)
(chi-squared
(3) . 3.00)
Axelson
et al .
(288)
(chi-squared
(1) - 5.9)
Dose
Measure
Upper
Bounds
Best
Estimates
Lower
Bounds
Upper
Bounds
Best
Estimates
Lower
Bounds
Potencies^ipjjm-yj-s)"1)
Lower Linitb MLE Upper Limitb
-1.0t»E-*»" 8.95E-5 i».28E-4
-*f.59E-'» S.g'tE-'f" 1.8SE-3
-2.39E-3 2.0* 1.67E-3
-1.«»3E-3 1.99C-3 7.«f1E-3"
°Based on the risk of oil malignant neoplasms.
b905t confidence limits ore shown.
"An asterisk marks the parameters used to derive RRD estimates.
2-235
-------
Table 2-112
RRD ESTIMATES0 FOR TRICHLOROETHYLENE (ppm)
Estimation
Study Method
Tola
at ol .
(291)
Axelson
at al .
(288)
1
2
1
2
RRDi
8.05E-6
1 .06E-5
1 .06E-5
1 . 40E-5
Level of Extra Risk
10~6 0.25
MLE RRDU RRDi MLE RRDU
2.00E-4 oo 2.01 5.00E+1 oo
2.63E-
-------
Vinyl Chloride
The halogenated hydrocarbon vinyl chloride (CAS No. 74-01-4) is used
primarily in the plastics industry. At one time it was used as a
refrigerant and as an extraction solvent (293). Vinyl chloride has been
found to be mutagenic in a number of test systems, even in the absence
of metabolic activation, and it induces chromosomal aberrations and
sister chromatid exchanges in vivo in mammalian species. In addition,
vinyl chloride was noted to have alkylated liver DNA of rats treated
in vivo (2). It has been suggested (294) that the carcinogenic moiety
of vinyl chloride may be the epoxide formed during metabolism.
The epidemiologic literature includes several articles describing the
toxic and carcinogenic effects associated with vinyl chloride exposure.
(295-302).
Although there are suggestions that vinyl chloride is linked to nervous
system and hematopoietic and lymphatic system tumors, it is most firmly
associated with angiosarcoma of the liver, a cancer rarely seen outside
of vinyl chloride-exposed populations, and in occupational settings with
respiratory cancer. In accordance with the guidelines outlined earlier
in this section, risk estimates have been developed for liver cancer
(since vital statistics or. angiosarcoma specifically are absent), for
respiratory cancer, and for all malignant neoplasms.
The studies by Barnes (305). Jones (304). and Aryonpur (305) describe
the industrial processes involving work with vinyl chloride and the
associated atmospheric concentrations of that chemical. The reports by
Nicholson, Henneberger, and Seidman (294) and Nicholson, Henneberger,
and Tarr (306) describe, in general, the hazards associated with employ-
ment in the VC-PVC industry and the trends in cancer mortality that may
be expected due to post exposures. Kuzmack and McGoughy (307) analyze
2-237
-------
exposures that may be encountered by populations living near vinyl
chloride monomer or polymerization facilities.
Severe'1, studies describe the cancer mortality experience of occupational
cohorts (508-316). Angiosarcomas of the liver were found in many of
these cohorts.
Unfortunately, the data on exposure to vinyl chloride necessary for a
quantitative assessment of risk are available only for those cohorts
described by Ott et al., Fox and Collier, Buffler et al., and Heldaas
at al. The study by Fox and Collier alone presents data on expected
numbers of liver cancers. This study forms the basis of the liver
cancer risk estimates. Ott et ol. do not provide respiratory cancer
data, so the other three studies are used for estimation of respiratory
cancer risk. All four studies have been used for calculation of risk of
any malignant neoplasm. The cohorts and the derivations of the risk
estimates are described below.
Ott et al. (311 4) studied the mortality experience of 594 employees with
potential exposure to vinyl chloride between 1942 and 1960. Follow-up
was through 1974. Although there were no deaths due to any liver malig-
nancy, the observed number of total malignancy deaths exceeded the
expected number in the high exposure category.
Some of the workers studied by Ott et al. were also exposed to arsenic,
which is a known human carcinogen. In order to avoid a possible
confounding effect of arsenic exposure, only analyses by Ott et al.
which omitted workers exposed to arsenic were utilized.
Based upon industrial hygiene data, each job was assigned an exposure
level of low (<25 ppm), intermediate (25-200 ppm) or high (>200 ppm).
This classification was based primarily upon measurements of time-
2-238
-------
weighted average concentrations for an 8-hour day. Ott et al. attempted
to take into account in the classification frequent short term exposures
to several thousand ppm encountered in some jobs. For risk assessment,
estimates of average exposures for each of these exposure categories are
needed. Except for the job classification of "coagulator" (which had
estimated 8-hour TWA's of 135-825 ppm), none of the job classifications
had estimated exposures in excess of 385 ppm; however, several cate-
gories had 120-385 or 95-350 ppm as an estimated range of concentra-
tions. Based on these and similar observations, it appears reasonable
to assign an average concentration of 300 ppm to Ott's high exposure
range (>200 ppm). For the intermediate exposure range (25-200 ppm) an
average of [(25)(200)J1/2 . 70 ppm was assumed and an average exposure
of 25/2 - 12.5 ppm was assumed for the low exposure range (<25 ppm).
Ott et al. assigned individuals to exposure groupings based on the
highest exposure experienced for one or more months. Numbers of workers
in these groupings by years of exposure are shown in Table 2-113. In
this table years of cumulative exposure are calculated in two ways: by
considering all years of exposure and by ignoring years of exposure at
lower levels. By this latter method, an individual exposed at high
levels for six months, and at lower levels for ten years, would be
classified in the high category with less than one year of exposure.
Table 2-114 shows the observed and expected deaths for all malignant
neoplasms by exposure level and according to whether exposure was less
than or greater than one year. Only data on all malignant neoplasms are
recorded in this way and consequently risk estimates were developed only
for this disease category. In these mortality analyses, duration of
exposure was separated into two categories: less-than-one-year and one-
year-or-longer. Exposures at lower levels were not considered in
assigning individuals to less-than-one-year or one-year-or-longer
categories (i.e., results in this table correspond to the rows in Table
2-239
-------
2-113 labeled "High levels only", "Intermediate levels only", and "Low
levels only").
In order to use the mortality results in Table 2-114 for making quanti-
tative estimates of risk, we need to estimate average exposures for each
category in ppm-years. This can be done, but is made more complicated
by the fact that Ott ot al. placed workers in exposure groups by the
highest exposure in any one- month period. Certain assumptions about
length of exposures are based on the pattern of employment outlined in
Table 2-113. First, average lengths of exposure for the four duration
of exposure categories are assumed to be 0.5, 5, 15 and 27 years. A
low-intermediate exposure is calculated as the average of the low and
intermediate levels estimates discussed above, i.e. 41 ppm. The 417 men
classified by exposure level are considered to have the following
exposure pattern:
High exposure group, 163 men:
20+ years of low-to-high exposure; 27 men:
19 with high exposure for 15 years and
low-intermediate exposure for 12 years,
8 with high exposure for 5 years and low-intermediate
exposure for 22 years.
10-19 years of low-to-high exposure, 39 men:
39 with high exposure for 5 years and low-intermediate
exposure for 10 years.
1-9 years of low-to-high exposure, 64 men:
24 with high exposure for 5 years,
40 with high exposure for 1/2 year and
low-intermediate exposure for 4.5 years.
<1 year of low-to-high exposure, 33 men:
33 with 1/2 year of high exposure.
2-240
-------
Intermediate exposure group, 73 men:
10-19 years of low-to-intermediate exposure, 6 men:
5 with 15 years of intermediate exposure,
1 with 5 years of intermediate exposure and 10 years
of low exposure.
1-9 years of low-to-intermediate exposure, 38 men:
35 with 5 years o* intermediate exposure,
3 with 1/2 year of intermediate exposure and *».5 years
of low exposure.
<1 year of low-to-intermediate exposure, 29 men:
29 with 1/2 year of intermediate exposure.
Low exposure group, 181 men:
1 with 27 years of low exposure,
3 with 15 years of low exposure,
50 with 5 years of low exposure,
127 with 1/2 year of low exposure.
These assumptions allow calculation of cumulative dose to accompany the
observed and expected numbers of cancer deaths given in Table 2-1T».
For example, the dose in the group exposed for >1 year to high exposures
is estimated as
[19-(15-300+12-M) + 8-(5-300+22-<»1) + 39- (5- 300+10- 41 >
+ 2«*-(5-300)]/[l9+8+39+2<»] - 2<»95 ppm-years.
The group exposed for less than a year to high levels has an average
dose of
[(*0(0.5-300+(».5-i»1) + 33(0.5-300)]/[<»0+33] • 251 ppm-years.
The remaining dose estimates are similarly derived (Table 2-115). Table
2-115 olso presents the bounds on the dose estimates. The factors that
2-241
-------
contribute to tha uncertainty in tho dcse estimates for this cohort are
given below, along with the corresponding numericol contribution to a
and 7:
1. Length of sxposure is brokun down into groups of width up to
ten years , No avercgas -vrs givon for the groups, hsnca
oc| - Tj - 0.2.
2. Although, exposure to vinyl chloride began as early as 1942,
occupational hygiene measurements began only in 1950. It is
likely that vinyl chloride concentrations may have been
somewhat higher prior to 1950 than after, so a portion of the
cohort may have experienced doses higher that those calculated.
Since it is a period of only 8 years and no drastic process
changes are documented, a relatively small valua for 73, namely
0.2, is selected. The factor a2 *s set equal to zero.
3. A continuous concentration analyzer has been in operation at
least one of the units since 1959. Measurements since 1950
have been fairly complete for the plant as a whole and,
apparently for the various job classifications, although this
is not clearly documented. Accordingly, aj and 73 are set
equal to 0.2.
it. The categories of exposure, "high", "medium", and "low", have
been broadly defined and no average values for the groups have
been given. Consequently, oe^ • 7^ • 0.3.
5. A serious recording bias results from classification by the
highest exposure experienced. An attempt at partial correction
is reflected in the assumptions about distribution of individ-
uals across exposure groups based on Table 2-113. The
uncertainty associated with this bias is still substantial,
however, and 05 has been assigned a value of 0.6. Since the
assumptions may possibly over correct the bias, 75 is 0.2.
2-242
-------
6. Since orea, not personal, samples are used to estimate
exposures, both ag and 75 have been given values of 0.05.
7. Some small amount of uncertainty is associated with the choice
of U.S. national mortality rates to represent the expected
mortality experience of the cohort. Both ag and -IQ have been
set equal to 0.05 to account for this effect.
The resulting values of a and i are 2.<» and 2.2, respectively, with the
resulting bounds on dose as represented in Table 2-115.
Fox and Collier (315) studied the mortality experience of over 7000
British workers exposed to vinyl chloride. The study covered persons
who mny have been exposed through industrial employment in Great Britain
between the years 19**0 and 197<*. Two factories employing 255t of the
cohort started production in 1969 and 1970; thus, follow-up was quite
short for a sizable fraction of the cohort. Two cases of liver angio-
sarcoma were found. Mortality from all malignant neoplasms was less
than expected, when compared to national rates. Dose response analyses
were presented by Fox ana Collier for all malignant neoplasms, liver
cancer, lung cancer, arid brain cancer.
Workers were classified into low, medium, and high exposure categories,
depending upon whether the time weighted average exposures in the job in
which they received their maximum exposure was less that 25 ppm, between
25 and 200 ppm, or greater than 200 ppm. Workers were also classified
as to whether their exposures were constant (most of the time) or
intermittent (occasional). No measurements of air concentrations made
before the mid-1960s were available.
This study involved workers from a number of different companies, and
each company classified exposures of its own employees No mention is
mode of the time required in the job with the maximum exposure level in
2-2<»3
-------
order for that job to be the basis for a worker's exposure classifica-
tion. These shortcomings make estimates of average exposures problema-
tical. Since the exposure ranges are the same as were used by Ott
et al . , and there is no evidence that the average exposures in the
various categories differed between the two studies, the same average
exposures are assumed for the Fox and Collier cohort as were assumed
earlier for the Ott et al. cohort — namely, averages of 12.5 ppm, 70 ppm,
and 300 ppm for 0-25 ppm, 25-200 ppm, and 200+ ppm constant exposures,
respectively. The corresponding concentrations for the intermittent
exposures are assumed to be 7 ppm, 30 ppm, and 150 ppm.
Table 2-116 displays the cohort, broken down by exposure intensity and
duration. Assumed average lengths of exposure are 5, 15, and 27 years
in the duration groups. Average exposures are estimated for the high,
medium, and low exposure categories, combining constant and intermittent
groups and the duration classes. As an example, the average cumulative
exposure for the high exposure group is estimated as
605 + 142 +34+60+35+11
• 2244 ppm-years.
Table 2-117 also contains the response data for all malignant neoplasms,
liver cancer, and lung cancer as well as the bounds on dose estimates.
The exposure estimation for the Fox and Collier cohort was even less
certain than for the group described by Ott et al . Specific issues are
discussed below:
1. Length of exposure is grouped, much the same as Ott et al .
grouped exposure duration, again without averages for the
groups. Moreover, "constant" and "intermittent" exposures are
not well-defined. The value for a-\ and TJ assigned in this
case is 0.3.
2-244
-------
2. Although the follow-up period began in 19<»0, exposure estimates
were available only since the mid-19601 s. The possibility for
ur.derestimction of early exposures is great, so 72 • 0.7
3. No documentation is providad on the completeness of the concen-
tration measurements done since the mid-1 960 's that form the
basis of the exposure estimates. A value of 0.3 is selected
for 0(3 and 73.
4. Once again, no averages for the exposure categories are given.
The same definitions were used as in the Ott et al . study, so
a^ and T*» ore the same also, namely 0.3.
