EPA/600/P-97/001F

April 1998

Carcinogenic Effects of Benzene:

An Update

National Center for Environmental Assessment–Washington Office

Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
 DISCLAIMER

This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.

ABSTRACT

The major issue addressed in this document involves the nature and magnitude of the
inhalation risk of cancer to humans exposed to low levels of benzene. Occupational studies
continue to provide the bulk of evidence of benzene’s carcinogenicity. Workers are exposed at
much higher levels than is the general public. This document reconfirms that benzene is a
“known” human carcinogen by all routes of exposure. This finding is supported by evidence from
human epidemiologic studies, animal data, and an improvement in understanding of mechanism(s)
of action. Human epidemiologic studies of highly exposed occupational cohorts have
demonstrated that inhalation exposure to benzene can cause acute nonlymphocytic leukemia and
other blood disorders, that is, preleukemia and aplastic anemia. Additionally, changes in blood
and bone marrow consistent with hematotoxicity are recognized in humans and experimental
animals. Currently, there is insufficient evidence to deviate from using an assumption of a linear
dose-response curve for benzene, hence, the Agency’s past approach of using a model with low-
dose linearity is still recommended. Of the various approaches employing a linear assumption,
utilizing the Pliofilm workers cohort, the inhalation risk at 1 ppm ranges from 7.1 × 10-3 to 2.5 ×
10-2. This reflects a modest change from the EPA’s 1985 interim risk assessment which provided
only a single estimate of risk (i.e., 2.6 × 10-2).

ii
 CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

PREFACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

AUTHORS, CONTRIBUTORS, AND REVIEWERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

GLOSSARY OF ACRONYMS AND ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

EXECUTIVE SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1. HISTORY OF THE 1985 INTERIM DOCUMENT . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. PROPOSED 1996 GUIDELINES FOR CARCINOGEN RISK ASSESSMENT . . 2

2. HAZARD ASSESSMENT AND CHARACTERIZATION . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1. HUMAN DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. LABORATORY ANIMAL DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3. MODE-OF-ACTION INFORMATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1. Metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2. Mutagenicity and Genotoxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.3. Epigenetic Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.4. Pathogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4. HAZARD CHARACTERIZATION SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . 28

3. DOSE-RESPONSE ASSESSMENT AND CHARACTERIZATION . . . . . . . . . . . . . . . . . 30

3.1. DESCRIPTION OF DIFFERENT RISK ASSESSMENTS . . . . . . . . . . . . . . . . . 32
3.2. SHAPE OF THE DOSE-RESPONSE FUNCTION AT LOW DOSES . . . . . . . . . 34
3.3. DOSE-RESPONSE CHARACTERIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4. CHILDREN’S RISK CONSIDERATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5. FUTURE RESEARCH NEEDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

6. REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

iii
 LIST OF TABLES

1. Relative risk as a function of cumulative exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2. Standardized mortality ratios for deaths from leukemia among

Pliofilm workers based on the estimated cumulative exposure
of the selected investigators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3. Estimated relative risks of leukemia derived by the proportional hazards

dose-response model according to the estimated cumulative
exposure (ppm-years) of the selected investigators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4. Risk estimates calculated on the basis of Pliofilm workers

by various investigators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

5. Evidence that benzene-induced leukemia is nonlinear at low doses . . . . . . . . . . . . . . . . 35

LIST OF FIGURES

1. Key metabolic activation pathways in benzene toxicity. . . . . . . . . . . . . . . . . . . . . . . . . 17

2. Schematic for mechanistic hypothesis of benzene pathogenesis (leukemogenesis) . . . . . 27

3. Schematic illustrating various options for modulation of bone marrow cell

populations by benzene metabolites that could result in the induction of aplastic
anemia, leukemia, and immunotoxicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

iv
 PREFACE

The National Center for Environmental Assessment-Washington Office, Office of

Research and Development, has prepared this document on the Carcinogenic Effects of Benzene:
An Update to serve as a source document for the Office of Air and Radiation, Office of Mobile
Sources, to support decision making regarding regulation of benzene as a hazardous air pollutant.
In the development of this document, the scientific literature has been reviewed, key
studies have been evaluated, and summary/conclusions have been prepared so that the
carcinogenicity and related information are qualitatively and quantitatively characterized. This
updated evaluation and review of benzene carcinogenicity was conducted under the standing
guidance of the 1986 cancer risk assessment guidelines, but with a recognition of the proposed
1996 cancer risk assessment guidelines emphasizing mode of action and dose-response analysis.
Relevant literature has been reviewed through July 1997.
The emphasis of this document is a detailed discussion of the relevancy of the 1985 cancer
unit risk assessment of benzene in light of new information.
This final document reflects a consideration of all comments received on an External
Review Draft dated June 1997 (EPA/600/P-97/001A) provided by an expert panel at a peer
review workshop (July 16, 1997) and comments received during a public review and comment
period (June - July 1997).

v
 AUTHORS, CONTRIBUTORS, AND REVIEWERS

This document was prepared by the National Center for Environmental Assessment-
Washington Office (NCEA-W) of EPA’s Office of Research and Development.

AUTHORS
David L. Bayliss, NCEA-W
Chao Chen, NCEA-W
Annie Jarabek, NCEA-W
Babasaheb Sonawane, NCEA-W
Lawrence Valcovic, NCEA-W

CONTRIBUTORS

NCEA
Robert McGaughy, NCEA-W
James Walker, NCEA-W

Outside Contributor
Martyn T. Smith (contractor), University of California, Berkeley, CA

REVIEWERS

NCEA
Michael Callahan, NCEA-W
David Cleverly, NCEA-W
James Cogliano, NCEA-W
William Farland, NCEA
Charlie Ris, NCEA-W
John Schaum, NCEA-W
Chon Shoaf, NCEA-RTP

Other EPA Offices

Linda Birnbaum, NHEERL, RTP
Pam Brodowicz, OMS, RTP

ACKNOWLEDGMENT
The benzene team would like to acknowledge Dr. Richard Williams for his assistance in
completing the final document.

vi
 GLOSSARY OF ACRONYMS AND ABBREVIATIONS

ALL acute lymphocytic leukemia

AML acute myeloid leukemia
AMML acute myeloid and monocytic leukemia
ANLL acute nonlymphocytic leukemia
ATSDR Agency for Toxic Substances and Diseases Registry
C.I. confidence interval
CFU-GM colony-forming unit-granulocyte/macrophage
CLL chronic lymphocytic leukemia
CML chronic myeloid leukemia
DNA desoxyribonucleic acid
FISH fluorescent in situ hybridization
GM-CSF granulocyte/macrophage-colony-stimulating factor
GPA glycophorin A
GSH glutathione
i.p. intraperitoneal
IARC International Agency for Cancer Research
IL-1 interleukin 1
IPCS International Programme on Chemical Safety
MA trans, trans-muconaldehyde
MDS myelodysplasic syndromes
MOE margin of exposure
MPO myeloperoxidase
NCEA National Center for Environmental Assessment (EPA)
NIOSH National Institute for Occupational Safety and Health
NQO1 NAD(P)H:P quinone oxidoreductase
OAQPS Office of Air Quality Planning and Standards (EPA)
OAR Office of Air and Radiation (EPA)
OMS Office of Mobile Sources (EPA)
OSHA Occupational Safety and Health Administration
PHA phytohemaglutin
POD point of departure
ppb parts per billion
ppm parts per million
RBC red blood cell
RH rubber hydrochloride
RR relative risk
RR rate ratio
SMR standard mortality ratio
TWA time weighted average
WBC white blood cell

vii
 EXECUTIVE SUMMARY

In 1992, the U.S. Environmental Protection Agency’s (EPA’s) Office of Air and
Radiation, Office of Mobile Sources, requested that the National Center for Environmental
Assessment (NCEA) provide an updated characterization of the cancer risk to humans from
inhalation exposure to benzene. The previous characterization of the carcinogenic risk of
exposure to benzene was done in 1985 by the Office of Health and Environmental Assessment
(the predecessor organization to NCEA). Additional scientific data relevant to the carcinogenicity
of exposure to benzene have been published in the literature since that time. This has brought into
question the relevancy of the earlier quantitative cancer risk estimates. The 1985 cancer unit risk
estimates were based on assumptions about the effects of low-level benzene exposure on humans
derived from occupational health studies. The scope of this report is limited to issues related to
the carcinogenicity of exposure to benzene. Specifically, this report evaluates and discusses
studies published since 1985 to ascertain if there has been sufficient new scientific information that
would significantly alter the 1985 interim benzene cancer unit risk estimate.
This updated evaluation and review of the benzene risk assessment is being conducted
under the standing guidance of the 1986 Guidelines for Carcinogen Risk Assessment but with a
recognition of additional areas of emphasis in the 1996 Proposed Guidelines for Carcinogen Risk
Assessment. Thus, this updated review of the benzene risk assessment contains a discussion of
how recent evidence on mode of action can be incorporated into hazard characterization and
dose-response approaches.
The major issue addressed in this document involves the magnitude of the risk of cancer to
humans exposed to low inhalation levels of benzene. Occupational exposure studies provide the
bulk of evidence of carcinogenicity, since workers were exposed at higher levels than is the
general public.
It has been clearly established and accepted that exposure to benzene causes acute
nonlymphocytic leukemia and a variety of other blood-related disorders in humans. The existing
EPA Group A classification of benzene based on the 1986 guidelines is replaced with a narrative
incorporating the “known/likely” descriptor under the 1996 proposed guidelines. The narrative
discusses the uncertainties about the following: the shape of the dose-response curve at low
doses, mode of action, and exposure in human studies. The study of Pliofilm rubber workers at

viii
three facilities in Ohio provides the best published set of data to date for evaluating human cancer
risks from exposure to benzene. Compared to other published studies, this cohort has the fewest
reported co-exposures in the workplace to other potentially carcinogenic substances that might
confound risk analysis for benzene. Since the 1985 interim assessment, this cohort has been
expanded to include workers who were employed for at least one day between 1940 and 1965. It
included an additional 6.5 years of follow-up from the earlier study as well as individual estimates
of personal exposure, which were not included in the previous study. One myeloblastic leukemia
was subsequently noted after the additional follow-up. However, because of the compensating
increase in expected deaths due to the additional person-years of follow-up, only a small change
occurred in the overall relative risk.
Several investigators have employed various assumptions to estimate occupational
exposure levels prior to 1950, when exposures were most intense. The estimates of exposure
made by Rinsky were generally the lowest, but there is no consistent pattern among the estimates
for particular years. Even with the differences in the exposure levels produced by utilizing these
sets of estimates of exposure for the employees, the cumulative SMRs differed by no more than a
factor of 3.
In the 1985 interim benzene document, a single overall unit risk estimate was calculated as
the geometric mean of four maximum likelihood unit risk estimates generated from the three
available exposed worker studies. The result was a probability of 2.6 × 10-2. Neither of the
chemical worker studies has sufficient power for independent calculations, and the net result of
discarding data from them has only a small effect on the unit risk estimate.
Recently, the National Cancer Institute, in cooperation with the Chinese Academy of
Preventive Medicine, published early results from a comprehensive study of 74,828 benzene-
exposed workers employed from 1972 to 1987 in 672 factories in 12 cities of China. This study,
one of the largest of its type ever undertaken, enabled its authors to claim detection of
significantly elevated risks at extraordinarily low levels of exposure. Their findings suggested that
workers exposed to benzene at average levels of less than 10 ppm are subject to a higher risk of
hematologic neoplasms. Although most of the attributes considered important in a long-term
retrospective study of this nature and magnitude were addressed by the authors, some
uncertainties and potential weaknesses still remain. The derivation of the cohort from many
different factories across China suggested the possibility that this cohort was exposed to mixtures