5. Exposure classification is based on maximum exposure and no
data is provided thcK Allows for factoring-in time spent at
other, lower-exposure jobs. Although the upper bound need not
be modified in this case (75 - 0), the lower bound is affected
to a much greater extent than in the Ott et al . uncertainty
estimation. Consequently, 05 has been set equal to 0.9.
6. Substantial uncertainty exists with respect to the applicabili-
ty of the reported exposures. Several different facilities
were studied and each classified its own exposures, possibly
affecting consistency over the whole cohort. It is not known
if area samples were used to define exposures, although one
would suspect that they or "best guesses" were employed. The
problem of the definition of "constant" and "intermittent"
exposure also influences uncertainty here. The value of ag and
76 is °-3-
7. National rates were used to calculate expected numbers of
deaths. Fairly wide time intervals for the early part of
follow-up (1940-55, 1956-65) were compared to death rates of
single years (1951 and 1961 respectively). The factors
-------
The uncertainty factors a and 7 have the values 3.2 and 3.0. The bounds
on dose obtained by applying these factors (Table 2-117) are used to
investigate the sensitivity of the analysis to possible misestimaticn,
particularly of the exposures.
Buffler et al. (310) studied individuals employed in a vinyl chloride
monomer production plant that began operations in 1948. A total of 481
males employed for at least two consecutive months between 1948 and 1975
were included in their cohort. Dr. Buffler and her colleagues were kind
enough to supply a copy of the cohort history which allowed inclusion of
all white males (504) in the following analyses.
Cjmulative exposures were calculable for each individual once concentra-
tions of vinyl chloride were assigned to each job code. The original
Buffler et ol. publication provided estimated time-weighted overage
levels for the jobs entailing exposure to vinyl chloride, but only for
the time period 1971-1975. The authors state that from 1948 to 1960
exposure to concentrations of 200-500 ppm were not uncommon. Through
1960, the threshold limit value was 500 ppm. In 1961, the company
reduced its standard for exposure to 50 ppm (TWA). It was the authors'
opinion that levels before 1971, though higher, would have been such
that the ratio of levels for any two jobs would be roughly the same as
the ratio after 1971. Using this information, time-dependent concen-
tration estimates have been derived and are shown in Table 2-118.
The resulting dose and response data for all malignant neoplasms and
lung cancer are presented in Table 2-119.
Uncertainty with respect primarily to exposure estimates in this
analysis ore as follows, discussed in cerms of the factors a^ and TJ.
2-246
-------
The uncertainty factors a ana i have the values 3.2 and 3.0. The bcunds
on dose obtained by applying these factors (Table 2-117) are used to
investigate the sensitivity of the analysis to possible misestimotion,
larticularly of the exposures.
Buffler et al. (510) studied individuals employed in a vinyl chloride
monomer production plant that began operations in 1948. A total of 481
males employed for at least two consecutive months between 1948 and 1975
were included in their cohort. Dr. Buffler and her colleagues were kind
enough to supply a copy of the cohort history which allowed inclusion of
all white males (504) in the following analyses.
Cumulative exposures were calculable for each individual once concentra-
tions of vinyl chloride were assigned to each job code. The original
Buffler et al. publication provided estimated time-weighted average
levels for the jobs entailing exposure to vinyl chloride, but only for
the time period 1971-1975. The authors state that from 1948 to 1960
exposure to concentrations of 200-500 ppm were not uncommon. Through
1960, the threshold limit value was 500 ppm. In 1961, the company
reduced its standard for exposure to 50 ppm (TWA). It was the authors'
opinion that levels before 1971, though higher, would have been such
that the ratio of levels for any two jobs would be roughly the same as
the ratio after 1971. Using this information, time-dependent concen-
tration estimates have been derived and ore shown in Table 2-118.
The resulting dose and response data for all malignant neoplasms and
lung cancer are presented in Table 2-119.
Uncertainty with respect primarily to exposure estimates in t'us
analysis are as follows, discussed in terms of thn foctors a^ and i^.
2-246
-------
1. No measurements of vinyl chloride concentrations were available
before 1971. Extrapolation to those periods between 1948 and
1970 has been accomplished with reference to certain assump-
tions about the conditions prevailing at that time, but this is
highly uncertain. The uncertainty affects both upper and lower
bounds on dose estimates, since the extrapolation may indeed
overestimate early exposure. Consequently, o<2 ar|d 12 nave both
been assigned o value of 0.8.
2. The completeness of the measurements of vinyl chloride concen-
tration is not documented. A few of the specific job titles
lacked estimates of associated exposure, and the original
authors had to rely in part on the judgorr,. it of plant
supervisors and others with experience at the facility to
derive estimates. A value of 0.3 is assumed for 03 and 73.
3. A minor amount of uncertainty is attributable to the question
of the applicability of the reported exposures. The value 0.05
is given to ag and Tg to reflect the use of area samples and
"classification groups" instead of personal samples.
<». Since one of the endpoints considered with this cohort is
respiratory cancer, it would be most appropriate to use
expected numbers of deaths derived by reference to smoking-
specific death rotes. This was not possible, but it is not
known if the smoking behavior of the cohort differed from
national patterns. Both 03 and IQ are assigned a value of 0.1.
All other factors (such as recording bias, grouping without average
»alues, length of exposure uncertainty) do not apply. The resulting
valuo for a and -,, 2.25, is used to derive reasonable bounds for the
dose variable (Table 2-119).
Heldaos et al . (316) described a cohort of workers employed by a
Norwegian, producer of vinyl chloride monomer and polyvinyl chloride.
2-247
-------
The manufacture of these products began in 1950. Of 1233 workers who
started work before 1975 and who worked more than one month, the authors
selected for study 454 males who were employed for at least one year and
who began before 1970. The follow-up period for which cancer incidence
was determined (the studies previously described used cancer mortality
as their endpoint), extended from 1953 to 1979. Norwegian notional
rates were used for comparison purposes.
According to the report, the plant design and process operation must
have entailed high concentrations of vinyl chloride, but no industrial
hygiene surveys were performed before 1974. Based on sporadic measure-
ments carried out with an "explosion-meter", interviews with workers,
and an odor threshold of about 500 ppm, the following vinyl chloride
monomer concentration history was assumed:
1950 to 1954: 2000 ppm
1955 to 1959: 1000 ppm
1960 to 1967: 500 ppm
1968 to 1974: 100 ppm
These concentrations, coupled with work histories of the cohort members,
allowed the authors to calculate cumulative exposures and classify each
member according to his cumulative exposure by the end of follow-up.
Table 2-120 displays the classification presented by Heldaas et al.
That table also presents the observed and expected incidence of all
The bounds on the doses have been derived by assuming a value for a and
1 of 2.8. That value is itself derived by consideration of the
following contributions to uncertainty.
1. Formal measurement of vinyl chloride concentrations began only
in 1974. The need to extrapolate back to 1950 introduces great
uncertainty. Since some guess was ventured as to the high
2-248
-------
values that might have been prevalent, uncertainty above and
below the best estimates is present. Both aj and 12 are given
a value of 0.8.
2. The exposure groups defined by the authors are extremely wide
and are presented without averages. Arithmetic averages have
been assumed for the groups (cf. Table 2-120) but, of course,
there is no way to know how accurate these may be. A value of
0.3 is assigned to a^ and 74.
3. Some uncertainty is associated with the presumed applicability
of the exposures reported. The authors presented information
indicating that certain jobs entailed greater vinyl chloride
exposure that others. Yet, their extrapolation was, for the
most part, based on time but not job classification. Whether
this would entail over- or under-estimation, in general, is not
known. Consequently, the value of 0.3 is assigned tc ag and
le-
ft-. The definition of the dose groups in Table 2-120 is based on
one measure for each cohort member, i.e. his total cumulative
exposure. In reality, as each member continues to work at the
plant, his cumulative exposure variable increases and passes
through one or more of the groups defined. Tt. would be better
to attribute person-years, as opposed to individuals, to the
groups and calculate expected values on that basis. Moreover,
the authors indicate that cancer incidence in the local area
surrounding the facility differ from the national rates.
Consequently, the factors accounting for uncertainty of the
expected response rate, ag and IQ, are given a value of 0.4 to
reflect these uncertainties and that relating to use of non-
smoking-specific rotes for comparison.
The dose-response model has been fit to the data from the four studies
of Ott ot ol., Fox and Collier, Buffler et al., and Heldaas et al. The
2-249
-------
resulting potency parameters are given in Table 2-121 .
The potency parameters have been used, in turn, to estimate RRDs. A
mixed bag of RRO estimates result (Table 2-122). Three of the four
studies, including the morbidity study, cannot rule out the hypothesis
that vinyl chloride has no effect on overall carcinogenesis. The same
mixed results are apparent for respiratory cancer. Only one study
contained data relevant to liver cancer risk (315) and that data appears
to indicate a definite risk associated with vinyl chloride exposure.
2-250
-------
Table 2-111
DURATION OF EXPOSURE BY LEVEL OF EXPOSURE FOR EACH
EXPOSURE GROUPING (ARSENIC WORKERS EXCLUDED),
VINYL CHLORIDE-EXPOSED COHORT0
Number of Employees
With Cumulative Exposure of:
Exposure Group and
Level of Exposure <1 yr 1-9 yr 10-19 yr 20+ yr
High-exposure group
Low-to-high levels 33 64 39 27
High levels only 73 71 19 0
Intermediate exposure group
Low-to-intermediate levels 29 38 6 0
Intermediate levels only 32 36 5 0
Low-exposure group
Low levels only 127 50 3 1
Unmeasured exposure group
Unmeasured or low levels 39 44 • 16 6
°0tt et al. (314)"
2-251
-------
Toble 2-1U
OBSERVED AND EXPECTED DEATHS BY EXPOSURE INTENSITY
AND DURATION OF EXPOSURE, 1942-1973, OTT ET AL. (314)
VINYL CHLORIDE-EXPOSED COHORT (ARSENIC WORKERS EXCLUDED)
Deaths Due to
Exposure Intensity Malignant Neoplasms
and Duration Observed Expected
High Exposure
< 1 yr 3 2.2
> 1 yr 6 2.9
Intermediate Exposure
< 1 yr 0 1 .4
> 1 yr 2 1.5
Low Exposure
< 1 yr 0 2.4
> 1 yr 1 1.7
2-252
-------
Toble 2-115
DOSE AND RESPONSE INFORMATION FOR THE VINYL
CHLORIDE-EXPOSED COHORT OF OTT ET AL. (514)
(ARSENIC WORKERS EXCLUDED)
Exposure Intensity
and Duration
Cumulative Exposure
(ppm-yeors)
Lower
Bound
Best
Estimate
Upper
Bound
Deaths Due to
Malignant Neoplasms
Observed Expected
High Exposure
< 1
> 1
yr
V
105
1040
251
2495
552
5489
Intermediate Exposure
2.2
2.9
< 1
> 1
yr
yr
16.7
182
40
88
964
0
2
1 .4
1.5
Low Exposure
< 1
> 1
yr
V
2.58
31.2
6.2
75
13.6
165
2.4
1.7
2-253
-------
Table 2-116
LEVELS OF EXPOSURE AND LENGTHS OF EXPOSURE FOR MEN IN THE
FOX AND COLLIER (513) VINYL CHLORIDE-EXPOSED COHORT
Length of
Exposure
(years)
0-9
10-19
20+
Total
(*)
Levels of
Constant
High
605
142
34
Medium
1202
117
92
781 1i»12
(10.5*)(19.0*)(
Low
890
263
114
1266
17.1*)
Exposure
Intermittent
High Medium
60
35
11
1857
210
130
Low
1094
351
202
106 2197 1647
(1.4*)(29.7*)(22.2*)
Total No. (*)
5708
(77. C*)
1118
(15.1*)
583
(7.9*)
7409
(100*)
Constant - Most of the time.
Intermittent • Occasionally.
High - greater than 200 ppm.
Medium • between 25 and 200 ppm.
Low - lass than 25 ppm.
2-234
-------
Table 2-117
DOSE AND RESPONSE INFORMATION FOR .'HE FOX AND
COLLIER (ill) COHORT OF VINYL CHLORIDE-EXPOSED WORKERS
Cumulative Exposure
(ppm-yeors) All Malignant
Lower Best Upper Neoplasms Liver Cancer Respiratory Cancer
Bound Estimate Bound Observed Expected Observed Expected Observed Expected
27.5 88 264 53 57.5 1 0.75 21 23.2
103 331 993 53 59.7 1 0.77 23 2<*.i»
70'. 22kk 6732 9 9.6 2 0.13 2 3.7
M
(/I
in
-------
Table 2-118
ESTIMATED VINYL CHLORIDE CONCENTRATIONS FOR
BUFFLER ET AL. COHORT, BY TIME AND JOB CLASSIFICATION
Concentration (ppm)
1948-
Job Classification 1960
Control Lab Personnel
Development Lab Personnel
Production Personnel
Control A
Control B
Control C
Class 3 Operators
Loaders and Plant Men
Class 1,2 Operators
Head Packaging Operator
Material Handling Operator
Packaging Operator,
Service Technician
Certain Supervisory Positions
Production Superintendent,
Assistant Production Superintendent,
Production Engineer, h & 0 ffngineer,
Engineer, Safety Engineer, Sr .
Production Engineer
Parts Technician, Sr .
Manager
Other
Maintenance Personnel
Boilermaker, Apprentice
Welder, Apprentice
Machinist, Apprentice, Helper,
Crew Leader
Pipefitter, Apprentice, Helper
Utility Man
Instrument Technician
Production Foreman, Shift Forenan
Maintenance Foreman
Utility Crew Leader, Rotating Shift
Foreman, Foreman, Other
Other Personnel
446
82
154
56
88
114
142
246
4
148
38
4
6
36
28
4
22
34
4
64
86
8
50
76
1961-
1970
44. G
8.2
15.4
5.6
8.8
11.4
14.2
24.6
0.4
14.8
3.8
0.4
0.6
3.6
2.8
0.4
2.2
3.<»
0.4
6.4
8.6
0.8
5.0
7.6
1971-
1975
22.3
4.1
7.7
2.8
4.4
5.7
7.1
12.3
0.2
7.4
1.9
0.2
0.3
1.8
1 .4
0.2
1 . 1
1 .7
0.2
3.2
4.3
0.4
2.5
3.8
2-256
-------
Table 2-119
DOSE AND RESPONSE DATA FOR THE VINYL CHLORIDE-EXPOSED
COHORT OF BUFFLER £T AL. (3_H))
Cumulative
Exposure (ppm-yrs)
All Malignant
Neoplasms
Respiratory Cancers
Observed Expected Observed Expocted
0-75
(1.7, 3.9, 8.8)°
75 - 150
. 106, 238)°
150 - 300
(91, 204, 459)°
300+
(378, 859, 1912)°
3.36
0.56
0.69
1.62
1 . 10
0.22
0.27
0.64
°In parentheses are the lower bound, best estimate, and upper bound on
dose, respectively.
2-257
-------
Toble 2-120
DOSE AND RESPONSE DATA FOR THE VINYL CHLORIDE-EXPOSED
COHORT OF HELDAAS £T AL. (516)°
Cumulative
Exposure (ppm-yrs)
All Malignant-
Neoplasms
Observed Expected
Respiratory Cancers
Observed Expected
0 - 500
(89, 250, 700)b
500 - 2500
(536, 1500, 4200)b
2500+
(1339, 3750, 10500)b
7.06
3.10
9.95
1.01
O.tO
°Incidence, not mortality, is presented for observed and expected
responses.
bln parentheses are the lower bound, best estimate, and upper bound on
dose, respectively, derived for the exposure groups. These values were
not provided by the authors.
2-258
-------
Table 2-121
POTENCY PARAMETER ESTIMATES FOR VINYL CHLORIDE
Potencies ((ppm-yrs)~^ )
Dose Lower Upper
Study Response Measure Limit0 MLE Limit0
Ott All Upper 3.52E-5* 1 . 99E-4 4.26E-4
et al. Malignant Bounds
(314) Mortality
(cni-squared Best 7.74E-5 4.38E-4* 9.38E-4
(5) • 4.3) Estimates
Lower 1.86E-4 1.05E-3 2.25E-3*
Bounds
Fox * All Upper -7.03E-5* 0.00 4.09E-5
Collier Malignant Bounds
(313) Mortality
(cni-squared Best -2.11E-4 0.00* 1.23E-**
(2) • 1.14) Estimates
Lower -6.75E-4 0.00 3.93E-4"
Bounds
Liver Cancer Upper 4.67E-4* 1.46E-3 3.09E-3
Mortality Bounds
(cni-squared
(2) - 0.67) Best 1.40E-3 4.38E-3* 9.28E-3
Estimates
Lower 4.49E-3 1.41E-2 2.98E-2"
Bounds
Respiratory Upper -1.29E-4* O.CO 5.16E-5
Cancer Bounds
Mortality
(cni-squared Best -3.87E-4 0.00" 1.55E-1*
(2) - 1.07) Estimates
Lower -1.24E-3 0.00 4.95E-4*
Bounds
2-259
-------
Toble 2-121 (continued)
POTENCY PARAMETER ESTIMATES FOR VINYL CHLORIDE
Potencies (
Dose Lower Upper
Study _ Response _ Meosure _ Limit0 _ MLE _ Limit0
Buffler All Upper -2.10E-4* 3.38E-<» 1.25E-3
at ol . Malignant Bounds
(310) Mortality
(chi-squared Best -*».70E-4 7.<»7E-V 2.78E-3
(3) • 5.8) Estimates
Lower -1.06E-3 1.71E-3 6.34E-3"
Bounds
Respiratory Upper 3.58E-U* 1.73E-3 3.92E-3
Cancer Bounds
Mortality
(chi-squored Best 7.91E-4 3.8<»E-3" 8.7.'E-3
(3) • 5.3) Estimates
Lower 1.81E-3 8.7<»E-3 1.98E-2"
Bounds
Heldaas All Upper -3.78E-6* 3.52E-5 8.35E-5
et ol. Malignant Bounds
(316) Morbidity
(chi-squored Best -1.06E-5 9.85E-5* 2.3(«E-<»
(2) • 0.31) Estimates
Lower -2.96E-5 2.76E-"* 6.55E-4"
Bounds
Respiratory Upper -1.53E-b" 9.01E-5 2.61E-'»
Cancer Bounds
Morbidity
(chi-squored Best -i*.28t-5 2.b2E-<»" 7.I2E-4
(2) < 1.4) Eitimatos
Lower -1.20E-U 7.06E-I- 2.0bE-3'
Bounds
confidence limits are shown
An asterisk marks the parameters used to derive HMD estimates.
2-260
-------
Table 2-122
RRD ESTIMATES FOR VINYL CHLORIDE (ppm)
Estimation
Study
Ott
ot al.
(ili)
Fox *
v Collier
w (3JL1)
en
Buffler
et al.
(114)
Response Method
All
Malignant
Mortality
All
Malignant
Mortality
Liver
Cancer
Mortality
Respiratory
Cancer
Mortality
All
Malignant
Mortality
Respiratory
Cancer
Mortality
1
2
1
2
1
2
1
2
1
2
1
2
RRD|_
3.50E-5
4.60E-5
2.00E-4
2.63E-4
4.72E-4
5.13E-4
5.49E-4
6.34E-4
1.24E-5
1.63E-5
1.37E-5
1.58E-5
Level of
10-6
MLE RRDU
1.80E-4 2.24E-3
2.37E-4 2.95E-3
CO 00
m oo
3.20E-3 3.01E-2
3.48E-3 3.27E-2
at CD
00 00
1.05E-4 »
1.39E-4 oo
7.09E-5 7.61E-4
8.18E-5 8.78E-4
Extra Risk
RROL
8.75
1 .47E+1
5.00E+1
8.43E+1
1.18E+2
1.61E+2
1.37E+2
1.95E-f2
3.10
5.23
.5.43
4.87
0.25
MLE RRDU
4.50E+1 5.60E+2
7.58E+1 9.43E+2
00 00
(JO. 00
8.01E+2 7.52E+3
1.09E+3 1.03E+4
CD 00
00 OD
2.64E+1 »
4.44E+1
1.77E+1 1.90E+2
2.52E+1 2.70E+2
-------
Table 2-122 (continued)
RSD ESTIMATES FOR VINYL CHLORIDE (ppm)
Level of Extra Risk
rvj
i
M
O)
rvj
Study
Heldaas
et al.
(316)
Estimation 10~6
Response Method RRD|_ MLE RRDU
All 1 5.74E-5 3.82E-4 »
Malignant
Morbidity 2 8.78E-5 B.S^E-* «
0.25
RRDt MLE RRDU
1.44E+1 9.55E+1 to
2.76E+1 1.8(tE+2 »
Respiratory 1
Cancer
J»lorbidi.ty 2_
I.ISE-'t 9.19E-4
1.32E-4 1.07E-3
2.30E+2
3.29E+2
-------
Summary of Results
Of the two basic methods described above, Method 2 more appropriately
considers the timing of exposure in the calculation of RRD estimates.
Note that the Method 2 RRD estimates are uniformly larger than the
corresponding Method 1 RRD estimates, though by only a small factor.
This is a result of the life-table aspect of Method 2; the probability
of getting cancer in the older age groups is discounted by the likeli-
hood of dying at earlier ages of other causes. Method 1 lacks this
feature. Consequently, this summary of results is framed in terms of
the Method 2 RRD estimates.
The summary also includes only the estimates corresponding to an extra
risk of 0.25. By emphasizing these estimates, we hope to avoid diffi-
culties associated with extrapolation of results to low levels of risk.
In most case*, the RRDs corresponding to an extra risk of 0.25 are not
far from the range of doses reported in the epidemiological studies
supplying the necessary quantitative data.
The Method 2 estimates of doses corresponding to an extra risk of 0.25
determined for all the responses observed in the selected studies are
displayed in Figure 2-1. The RRD estimates have been converted to the
units of mg/kg/day and correspond to the specific scenario chosen, *»5
years of exposure (2^0 days per year) starting at age 20. Figure 2-1
indicates the particular responses chosen to represent the RRD estimates
for each chemical; Table 2-123 summarizes the selected values. These
are the values that define the human intervals and point estimates to
which the bioossay results are compared.
2-263
-------
DISCUSSION
For the most part, prospective cohort studies and case-control reports
have supplied the data necessary for the quantitative approach adopted
here. Both forms of study have limitations that influence our ability
to derive risk estimates. Case-control studies concentrate on a single
carcinogenic response, and look at differences in exposure between the
cases and the controls. It is possible that other carcinogenic' effects
of the same exposure are missed with this format, perhaps biasing the
risk estimates. Since we opted to examine only single responses (or all
malignant neoplasms when a specific response was not suggested) this may
not be a particular problem in the current analysis. However, case-
control studies present some difficulties with respect to sequencing of
events; it is not easy to determine in some cases which exposures
preceded the disease or which exposures may have occurred or been
modified because of the disease (cf. the discussion of estrogens).
Moreover, when reliance on patients recall of timing and intensity of
exposure has been necessary, uncertainty is compounded.
Prospective studies suffer from many of the same problems. Exposure in
the less recent past is often highly uncertain. Observed responses may
have nothing to do with exposure to the suspected carcinogen, a fact
that one attempts to reflect in the calculation of expected (without
exposure) numbers of careers. But, once again, uncertainty is asso-
ciated with those estimates of expected numbers. Finally, incomplete
follow-up means incomplete information on the carcinogenic potential of
the chemical in question.
Many of these uncertainties, particularly those related to exposure and
dose calculations, are included in the estimation of reasonable bounds
on the RRDs. The subfactors used to calculate the uncertainty factors,
-------
uncertainty represented by a subfactor might affect the estimation of
the "true" dose or exposure. So, for example, the subfactors relating
to categorization of exposure (o^ and 7^) are selected from the interval
[0, 0.3] whereas those relating to recording biases (05 and 75) are
chosen from [0, 1.0]. The possible effect on dose estimation related to
the latter factor (which could --esult from classification of workers by
their maximal exposure) is deemed to be greater than the possible effect
on dose estimation related to the former factor (which is primarily the
estimation of an average value for a range of exposures). On the other
hand, specific cases exist in which the uncertainty associated with a^
and 74. is greater than that associated with ag and 75, because, under
the the guidelines for selecting the subfactors, we are at liberty to
select them so that 05 < a^ < 0.3 and 75 < 7^ < 0.3.
If, for example, the only uncertainty is associated with the categoriza-
tion of exposure (a^ and 7^), in that averages had to be estimated for a
range of exposures, and if the values of a^ « 7^ - 0.3 were selected,
then this implies that the "true" average cumulative exposure may be as
much as a factor of 1 .3 larger or smaller than the estimated average.
Obviously, any such selections are themselves estimates, and, moreover,
somewhat arbitrary estimates.
Thus, in this analysis, specific ranges have been defined and, for
subfactors representing specific elements of uncertainty, values have
been selected from those ranges. A specific value is selected somewhat
arbitrarily and subjectively as providing a "reasonable" multiplicative
factor for reflecting that uncertainty in the dose estimates. It has
not been possible to investigate the effect on the analyses (comparisons
with the bioassay-based results) of choosing other reasonable ranges or
particular values for the uncertainty subfactors. A simulation study of
that issue might be informative.
2-265
-------
In the meantime, it is appropriate to view the uncertainty bounds for
cumulative exposure in two ways. First, within the epidemiologicol
analysis, one can see the bounds as providing information on the sensi-
tivity of the analysis to reasonable changes in the exposure data.
Second, in the context of comparisons of the bioassay-based estimates to
the epidemiologically derived estimates (cf. Volume 3 of this report),
the intervals defined by the combination of statistical and exposure
uncertainty (those shown in Table 2-123) are important in determining
the degree of similarity (the fit) of the animal estimates and the human
estimates. Hence, as long as the uncertainty estimates are consistent
from study to study and chemical to chemical, even if they are somewhat
arbitrary, then the intervals defined in part by those uncertainty
factors provide a consistent measure of the weight that should be
attached to any given chemical when the comparison is made. Chemicals
with wider intervals (more uncertainty) should not influence the evalua-
tion of how well the animal-based results predict the human-based
results as much as a chemical with a shorter interval. It is in this
sense that the uncertainty factors and the intervals defined by
reference to the range of reasonable doses contribute most heavily to
the present analysis.
In closing, one further feature of the epidemiological analysis requires
comment. For comparison with the bioassay-based results, a single
triple of Rh.O estimates has been selected (Table 2-123). That triple is
selected from among all the responses and studies analyzed, but it is
not the result of combining responses or studies. That procedure is
''ollowod for several reasons. First, it is not clearly appropriate to
combine studies that may differ in their definition of cohort member-
ship, latency period, cumulative exposure classes, control series, or
even endpoint. The methodology for combining case-control study results
is unknown and even for prospective studies the appropriate method is
not clear (317). Second, the studies for which quantitative data appro-
2-266
-------
priote for the approach adopted here are usually a small subset of the
whole set of epidemiological information. While we wanted to derive
quantitative estimates, we also wished to have the flexibility to select
as our final answers numbers that were as reflective of the consensus
from the epidemiological literature as possible. This entails selecting
a particular carcinogenic response that is generally agreed to be
associated with exposure to the chemical of interest and selecting a
study yielding results consistent with the consensus, at least in terms
of having high potency, low potency, or (perhaps) no carcinogenic effect
in humans. To avoid bias when comparing results in animals and humans,
all such selections were made independently of the animal results.