ix
of many different chemicals. Exposure to other carcinogens in the workplace could have
produced confounding effects, especially if exposures were to chemicals that increase the risk of
leukemia. In addition, the derivation of early exposure estimates to benzene may be biased and
life-style and socioeconomic factors may have had impacts. It is clear that this study contains
thought-provoking new findings and conclusions. It is, however, premature to assume that the
Chinese data should now replace Rinsky’s Pliofilm workers cohort in the derivation of risk
estimates.
At present, identification of the mechanisms by which exposure to benzene and its
metabolites exerts their toxic and carcinogenic effects remains elusive. Animal studies showed
that benzene itself was unlikely to be the actual toxicant, but rather that metabolism (i.e.,
metabolic transformation) was required for toxic effects to occur. The central issues in
integrating the mechanistic data from the laboratory animal experiments with the occupational
epidemiologic data to estimate risk of the anticipated ambient low-level human scenario are
establishing whether the mechanisms that are operative in laboratory animals are similar to
mechanisms operative in humans and accounting for the dose dependency of those mechanisms.
Two pathways have been commonly postulated as being responsible for benzene toxicity. The
first involves the major hepatic metabolites of benzene—phenol, catechol, and hydroquinone—
while the second involves ring-opened forms of benzene. The majority of data suggest that the
hepatic conversion of benzene to phenolic metabolites is an important primary event in benzene
toxicity. Catechol and hydroquinone have been shown to persist in bone marrow after benzene
exposure and the bone marrow is rich in peroxidase activity, and phenolic metabolites of benzene
can be activated by peroxidases to reactive quinone derivatives. Although much less is known
about the metabolism of benzene in humans than in laboratory animals, existing studies indicated
that both metabolize benzene along similar pathways. The rates of some reactions, however, may
differ. A recent series of investigations on Chinese workers highly exposed to benzene provides
insights into particular enzymes underlying the hematotoxicity of benzene in humans. Overall,
individuals with the rapid hydroxylator phenotype exhibited a 2.5-fold increased risk of benzene
poisoning as compared to those with the slow hydroxylator phenotype.
Benzene affects bone marrow cells in several different ways. Based on our current
understanding, these effects are produced by the interactive effects of multiple metabolites.
Genotoxic effects are a critical component of the leukemogenic properties of benzene. As more

x
information becomes available about the epigenetic effects of benzene and the role these effects
play in the leukemogenic process in general, it is likely that these will be shown to have an
important role. Evidence supports the hypothesis that more than one toxic effect contributes to
the leukemogenic process, especially because benzene metabolic products may be able to cause
general disruption of protein functions in bone marrow cells. Protein damage is likely to result in
pleiotropic effects, including general toxicity, alteration of growth factor responses, and DNA
damage. Therefore, the overall picture of benzene-induced leukemogenesis is an increased rate of
genetic damage to hematopoietic cells that occurs in the context of disrupted bone marrow
function. This situation could encourage not only the production of cells with key genetic
changes, but also the selection and expansion of such cells due to the abnormal marrow.
However, data are not sufficient at this time to state precisely which of the various documented
effects, genotoxic or otherwise, are the critical ones for benzene-induced leukemogenicity.
In the 1985 benzene risk assessment, the lifetime leukemia risk due to benzene was
developed using the geometric mean of risk estimates that were calculated on the basis of data
from three studies of exposed workers. Subsequently, several risk assessments based on Rinsky's
cohort have become available with individual risk estimates using varying assumptions and/or
models, with outcomes ranging more than six orders of magnitude. Some recent evidence
suggests the possibility that the low-dose curve could be supralinear since the formation of toxic
metabolites plateaus above 25 ppm benzene in air. Thus, it is possible that the unit risk is
underestimated if linearity is assumed at low doses. However, none of the approaches can be said
to have greater scientific validity than any other; hence there is no clear basis for choosing a single
best estimate. Rather, the set of risk estimates reflects both the inherent uncertainties in the risk
assessment of benzene and the limitations of the epidemiologic response and exposure data.
While the risk estimates would be significantly different if a nonlinear exposure response model
was found to be more plausible, the shape (i.e., the nonlinearity) of the exposure-response curve
cannot be determined without a better understanding of the biological mechanism of benzene-
induced leukemia. The arguments made in favor of benzene-induced leukemia being nonlinear at
low doses can be matched by arguments opposing this as a most likely occurrence. Thus, there is
not sufficient evidence currently to reject a linear dose-response curve for benzene in the low-
dose region, nor is there sufficient evidence to demonstrate that benzene is, in fact, nonlinear in its
effects. Since this knowledge is not available at the present time, the EPA default approach of

xi
using a model with low-dose linearity is still recommended. Of the various approaches employing
a linear assumption, the risk at 1 ppm ranges from
7.1 × 10-3 to 2.5 × 10-2 , within which any calculated unit risk estimate would have equal scientific
validity.
The effects from exposure to benzene could be potentially different among
subpopulations, including children. However, there is insufficient data on differential
susceptibility from environmental exposure to benzene, and it is not possible to make quantitative
adjustments for these factors at this time.
Data insufficiencies in several areas are noted and research in these areas ultimately should
provide a better understanding of how benzene causes cancer, particularly the mechanism of
benzene-induced leukemia. Specific measures of early genetic damage in humans with known
exposure to benzene will help define the biological events leading up to the disease by providing
internal markers of its progression. Such information may be forthcoming in the near future from
a large cohort of benzene-exposed workers under study in China. A need exists to further
validate toxicokinetic models and to assess metabolic susceptibility factors in human subjects.
Continued basic research in hematopoiesis and leukemia is critical for identifying the mechanisms
of leukemogenesis. There remain important unanswered questions about the cell population that
contains targets for leukemic transformation, such as cell number and rate of division, quiescence
patterns, maturation, regulation, and apoptotic behavior. Particular emphasis should be placed on
research on those sensitive subpopulations who are believed to be at increased risk (e.g., infants
and children, the elderly).

xii
 1. INTRODUCTION

In 1992, the U.S. Environmental Protection Agency’s (EPA’s) Office of Air and Radiation
(OAR), Office of Mobile Sources (OMS) requested that the National Center for Environmental
Assessment (NCEA) provide an updated characterization of the cancer risk to humans of
inhalation exposure to benzene. The previous interim characterization of the carcinogenic risk of
exposure to benzene was done in 1985 by the Office of Health and Environmental Assessment
(the predecessor to NCEA). Additional scientific data relevant to the carcinogenicity of benzene
exposure have been published since that time. This has brought into question the relevancy of the
earlier quantitative cancer risk estimates. The 1985 estimates were based on assumptions about
the effects of low-level benzene exposure on humans derived from occupational health studies.
The regulatory authority (Clean Air Act Amendments, 1990) for controlling fuel emissions
from vehicles resides in OAR. OMS has asked NCEA to provide a scientific support document
based on health implications of continued exposure to benzene.
The scope of this report is limited to issues related to the carcinogenicity of exposure to
benzene. Specifically, this report evaluates and discusses studies published since 1985 to ascertain
if there has been sufficient new scientific information that would significantly alter the 1985
interim benzene cancer unit risk estimate.

1.1. HISTORY OF THE 1985 INTERIM DOCUMENT

In 1985, the Office of Research and Development prepared estimates of the inhalation unit
risk for benzene (U.S. EPA, 1985) at the request of Office of Air Quality Planning and Standards
(OAQPS). The previous cancer risk assessment of benzene by the Agency was completed in
January 1979 (U.S. EPA, 1979). Subsequently, this assessment became out of date as new
scientific information became available. In response to the need to update the 1979 assessment,
the 1985 Interim Quantitative Cancer Unit Risk Estimate Due to Inhalation of Benzene was
developed. It reviewed and incorporated information from three epidemiologic studies at the time
(Rinsky et al., 1981; Ott et al., 1978; Wong et al., 1983). In addition, animal inhalation studies in
male rats and mice (Goldstein et al., 1982) and in male and female rats (Maltoni et al., 1983) were
added to the information base.
Data from the occupational cohorts of Rinsky et al. (1981), Ott et al. (1978), and Wong et
al. (1983) were pooled and analyzed by Crump and Allen (1984) to provide exposure (cumulative
dose) estimates for use in the development of a benzene cancer risk assessment for the
Occupational Safety and Health Administration (OSHA, 1987) independently of EPA. These
exposure estimates were available for use by the Agency. Crump and Allen (1984) made their
exposure estimates using three separate approaches (cumulative, weighted cumulative, and

1
window) and two risk models (absolute and relative). The cumulative dose approach assumes
that the risk depends on the air concentration times duration of exposure. The weighted
cumulative dose approach assumes that the contribution of an exposure to risk varies depending
on when exposure occurred. The window approach assumes that benzene exposure for longer
than 15 years induces no additional risk, but that exposure between 5 and 10 years induces a risk
proportional to the air concentration and exposure duration. All three approaches assume a
latency period that begins at the beginning of exposure, during which there is assumed to be no
increased risk. An absolute risk model assumes that the risk from exposure is independent of the
background risk of disease, whereas a relative risk model assumes that the risk from exposure is
proportional to the background incidence of the disease (see Section 3.1).
The Agency concluded that the cumulative and the weighted cumulative exposure
estimates were both valid and preferable to the window approach. EPA also concluded that the
absolute and relative risk models had equal validity. It was decided to calculate the geometric
mean of the four resulting estimates derived from the different exposure estimates and risk models
and then multiply this by a correction factor based on the epidemiologic data of Wong et al.
(1983). This correction factor (1.23) was the ratio of risk estimates (under the relative risk model
and cumulative exposure estimate) when all three studies (Rinsky et al., 1981; Ott et al., 1978;
Wong et al., 1983) were used to the risk estimate generated when only the Rinsky et al. (1981)
and Ott et al. (1978) studies were used. The Wong et al. (1983) study was not used under the
absolute risk model because its information was considered insufficient by Crump and Allen
(1984), whose study formed the basis for EPA risk numbers. The resulting quantitative cancer
unit risk of 2.6 × 10-2 per ppm air concentration was about 10 times greater than the human risk
estimate based on the three animal inhalation studies and 1.5 times higher than the pooled
estimates from the three gavage studies. This estimate compared well with the original estimate
from the 1979 benzene risk document (U.S. EPA, 1979) of 2.41 × 10-2, which was based on the
geometric mean of three unit risk estimates derived from the occupational cohort studies of
Infante et al. (1977), Aksoy (1976, 1977), Aksoy et al. (1974), and Ott et al. (1977).