In general, the RRD estimates selected are reflective of the whole of
the epidemiological data. In some instances, very similar estimates
were obtained from each study analyzed (cf. Figure 2-1, ethylene oxide
or methylone chloride). We have chosen estimates from particular
studies, but some other selection may also be appropriate. We have
not investigated the effect such choices may have on the comparisons
with the bioassay-based estimates, although we expect it to be small.
2-267
-------
Toble 2-123
RRD ESTIMATES0 SELECTED FROM THE EPIDEMIOLOGIC DATA
Chemical
RRDn.1
RRDH
RRDH.u
Af latoxin
Arsenic
Asbestos
Benzene
Benzidine
Cadmium
Chlorambucil
Chromium
Cigarette smoke
DES
Epichlorohydrin
Estr ogens
Ethyleno oxide
Isoniazid
Melphalan
Methylene chloride
Nickel
PCB's
Phenacetin
Reserpine
Saccharin
Trichloroethylene
Viny1. chloride
4.30E-4
2.05E-2
1.20E-2
1.20E-H
1.20E-4
6.00E-2
1.57E-2
4.41E-4
2.79E+2
4.63E-18
1 .79E-1
1.67E-3
4.69E-1
1 .45
1 . 04E-4
8.63E-H
2.46E-1
2.24E-2
6.91
4 . 08E-4
1 . 90E+1
2.61
5.90E+1
1.6«fE-3
5.84E-2
7.00E-2
2.53E+1
6.^3E-if
1 .74E-1
3.91E-2
3.89E-3
6.41E+2
Hf7E-3
X
5.13E-3
2.24
6.78E+1
3.15E-1*
00
6.80E-1
1 .51E-1
2.00E+1
1.91E-3
00
6.48E+1
4.00E+2
1.08E-i
1 .57E-1
«*.10E-1
7.61E+1
2.56E-3
7.05E-1
1.05E-1
1.15E-1
1.47E+3
4.61E-2
00
1.89E-2
1 .59E-H
CO
6.86E-4
ac
1 .36E+1
X
6.50E+1
03
DO
OC
3.76E+3
°The estiamtes correspond to an extra risk of 0.25 and to the exposure
scenario selected, exposure 240 days per year for 45 years starting at
age 20. The estimates are expressed in terms of mg/kg/day.
2-268
-------
Aflatcxi:
Arsenic
..: ver
rerr: ratorv
Asbestos
• • • * all
• lunf
r.esothelior.a
Eenzene
zi dine
* bladder (".or: : Ji tv ';
re.-er.t at i CM of :-:;: ••.•-* >.;i* r-;- btn;:;e1 fcr
Cb.e~.icai." ar;J Kdcr. ;'ut':* iv" ,'ite cf Actici.
2-269
-------
Liiror.iu.T.
all t-
all (morbidity) •-
-**»
respiratory •*•
(morbidity)
Cigarette Smoke
DES
gastrointestinal
all •-
lune *
Epi chlorohydri n
estrogens
breart (r.orbiditv)
all
16*
-3
Fir-re T-l (continued)
Represent at i or: if :7^ Estimates Obtained for Al 1 •
Cher-icals ar;-i iincii Putative Site of Action
2-270
-------
* leuXeria (r.orbi dity }
Nickel
resri ratcrv
•ii ••ertive *»-
iver
2-271
-------
resri rat crv
2-272
-------
REFERENCES
1. Breslow, N. and Day, N. (1980). Statistical Methods in Cancer
Research Volume 1. The Analysis of Case-Control Studies.
International Agency for Research on Cancer. Lyon.
2. International Agency for Research on Cancer (1982). Chemicals, •
industrial processes and industries associated with cancer in
humans. IARC Monographs on the Carcinogenic Risk of Chemicals to
Humans Supplement 4.
3. Sankale, M., Gendron, Y., and Courbil, L. J. (1983). Le Cancer
primitif du foie en miliuu tropical. Medecine Tropicale
43(3):239-252.
4. Wang, Y., Lan, L., Ye, B. , Xu, Y., Liu, Y. and Li, W. (1983).
Relation between geographical distribution of liver cancer and
climiate-af latoxin B-| in China. Scientia Sinico 26( 11): 1166-1175 .
5. Store, C., Dvorackova, I., and Ayraud, N. (1981). Characterization
of aflatoxin B-\ (AFB) in human liver cancer. Research
Communications in Chemical Pathology and Pharmacology 31(1):77-84.
6. Stoloff, L. (1983). Aflatoxin as a cause of primary liver-cell
cancer in the United States: A probability study. Nutrition and
Cancer 5(3-4):165-186.
7. Van Rensburg, S. J., Kirsipuu, A., Coutinho, L. P., and Van Der
Watt, J. J. (1975). Circumstances associated with the contami-
nation of food by aflatoxin in a high primary liver cancer area.
South African Medical Journal 49:877-883.
8. Keen, P. and Martin, P. (1971a). Is aflatoxin carcinogenic in man?
The evidence in Swaziland. Tropical and Geographical Medicine
23:44-53.
9. .Nizami, H. M. and Zuberi, S. J. (1977). Aflatoxin and liver
cancer in Karachi, a preliminary survey. Journal of the Pakistan
Medical Association June 351-352.
Keen, P. and Martin, P. (1971b). The toxicity and fungal infesta-
tion of foodstuffs in Swaziland in relation to harvesting and
storage. Tropical and Geographical Medicine 23:35-43.
Hayes, R. B., Van Nieuwenhuize, J. P., Raatgever, J. W. , and Ten
•'ate, F. J. W. (1984). Aflatoxin exposures in the industrial
••tting: An epidemiological study of mortality. Food and
-"ical Toxicology 22(1):39-43.
2-273
-------
12. Bulatoo-Jayme, J., Almero, E. M. , Castro, M. C. A., Jardeleza, M.
T. R. and Salamat, L. A. (1982). A case-control dietary study of
primary liver cancer risk from aflatoxin exposure. International
Journal of Epidemiology 11(2 ): 1 12-119 .
13. Linsell, C. A. and Peers, F. G. (1977). Aflatoxin and liver cell
cancer . Transactions of the Royal Society of Tropical Medicine
and Hygiene 71(6) : ^7
1U. Carlborg, F. V). (1979). Cancer, mathemathical models and
aflatoxin. Food and Chemical Toxicology 17:159-166.
15. Keyfitz, F. (1968). World Population - An Analysis of Vital Data.
University of Chicago Press, Chicago.
16. Shank, R. C. , Bhamarapravati , N. , Gordon, J. E., and Wogan, G. N.
(1972). Dietary aflatoxin and human liver cancer. IV. Incidence
of primary liver cancer in two municipal populations of Thailand.
Food and Chemical Toxicology 10:171-179.
17. Peers, F. G. and Linsell, C. A. (1977). Dietary aflatoxins and
human primary liver cancer. Annals of Nutrition and Metabolism
31:1005-1018.
18. Peers, F. G., Gilmon, G. A., and Linsell, C. A. (1976). Dietary
aflatoxin and human liver cancer. A scudy in Swaziland.
International Journal of Cancer 17:167-176.
19. Department of Health and Human Services (1982). Vital Statistics
of the United States, 1978. Volume II - Mortality, Port A.
National Center for Health Statistics, Hyattsville, Maryland.
20. International Agency for Research on Cancer (1980). Some metals
and metallic compounds. IARC Monographs on the Evaluation of the
Carcinogenic Risk of Chemicals to Humans 23:39-1«t1.
21. Neubauer, 0. (19^7). Arsenical cancer: A review. British
Journal of Cancer 1:192-251.
22. Lander, J. J., Stanley, R. J., Sumner, H. W. , Boswell, D. C., and
Aach, R. D. (1975). Angiosarcoma of the liver associated with
Fowler's solution (potassium arsenite). Gostroenterology 68:
1582-1586.
23. Root, J. W., Wald, A., Mendelow, H., and Pataki, K. I. (1982).
Hepatic angiosarcoma associated with short-term arsenic ingestion.
The American Journal of Medicine 73:933-936.
2-27^
-------
24. Tseng, W. (1977). Effects and dose-response relationships of skin
cancer and blackfoot disease with arsenic. Environmental Health
Perspectives 19:109-119.
25. Tseng, W., Chu, H., How, S., Fong, 0., Lin, C. and Yen, S. (1988).
Prevalence of skin cancer in an endamic area of chronic arsenicism
in Taiwan. Journal National Cancer Institute 40:453-462.
26. Paton, G. R. and Allison, A. C. (1972). Chromosome damage in
human cell cultures induced by metal salts. Mutation Research
16:332-336.
27. Petres, J., Baron, D., and Hagedorn, M. (1977). Effects of
arsenic cell metabolism and cell proliferation: Cytogenetic and
biochemical studies. Environmental Health Perspectives 19:
223-227.
28. Bencko, V. (1977). Carcinogenic, teratogenic, and mutagenic
effects of arsenic. Environmental Health Perspectives 19:179-182.
29. Environmental Protection Agency (1984). Health Assessment
Document for Inorganic Arsenic. EPA-600/8-83-021F.
30. Hill, A. B. and Faning, E. L. (1948). Studies in the incidence of
cancer in a factory handling inorganic compounds of arsenic.
British Journal of Industrial Medicine 5:1-6.
31. Kuratsune, M., Tokudome, S., Shirakusa, T., Yoshida, M. ,
Tokumitsu, Y. , Hayano, T., and Seita, M. (1974). Occupational
lung cancer among copper smelters. Internotionol Journal of
Cancer 13:552-558.
32. Tokudome, S. and Kuratsune, M. (1976). A cohort study on
mortality from cancer and other causes among workers at a metal
refinery. Internotionol Journal of Cancer 1^:310-317.
33. Ott, M. G., Holder, 8. B., and Gordon, H. L. (1974). Respiratory
cancer and occupational exposure to arsenicals. Archives of
Environmental Health 29:250-255.
34. Enterline, P. E. and Marsh, G. M. (1981). Mortality Among
Workers Exposed to Arsenic and Other Substances in a Copper
Smelter. Presented at the Arsenic Symposium, November 4-6, 1981.
35. Pinto, S. and Bennett, B. (1963). Effect of arsenic trioxide
exposure on mortality. Archives of Environmental Health 7:
583-591.
2-275
-------
36. Pinto, S. S., Enterline, P. E., Henderson, V., and Varner, M. 0.
(1977). Mortality experience in relation to a measured arsenic
trioxide exposure. Environmental Health Perspectives 19:127-130.
37. Enterline, P. E. and Marsh, G. M. (1980). Mortality studies of
smelter worKers. American Journal of Industrial Medicine
1:251-259.
38. Lee, A. M. and Fraumeni, J. F. (1969). Arsenic and respiratory
cancer in man: An occupational study. Journal of the Notional
Cancer Institute 42(6):1045-1052.
39. Lubin, J. H., Pattern, L. M., Blot, W. J., Tokudome, S., Stone, B.
J., and Fraumeni, J. F. (1981). Respiratory cancer among copper
smelter workers: Recent mortality statistics. Journal of
Occupational Medicine 23(11):779-78*.
40. Higgins, I., Welch, K., Oh, M., Bond, G.. and Hurwitz, P. (1981).
Influence of arsenic exposure and smoking on lung cancer among
smelter workers: A pilot study. American Journal of Industrial
Medicine 2:33-41.
41. Welch, K., Higgins, I., Oh, M., and Burchfiel, C. (1982). Arsenic
exposure, smoking, and respiratory cancer in copper smelter
workers. Archiejs of Environmental Health 37(6):325-335.
42. Lee-Feldstein, A. (1983). Arsenic and respiratory cancer in
humans: Follow-up of copper smelter employees in montana.
Journal of the Notional Cancer Institute 70(4):601-610.
43. Crump, K. S. and Ng, T-H. (1983). Quantitative Risk Assessment
for Environmental Exposure to Inorganic Arsenic. Unpublished
report.
44. Consumer Product Safety Commission (1983). Chronic Hazard
Adviso-y Panel on Asbestos. U.S. Consumer Product Safety
Commission, Directorate for Health Sciences, Washington, D.C
45. Occupational Safety one* Health Administration (1983). Quantita-
tive Risk Assessment 'or Asbestos Related Cancers. Office of
Carcinogen Standards. OSHA Contract J-9-F-2-0074.
46. Environmental Protection Agency (1985). Asbestos Health Effects
Update. Office of Health and Environmental Assessment,
Washington, O.C.
47. Ontario Royal Commission (1984). Report of The Royal Commission
on Matters of Health and Safety Arising from the Use of Asbestos
in Ontario. Ontario Ministry of the Attorney General.
2-276
-------
48. National Academy of Science (1984). Asbestiform Fibers:
Nonoccupational Health Risks.
49. Doll, R. and Peto, J. (1985). Effects on Health of Exposure to
Asbestos. Prepared for Health and Safety Commission. London.
50. McDonald, J. C. and McDonald, A. D. (1977). Epidomiology of meso-
thelioma from estimated incidence. Preventive Meoicine 6:426-446.
51. Hammond, E.G., Selikoff. I.J., and Seidman, H. (1979). Asbestos
Exposure, Cigarette Smoking, and Death Rates. In: Health Hazards
of Asbestos Exposure. Selikoff, I.J. and Hammond E.C. (eds.).
New York Academy of Sciences. New York.
52. McDonald, J. C., Liddel, F. D. K., Gibbs, 6. W. , Eyssen, G. E.,
and McDonald, A. D. (1980). Dust exposure and mortality in
chrysotile mining, 1910-75. British Journal of Industrial
Medicine 57:11-24,
53. Nicholson, W. J., Selikoff, I. J., Seidmon, H., Lilis, R., and
Formby, P. (1979). Long-term mortality experience of chrysotile
miners and millers in Thetford mines, Quebec. Annals New York
Academy of Sciences 330:11-21.
54. Rubino, G., .Piolatto, G., Newhouse, M., Scansetti, G., Aresini,
G., amd Murray, R. (1979). Mortality of chrysotile asbestos
workers at the Balangero Mine, Northern Italy. British Journal
of Industrial Medicine 36:187-194.