1.2. PROPOSED 1996 GUIDELINES FOR CARCINOGEN RISK ASSESSMENT

The Agency recently published its Proposed Guidelines for Carcinogen Risk Assessment
(U.S. EPA, 1996). When final, these guidelines will supersede the existing Guidelines for
Carcinogen Risk Assessment (U.S. EPA, 1986). The 1996 proposed cancer risk assessment
guidelines include a number of changes that encompass a more detailed understanding of the
carcinogenic processes and provide a framework for the use of mechanistic data. It should be
noted, however, that the results of an assessment under the new guidelines will not differ greatly

2
from those under the 1986 guidelines, unless new kinds of information are forthcoming from
research on mechanisms and toxicokinetics.
The proposed guidelines are intended to provide for greater flexibility in incorporating the
rapidly increasing data in decisions to implement the Agency’s regulatory authority. Risk
characterizations are important components of the new guidelines and serve to explain the key
lines of evidence and conclusions, discuss the strengths and weaknesses of the evidence, present
alternative conclusions, and point out significant issues and uncertainties deserving serious
consideration. A risk characterization summary would integrate technical characterizations of
exposure, hazard, and dose response to form the overall synthesis and conclusions about human
health risk. This document is limited to discussions of the hazard and dose-response
characterizations.
The hazard assessment component emphasizes use of information about an agent’s mode
of action to reduce the uncertainty in describing the likelihood of harm and to provide insight into
appropriate extrapolation procedures. Mode of action is defined as the agent’s influence on
molecular, cellular, and physiological functions. Because it is the sum of the biology of the
organism and the chemical properties of an agent that leads to an adverse effect, evaluation of the
entire range of data (i.e., physical, chemical, biological, and toxicological) permits a reasoned
judgement of an agent’s mode of action. Although cancer is a complex and diverse process, a risk
assessment must analyze the presumed critical events, at least those that can be measured
experimentally, to derive a reasonable approximation of risk. Understanding the mode of action
helps interpret the relevancy of the laboratory animal data and guides the dose-response
extrapolation procedure, i.e., it helps to answer the question of the shape of the dose-response
function at low doses. The conditions (i.e., route, duration, pattern, and magnitude of exposure)
under which the carcinogenic effects of the agent may be expressed should also be considered in
the hazard characterization.
The weight-of-evidence narrative for the hazard characterization includes classification
descriptors. Three standard descriptors (“known/likely,” “cannot be determined,” and “not
likely”) were proposed to replace the six letter categories used in the 1986 guidelines (i.e., A-E).
Because of the wide variety of data sets encountered, these descriptors are not meant to stand
alone; rather, the narrative context is intended to provide a transparent explanation of the
biological evidence and how the conclusions were derived.
The dose-response assessment under the new guidelines is a two-step process. In the first
step, the response data are modeled in the range of empirical observation. Modeling in the
observed range is done with biologically based or appropriate curve-fitting models. The second
step, extrapolation below the range of observation, is accomplished by modeling if there are
sufficient data or by a default procedure.

3
 This updated evaluation and review of the benzene risk assessment is being conducted
under the standing guidance of the 1986 cancer risk assessment guidelines, but with a recognition
of these areas of emphasis in the 1996 proposed cancer risk assessment guidelines. Thus, this
updated review contains a discussion of how recent evidence on mode of action can be
incorporated into hazard characterization and dose-response approaches. Earlier dose-response
assumptions or alternative approaches will be discussed in this context.
The major issue addressed in this document involves the magnitude of the risk of cancer to
humans exposed to low levels of benzene. Occupational exposure studies provide the bulk of
evidence of carcinogenicity, since workers are exposed at much higher levels than is the general
public. The 1996 proposed cancer risk assessment guidelines recommend a detailed discussion of
the basis for developing the quantitative unit risk estimate, drawing on mode-of-action,
metabolism, and pharmacokinetics information replete with uncertainty discussions as appropriate.
The 1985 interim risk estimate calculation for benzene was based on science policy using a
procedure incorporating the geometric mean of maximum likelihood estimates because little
information was available regarding carcinogenicity at low exposure levels.

2. HAZARD ASSESSMENT AND CHARACTERIZATION

The “known/likely” category of the proposed 1996 cancer risk assessment guidelines
includes agents for which adequate epidemiologic evidence (known) or a combination of
epidemiologic and experimental evidence demonstrates an association between human exposure
and cancer.
It has been clearly established and accepted that exposure to benzene causes acute
nonlymphocytic (myelogenous) leukemia (ANLL) and a variety of other blood-related disorders
in humans (ATSDR, 1997; IARC, 1982; U.S. EPA, 1979, 1985). The existing Group A
classification of benzene based on the 1986 guidelines would be replaced with a narrative
incorporating the “known/likely” descriptor under the 1996 proposed guidelines. The narrative
discusses the uncertainties about the following: the shape of the dose-response curve at low
doses, mode of action, and exposure in human studies; these topics are addressed in this section.

2.1. HUMAN DATA

Epidemiologic studies and case studies provide clear evidence of a causal association
between exposure to benzene and leukemia, especially ANLL and, to a lesser extent, chronic
nonlymphocytic leukemia as well as chronic lymphocytic leukemia (CLL) (Vigliani and Saita,
1964; Aksoy et al., 1974; Aksoy, 1976, 1977; Infante et al., 1977; Rinsky et al., 1981, 1987;

4
IARC, 1982; ATSDR, 1997). A number of studies, including the Pliofilm cohort, have indicated
that benzene exposure is associated with various types of lymphohematopoietic neoplasia other
than acute nonlymphocytic leukemia (ANLL) (Savitz and Andrews, 1996). However, the specific
types associated with benzene exposure remain unidentified. Lymphocytic leukemia, commonly
found in children, may have a genetic component as well as an environmental exposure
component (Linet, 1985). Exposure to benzene and other environmental chemicals cannot be
ruled out. A higher risk of multiple myeloma was once thought to be associated with exposure to
benzene (DeCouflé et al., 1983; Rinsky et al., 1987). However, later studies have failed to
confirm this (Hayes et al.,1996, 1997). One new site-specific cancer, Hodgkin’s lymphoma,
appears to be associated with exposure to benzene as well as with hematologic neoplasms in
general, which includes AML and related myelodysplasic syndromes (Hayes et al., 1997).
The study of Pliofilm rubber workers at three facilities in Ohio (Rinsky et al., 1981)
provides the best published set of data to date for evaluating human cancer risks from exposure to
benzene. Compared to other published studies (Hayes et al., 1996; Bond et al., 1986a; Wong,
1987; Schnatter et al., 1996a; Rushton et al., 1997), Rinsky et al. (1981) has the fewest reported
co-exposures in the workplace to other potentially carcinogenic substances that might confound
risk analysis for benzene. Except for the cohort studied by Bond et al. (1986a), the Pliofilm
workers, furthermore, experienced a greater range of estimated exposure to benzene than the
cohorts of other studies in which efforts were made to estimate individual exposures. The value
of Bond et al. (1986a) for analysis of the effects of exposure to benzene was diminished by
reported coexposures to styrene, arsenic, and other potentially carcinogenic substances. Hence,
the Pliofilm workers are the preferred population for estimating the effects of exposure to
benzene. Since the 1985 interim assessment, this cohort has been expanded (Rinsky et al., 1987)
to include workers who were employed at least 1 day between January 1, 1940, and December
31, 1965. (In the previous study, employment after December 31, 1950, was not considered.)
Three questions have been raised concerning the impact of these more recent data on the
present updated assessment of benzene and its use in a quantitative risk assessment. First, does
the update lead to any substantial changes in the estimated relative risk ratios that were derived in
the 1981 study? Second, do the various approaches used to estimate exposure in the those early
years lead to risk estimates that differ by a substantial amount? One of the major problems with
exposure estimates used by Rinsky et al. (1981, 1987) and others in deriving relative risk
estimates for use in developing quantitative unit risk estimates is that no ambient air
measurements of exposure to benzene in the Pliofilm workplace were taken before 1946, and in
that year there were only four samples measured. The absence of earlier definitive ambient air
measurements has led to many quantitative risk estimates by numerous investigators over the past
several years that have differed from each other partially on the basis of differences in the

5
assumptions made about what those earlier exposures to benzene were. Third, because the
Rinsky et al. (1987) Pliofilm cohort is currently the best set of data available for estimating
exposure and the risk of leukemia, would it be advisable to calculate the quantitative unit risk
estimates utilizing that cohort only; and what would the effect be if the Ott et al. (1978) and
Wong et al. (1983) epidemiologic studies were included in the calculation of a unit risk estimate?
To answer the first question, the first study (Rinsky et al., 1981) of Pliofilm workers in the
rubber industry covered three facilities in Ohio and consisted of 1,165 male workers who had
been employed between 1940 and 1965 and followed through 1981. The second study (Rinsky et
al., 1987) included an additional 6.5 years of follow-up from the earlier study. It also included
individual estimates of personal exposure, which were not included in the previous study.
Duration of employment and personal exposure estimates during that employment were used to
generate risk estimates based on grouped data. The updated follow-up made it possible to
evaluate dose-response relationships and estimate risks at low exposure levels in terms of ppm-
years of exposure. One myeloblastic leukemia was subsequently noted after the additional follow-
up. However, because of the compensating increase in expected deaths due to the additional
person-years of follow-up, only a small change occurred in the overall relative risk. Altogether, 9
leukemias were observed versus 2.66 expected in this cohort by December 31, 1981 (Rinsky et
al., 1987). The relative risks were found to increase with cumulative exposure as shown in Table
1.
Cumulative exposure expressed in terms of ambient respirable benzene multiplied by
length of exposure (parts per million times years exposed) is a variable that has been used by
researchers to measure individual dose in most epidemiologic studies. However, some recent
limited epidemiologic evidence supports certain alternative dose measures as perhaps better than
the measurement of cumulative dose. These alternative measures may or may not more closely