55. Crump, K. S. (1984). Expert Testimony on Environmental Protection
Agency Regional Hearings on Asbestos. Boston. June 1984.
56. Henderson, V. L. and Enterline, P. E. (1979). Asbestos exposure:
factors associated with excess cancer and respiratory disease
mortality. Annals New York Academy of Sciences 330:117-126.
57. Weill, H., Hughes, J., and Waggenspack, C. (1979). Influence of
dose and fiber type of respiratory malignancy risk in asbestos
cement manufacturing. American Review of Respiratory Disease
120:345-354.
58. Dement, J. M., Harris, R. L., Symons, J. M., and Shy, C. (1982).
Estimates of dose-recponse for respiratory cancer among chrysotile
asbestos textile workers. In: Inhaled Particles. Walton, W.H.
(ed.). Volume V. Oxford. Pergamon. pp. 869-887.
59. Berry, G. and Newhouse, M.L. (1983). Mortality of workers manu-
facturing friction materials using asbestos. British Journal of
Industrial Medicine 40:1-7.
2-277
-------
60. Peto, J. (1980). Lung Cancer Mortality In Relation to Measured
Dust Levels in an Asbestos Textile Factory. In: Bioloqicol
Effects of Mineral Fibres. Volume 2. Wagner, J. (ed.).
International Agency for Research on Cancer. IARC Scientific
Publications No. 30.
61. Finkelstein, M. M. (1983). Mortality in Asbestos Cement Factory
Workers. (In publication).
62. Finkelstein, M. M. (1983). Mortality among long-term employees
of on Ontario asbestos cement factory. British Journal of
Industrial Medicine 40:138-144.
63. Seidman. H., Lilis, R. and Selikoff. I. (1979). Short-term
asbestos work exposure and long-term observation. Annals New York
Academy of Sciences 330:61.
64. Selikoff, I., Hammond. E. and Seidman, H. (1979). Mortality
experience of insulation workers in the United States and Canada.
Annals New York Academy of Sciences 300:91.
65. International Agency for Research on Cancer (1979). Benzene.
IARC Monographs on the Evaluation of the Carcinogenic Risk of
Chemicals to Humans 7:203-221.
66. Hunter, C. G. (1968). Solvents with reference to studies on the
pharmacodynamics of benzene. Proceeding of the Royal Society
Medicine 61:913-915.
67. Hunter, C. G. and Blair, D. (1972). Benzene: Pharmacokinetic
studies in man. Annals of Occupational Hygiene 15:193-199.
68. Docter, H. J. and Zielhuis, R. L. (1967). Phenol excretion as a
measure of benzene exposure. Annals of Occupational Hygiene
10:317-326.
69. Berlin, M. (1984). Low Level Benzene Exposure in Sweden: Effect
on Blood Elements and Body Burden of Benzene. Institute of
Environmental Health, University of Lund. MARC, Chelsea College,
London.
70. McMichael, A J., Spirtas, R., Kupper, L. L. and Gamble, J. F.
(1975). Solvent exposure and leukemia among rubber workers: An
epidemiologic study. Journal of Occupational Medicine 17(4):
234-239.
71. Arp, E. W., Wolf, P. H., and Checkoway, H. (1983). Lyirphocytic
leukemia and exposure to benzene and other solvents in the rubber
industry. Journal of Occupational Medicine 25(8):598-602.
2-278
-------
72. Checkoway. H., Wilcosky, T., Wolf, P., and Tyroler, H. (1984). An
evaluation of the associations of leukemia and rubber industry
solvent exposures. American Journal of Industrial Health
5:239-249.
73. Infante, P. and White, M. (1983). Benzene: Epidemiologic observa-
tions of leukemia by cell type and adverse health effects
associated with low-level exposure. Environmental Health
Perspectives 52:75-82.
74. Vigiani, E. C. and Saita, G. (1964). Benzene and leukemia. The
New England Journal of Medicine 271(17):872-876.
75. Vigliani, E. C. (1976). Leukemia associated with benzene
exposure. Annols of the New York Academy of Sciences 143-151.
76. Aksoy, M. , Erdem, S., and OinCol, G. (1974). Leukemia in shoe-
workers exposed chronically to benzene. Blood 44(6):837 841.
77. Aksoy, M., Erdem, S., DinCol, K., Hepyuksel, T., and Dincol, G.
(1974). Chronic exposure to benzene as a possible contributory
etiologic factor in Hodgkin's disease. Blut 28:293-298.
78. Aksoy, M.,, Erdem, S. , Erdogan, G., and Dincol, G. (1974). Acute
leukaemia in two generations following chronic exposure to
benzene. Human Herdeity 24:70-74.
/9. Aksoy, M., Erdem, S., Erdogan, G., and Dincol, G. (1976).
Combination of genetc factors and chronic exposure to benzene in
the aetiology of leukaemia. Human Heredity 26:149-153.
80. Aksoy, n. (1980). Different types of malignancies due to occupa-
tional exposure to benzene: A review of recent observations in
Turkey. Environmental Research 23:181-190.
81. Aksoy, M. (1981). Problems with benzene in Turkey. Regulatory
Toxicology ond Pharmacology 1:147-155.
82. Aksoy, M. (1983). Malignancies Due to Occupational Exposure to
Benzene. Unpublished report.
83. Aksoy. M., Erdem, S., DinCol, G., Kutlar, A., Bakioglu, I.,
Hepyuksel, T. (1984). Clinical observations showing the role of
some factors in the etiology of multiple myeloma. Acto Haematol
71:116-120.
8
-------
85. Holmerg, B. and Lundberg, P. (1984). Benzene: Standards,
Occurrence and Exposure. Unit of Occupational Toxicology,
Research Department, National Board of Occupational safety and
Health. S-17184 SOLNA, Sweden.
86. Ott, M. G., Townsend, J. C., Fishbeck, W. A., and Langner, R. A.
(1978). Mortality among individuals occupationally exposed to
benzene. Archives of Environmental Health 33:3-10.
87. Rinsky, R. A., Young, R. J., and Smith, A. B. (1981). Leukemia in
benzene workers. American Journal of Industrial Medicine 2:
217-245.
88. Wong, 0. (1983). An Industry-Wide Mortality Study of Chemical
Workers Occupationolly Exposed to Benzene. Chemical Manufacturers
Association. Environmental Health Associates, Inc. Berkley, CA.
89. Crump, K. and Allen, B. (1984). Quantitative Estimates of Risk of
Leukemia From Occupational Exposure to Benzene. Prepared for U.S.
Occupational Safety and Health Administration.
90. Tsuchiya, K., Okubo, T., and Ishi*u, S. (1975). An epidemiologi-
cal study of occupational bladder tumours in the dye industry of
Japan. British Journal of Industrial Medicine 32:203-209.
91. Mancuso, T. F. and El-Attar, A. A. (1967). Cohort Study of
Workers Exposed to Betanaphthylamine and Benzidine. Journal of
Occupational Medicine 9(6):277-285.
92. Morinaga, K., Optima, A., and Hara, I. (1982). Multiple Primary
Cancers Following Exposure to Benzidine and Beta Naphthylamine.
American Journal of Industrial Medicine 3:243-246.
93. Goldwater, L. J., Rosso, A. J., and Kleinfeld, M. (1965). Bladder
tumors in a coal tar dye plant. Archives of Environmental Health
11:814-817.
94. Case, R. A, M., Hosker, M. E., McDonald, D. B., and Pearson, J. T.
(1954). Tumours of the Urinary Bladder in Workmen Engaged in the
Manufacture and Use of Certain Dyestuff Intermediates in the
British Chemical Industry. British Journal of Industrial Medicine
11:75-104.
95. Rubino, G. F., Scansetti, G. . Piolatto, G., and Pira, E. (1982).
The carcinogenic effect of aromatic amines: An epi-lemiological
study on the role of o-toluidine and 4,4' methylene bis (2-
methylaniline) in inducing bladder cancer in man. Environmental
Research 27:241-254.
2-280
-------
96. Walker, B., Jr., and Gerber, A. (1981). Occupational exposure to
aromatic amines: Benzidene and benzidene-based dyes. Notional
Cancer Institute Monographs 58:11-13.
97. Horton, A. W. and Binghom, E. U. (1977). Risk of bladder tumors
among benzidene workers and their serum properdin levels. Journol
of the Notional Cancer Institute 58:1225-1228.
98. Lower, G. M., Jr., Nilsson, T., Nelson, C. E., Wolf, H., Gamsky,
T. E., and Bryan, G. T. (1979). N-Acetyltransferase phenotype and
risk in urinary bladder cancer: Approaches in molecular epidemi-
ology. Preliminary results in Sweden and Denmark. Environmental
Health Perspectives 29:71-79.
99. Zavon, M. R., Hoegg, U., and Bingham, E. (1973). Benzidene
exposure as a cause of bladder tumors. Archives of Environmental
Health 27:1-7.
100. National Institutes of Health (1981). Surveillance, Epidemiology,
and End Results: Incidence and Mortality Data, 1973-77. Notional
Cancer Institute Monographs 57.
101. Rinde, E., and Troll, W. (1975). Metabolic Reduction of Benzidine
Azo Dyes to Benzidine in the Rhesus Monkey. Journal of the
Notional Cancer Institute 55:181-187.
102. National Academy of Sciences Executive Committee (1975).
Contemporary Pest Control Practices and Prospects. Pest Control:
An Assessment of Present and Alternative Technologies, Vol. 1.
103. International Agency for Research on Cancer (1975). Cadmium and
cadmium compounds. IARC Monographs on the Evaluation of
Carcinogenic Risk of Chemicals to Man 11:39-41.
104. Piscator, M. (1981). Role of cadmium in carcinogenesis with
special reference to cancer of the prostate. Environmental Health
Perspectives 40:107-120.
105. Welinder, H., Skerfving, S., and Henriksen, 0. (1977). Cadmium
metabolism in man. British Journol of Industrial Medicine 34:
221-228.
106. Flanagan, P. R. , McLellon, J. S., Haist, J., Cherian, M. G.,
Chamberlain, M. J., and Valberg, L. S. (1978). Increased dietary
cadmium absorption in mice and human subjects with iron
deficiency. Costrpenterology 74:841-846.
107. Kolonel, L. N. (1976). Association of cadium with renal cancer.
Cancer 37:1782-1787.
2-281
-------
108. Boko, 0.. Smith, E. S. 0., Hanson, J., and Dewar, R. (1982). The
geographical distribution of high cadmium concentrations in the
environment and prosta-e cancer in Alberta. Canadian Journal of
Public Health 73:92-94.
10S. Berg, J. W. and Burbank, F. (1972). Correlations between
carcinogenic trace metals in water supplies and cancer mortality.
Annals of the New York Academy of Sciences 199:249-261.
110. Potts, C. L. (1965). Cadmium proteinuria: The health of battery
workers exposed to cadmium oxide dust. Annuls of Occupational
Hygiene 8:55-61.
111. Kipling, M. D. and Waterhouse, J. A. H. (1967). Cadmium and
prostatic carcinoma. The Lancet April 1:730-731.
112. Lemen, R. A., Lee, J. S., Wagoner, J. K., and Blejer, H. P.
(1976). Cancer mortality among cadmium production workers.
Annals of the New York Academy of Sciences 271:273-279.
113. Kjellstrom, T., Friberg, L., and Rahnster, B. (1979). Mortality
and cancer morbidity among cadmium-exposed workers. Environmental
Health Perspectives 28:199-204.
114. Elinder, C-G., Kjellstro/n, T., Hogstedt, C., Andersson, K. and
Spang, G. (1985). Cancer mortality of cadmium workers. British
Journal of Industrial Medicine 42:651-655.
115. Armstrong. B. G. and Kazantzis, G. (1983). The mortality of
cadmium workers. The Lancet June 25:1425-1427.
116. Sorohan, T. and Waterhouse, J. A. H. (1983). Mortality study of
nickel-cadmium battery workers by method of regression models in
life tables. British Journal of Industrial Medicine 40:293-300
117. Thun, M. J., Schnorr, T. M., Smith, A. B., Halperin, W. £., and
Lemen, R. A. (1985). Mortality omong a cohort of U. S. cadmium
production workers--on update. Journal of the Notional Cancer
Institute 74(2) :325-333.
118. International Agency for Research on Cancer (1981). Chlorombucil.
IARC Monographs on the Evaluation of Carcinogenic Risk of
Chemicals to Man 26:115-136.
119. Rieche, K. (1984). Carcinogenicity of ontineoplostic agents in
man. Cancer Treatment Reviews 11:39-67.
120. Reimer, R. R., Hoover, R., Fraumeni, J. F., and Young, R. C.
(1977). Acute leukemia ofte,* olkyloting-agent therapy of ovarian
cancer. The New England Journal of Medicine 297(4):177-181.
2-282
-------
121. Greene, M. H., Boice, 0. D., Groer, B. E., Blessing, J. A., and
Dembo, A. J. (1982). Acute nonlymphocytic leukemia after therapy
with alkylating agents for ovarian cancer. The New England
Journal of Medicine 307(23) : 1416-1421.
122. Berk, P. D. , Goldberg, J. D., Silverstein, M. N., Weinfeld, A.,
Donovan, P. B., Ellis, J. T., Landaw, S. A., Laszlo, J., Najean,
Y., Pisciotta, A. V. and Wasserman, L. R. (1981). Increased
incidence of acute leukemia in polycythemia vera associated with
chlorambucil therapy. The New England Journal of Medicine
304(8):441-447.
123. International Atjency for Research on Cancer (1980). Chromium and
chromium compounds. IARC Monographs on the Evaluation of the
Carcinogenic Risk of Chemicals to Humans 23:205-323.
124. Hayes, R. B. (1979). Cancer and Occupational Exposure to Chromium
Chemicals. In: Reviews in Cancer Epidemiology, Vol. 1.