Table 1. Relative risk as a function of cumulative exposure

Cumulative exposure
(ppm-years) Relative risk
0-40 1.1
40-200 3.2
200-400 11.9
More than 400 66.4

6
approximate the actual dose that induces leukemia. It is uncertain whether it is benzene or an
active metabolite of benzene (or some combination of metabolites with or without benzene),
either delivered to the hematopoietic system or formed within the stem cells, that is responsible
for leukemogenesis.
Hayes et al. (1997) suggested that recent exposure to benzene may be more relevant to
the risk of ANLL/MDS (myelodysplasic syndromes) than distant past exposures, although the
earliest exposure to benzene may have been the most intense. However, measurements of these
earliest exposures are almost nonexistent. Recent epidemiologic studies of petroleum workers
suggest that there are few or no apparent effects to workers exposed to exceedingly low levels of
benzene for long periods of time (Raabe and Wong, 1996). However, some possible effects can
be detected when these data are subjected to alternative dose metrics. In a case control study,
Rushton and Romaniuk (1997) examined 91 cases of leukemia and matched controls (4 to 1) on
the basis of age and controls alive at the time of case occurrence. These authors applied a number
of dose metrics to determine which was the most appropriate for determining the risk of specific
leukemia types in workers exposed to benzene. At low levels of exposure (generally 80% had
cumulative exposure of less than 5 ppm-yrs.), no elevated risks could be detected. The authors
found no relationship to any of several alternative exposure scenarios for acute
lymphocytic leukemia (ALL) and chronic myeloid leukemia (CML). However, for CLL, there
was a nonsignificant increase in risk, which tended to increase with duration of employment for
white-collar employees who had only background exposure to benzene. On the other hand, for
acute myeloid and monocytic leukemia (AMML), a significantly increased risk tended to be
associated with “peaked” exposures to benzene but not with cumulative exposure, maximum
intensity of exposure (which was not all that intense; levels for the most part never exceeded 0.4
ppm, time-weighted average, 8-h day), or mean intensity of exposure if treated as a continuous
variable. This could be important when using the dose metric, intensity times time (ppm-years).
This analysis suggests that a person exposed to 40 ppm of benzene for 1 year might be at a
greater risk of leukemia than a person exposed to 1 ppm for 40 years.
Utilizing a nested case-control design, Schnatter et al. (1996a) conducted a similarly
designed study of Canadian petroleum distribution workers exposed to benzene. Fourteen
leukemia cases were matched 4 to 1 on birth date and time at risk with members from the same
cohort. Average benzene concentrations ranged from 0.01 to 6.2 ppm during the history of the
plant. The authors reported that at those exposure levels no significant increase in the risk of
leukemia was detected utilizing either cumulative exposure or other dose metrics. The authors
used several dose metrics, i.e., intensity, duration of exposure, family history of cancer, cigarette
smoking, and years of exposure at certain levels of intensity (+0.5 ppm or +1.0 ppm). Only family
history and whether the victim smoked cigarettes produced an elevated but still nonsignificant risk

7
of leukemia. The study, however, had little power to detect an elevated risk of leukemia because
it was based on only 14 cases.
As was pointed out by Raabe and Wong (1996), few employees in the petroleum industry
reported large amounts of exposure to benzene measured in terms of ppm-years, although many
employees worked in the industry for long periods of time. In their meta-analysis of 208,000
workers combined from 19 different cohorts, only a small fraction of the workers accumulated as
much as 200 ppm-yrs., the level the authors suggested was the “threshold” for carcinogenesis. In
fact, Raabe and Wong (1996) reported that the average benzene level was only 0.22 ppm based
on 14,824 samples from industrial hygiene surveys.
It has been suggested also that metabolic saturation may have taken place in the early
years of the Pliofilm cohort, if the Paustenbach et al. (1993) exposure estimates are considered.
But based upon a discussion of their estimates by Utterback and Rinsky (1995), it is unlikely that
such high levels of exposure could have occurred to these Pliofilm workers. Thus, adjusting for
metabolic saturation may be an irrelevant issue. Furthermore, since the exact mechanism(s) of
leukemogenesis in benzene-exposed workers is not known, an adjustment to the exposure metric
is not possible at this time.
Also, almost all of the 9 leukemia victims in the Pliofilm cohort received long intervals of
exposure and 7 experienced latent periods of 15 years or longer. One subject succumbed to AML
in 1954 after only 3-1/2 years of exposure. The shortest latent period is that of a Pliofilm worker
with CML who died in 1950 after 2 years’ exposure. The next earliest death from leukemia was
in 1957 after a latent period of 15 years. There was little evidence of “recent” exposure being
responsible for the leukemia deaths seen in the Rinsky cohort. Suppression of the hematopoietic
system from exposure to benzene, which was suggested by the rising blood count data in the early
years (1940-1946) of the Pliofilm cohort, appears not to have facilitated the diagnosis of leukemia
at an early stage in such workers, based upon latent factors. Furthermore, after 1946 some
measurements were available that made it possible to calculate rough estimates of personal
cumulative exposure for each member of the cohort. These estimates tended to be similar among
different investigators. However, Rinsky et al. (1981, 1987), Crump and Allen (1984), and
Paustenbach et al. (1992, 1993) employed various assumptions to estimate personal exposure
levels before 1950, when exposures were most intense. The estimates of exposure made by
Rinsky et al. (1981, 1987) were generally the lowest of the three sets, thus giving rise to the
highest cancer unit risk estimates.
Paustenbach et al. (1992, 1993) used a variety of assumptions to derive the highest
estimates of personal exposure of any of the investigators. They cited seven factors that
influenced their estimates: (1) inaccuracy of devices used for monitoring airborne concentrations
of benzene; (2) length of the work week; (3) rubber shortages during World War II; (4) installa-

8
tion of local exhaust systems to reduce airborne concentrations of benzene; (5) additional
exposure to benzene by skin contact; (6) ineffectiveness of respiratory devices; and (7) medical
evidence of overexposure of workers to benzene. The authors concluded that these pliofilm
workers were exposed to the highest levels during the early years of exposure.
These estimates, however, were severely criticized by Utterback and Rinsky (1995), who
observed that the Paustenbach et al. (1992) exposure estimates were based upon worst-case
assumptions concerning actual exposure levels that may have existed during the earlier years of
the rubber hydrochloride (RH) cohort. Utterback and Rinsky (1995) claimed that Paustenbach et
al. used selected information, improperly cited, to inflate estimates of exposure and produce risk
estimates that were off by an order of magnitude. Perhaps one of the more important examples of
this was the conclusion of Paustenbach et al. (1992) that detector tube readings underestimated
benzene concentrations in RH plants. They cited the “validity studies” of Hay (1961) to support
this position. To compensate for this supposed inaccuracy in exposure estimates made with the
instrument, Paustenbach et al. apparently inflated the readings by about 50% and used these
inflated estimates as the basis for benzene exposure received by members of the RH cohort.
However, Utterback and Rinsky (1995) pointed out that Hay (1961) indicated in his paper that
there were sufficient problems in determining the accuracy of these detector tube kits that the
conclusions presented in that paper (Hay, 1961) might not be correct. Hay (1961) additionally
stated that “the results of this study show all of these (detector tube) kits to be unsatisfactory in
terms of the quantitative results obtained. Their value seems to be limited to semiquantitative
indications of benzene concentrations in environments to which the tubes have been calibrated.”
Utterback and Rinsky (1995) further noted that prolonged exposure to the elevated levels of
benzene postulated by Paustenbach et al., 100-200 ppm, for as long as a decade would have
produced an “epidemic of serious benzene poisoning” that would have resulted in much sickness
and many more deaths than actually seen in the RH cohort.
Rinsky et al. (1981, 1987), on the other hand, after analyzing data from various sources
(Industrial Commission of Ohio in 1946 and 1955, Ohio Department of Health in 1956, the
University of North Carolina in 1974, NIOSH in 1976, and company surveys from 1946 to 1950
and 1963 to 1976), assumed that the levels of benzene, as measured by the 8-h time-weighted
average (TWA) exposure of the workers, were close to the recommended standards for specific
years as follows: 100 ppm (1941), 50 ppm 8-h TWA (1947), 35 ppm 8-h TWA (1948), 25 ppm
8-h TWA (1957 and 1963), and 10 ppm TWA (1969). This analysis produced the lowest set of
estimates.
Crump and Allen (1984) developed a third set of exposure estimates based on the concept
that benzene levels declined progressively as more restrictive standards were implemented in the
workplace. Their estimates lie somewhere between those of Rinsky et al. (1981, 1987) and of

9
Paustenbach et al. (1993). These estimates were used in deriving the quantitative unit risk
estimates in EPA’s Interim Quantitative Cancer Unit Risk Estimates Due to Inhalation of Benzene
(U.S. EPA, 1985). Even with the differences in the these three sets of estimates of exposure for
employees, the cumulative standardized mortality ratios (SMRs) differed from the Crump and
Allen estimates by no more than a factor of 2 (Table 2).
During the period 1940 through 1975, Kipen et al. (1988) studied over 17,000 peripheral
blood counts that were collected from 459 benzene-exposed Pliofilm workers from the same
company studied by Rinsky et al. (1987). The most important observation from this study (as
evidenced by WBCs and RBCs) was that the mean blood counts were somewhat lower in the
early 1940s and gradually rose until around 1948, when mean levels attained the levels of the
most recent measurements. The authors attributed this phenomenon to a reduction in the ambient
airborne exposure levels of benzene received in the workplace. These authors assumed that the
mean estimated exposure to workers in the plants from 1940 to 1948 was 75 ppm, based upon the
Crump and Allen estimates of exposure (Crump and Allen, 1984). During the period from 1940
to 1948, Crump and Allen assumed the estimated exposure was declining from a mean of 137
ppm in 1940 to a mean of 32 ppm in 1948. From 1948 to 1975, the average estimated mean
exposure was 15 to 20 ppm.
This assumed reduction in estimated mean levels of benzene in the air, from a high in 1940
to a low in 1948, based upon the educated but entirely subjective guesswork of Crump and Allen,
was paired with the mean blood count data from 1940 to 1948 by year. Significant inverse
correlations of the estimated exposure levels of benzene with WBC and RBC measurements were
produced over that period. This calculation tends to support the theory that elevated levels of
exposure to benzene may indeed suppress the body’s ability to produce red and white blood cells
if the increase is not temporal or if it is not due to changes or improvements in diagnostic
procedures.

Table 2. Standardized mortality ratios for deaths from leukemia among

Pliofilm workers based on the estimated cumulative exposure of the selected
investigators

Investigators 0-5 ppm-yrs 5-50 ppm-yrs 50-500 ppm-yrs >500 ppm-yrs

Rinsky et al., 1981, 1987 2.0 2.3 6.9 20

Crump and Allen, 1984 0.9 3.2 4.9 10.3

Paustenbach et al., 1993 1.3 1.8 2.8 11.9

Source: Adapted from Paxton, 1996.

10
 However, a significantly negative correlation does not necessarily support any
assumptions about the actual levels in those early days. This point is made clearly by Kipen et al.
(1989). Their purely mathematical calculation provides no information on what the early levels
actually were. A similar association could also be achieved if one assumes lower levels of
benzene in 1940, as was done by Rinsky et al. (1981, 1987), that fall to even lower levels by 1948
(although still above the levels seen after 1948). These calculations suggest only that, at some
elevated level of exposure to benzene, the body’s ability to produce red and white blood cells is
compromised.
In a later study of the same blood count data, Ward et al. (1996) came to a similar
conclusion through the use of the Rinsky et al. (1987) exposure estimates, which they also
assumed to be elevated during the early years of exposure, although not by as much as the Crump
and Allen (1984) exposure estimates. Both groups of researchers agree that elevations in
exposure to benzene tend to suppress blood counts. They do not, however, agree on the exact
level at which this happens. Both find support for their choice of estimates of exposure during the
period 1940 to 1948. Differences in these estimates result mainly from the choice of methodology
utilized. As a result, the blood count data cannot be used to determine which set of exposure
estimates is more appropriate in determining the unit risk.
With the proportional hazards dose-response model, as used by Paxton et al. (1992) and
Paxton (1996), the estimated relative risks differed by no more than a factor of 4 from the Crump
and Allen (1984) estimates within each cumulative dose-response category (Table 3). Hence, the
use of Rinsky et al. (1981, 1987) or Paustenbach et al. (1993) exposure estimates would have
little effect on the quantitative risk estimate.
More recently, Schnatter et al. (1996b) provided a new exposure analysis of the Pliofilm
cohort that used the median of the sets of exposure estimates described above to develop a new
set of indices of exposure per person. This technique differed from the standard method of

Table 3. Estimated relative risks of leukemia derived by the proportional

hazards dose-response model according to the estimated cumulative exposure
(ppm-years) of the selected investigators

Investigators 4.5 ppm-yrs 45 ppm-yrs 90 ppm-yrs 450 ppm-yrs

Rinsky et al., 1981, 1987 1.02 1.19 1.41 5.5
Crump and Allen, 1984 1.00 1.04 1.07 1.43
Paustenbach et al., 1993 1.01 1.07 1.14 1.96

Source: Adapted from Paxton et al., 1992.

11
measuring total exposure to benzene (i.e., cumulative exposure = length of exposure ×
concentration) in that an “average” total concentration per person was determined from the job
category with the greatest exposure (maximally) and of longest duration. This method made it
possible to isolate subgroups with less exposure to specified concentrations of benzene and then
to calculate the risk of leukemia in those subgroups. In theory, these subgroups were unlikely to
be exposed to concentrations greater than a specified concentration.
The results of the Schnatter et al. (1996b) analysis indicated that for the lowest exposure
estimates (Rinsky et al., 1981, 1987), the “critical” concentration is between 20 and 25 ppm for
the risk of acute myelogenous leukemia (AML) “to be expressed,” and for the median, the risk is
between 50 and 60 ppm, although there appeared to be little power to detect a significant effect in
these risk estimates. Interestingly, for total leukemia, the “critical” concentrations for the median
were lower and appeared to fall in the range of 35 to 40 ppm, and the risk estimates appeared
somewhat less erratic.
These risk estimates are not inconsistent with estimates of Wong (1995), who utilized
cumulative exposure to estimate the risk of AML in Pliofilm workers. The Schnatter et al. (1996)
analysis suffers from the same problems as the Wong (1995) and the Rinsky et al. (1987) studies
of Pliofilm workers: the data lack sensitivity to detect the effects of benzene exposure at low
levels. To assume from this data set that a critical concentration exists at the levels indicated and
from which a “threshold” could be inferred is unwarranted. In fact, the lower estimates of the
critical concentration based on the sum total of leukemia deaths versus just those deaths from
AML seem to suggest that there might be a lower critical region for AML, if a larger data set
were available.
The National Cancer Institute, in cooperation with the Chinese Academy of Preventive
Medicine, has been conducting a comprehensive study of 74,828 benzene-exposed workers
employed from 1972 to 1987 in 672 factories in 12 cities of China (Dosemeci et al., 1994; Hayes
et al., 1996, 1997; Yin et al., 1987, 1989, 1994, 1996). A comparison group of workers
consisting of 35,805 employees was assembled from non-benzene-exposed units of 69 of the
above factories and 40 factories elsewhere. A variety of job categories were studied in the
painting, printing, footwear, rubber, and chemical industries. Workers in both groups were
followed for an average of slightly less than 12 years. Less than 0.3% were lost to a follow-up in
both the exposed and the unexposed group. Work histories were utilized to link benzene-
exposure data to individual time-specific estimates for each worker (Dosemeci et al., 1994).
This study, one of the largest of its type ever undertaken, enabled its authors to claim
detection of significantly elevated risks at extraordinarily low levels of exposure. Their findings
suggested that workers exposed to benzene at average levels of less than 10 ppm are subject to a
higher risk of hematologic neoplasms (RR = 2.2, 95% C.I .= 1.1-4.2). A combination of ANLL

12
and MDS produced a relative risk of 3.2 (95% C. I. = 1.0-10.1). For exposure to a sustained
concentration of 25 ppm benzene, the risk of ANLL and MDS increased to 7.1 (95% C.I. = 2.1-
23.7). These risks were associated with more recent exposure to benzene (less than 10 years).
The risk of other leukemias (other than ANLL), including chronic myeloid and monocytic
leukemia, was also elevated (RR = 2.0), although not significantly so. Additionally, the risk of
non-Hodgkin’s lymphoma was significantly elevated (RR = 4.2 with 95% C.I. = 1.1-15.9) for
those with a sustained exposure to benzene that occurred 10 years prior to diagnosis. The
authors concluded that benzene exposure “is associated with a spectrum of hematologic
neoplasms and related disorders in humans and that risks for these conditions are elevated at
average benzene-exposure levels of less than 10 ppm.”
Although most of the attributes considered important in a long-term retrospective study of
this nature and magnitude were addressed by the authors, some uncertainties and potential
weaknesses still remain. The derivation of the cohort from many different factories across China
suggested the possibility that this cohort was exposed to mixtures of many different chemicals.
Exposure to other carcinogens in the workplace could have produced confounding effects,
especially if exposures were to chemicals that increase the risk of leukemia. Although not
specifically stated, concurrent exposures to many other chemicals, some hazardous, must have
occurred because benzene was used as a solvent for paints, varnishes, glues, coatings, and other
products (Dosemeci et al., 1994) that were part of the occupational environment for this cohort.
These products contain a myriad of chemicals, of which some undoubtedly were carcinogens.
The authors claim that since there was little movement of the workforce from one job to
another (i.e., 1.4 jobs held during an average work history), little confounding could have taken
place. In the absence of information on exposure to other chemicals, the conclusions of the
authors are questionable. It would seem nonmobility among the workforce would ensure that
longer exposure to a few confounders could occur, rather than shorter exposures to many
confounders. To reduce the uncertainties of confounding under these circumstances, it would be
necessary to eliminate from the cohort all jobs where exposure to other hazardous substances
occurred. This would be helpful in determining more precisely the impact on the risk from any
one confounder, since multiple confounders would not then be present.
The second potential problem was the development of exposure estimates for the 74,828
benzene-exposed workers and the 35,805 unexposed workers. According to the authors
(Dosemeci et al., 1994), only 38% of the 18,435 exposure estimates were based upon actual
measurements of benzene concentrations; the remainder were numbers generated by factory
industrial hygienists based upon their estimates of benzene concentrations. During the earliest
period, only 3% of the exposure estimates were based on actual measurements. Accuracy and
precision of these subjective estimates was unknown. As a consequence, derivations of dose per

13
individual may have been subject to random error and to bias. Such bias, if present, might have
contributed to the supralinearity at higher benzene concentrations evident in the results of this
study.
It is clear that this study contains thought-provoking new findings and conclusions. These
include derivation of stable estimates of risk at lower exposure levels owing to the massive size of
the cohort, the suggestion that non-Hodgkin’s lymphoma may be a consequence of exposure to
benzene, the possibility that elevated risk of non-AML (which presumably includes CML) also
occurs as a consequence of exposure to benzene, and the finding of some evidence for
supralinearity in the mode of action, since the risk ratios based on these data are significantly
elevated at 10 ppm and tend to plateau as the dose increases to somewhat higher levels. It is,
however, premature to assume that the Chinese data (Hayes et al., 1997) should now replace
Rinsky’s Pliofilm workers cohort in the derivation of risk estimates. Further work must be
accomplished to identify the other potentially hazardous chemicals in addition to benzene to which
the Chinese workers were exposed. In addition, it is important to analyze lifestyle and
socioeconomic factors impacting this cohort and to determine how they differ from those of
workers in similar occupational settings in Western countries.
To answer the question posed earlier (p. 6), the net result of discarding data from the Ott
et al. (1978) and the Wong et al. (1983) studies would have little effect on the unit risk estimate.
The Ott et al. (1978) cohort and its later update (Bond et al., 1986a) rely on a smaller data set.
Both the Ott study and its update by Bond have insufficient power to detect a risk of leukemia at
low doses. Furthermore, Bond et al. (1986a) also stated that their data for risk assessment
purposes should not be used for determining unit risk estimates, for several reasons (i.e., small
number of events, competing exposures to other potentially hazardous materials, and the
uncertain contribution of unquantified brief exposures).
Even though the unpublished Wong et al. (1983) cohort has ample power to detect a risk
of leukemia, and the published version (Wong, 1987) includes estimates of personal exposure to
benzene, the estimates apparently were not reliable. Wong (1987) stated that the estimated
historical industrial hygiene data were not precise enough for absolute quantitative risk
assessment. The Rinsky et al. (1981, 1987) cohort has ample power, latency, and better estimates
of later exposure to airborne benzene. However, during certain time frames (i.e., levels of
ambient air benzene before 1950), the actual airborne measurements of benzene in the workplace
were either meager or nonexistent.
In the 1985 interim benzene document (U.S. EPA, 1985), a single overall unit risk
estimate was calculated as the geometric mean of four maximum likelihood unit risk estimates
generated from the Ott et al. (1978) and Rinsky et al. (1981, 1987) studies using both absolute
and relative risk models, and then “correcting” this mean by multiplying it by the ratio of the

14
largest unit risk estimate from the four separate unit risk numbers above to the unit risk estimate
calculated from the Wong (1987) cohort. The result was a probability of 2.6 × 10-2, which is
close to that calculated by Crump (1992) assuming a similar approach, that is, a linear model and
Crump and Allen (1984) exposure estimates but excluding Ott at al. (1978), Bond et al. (1986a),
and Wong (1987). These risk estimates range from 1.1 × 10-2 to 2.5 × 10-2 and can be found in
Section 3.1 (Table 4). By inspection, the inclusion of data from Ott et al. (1978), Bond et al.
(1986a), or Wong (1987) changes these unit risk estimates little.
It is apparent that the calculation of a new unit risk estimate based on a reordering of the
assumptions about what the earlier distribution of ambient air measurements of benzene might
have been, or from the elimination of data sets that add little to the knowledge of risk at low
doses, has questionable validity. Such a recalculation likely would result in little change from
EPA’s 1985 interim quantitative unit risk estimate, based on the current epidemiologic data and
associated uncertainties.