Lilienfeld, A. M. (ed.). Elsevier/North-Holland Publisher. New
York, New York.
125. Royle, H. (1975). Toxicity of chromic acid in the chromium
plating industry (1). Environmental Research 10:39-53.
126. Baetjer, A. M. (1950). Pulmonary carcinoma in chromate workers.
I. A review of the literature and report of cases. Archives of
Industrial Hygiene and Occupational Medicine 2(5):487-504.
127. Enterline, P. E. (1974). Respiratory cancer among chromate
workers. Journal oT Occupational Medicine 16(8):523-526.
128. Stern, R. M. (1983). Assessment of risk of lung cancer for
welders. Archives of Environmental Health 38f3):148-155.
129. Taylor, F. H. (1966). The relationship of mortality and duration
of employment as reflected by a cohort of chromate workers.
American Journal of Public Health 56(2):218-229.
130. Royle, H. (I975b). Toxicity of chromic acid in the chromium
plating industry (2). Environmental Research 10:141-163.
131. Ohsaki, Y., Abe, S., Kimura, K., Tsuneta. Y., Mikami, H., and
Murao, M. (1978). Lung cancer in Japanese chromate workers.
Thorax 33:372-374.
132. Hayes, R. B., Lilienfeld, A. M., and Snell, L. M. (1979).
Mortality in chromium chemical production workers: a prospective
study. International Journal cf Epidemiology 8(4):365-374.
2-283
-------
133. Dalager, N. A., Mason, T. J., Fraumeni, J. F., Hoover, R., and
Payne, W. W. (1980). Cancer mortality among workers exposed to
zinc chromate paints. Journal of Occupational Medicine
22(1):25-29.
134. Alderson, M. R., Rattan, N. S., and Bidstrup, L. (1981). Health
of workmen in the chromate-producing industry in Britain. British
Journal of Industrial Medicine 38:117-124.
135. Sheffet, A., Tnind, I., Miller, A. M.. and Louria, 0. B. (1982).
Cancer mortality in a pigment plant utilizing lead and zinc
chromates. Archives of Environmental Health 37(1): 44-52.
136. Davies, J. M. (1984). Lung cancer mortality among workers making
lead chromate and zinc chromate pigments at three English
factories. British Journal of Industrial Medicine 41:158-169.
137. Hernberg, S., Westernolm, P., Schultz-Larsen, K., Oegerth, R.,
Kuosma, E., Englund, A., Engzell, U., Hansen, H. S., and Mutanen,
P. (1983). Nasal and sinonasal cancer. Connection with occupa-
tional exposure in Denmark, Finland and Sweden. Scandinavian
Journal of Work Environment and Health 9:315-326.
138. Langard, S. and Norseth, S. (1979). Cancer in the gastrointes-
tinal tract in chromate pigment workers. Arhiv Za Hiqijenu Rodo I
Toksikologiju 30(Suppl)-. 301-304 .
139. Langard, S. and Norseth, T. (1975). A cohort study of bronchial
carcinomas in workers producing chromate pigments. British
Journal of Industrial Medicine 32:62-65.
140. Langard, S. and Vigander, T. (1983).
workers producing chromium pigments.
Industrial Medicine 40:71-74.
Occurrence of lung cancer in
British Journal of
141. Longard, S., Andersen, A., and Gylseth, B. (1980). Incidence of
cancer among ferrochromium and ferrosilicon workers. British
Journal of Industrial Medicine 37:114-120.
142. Axelsson, G. and Rylander, R. (1979). Cancer mortality among
ferrochromium workers. Arhiv Zo Hiqijenu Rodo I Toksikoloqiju
30(Suppl):305-310.
143. Axelsson, G., Rylander, R., and Schmidt, A. (1980). Mortality and
incidence of tumours among ferrochromium workers. British Journal
of Industrial Medicine 37:121-127.
144. Department of Health, Education, and Welfare (1979). Smoking and
Health. A Report to the Surgeon General. No. (PHS) 79-50066.
2-284
-------
1*5. Loeb, L. A., Ernster, V. L.. Warner, K. E., Abbotts, J. and
Loszio, J. (1984). Smoking and lung cancer: an overview. Cancer
Research 44:5940-5958.
146. Doll, R. and Peto, R. (1976). Mortality in relation to smoking: 20
years' observations on male British doctors. British Medical
Journal 2:1515-1536.
147. Doll, R. and Peto, R. (1978). Cigarette smoking and bronchial
carcinoma: dof» and time relationships among regular and lifelong
non-smokers. Journal of Epidemiology one* Community Health
32:303-313.
148. Crump, K. S. , and Allen, 8. C. (1985). Methods for quantitative
risk assessment using occupational studies. The American
Statistician 4(4) :442-450.
149. Lanier, A. P, Noller, K. L., Decker, D. G., Elveback, L. R., and
Kurland, L. T. (1973). Cancer and stilbestrol: A follow-up of
1,719 persons exposed to estrogens in utero and born 1943-1959.
Mayo Clinic Proceedings 48:793-799.
150. Dieckmann, W. J., Davis, M. E., Rynkiewicz, L. M. , and Pottinger,
R. E. (1953). Does the administration of diethylstilbestrol
during pregnancy have theroyt'jtic value. American Journal of
Obstetrics and Gynecolog/ 66:1062-1081.
151. Robboy, S. J., Scully, R. E., Welch, W. R., and Herbst, A. L.
(1977). Intrauterine diethylstilbestrol exposure and its conse-
quences. Archives of Pathology and Laboratory Medicine 101:1.
152. Ulfelder, H. (1976). DES - transplocental teratogen - and
possibly also carcinogen. Teratology 13:101-104.
153. Fu, Y. S., Reagan, J. W., Richart, R. M., and Townsend, D. E.
(1979). Nuclear DNA and histologic studies of genital lesions in
diethylstilbestrol-exposed progtny. American Journal of Clinical
Pathology 72(4) :515-520.
154. Gill, W. B., Schumacher, G. F. B., and Bibbo, M. (1976).
Structural and functional abnormalities in the sex organs of male
offspring of mothers treated with diethylstilbestrol. The Journal
of Reproductive Medicine 16(4) :147-153.
155. Gill, W. B.. Schumacher, G. F. B., and Bibbo, M. (1977).
Pathological semen and anatomical abnormalities of the genital
tract in human male subjects exposed to diethylstilbestrol in
utero. The Journal of Urology 117:477-480.
2-285
-------
156. Bellet, R. E. and Squitieri, A. P. (1974). Estrogen-induced
hypernephroma. The Journal of Urology 112:160-161.
157. Greenberg, E. R., Barnes, A. B., Resseguie, L., Barrett, J. A.,
Burnside, S., Lanza, L. L., Neff, R. K., Stevens, M., Young, R.
H., and Colton, T. (1984). Breast cancer in mothers given
diethylstilbestrol in pregnancy. The New England Journal of
Medicine 311:1393-1398.
158. Bibbo, M., Haenszel, W. M., Wied, G. L., Hubby, M. , and Herbst, A.
L. (1978). A twenty-five-year follow-up study of women exposed to
diethylstilbestrol during pregnancy. The New England Journal of
Medicine 298(14) :763-767.
159. Hadjimichael, 0. C., Meigs, J. W., Falcier, F. W., Thompson, W.
0., and Flannery, J. T. (1984). Cancer risk among women exposed
to exogenous estrogens during pregnancy. Journal of the Notional
Cancer Institute 73(4):831-834.
160. Greenwald, P., Barlow, J. J., Nasca, P. C., and Burnett, W. S.
(1971). Vaginal cancer after maternal treotment with synthetic
estrogens. The New Engjond Journol of Mecj-ic.'ne 285(7) : 390-392.
131. Greenwald, P. and Nasca, P. C. (1973), Transplacental induction
of vaginal cancer by synthetic estrogens. Progress in Clinical
Cancer 5:13-19.
162. Mangan, C. E., Giuntoli, R. L.. Sedlacek, T. V., Rocereto, T.,
Rubin, E., Burtnett, M., and Mikuta, J. J. (1979). Six years'
experience with screening of a diethylstilbest'-ol exposed
population. American Journol of Obstetrics and Gynecology
134:860-865.
163. Herbst, A. L., Cole, P. C., Colton, T., Robboy, S. J., and Scully,
R. E. (1977). Age-incidence and risk of diethylstilbestrol-
related clear cell adenocarcinoma of the vagina and cervix.
American Journal of Obstetrics and Gynecology 128(1):43-50.
164. Herbst, A. L., Ulfeloer, H., and Poskanzer, D. C. (1971) Andeno-
carcinoma of the vagina. Association of maternal stilbestrol
therapy with tumor appearance in young women. The New England
Journol of Medicine 284(16):878-881.
165. Herbst, A. L., Kurman, R. J., Scully, R. E., and Poskanzer, D. C.
(1972). Clear-cell adenocarcinoma of the genital tract in young
females. The New England Journol of Medicine 287(25):1259-1264.
2-286
-------
-, D. C., Robboy, S. J., Friedlonder, L.,
). Prenatal exposure to stilbestrol. A
if exposed female offspring with unexposed
ournol of Medicine 292(7):334-339.
on Chemical Safety (1984). Epichloro-
health criteria 33. World Health
J:1-51.
.nsky, R. A. (1979). Reanalysis of the Shell
tudy. U. S. Department of Health, Education and
ion of Surveillance, Hazard Evaluations and Field
- E. (1982). Importance of sequential exposure in the
f epichlorohydrin and isopropanol. The Annals of the
of Sciences 381:344-349.
•>r, R. J., McClimans, C. D. , Ott. M. G. , Flake, R. E.,
, R. L. (1979). An Evolution of the Mortality
» of Employees with Potential for Exposure to
'hydrin. Midland, Michigan, Dow Chemical Toxicology
Laboratory.
n, J. P., Bos, 6. D. , Craigen, A. A., Jacquet, Q., Kueng,
anouziere-Simon , C. , and Pierre, C. (1983). Mortality in
jpean cohort occupationally exposed to epichlorohydrin .
ationol Archives of Occupational and Environmental Health
^25-326,
tioral, V. (1980). Hormones and Cancer. In: Reviews of Concer
Epidemiology. Lilienfeld, A. (ed.). Elsevier Press. New York.
i. 'Hoover, R., Gray, L. A., Cole, P., and MacMahon, B. (1976). The
New England Journal of Medicine 295:401-405.
174. Ross, R. K. , Paganini-Hill, A., Gerkins, V. B., Mock, T. M. ,
Pfeffer, R., Arthur, M. , and Henderson, B. E. (1980). A case
control stud of menopausal estrogen therapy and breast cancer.
Journal of the American Medicol Association 243:1635-1639.
175. Horwitz, R. I. and Stewart, K. R. (1984). Effect of clinical
features on the association of estrogens and breast cancer. The
American Journal of Medicine 76:192-198.
176. Smith, D. C., Prentice, R., Thompson, D. J., and Herrman, W. L.
(1975). Association of exogenous estrogen and endometrial carci-
noma. The New England Journal of Medicine 293:1164-1167.
2-287
-------
177. Mack, T. M., Pike, M. C., Henderson, B. E., Pfeffer, R. I.,
Gerkins, V. R., Arthur, M,, and Brown, S. E. (1976). Estrogens
and endometrial cancer in a retirement community. The New England
Journal of Medicine 294:1262-1267.
178. McDonald, T. W. , Annegers, J. F., O'Follon. W. M., Dockerty. M.
B., Malkasian, G. D., and Kurland, L. T. (1977). Exogenous estro-
gen and endometrial carcinoma: Case-control and incidence study.
The American Journal of Obstetrics and Gynecoloqy 127:572-579.
179. Gray, L. A., Christopherson, W. M., and Hoover. R. N. (1977).
Estrogens and Endometrial Carcinoma. Obstetrics and Gynecology
49:385-389.
180. Antunes, C. M. F., Stolley, P. D., Rosenshein, N. B., Davies, J.
L., Tonascia, J. A., Brown, C., Burnett, L., Rutledge, A.,
Pokempner, M., and Garcia, R. (1979). Endometrial cancer and
estrogen use. The New England Journal of Medicine 300:9-13.
181. Hammond, C. B., Jelovsek, F. R., Lee, K. L., C. aosman, W. T., and
Parker, R. T. (1979). Effects of long-term estrogen replacement
therapy II. Neoplasic. The American Journal of Obstetrics and
Gynecology 133:537-547.
182. Hulko, B. S., Fowler, W. C.. Kaufman, D. G.. Crimson, R. C..
Greenberg, B. G., Hague, C. J. R., Berger, G. S., and Pulliam, C.
C. (1980). Estrogen and endometrial cancer: Cases and two control
groups from North Ca-olina. The Americon Journal ofObstetrics
and Gyneeoloqy 137:S2-101.
183. Shapiro, S., Kaufman, D. W., Slone, D., Rosenberg, L., Miettinen,
0. S., Stolley, P. D., Rosenshein, N. B., Watring, W. G. ,
Leavitt, T., and Knapp, R. C. (1980). Recent and past use of
conjugated estrogens in relation to adenocarcinoma of the endome-
trium. The New England Journal of Medicine 303:485-489.
184. Spengler, R. F., Clarke, E. A., Woolever, C. A., Newman, A. M.,
and Osborn, R. W. (1981). Exogenous estrogens and endometrial
cancer: a case-control study and assessment of potential biases.
American Journal of Epidemiology 114:497-506.
185. Morgan, R. W. , Claxton, K. W,, Divine, B. J., Kaplan, S. D., and
Harris, v. B. (1981). Mortality among ethylene oxide workers.
Journal of Occupational Medicine 23( 11 )-.767-770.
186. Korpela, D. B., McJilton, C. E., and Howkinson, T. E. (1983).
Ethylene oxide dispersion from gas sterilizers. American
Industrial Hygiene Association Journal 44(8):589-591.
2-288
-------
187. Garry, V. F., Hozier, J., Jacobs, D., Wade, R. L., and Gray, D. G.
(1979). Ethylene Oxide: Evidence of human chromosomal effects.
Environmental Mutoqenesis 1:375-382.
188. Vainio, H. (1982). Inhalation anesthetics, anticancer drugs and
sterilants as chemical hazards in hospitals. Scondonovion Journal
of Work, Environmental, and Health 8:9^-107.
189. Landrigan, P. J., Meinhardt, T. J., Gordon, J., Lipscomb, J. A.,
Burg, J. R., Mazzuckelli, L. F., Lewis, T. R., and Lemen, R. A.
(198*). Ethylene oxide: An overview of toxicologic and epidemic-
logic research. American Journal of Industrial Medicine 6:
103-115.
190. Hogstedt, C., Malmqvist, N., and Wadmon, B. (1979). Leukemia in
workers exposed to ethylene oxide. Journal of the American
Medical Association 241<[ 11 ): 1132-1133.
191. Hogstedt, C., Rohlen, 0., Berndtsson, B. S., Axelson, 0., and
Ehrenberg, L. (1979). A cohort study of mortality and cancer
incidence in ethylene oxide production workers. British Journal
Industrial Medicine 36:276-280.
192. Ferebee, S. H. (1970). Controlled chemoprophylaxis trials in
tuberculosis. A general review. Advances in Tuberculosis
Research 17:28-106.
193. Costello, H. D. and Snider, D. E. (1980). The incidence of
cancer among participants in a controlled, randomized isoniazid
preventive therapy trial. American Journal of Epidemiology
3(1):67-7*.
19*. International Agency for Research on Cancer (197*). Isonicotinic
acid hydrazide. IARC Monographs on the Evaluation of the
Carcinogenic Risk of Chemicals to Man *:159-172.
195. Hammond, E. C., Selikoff, I. J., and Robitzek, E. H. (1967).
Isoniazid therapy in relation to later occurrence of cancer in
adults and in infants. British Medical Journal 2:792-795.
196. Stott, H., Peto, J., Stephens, R., Fox, W., Sutherland, I.,
Foster-Carter, A. F., Teare, H. D., and Penning J. (1976). An
assessment of the carcinogenicity of isoniazid in patients with
pulmonary tuberculosis. Tubercle 57:1-15.
197. Glassroth, J. L., White, M. C., and Snider, D. E. (1977). An
assessment of the possible association of isoniazid with human
cancer deaths. American Review of Respiratory Disease
116:1065-107*.
2-289
-------
198. Miller, C. T., Neutel, C. I., Noir, R. C., Marrett, L. D., Last,
J. M., and Collins, W. E. (1978). Relative importance of risk
factors in bladder carcinogenesis. Journal of Chronic Diseases
31:51-56.
199. Sanders, B. M, and Draper, G. J. (1979). Childhood cancer and
drugs in pregnancy. British Medical Journal 1:718-719.
200. Clemmesen, J. and Hjalgrim-Jensen, S. (1979). Is isonicotinic
acid hydrazide (INH) carcinogenic to man? A 24-year follow up of
3371 tuberculosis cases. Ecotoxicology and tinvironmentol Safety
3:439-450.
201. Howe, G. R., Lindsay, J., Coppock, £., and Miller, A. B. (1973).
Isoniazid exposure in relation to cancer incidence and mortality
in a cohort of tuberculosis patients. International Journal of
Epidemiology 8(4):305-312.
202. Boice, J. D. and Fraumeni, J. F. (1980). Late effects following
isoniazid therapy. American Journal of Public Health 70(9):
987-989.
203. International Agency for Research on Cancer (1975). Melphalan,
Medphalan and Merpholan. IARC Monographs on the Evaluation of the
Carcinogenic Risk of Chemicals to Man 9:167-180.
204. Gori, S., Donti, E., Venti, G., Mecucci, C., Crino, L., and
Tonato, M. (1983). Acute myeloblastic leukemia after adjuvant
chemotherapy with melphalan in breast cancer case report with
cytogenetic analysis. Tumori 69:117-122.
205. Law, I. P. and Blom, J. (1977). Second malignancies in patients
with multiple myeloma. Oncology 34:20-24.
206. Einhorn, N. (1978). Acute leiihssMe after chemotherapy
(Melphalan). Cancer 41:444-447.
207. Friedlander, B. R.. Hearne, T., and Hall. S. (1978). Epidemic-
logic investigation of employees chronically exposed to methylene
chloride. Journal of Occupational Medicine 20(10):657-666.
208. Riley, E. C., Fassett, D. W.. and Sutton, W. L. (1966). Methylene
chloride vapor in expired air of human subjects. American
Industrial Hygiene Journal 27:341-348.
209. Ott, M. G., Skory, L. K., Holder, B. B., Bronson, J. M., and
Williams, P. R. (1983). Health evaluation of employees occupa-
tionally exposed to methylene chloride. General study design and
environmental considerations. Scondinovion Journal of Work,
Environment and Health 9(Suppl 1):1-7.
2-290
-------
210. Hearne, F. T. and Friedlonder, B. R. (1981). Follow-up of
methylene chloride study. Journal of Occupotionol Medicine
23:660.
211. Ott, M. G., Skory, L. K., Holder, B. B., Branson, J. M., and
Williams, P. R. (1983). Health evaluation of employees occupa-
tionally exposed to methylene chloride. Mortality. Scandinavian
Journal of Work, Environment and Health 9(Suppl 1):8-16.
212. Environmental Protection Agency (1983). Health Assessment
Document for Nickel. EPA-600/8-83-012.
213. NIOSH. (1977). Criteria for a Recommended Standard. Occupational
Exposure to Inorganic Nickel. U.S. Government Printing Office,
Washington D.C.
214. Sunderman, F. W. (1968). Nickel carcinogenesis. Diseases of the
Chest 54(6):41-48.
215. Sunderman, F. W. (1981). Recent research on nickel carcino-
genesis. Environmental Health Perspectives 40:131-141.
216. Doll, R. (1958). Cancer of the lung and nose in nickel workers.
British Journal of Industrial Medicine 15:217-223.
217. Doll, R., Mathews, J. D., and Morgan, L. G. (1977). Cancer of the
lung and nasal sinuses in nickel workers: a reassessment of the
period of risk. British Journal of Industrial Medicine 34:102-
105.
218. Morgan, J. G. (1958). Some observations on the incidence of
respiratory cancer in nickel workers. British Journal of
Industrial Medicine 15:224-234.
219. Norseth, T. (1980). Cancer hazards caused by nickel and chromium
exposure. Journal of Toxicology and Environmental Health 6:1219-
1227.
220. Sutherland, R. B. (1959). Respiratory Cancer Mortality in Workers
Employed in an Ontario Nickel Refinery (1930-1957). Division of
Industrial Hygier^. Department of Health for Ontario.
221. Sutherland, R. B. (1969). Mortality Among Sinter Workers.
International Nickel Company of Canada, Limited. Cooper Cliff
Smelter. Environmental Health Services Branch. Ontario
Department of Health.
2-291
-------
222. Chovil, A., Sutherland, R. B.. and Halliday, M. (1981).
Respiratory cancer in a cohort of nickel sinter plant workers.
British Journal of Industrial Medicine 38:327-333.
223. Pedersen, E., Hogetveit, A. C., and Andersen, A. (1973). Cancer
of respiratory organs among workers at a nickel refinery in
Norway. International Journal of Cancer 12:32-41.
224. Kreyberg, L. (1978). Lung cancer in workers in a nickel refinery.
British Journal of Industrial Medicine 35:109-116.
225. Magnus, K., Andersen, A., and Hogetveit, A. C. (1982). Cancer of
respiratory organs among workers at a nickel refinery in Norway.
International Journal of Cancer 30:681-685.
226. Lessard, R., Reed, D., Maheux, B., and Lambert, J. (1978). Lung
cancer in New Caledonia, a nickel smelting island. Journal of
Occupational Medicine 20(12) :815-229.
227. Silverstein, M., Mirer, F., Kotelchuck, D., Silverstein, B., and
Bennett, M. (1981). Mortality among workers in a die-casting and
electroplating plant. Scondinovion Journal of Work. Environment
and Health 7(Suppl 4):156-165.
228. Bernacki, E. J., Parsons, G. E., and Sunderman, F. W. (1978).
Investigation of exposure to nickel and lung cancer mortality.
Annuls of Clinical ond Laboratory Science 8(3):190-194.
229. Godbold, J. H. and Tompkins, E. A. (1979). A long-term mortality
study of workers occupotionally exposed to metallic nickel at the
Oak Ridge gaseous diffusion plant. Journal of Occupotionol
Medicine 21(12):799-806.
230. Olsen, J. and Sabroe, S. (1984). Occupational causes of laryngeal
cancer. Journal of Epidemiology ond Community Health 38:117-121.
231. Cox, J. E.. Doll. R., Scott, W. A., and Smith, S. (1981).
Mortality of nickel workers: Experience of men working with
metallic nickel. British Journal of Industrial Medicine
38:235-239.
232. Poledrak, A. P. (1981). Mortality among welders, including a
group exposed to nickel oxides. Archives of Environmental Health
36(5).235-242.
233. Enterline, P. E. and Marsh, G. M. (1982). Mortality among workers
in a nickel refinery and alloy manufacturing plant in West
Virginia. Journal of the Notional Cancer Institute 68(6):925-933.
2-292
-------
234. International Agency for Research on Cancer (1978). Polychlori-
nated Biphenyls. IARC Monographs on the Evaluation of the
Carcinogenic Risk of Chemicals to Man 18:43-103.
235. Brown, D. P. and Jones, M. (1981). Mortality and industrial
hygiene study of workers exposed to polycMurinated biphenyls.
Archives of Environmental Health 36(3):120-129.
236. Safe, S. (1984). Polychlorinate biphenyls (PCBs) and polybro-
minated biphenyls (PBBs): Biochemistry, toxicology, and
mechanism. Criticols Reviews in Toxicology 13(4):319-396.
237. Cordle, F.. Corneliussen, P., Jelinek, C., Hackley, B., Lehman,
R., McLaughlin, J., Rhoden, R., and Shapiro, R. (1978). Human
exposure to polychlorinated biphenyls and polybrominated
biphenyls. Environmental Health Perspectives 24:157-172.
238. Unger. M. 0. and Nordberg, G. F. (1979). Distribution of PCB and
DDT among human tissues. Arhiv Zo Higijenu Rodo I Toksikologiju
30:537-543.
239. Moroni, M.. Colombi, A., Cantoni, S., Ferioli, and Foa, V. (1981).
Occupational exposure to polychlorinated biphenyls in electrical
workers. I. Environmental and blood polychlorinated biphenyls
concentrations. British Journol of Industrial Medicine 38:49-54.
240. Moroni, M., Colombi, A., Cantoni, S., Ferioli, and Foa, V. (1981).
Occupational exposure to polychlorinated biphenyls in electrical
workers. II. Health effects. British Journol of Industrial
Medicine 38:55-60.
241. Bahn, A. K., Rodenwaike, I., Herrmann, N., Grover, P., Stollman,
J., and O'Leary, K. (1976). Melanoma after exposure to PCBs. The
New England Journol of Medicine 295:450.
242. Maroni, M., Colombi, A., Arbosti, G., Cantoni, S., and Foo, V.
(1081). Occupational exposure to polychlorinated biphenyls in
electrical workers. II. Health effects. British Journol of
Industrial Medicine 38:55-60.
243. Maroni, N., Colombi, A., Cantoni, S., Ferioli, E., and Foa, V.
(1981). Occupational exposure to polychlorinated biphenyls in
electrical workers. I. Environmental and blood polychlorinated
biphenyls concentrations. British Journol of Industrial Medicine
38:49-54.
244. Bertozzi, P. A., Zocchetti, C. , Guercilena, S., Foglia, M. D.,
Pesatori, A., and Riboldi, L. (1S81). Mortality study of male
and female workers exposed to PCB's. Prevention of Occupational
Cancer - International Symposium pp. 242-248. Helsinke, Finland.
2-293
-------
245. Bertozzi, P., Riboldi, L., Pesotori, A., Rodice, L.. and
Zocchetti, C. (1985). Cancer Mortality of tlectrochemicol Workers
Exposed to PCS. Unpublished report.
246. International Agency for Research on Cancer (1980). Phenacetin.
IARC Monographs on the Evaluation of Corcinoqenic Risk of
Chemicals to Man 24:135-161.
247. Hultengren, N., Lagergren, C., and Ljungqvist, A. (1965). Carci-
noma of the renal pelvis in renal papillary necrosis. Acto
Chirurgico Scondinovico 130:314-320.
248. Adam, W. R., Dawborn, J. K.. Price, C. G., Riddell, J.. and Story,
H. (1970). Anoplastic transitional-cell carcinoma of the renal
pelvis in association with analgesic abuse. The Medical Journal
of Australia May 30:1108-1109.
249. Burnett, K. R.. Miller. J. B., and Greenbaum, E. I. (1980).
Transitional cell carcinoma: Rapid development in phenacetin
abuse. AJR 134:1259-1261.
250. Bengtsson, U., Angervall, L., Ekman, H., and Lehmann, L. (1968).
Transitional cell tumors of the renal pelvis in analgesic abusers.
Scandinavian Journal of Urology and Nephroloqy 2:145-150.
251. Angervall, L., Bengtsson, U., Zetterlund, C. G., and Zsigmond, M.
(1969). Renal pelvic carcinoma in a Swedish district with abuse
of a phenacetin-containing drug. British Journal of Urology
41:401-409.
252. Johansson, S. . Angervall, L., Bengtsson, U., and Wahlqvist, L.
(1974). Uroepithelial tumors of the renal pelvis associated with
abuse of phenacetin-contoining analgesics. Cancer 33:743-753.