2.2. LABORATORY ANIMAL DATA

Studies on the carcinogenicity of benzene in rodents include inhalation exposures to
Sprague-Dawley rats, C57BL/6 mice, AKR mice, CD-1 mice, and CBA mice, and gavage
treatment of Sprague-Dawley rats, Wistar rats, F344 rats, RF/J mice, Swiss mice, and B6C3F1
mice (Cronkite et al., 1989; Goldstein et al., 1982; Huff et al., 1989; Maltoni et al., 1983, 1988;
NTP, 1986; Snyder et al., 1980, 1982, 1984; Farris et al., 1993). Inhalation concentrations
ranged from 0 to 1,000 ppm and gavage doses ranged from 0 to 200 mg/kg. Upon exposure via
inhalation, benzene was found to be carcinogenic in rats and mice in multiple target organs,
including oral and nasal cavities, liver, forestomach, preputial gland, lung, ovary, and mammary
gland. It is noted that in humans the cancer induced by benzene exposure is predominantly acute
nonlymphocytic leukemia, while in rodents lymphocytic leukemia was observed in two series of
experiments in C57BL/6 mice (Snyder et al., 1980) and CBA/Ca mice (Cronkite et al., 1989).
Although the reason for the difference in lineage of hematopoietic cancers induced in mice
and humans is not fully understood, it may be related to differences in hematopoiesis.
Lymphocytes make up a larger portion of the nucleated cells in mouse bone marrow than in
human bone marrow (Parmley, 1988) and could simply represent a larger target cell population
for benzene metabolites. The target organs for benzene carcinogenicity in rodents are rich in
enzymes that may confer tissue sensitivity to benzene, as is human bone marrow (Low et al.,
1989, 1995). The bone marrow, Zymbal gland, and Harderian gland all contain peroxidases,
which can activate phenols to toxic quinones and free radicals. Sulfatases, which remove
conjugated sulfate and thus reform free phenols, are also present at high levels in these target
organs. The selective distribution of these two types of enzymes in the body may explain the

15
accumulation of free phenol, hydroquinone, and catechol in the bone marrow and the target organ
toxicity of benzene in humans and animals. Therefore, the animal bioassay results may have some
relevance to human leukemia, but it should be emphasized that there is no demonstrated and
reproducible animal model for leukemia resulting from benzene exposure. Thus, the mechanism
of leukemia development following benzene exposure is not well understood.

2.3. MODE-OF-ACTION INFORMATION

The mechanisms by which exposure to benzene and its metabolites exerts its toxic and
carcinogenic effects remain elusive. Animal studies showed that benzene itself was unlikely to be
the actual toxicant but rather required metabolism to exert its toxic effects (Andrew et al., 1997;
Gad-El Karim et al., 1986; Sammett et al., 1979; Sawahata et al., 1985). The central issues in
integrating the mechanistic data from the laboratory animal experiments with the occupational
epidemiologic data to estimate risk are establishing whether the mechanisms in laboratory animals
are similar to mechanisms in humans and accounting for the dose dependency of those
mechanisms. That is, understanding the mode of action permits rational extrapolation across
species and from high to low doses. Characterization of dosimetry, i.e., description of the uptake,
internal disposition, and translation of an exposure concentration to the effective dose at the
target site, is necessary. Processes such as altered gene regulation, cytotoxicity, and cell
proliferation are thought to be important for benzene leukemogenesis.

2.3.1. Metabolism
The primary pathways involved in benzene metabolism are shown in Figure 1; the majority
of these metabolites were isolated from in vivo studies. Phase II conjugation pathways (not
shown in Figure 1) are also extremely important in benzene metabolism and toxicity: extensive
glucuronidation and sulfation of phenols were reported in the original studies of Parke and
Williams (1953) and have been demonstrated in human systems (Seaton et al., 1995). Because
many reactive metabolites may be formed during benzene metabolism, it is difficult to elucidate
the pathway responsible for benzene toxicity. Indeed, it is certainly a possibility that more than
one metabolic pathway is responsible and that the mechanisms underlying benzene toxicity may be
multifactorial.
Two pathways have been commonly postulated as being responsible for benzene toxicity.
The first involves the major hepatic metabolites of benzene (phenol, catechol, and hydroquinone),
whereas the second involves ring-opened forms of benzene (Figure 1). Cytochrome P450 2E1
has been shown to metabolize benzene to phenol (Johansson and Ingleman-Sundberg, 1988). The
majority of data suggest that the hepatic conversion of benzene to phenolic metabolites is an
important primary event in benzene toxicity (Smith et al., 1989).

16
Figure 1. Major metabolic pathways in benzene toxicity.

Source: Ross, 1996.

17
Catechol and hydroquinone have been shown to persist in bone marrow after benzene exposure
(Rickert et al., 1979). The bone marrow is rich in peroxidase activity (Bainton et al., 1971), and
phenolic metabolites of benzene can be activated by peroxidases to reactive quinone derivatives
(Subrahmanyam et al., 1991). Peroxidase activation of hydroquinone is known to result in
covalent binding to protein (Subrahmanyam et al., 1989; Ganousis et al., 1992) and the formation
of DNA adducts, as detected by 32P-post-labeling (Lévay et al., 1993), in both murine and human
bone marrow in vitro. Peroxidases reported to be present in bone marrow include
myeloperoxidase (MPO) (Bainton et al., 1971), prostaglandin synthase (Gaido and Weirda,
1987), and eosinophil peroxidase; of these, MPO is known to be present in bone marrow in high
concentrations, and such a peroxidase can readily bioactivate benzene-derived phenolics to
reactive quinones (Figure 1) in situ in bone marrow (Bhat et al., 1988).
The major problem with the phenolic hypothesis, however, is that phenol does not
reproduce the myelotoxicity associated with benzene (Tunek et al., 1981). Kenyon and Medinsky
(1995), have suggested the inability of phenol to induce myelotoxicity may be related to its
preferential conjugation in the periportal region of the liver, whereas benzene is metabolized to
phenol and further to hydroquinone in the pericentral region. Preferential removal of phenol by
conjugation in the periportal region of the liver could explain why more hydroquinone is produced
from benzene than from phenol. An interaction of phenol and hydroquinone has been reported to
reproduce the myelotoxicity of benzene (Eastmond et al., 1987), and increased covalent binding
of 14C-hydroquinone has been observed in bone marrow when a combination of both compounds
was administered to mice (Subrahmanyam et al., 1990). Catechol also has been reported to
markedly stimulate peroxidatic bioactivation of hydroquinone in murine stroma (Ganousis et al.,
1992), and a synergistic genotoxic effect of catechol and hydroquinone has been observed in
human lymphocytes (Robertson et al., 1991). These data support the hypothesis that an
interaction of phenolic metabolites may induce or at least contribute to benzene toxicity.
The second metabolic mechanism commonly proposed to explain benzene toxicity
involves ring opening of benzene to reactive muconate derivatives. When benzene is given in
vivo, trans, trans-muconic acid can be detected as a urinary metabolite. A precursor of trans,
trans-muconic acid (trans, trans-muconaldehyde [MA]), has been suggested to be the ultimate
toxic species obtained during benzene metabolism (Latriano et al., 1986). The major problem
with this hypothesis is that MA has never been demonstrated as an in vivo metabolite of benzene,
although formation of MA has been reported from benzene in mouse liver microsomes. The
reactivity and rapid further metabolism of MA may preclude its isolation in vivo. The proposed
mechanism of formation of MA has been suggested to involve an iron-catalyzed ring opening of
the benzene epoxide (Figure 1) and appears not to be derived from the dihydrodiol (Zhang et al.,
1993b).

18
 The metabolism of benzene was more rapid and extensive in the mouse than in the rat, and
this observation appears to correlate with the greater sensitivity of the mouse to benzene’s
myelotoxic and genotoxic effects (Henderson et al., 1992; Sabourin et al., 1988). Irons and co-
investigators showed that very little metabolism of benzene occurs in the bone marrow (Irons et
al., 1980; Sawhata et al., 1985). The combination of metabolism in the liver and toxicity in the
bone marrow suggests that one or more relatively stable metabolites formed in the liver is
transported to the bone marrow, where it exerts its toxic effects.
Although much less is known about the metabolism of benzene in humans than in
laboratory animals, existing studies indicate that both metabolize benzene along similar pathways
(Cooper and Snyder, 1988; Inoue et al., 1988a,b; IPCS, 1993; Sabourin et al., 1989). The rates
of some reactions, however, may differ (Henderson et al., 1989; IPCS, 1993; Sabourin et al.,
1989). A recent series of investigations on Chinese workers highly exposed to benzene provides
insights into particular enzymes underlying the hematotoxicity of benzene in humans. Overall,
individuals with the rapid hydroxylator phenotype exhibited a 2.5-fold increased risk of benzene
poisoning as compared to those with the slow hydroxylator phenotype (Rothman et al., 1995). In
the Chinese worker study, DNA was collected to investigate the potential role of genetic
polymorphisms affecting NAD(P)H:P quinone oxidoreductase (NQO1) in the susceptibility of
workers to benzene (Rothman et al., 1996a). NQO1 (also known as DT-diaphorase) is an
enzyme that catalyzes a two-electron reduction of quinones to hydroquinones and, on the basis of
several in vitro and animal experiments, appears to provide protection against quinones formed
during benzene metabolism (Ross et al., 1990; Smart and Zannoni, 1984; Zhu et al., 1995). A
point mutation in the NQO1 gene results in a loss of enzyme activity in homozygous individuals.
Analyses performed on 38 workers and 35 controls showed that individuals homozygous for this
mutation had a 3.2-fold increased risk of myelotoxicity. These results, which were consistent with
the results of previous animal and in vitro studies, identified groups of individuals with increased
susceptibility to benzene’s hematopoietic effects (Seaton et al., 1994; Smart and Zannoni, 1984).
Recent biochemical studies also suggest a role for NQO1 in the greater sensitivity of mice than
rats to the myelotoxic effects of benzene (Zhu et al., 1995).
Recent biochemical studies also suggested a role for NQO1 in the greater sensitivity of
mice than rats to the myelotoxic effects of benzene (Zhu et al., 1995). Metabolic activation by
P450 2E1 and detoxification of benzene-derived quinones appeared to be important steps in
benzene hematotoxicity. Levels of P450 2E1, which are known to vary substantially between
individuals and between ethnic groups and may partially explain the differences in response seen in
different studies (Stephens et al., 1994). Recent studies showed that allelic differences at the
CYP2E1 and NQO1 loci were associated with susceptibility to benzene hematotoxicity (Rothman
et al., 1997). The role of hematotoxicity in benzene-induced leukemia, however, still remains

19
unclear. Furthermore, P450 2E1 is involved in the metabolism of ethanol and is readily inducible,
indicating that alcohol consumption may affect a person’s susceptibility to benzene poisoning
(Koop et al., 1989; Stephens et al., 1994).
There has been a considerable amount of progress in understanding and quantifying the
factors that contribute to the distribution and metabolism of benzene and its metabolites in
experimental animal species (Schlosser et al., 1993, 1995; Medinsky et al., 1994; Low et al.,
1995). The quantity of benzene metabolites produced in different species result from the subtle
interplay of oxidation and conjugation pathways and the distribution of enzyme systems in the
liver and other organs as well as relative rates of perfusion in different organs and different
species. These differences have been explored using a physiologically based pharmacokinetic
model (Schlosser et al., 1995; Medinsky et al., 1996), but their application in predicting
metabolism and dosimetry in humans remains a subject of considerable debate. Recent studies
using genetically engineered animals (transgenic CYP2E1 knockout mice) indicate that CYP2E1
is the primary isozyme responsible for benzene metabolism in vivo and that metabolic activation of
benzene is required for the development of both cytotoxicity and genotoxicity following benzene
exposure. These observations are important because humans vary in their expression of CYP2E1
activity, as well as in their ability to metabolically activate benzene (Lee et al., 1996; Valentine et
al., 1996). Although, there is a scientific consensus that metabolism of benzene is required for
resultant toxicity and carcinogenic response, the role of a metabolite or metabolites of benzene
responsible for these adverse effects is controversial and more research data is needed to better
define sequelae of pathogenesis following exposure to benzene and its metabolites.
In summary, it is generally agreed that benzene toxicity in both experimental animals and
humans results from the biotransformation of the parent compound to reactive species.
Furthermore, current evidence indicates that benzene-induced myelotoxicity and genotoxicity
results from a synergistic combination of phenol with hydroquinone, muconaldehyde, or catechol.
The pathways for benzene oxidative metabolism are generally understood and involve the
cytochrome P-450 family of enzymes (CYP2E1). Phenol, hydroquinone, catechol, and trans,
trans-muconic acid are the major metabolites produced in experimental animals and in humans (in
vitro and in vivo). The present studies further demonstrate that benzene oxidative metabolism
correlates with observed genotoxicity and cytotoxocity in bone marrow, blood, and lymphoid
tissues after benzene exposure. Recent studies also demonstrate the importance and magnitude of
CYP2E1 in benzene metabolism and toxicity and that CYP2E1 expression may play a significant
role in human variability, genetic polymorphism, and resultant differential risk from benzene
exposure.

20
2.3.2. Mutagenicity and Genotoxicity
The literature on the genotoxic effects of benzene is extensive, with more than 220
publications with original data. Reviews of the earlier literature (Dean, 1978, 1985) present clear
evidence that benzene exposure results in chromosome aberrations in a variety of in vitro and in
vivo assays and in persons occupationally exposed to benzene over long periods of time. Benzene
generally has yielded negative results in gene mutation assays in bacteria or in vitro mammalian
cell systems (Ashby et al., 1985; Oberly et al., 1984, 1990). However, Ward et al. (1992)
reported dose-related increases in mutations at the hprt locus in lymphocytes of CD-1 mice
exposed to benzene (40, 100, and 1,000 ppb) by inhalation for 6 weeks (22 h/day, 7 days/week).
Also, Mullin et al. (1995) detected increased mutant frequencies in the lacI transgene from lung
and spleen but not liver from C57BL/6 mice exposed to 300 ppm benzene for 6 h/day, 5
days/week for 12 weeks.
Benzene exposure results in a variety of both structural and numerical chromosome
damage (see cited reviews). Experiments in rodents have provided consistent evidence from a
number of studies that benzene exposure causes increased frequency of micronucleated cells
(summarized in ATSDR, 1997). Micronuclei also are seen in human cells exposed in vitro to
various metabolites and combinations of metabolites (Zhang et al., 1993a; Eastmond, 1993; Yager
et al., 1990; Hogstedt et al., 1991; Robertson et al., 1991). Synergistic increases in micronuclei
were induced by catechol and hydroquinone, but not catechol and phenol or phenol and
hydroquinone (Robertson et al., 1991). However, in mice treated intraperitoneally with binary or
tertiary mixtures of these three metabolites, synergistic effects resulted only from mixtures of
phenol and hydroquinone (Marrazzini et al., 1994); adding catechol to the mixture was no more
effective than hydroquinone alone in inducing micronuclei. Chen and Eastmond (1995a)
corroborated the phenol and hydroquinone synergy. Using an antikinetichore-specific antibody
and fluorescent in situ hybridization (FISH), they demonstrated that both chromosome breakage
and loss were induced and that the relative frequency of these events was indistinguishable
whether mice were treated with benzene (440 mg/kg) or the binary mixture of hydroquinone and
phenol (60/160 mg/kg).
The evidence that human exposure to benzene produces the types of chromosomal
rearrangements associated with AML and MDS, such as interstitial deletions, inversions, or
translocations, continues to accumulate. Earlier studies of patients with benzene-induced
hematopoietic disorders demonstrated increased chromosome aberrations in lymphocytes and
bone marrow cells (Dean, 1985). The rearrangements observed included stable and unstable
aberrations (Aksoy, 1989; Forni, 1971, 1994; Sarto et al., 1984; Sasiadek, 1992; Van den Berghe
et al., 1979). Tompa et al. (1994) performed cytogenetic analyses on groups of workers
occupationally exposed to benzene. Improved working conditions over a 3-year period resulted

21
in reducing average peak concentrations from 68.7 mg/m3 in 1990 to 27.1 mg/m3 in 1991 and 18.4
mg/m3 in 1992. Striking decreases in chromosome aberrations were seen by 1992 in those
workers with less than 10 years’ exposure. Workers exposed for more than 10 years showed
lesser reductions. Recently, Rothman et al. (1995) used the glycophorin A (GPA) gene mutation
assay to examine the type of mutations produced by benzene exposure in human bone marrow
among 24 workers heavily exposed to benzene and 23 matched controls. The assay detects a
spectrum of mutational mechanisms, and the results indicated significant increases in NN-type
variants but not in NO-type variants. The NN-type cells are presumed to be consequences of
mitotic recombination or gene conversion processes, while the NO-type are derived from point
mutations or deletion events.
Aneuploidy, the loss and gain of whole chromosomes, is also found in some cases of
myeloid malignancy. Patients with benzene-induced leukemia, rodents, and human cells treated in
vitro display increased aneuploidy. Numerical changes in the C-group chromosomes 6-12 and X
have been detected in the blood and bone marrow of patients with benzene-induced myelogenous
leukemia, myelodysplasic syndrome, and pancytopenia (Vigliani and Forni, 1976). A recent
report by Zhang et al. (1996) showed that the induction of aneuploidy of chromosome 9 as
measured by FISH in interphase lymphocytes from benzene-exposed workers is significantly
elevated only at high levels of exposure (>31 ppm in air). The human evidence for aneuploidy
induction also is supported by in vitro experiments. Hydroquinone and 1,2,4-benzenetriol induce
aneuploidy of chromosomes 7 and 9 in human cells (Zhang et al., 1994; Eastmond et al., 1994).
Eastmond and co-workers also have reported that micronuclei containing centromeres are formed
in bone marrow and spleen cells following oral benzene exposure in mice (Chen et al., 1994; Chen
and Eastmond, 1995a). Centromere-containing micronuclei are thought to be formed when a
whole chromosome is lost during mitosis. Thus, considerable evidence supports the assertion that
exposure to benzene produces aneuploidy in a variety of systems.
DNA adducts of phenol, hydroquinone, or benzoquinone have been reported in a number
of in vitro systems (Reddy et al., 1990; Lévay et al., 1993; Bodell et al., 1993). Reddy et al.
(1990) did not detect DNA adducts in rat bone marrow, Zymbal gland, liver, or spleen after four
daily gavage treatments of phenol or a 1:1 mixture of phenol and hydroquinone. Subsequently,
the same group (Reddy et al., 1994) did not detect DNA adducts in liver, bone marrow, or
mammary glands of mice sacrificed after receiving four daily intraperitoneal (i.p.) injections of 500
mg/kg benzene. Using the same P1-enhanced P32-postlabeling procedure, Pathak et al. (1995)
performed a series of in vivo experiments using concentrations ranging from 25 to 880 mg/kg and
a variety of injection schedules for periods up to 14 days, as well as in vitro experiments with
hydroquinone or 1,2,4-benzenetriol. One major and two minor DNA adducts were detected in
the bone marrow of mice receiving i.p. injections of 440 mg/kg of benzene twice a day for 3 days.

22
No adducts were seen with any treatment regimen involving only a single injection per day, even
at 880 mg/kg for 3 days. Co-chromatography indicated that the adducts were identical to those
seen after in vitro treatment of bone marrow with hydroquinone. Using the same treatment
regimen, the same adducts were detected in white blood cells of mice (Lévay et al., 1996).
The single-cell gel electrophoresis assay (comet assay) is a rapid and simple method for
detecting single-strand breaks in DNA induced by oxidative damage as well as topoisomerase II
inhibitors (McKelvey-Martin et al., 1993). Plappert et al. (1994) exposed mice to 100, 300, and
900 ppm of benzene 6 h/day, 5 days/week for 4 weeks with samples taken at 3 days and weekly
thereafter. DNA damage measured as increased tail moment was seen in liver and bone marrow
after 5 days at 100 ppm. Increased damage in peripheral blood was not recorded until 4 weeks at
100 ppm. Maximal damage was seen after 5 days at 300 ppm, with little increased or decreased
damage at longer exposures. Damage returned to or approached control levels if animals were
allowed 24 or 48 h recovery time after the cessation of benzene exposure. Tuo et al. (1996)
treated mice by gavage with benzene at 40, 200, and 450 mg/kg and detected dose-related
increases in DNA tail length in both peripheral lymphocytes and bone marrow. Pretreatment with
the CYP2E1 inhibitor propylene glycol reduced the damage by almost half at all concentrations.
The comet assay was used to detect damage induced by benzene and several metabolites in
cultured human lymphocytes (Anderson et al., 1995). Increased tail length was seen after 0.5 h
treatment with benzenetriol and catechol, after 1 h with benzene, and after 2 h with muconic acid,
hydroquinone, and benzoquinone. Recently, Andreoli et al. (1997) reported increased DNA
damage in gasoline station attendants monitored for 1 year with breathing zone air samplings.
Using tail moment as the measure of DNA damage, the mean for exposed workers was twice that
of controls matched for age and smoking habits. They also exposed peripheral lymphocytes from
unexposed donors to hydroquinone, benzenetriol, and benzoquinone, and positive responses were
recorded with all three. Hydroquinone-exposed cells stimulated to divide with PHA were more
than 10-fold less sensitive than resting cells. Further, the DNA repair inhibitor ara-C dramatically
increased the damage in hydroquinone-treated cells.