253. Bengtsson, U., Johansson, S., and Angervall, L. (1978).
Malignancies of the urinary tract and their relation to analgesic
abuse. Kidney International 13:107-113.
254. Begley, M., Chadwick, J. M., and Jepson, R. P. (1970). A possible
case of analgesic abuse associated with transitional cell carci-
noma of the bladder. The Medical Journal of Australia 12: 1133-
1134.
255. Rothert, P., Melchior, H., and Lutzeyer, W. (1975). Phenacetin:
A carcinogen for the urinary tract. Journal of Urology 113:
653-657.
2-294
-------
256. Gonwa, T. A., Corbett, W. T. , Schey, H. M., and Buckalew, V. M.
(1980). Analgesic-associated nephropathy and transitional cell
carcinoma of the urinary tract. Annals of Internal Medicine
93(2).-249-252.
257. Fokkens, W. (1979). Phenacetin abuse related to bladder cancer.
Environmental Research 20:192-198.
258. McCredie, M. , Ford, J. M., Taylor, J. S., and Stewart, J. H.
(1982). Analgesics and cancer of the renal pelvis in New South
Wales. Cancer 49(12):2617-2625.
259. McCredie, M., Stewart, J. H., Ford, J. M., and MocLennan, R. A.
(1983). Phenacetin-containing analgesics and cancer of the
bladder or renal pelvis in women. British Journal of Urology
55:220-224.
260. International Agency for Research on Cancer (1980). Reserpine.
IARC Monographs on the Evaluation of Carcinogenic Risk of
Chemicals to Won 24:211-241.
261. Danielson, D. A., Jick, H., Hunter, J. R., Stergachis, A., and
Madsen, S. (1982). Nonestrogenic drugs and breast cancer.
American Journal of Epidemiology 116(2):329-332.
262. Ross, R. K., Paganini-Hill, A., Kroilo, M. D., Gerkins, V. R.,
Henderson, B. E., and Pike, M. C. (198**). Effects of reserpine on
prolactin levels and incidence of breast cancer in postmenopausal
women. Cancer Research 44:3106-3108.
263. O'Fallon, W. M. , Lobarthe, D. R.. and Kurland, L. T. (1975).
Rauwolfia derivatives and breast cancer. The Lancet Aug. 16:
292-296.
264. Laska, E. M. , Siegel, C., Meisner, M., Fischer, S., and
Wanderling, J. (1975). Matched-pairs study of reserpine use and
breast cancer. The Lancet Aug. 16:296-300.
265. Xodlin, D. and McCarthy, N. (1978). Reserpine and breast cancer.
Cancer 41(2):761-768.
266. Kewitz, H., Jesdinsky, H. J., Kreutz, G., and Schulz, R. (1980).
Reserpine and breast cancer. Archives Internationales de
Phormocodynomie et de Theropie 246:22-24.
267. Labarthe, 0. R. and O'Fallon, W. M. (1980). Reserpine and breast
cancer: A community-based longitudinal study of 2,000 hyperten-
sive women. Journal of the American Medical Association
243(22):2304-2310.
2-295
-------
268. Lilienfeld, A. M., Chang, L., Thomas, D. B., and Levin, M. L.
(1976). Rauwolfia derivatives and breast cancer. The John
Hopkins Medicol Journal 139(2):41-50.
269. Curb, J. D., Hardy, R. J., Labarthe, D. R., Borhani, N. 0., and
Taylor, J 0. (1982). Reserpine and breast cancer in the hyper-
tension detection and follow-up program. Hypertension 4:307-311.
270. Arnold, 0. L., Moodie, C. A., Grice, H. C., Charbonneau, S. M.,
Stavric, B., Collins, B. T., McGuire, P. F., Zawidzka, Z. Z., and
Munro, I. C. (1980). Long-term toxicity of ortho toluene-
sulfonomide and sodium saccharin in the rat. Toxicology and
Applied Pharmacology 52:113-152.
271. Colburn. W. A., Bekersky, I., and Blumenthol, H. P. (1981).
Dietary saccharin kinetics. Clinical Pharmacology and
Therapeutics 30(4):558-563.
272. Morgan, A. and Jain, M. (1974). Bladder cancer: smoking,
beverages and artificial sweeteners. Conodion Medical Associotion
Journal 111:1067-1070.
273. Howe, G. R., Burch, J. D., Miller, A. B., Morrison, B., Gordon,
P., Weldon, L., Chambers, L. W., Fodor, G., and Winsor, G. M.
(1977). Artificial sweetners and human bladder cancer. The
Lancet 17:578-581.
274. Morrison, A. S. and Buring, J. E. (1980). Artificial sweeteners
and cancer of the lower urinary tract. The New England Journal of
Medicine 302(10):537-541.
275. Hoover, R. and Strosser, P. H. (1980). Artificial sweeteners and
human bladder cancer: Preliminary results. The Lancet 19:837-
840.
276. Wynder. E. L. and Stellmon, S. 0. (1980). Artificial sweetener
use and bladder cancer: A case-control study. Science 207: 1214-
1216.
277. Moller-Jensen, 0., Knudsen, J. B., Sorensen, B. L., and Clemmesen,
J. (1983). Artificial sweeteners and absence of bladder cancer
risk in Copenhagen. Internotionol Journal of Cancer 32:577-582.
278. Armstrong, B., Lea, A. J.. Adelstein, A. M., Donovan, J. W.,
White, G. C. and Ruttle, S. (1976). Cancer mortality and
saccharin consumption in diabetics. British Journal of Preventive
and Social Medicine 30:151-157.
2-296
-------
279. International Agency fcr Resenrch on Cancer (1979). Trichloro-
ethylene. IARC Monographs on the Evaluation of Carcinogenic Risk
of Chemicals to Man 20: 545-572 .
280. Page, N. (1979). Assessment of trichloroethylene as an occupa-
tional carcinogen. IARC Scientific Publication 25:75-79.
281. Stewart, R. D. and Dodd, H. C. (1964). Absorption of carbon
tetrachloride, trichloroethylene, tetrachloroethylene, methylene
chloride, and 1, 1, 1-tricnloroethane through the human skin.
American Industrial Hygiene Associotion Journal 25:439-446.
282. Stewart, R. D.. Dodd, H. C.. Gay, H. H., and Erley. D. S. (1970).
Experimental human exposure to trichloroethylene. Archives of
Environmental Health 20:64-71.
283. Stewart. R. D., Gay, H. H., Erley. D. S.. Hake, C. L., and
Peterson, J. E. (1962). Observations on the concentrations of
trichloroethylene in blood and expired air following exposure of
humans. American Industrial Hygiene Associotion Journal 23:
167-170.
284. Ikeda, M. . Ohtsuji, H , Imamuro, T., and Komoike, Y. (1972).
Urinary excretion of total trichloro-compounds, trichloroethanol,
and trichloroacetic acid as a measure of exposure to trichloro-
ethylene and tetrachloroethylene. British Journal of Industrial
Medicine 29:328-333.
285. Kimmerle. G. and Eben, A. (1973). Metabolism, excretion and toxi-
cology of trichloroethylene after inhalation. 2. Experimental
human exposure. Archives of Toxicology 30:127-138.
286. Sato, A., Nakajima, T., Fujiwara, Y., and Murayamc. N. (1977). A
pharmacokinetic model to study the excretion of tricMoroethylene
and its metabolites after an inhalation exposure. British Journal
of Industrial Medicine 3*:56-63.
287. Smith, G, F. (1978). Trichlorethylene - relationship of metabo-
lite levels to atmospheric concentrations: preliminary co-nmuni-
cation. Journal of the Royal Society of Medicine 71:591-595.
288. Axelson, 0., Andersson, K., Hogstedt, C., Holmbe.g, b., fiolina,
G., and de Verdier, A. (1978). A cohort study on trichloro-
ethylene exposure and cancer mortality. Journjl of Occupational
Medicine 20(3 ):194-196.
289. Blair, A. (1980). Mortality among workers in the metal polishing
and plating industry, 1951-1969. Journal of Occupational Medicine
22(3):158-162.
2-297
-------
290. Blair, A., Decoufle, P., and Grauman, 0. (1979). Causes of death
among laundry and dry cleaning workers. American Journal of
Public Health 69(5):508-511.
291. Tola, S., Vilhunen, R., Jarvinen, t. , and Korkala, M. L. (1980).
A cohort study on workers exposed to trichloroethylene. Journal
of Occupational Medicine 22( 11 ): 737-740.
292. Paddle, G. M. (1983). Incidence of liver cancer and trichloro-
ethylene manufacture: joint study by industry and a cancer
registry. British Medical Journal 286(846):836.
293. International Agency for Research on Cancer (1979). Vinyl
chloride, polyvinyl chloride and vinyl chloride-vinyl acetate
copolymers. IARC Monographs on the Evaluation of Carcinogenic
Risk of Chemicals to Man 19:377-438.
294. Nicholson, W. J., Henneberger, P. K., and Seidman, H. (1984).
Occupational hazards in the VC-PVC industry. Progress in
Clinical end Biological Research 1*1:155-175.
295. Suciu, I., Prodan, L., Ilea, E., Paduraru, A., and Pascu, L.
(1975). Clinical manifestations in vinyl chloride poisoning.
Annals of the New York Academy of Sciences 246:53-69.
296. Brady, J., Liberatore, F., Harper, P., Greenwald, P., Burnett, W.,
Davies, J. N. P.. Bishop, M., Polan, A., and Vienna, N. (1977).
Angiosarcomo of the liver: An epidomiologic survey. Journal of
the Notional Cancer Institute 59(5):1383-1385.
297. Delorme, F. and Theriault, G. (1978). Ten coses of angiosarcoma
of the liver in Shawinigan, Quebec. Journal of Occupational
Medicine 20(5):338-340.
298. Popper. H.. Thomas, L. B., Telles, N. C., Folk, H., and Selikoff,
I. J. (1978). Development of hepatic angiosarcoma in man induced
by vinyl chloride, thorotrast, and arsenic. American Journal of
Pathology 92(2):349-376.
299. Spirtas, R. and Kaminski, R. (1978). Angiosarcoma of the liver in
vinyl chloride/polyvinyl chloride workers. Journal of
Occupational Medicine 20(6):427-429.
300. Wagoner, J. K. (1978). Vinyl chloride and pulmonary coricer.
Journal of Environmental Pathology and Toxicology 1:Io1-362.
301. Infante, P. F. (1981). Observations of the site-specific carcino-
genicity of vinyl chloride to humans. Environmental Health
Perspectives 41:89-91.
2-298
-------
302. Moltoni, C., Lefemine, G., Ciliberti, A., Cotti, G., ond Carretti, D.
(1984). Experimental research on vinyl chloride carcinogenesis.
In: Archives of Research on Industrial Corcinogenesis 2:27-35.
Maltoni, C. ond Mehlman, M. A. (eds.) Princeton Scientific
Publishers, Inc., Princeton, New Jersey.
303. Barnes, A. W. (1976). Vinyl chloride and the production of PVC.
Proceeding of the Royal Society of Medicine 69:277-281.
304. Jones, J. H. (1981). Worker exposure to vinyl chloride and
poly(vinyl chloride). Environmental Health Perspectives 41:
129-136.
305. Aryanpur, J. (1977). Vinyl chloride: Its impact on occupational
medicine practice. Journal of Occupotionol Medicine 19(10):
689-692.
306. Nicholson, W. J., Henneberger, P. K., and Tarr, D. (1984). Trends
in cancer mortality among workers in the synthetic polymers
industry. Progress in Clinical ond Biological Research 141:65-78.
307. Kuzmack, A. M. ond McGaughy, R. E. (1975). Quantitative Risk
Assessment for Community Exposure to Vinyl Chloride. Office of
Health and Ecological Effects. U. S. Environmental Protection
Agency.
308 Nicholson, W. J., Hammond, E. C., Seidman, H., and Selikoff, I. J.
(1975). Mortolity experience of a cohort of vinyl chloride poly-
vinyl chloride workers. Annals of the New York Academy of
Sciences 246:225-230.
309. Tabershaw, I. R. and Gaffey, W. R. (1974). Mortality study of
workers in the manufacture of vinyl chloride and its polymers.
Journal of Occupational Medicine 16(8):509-51S.
310. Buffler, P. A., Wood, S., Eifler, C., Suarez, L., ond Kilian D. J.
(1979). Mortality experience of workers in a vinyl chloride
monomer production plant. Journal of Occupational Medicine
21(3) :195-2U3.
311. Waxweiler, R. J.. Stringer, W. , Wagoner, J. K., Jones, J., Folk,
H. 'ind Carter, C. (1976). Neoplastic risk among workers exposed
to vinyl chloride. Annols of the New York Academy of Sciences
271:40-48.
312. Byren, D., Engholm, G., Englund, A., and Westerholm, P. (1976).
Mortality nnd cancer morbidity in a group Swedish VCM and PCV
production workers. Environmental Health Porspoctives 17:167-170.
2-299
-------
313. Fox, A. J. and Collier, P. F. (1977). Mortality experience of
workers exposed to vinyl chloride monomer in the manufacture of
polyvinyl chloride in Greet Britain. British Journal of
Industrial Medicine 34:1-10.
3U. Ott,.M. G., Langner. R. R., and Holder, B. B. (1975). Vinyl
chloride exposure in a controlled industrial environment.
Archives of Environmental Health 30:333-339.
315. Duck, B. W., Carter, J. T., Coombes, E. J. (1975). Mortality
study of workers in a polyvinyl-chloride production plant. The
Lancet Dec. 13:1197-1199.
316. Heldaas, S. S., Langard, S. L., Andersen, A. (1984). Incidence
of cancer among vinyl chloride end polyvinyl chloride workers.
British Journal of Industrial Medicine 41:25-30.
317. Goldsmith, J. and Beeser, S. (1984). Strategies for pooling data
in occupational epidemiological studies. Annals Academy of
Medicine 13(2)(suppl) :297-310.
2-300
------- |