2.3.3. Epigenetic Effects

While the evidence is unambiguous that chromosome rearrangements are detected in
leukemia and are observed following benzene exposure, it is also evident that such effects are only
part of the complex process of leukemogenesis. Normal hematopoesis is a complex process in
which differentiation and proliferation are coordinately linked, and several investigations have
demonstrated that exposure to benzene and several metabolites adversely affects this process.
Benzene has long been recognized as a hematotoxicant, causing bone marrow suppression and
aplastic anemia (Goldstein, 1988; Aksoy, 1988; Kipen et al., 1988) In humans, lymphocytes

23
appear to be the peripheral blood cell population most sensitive to toxicity (Aksoy et al., 1971;
Moszczynski and Lisiewicz, 1982). A recent investigation of Chinese workers reported a highly
significant difference in absolute lymphocyte count between exposed and unexposed workers
(Rothman et al., 1996b). Decreased lymphocyte counts correlated with increased levels of urinary
benzene metabolites. Lymphocytes are the source of many regulatory cytokines, such as IL-3;
depression of marrow lymphocyte populations might disrupt normal production of these
cytokines. An abnormal regulatory environment could then contribute to bone marrow
suppression or leukemogenesis. In addition to lymphocytopenia, erythrocytic populations are
suppressed in cases of chronic benzene exposure. Peripheral granulocytic cells may be suppressed
or increased in number. These effects on peripheral populations are, in general, borne out in
experiments on rodents, although exposures and exposure rates are difficult to compare to those
experienced by humans.
There has been debate about whether hematotoxicity, evidenced as depressed white blood
cell counts, is a required precursor to benzene-induced leukemia. There have been reports of
benzene-induced leukemia without proof of prior bone marrow suppression (Yin et al., 1989), and
myeloid malignancy does not always arise in a setting of clinically detectable bone marrow
suppression. However, hematotoxicity is probably involved in a large percentage of cases and
probably increases the risk of malignancy. Extensive cell death in progenitor populations,
combined with abnormal regulatory signals, could assist an abnormal clone to gain bone marrow
dominance. Clonal hematopoiesis is thought to be a step toward MDS and/or AML. In addition,
cell death and cytopenia are likely to promote expansion of remaining stem cells, which increases
the chance that a cell carrying a genetic abnormality will be recruited into cycle. Irons and co-
workers have shown that in clonogenic assays hydroquinone pretreatment increased the number
of colonies formed by recombinant GM-CSF-induced CFU-GM at concentrations as low as 10-9
molar (Irons et al., 1992). This suggests that hydroquinone can increase recruitment of
hematopoietic progenitor cells into the granulocyte-macrophage pathway, or increase the number
of resting cells entering the cell cycle. Since myeloid progenitors are thought to be sensitive to
genotoxicity through expression of myeloperoxidase, both scenarios could result in increased
genetic damage occurring in the bone marrow by increased numbers of MPO-positive cells.
Alternatively, increased recruitment could simply increase the probability of recruiting a
previously genetically altered cell into cycle. Enhanced response to GM-CSF could occur by a
number of different biochemical mechanisms, including alteration of signal transduction pathways.
The phenomenon of increased clonogenicity in the presence of GM-CSF is also observed with
other chemicals that cause myeloid leukemias, suggesting that it may play an important role in
leukemogenesis (Irons and Stillman, 1993).

24
 Kalf and co-workers have reported that benzene toxicity to the bone marrow can be
prevented by IL-1 administration (Renz and Kalf, 1991). They postulate that hydroquinone is
oxidized to benzoquinone, which then inhibits the conversion of pre-IL-1 to the active form in
stromal macrophages (Niculescu et al., 1995). The role of IL-1 in early hematopoiesis is not
entirely clear, but this cytokine is able to stimulate fibroblasts and endothelial cells to secrete other
factors that, in turn, act on hematopoietic progenitors. Stromal macrophages also produce
growth factors that directly regulate myelopoiesis. In long-term culture, developing granulocytes
are found clustered near these cells, presumably because they are receiving either instructive or
permissive signals necessary for their proper development. Therefore, disruption of IL-1
production and/or other stromal effectors of hematopoiesis could act as a selective pressure for
the development of abnormal clones that have acquired the ability to grow under adverse
conditions.

2.3.4. Pathogenesis
Lymphohematopoietic neoplasia can be defined as uncontrolled proliferation or expansion
of lymphohematopoietic cells that no longer have the capacity to differentiate normally to form
mature blood cells. Clones derived from the myeloid lineage are designated as chronic or acute
leukemias. Within these general classes, leukemias represent a heterogeneous group of diseases.
Heterogeneity is apparent even within the group classified as AML. MDS consist of a group of
blood disorders with defects in hematopoietic maturation. They are considered as preleukemic
because a significant portion of these progress to frank leukemia (Wright, 1995). Consistent with
present models for the origin and progression of neoplasia, development of leukemia is thought to
be a multistep process that involves several independent genetic and epigenetic events. Cell
survival, differentiation, and proliferation are regulated processes under coordinated control by
multiple factors in normal hematopoiesis. Particularly challenging to understanding the
pathogenesis is having to discern the role of altered regulation of cell growth as it is superimposed
on more normal hematopoiesis and cell population dynamics involving survival, proliferation, and
differentiation.
Irons and Stillman (1996) have summarized much of the extensive literature relating to
secondary leukemia involving either therapy or occupational exposures. Clonal chromosome
aberrations involving more than 30 different abnormalities have been identified in the majority of
patients diagnosed with AML (Caligiuri et al., 1997). In secondary leukemias associated with
alkylating agent antineoplastic therapy, loss of genetic material from chromosomes 5 and 7 is
found in the great majority, whereas leukemias following topoisomerase II inhibitory drugs more
frequently involve aberrations involving chromosome band 11q23 (Pedersen-Bjergaard et al.,
1995). Several interleukin genes (IL-3, IL-4, IL-5), granulocyte/macrophage-colony-stimulating

25
factor (GM-CSF), and other regulatory genes are tightly linked on chromosome 5. Irons and
Stillman (1996) described a model for benzene-induced leukemia based on the disrupted functions
of these genes. Young and Saha (1996) discuss several different translocations, all involving
11q23. The gene at this location has been sequenced and has been designated MLL (mixed-
lineage leukemia), and while the normal function of this gene has yet to be determined, it shares
homology with the Drosophila trx gene that regulates transcription of genes for normal
development. Although many leukemias have one chromosomal rearrangement in all cells,
cytogenetically unrelated clones are more frequently found in secondary leukemias than in de
novo leukemias (Heim, 1996). Despite these complexities, a growing knowledge of the function
and role of cytokines, their receptors, protooncogenes, and suppressor genes can provide a useful
framework for analysis of the respective roles of altered cell growth and differentiation in
leukemogenesis.

2.3.5. Summary
Figure 2 presents a general schematic of leukemia induced by benzene exposure as
proposed by Smith (1996) that can be instructive to consider in view of the evidence in previous
sections on the mode of action. As discussed in Section 2.3.1, it is generally agreed that the
toxicity of inhaled benzene results from its biotransformation to reactive species. Benzene is
metabolized in the liver by cytochrome P450 2E1 (CYP2E1) to its major metabolites, phenol,
hydroquinone, and catechol. The intermediate benzene oxide can also undergo ring opening to
trans, trans-muconic acid (Figure 1). The selective toxicity of any one of these metabolites to
blood and bone marrow has been difficult to explain, so these metabolites are now viewed as
proximate, with secondary activation to toxic quinones and free radicals as ultimate metabolites
postulated to take place by peroxidase enzymes in the bone marrow. The hypothesis is further
supported by the fact that the target organs for toxicity in rodents are rich in both peroxidase and
sulfatase enzymes, i.e., the selective distribution of these two types of enzymes in the body may
explain the accumulation of free phenol, hydroquinone, and catechol in the bone marrow and
laboratory animal target tissues. Evidence that multiple metabolites are important in benzene
toxicity has also increased in recent years. The possible mechanisms of interaction were discussed
in Section 2.3.1.
Molecular targets for the action of these metabolites, whether acting alone or in concert,
include tubulin, histone proteins, topoisomerase II, and other DNA-associated proteins. Damage
to these proteins would potentially cause DNA strand breakage, mitotic recombination,
chromosomal translocations, and malsegregation of chromosomes to produce aneuploidy.
Evidence to support aspects of each of these exists to some degree, as described in Section 2.3.2.
If these effects took place in stem or early progenitor cells, a leukemic clone with selective

26
Figure 2. Schematic for mechanistic hypothesis of benzene pathogenesis
(leukemognesis).

Source: Smith, 1996.

27
advantage to grow could arise as a result of protooncogene activation, gene fusion, and
suppressor-gene inactivation. Epigenetic effects of benzene metabolites on the bone marrow
stroma, and perhaps the stem cells themselves, could then foster development and survival of a
leukemic clone. Data supporting a role for epigenetic effects were described in Section 2.3.3.
Within the bone marrow are a number of cell populations that can serve as potential
targets, including the hemopoietic and lymphopoietic stem cells, committed progenitors,
immature hemopoietic precursors, mature functional blood cells, and the various cells that
comprise the bone marrow stromal microenvironment. As depicted in Figure 3, the altered
function of different populations could result in the manifestation of different types of toxicities,
e.g., immunotoxicity, aplastic anemia, or leukemia. It is already recognized that persistent
cytopenias and other blood dyscrasias, including dyserythropoiesis, dysgranulopoiesis, and
dysmegakaryopoiesis, frequently precede the onset of leukemia in patients developing AML
secondary to exposure to benzene or alkylating agents (Irons and Stillman, 1996). Thus, although
the exact mechanisms remain to be further examined and elucidated, as yet unexplored
perturbations within these populations may ultimately prove to be mechanistically related and
provide a unified framework to comprehensive understanding of benzene toxicity, i.e., both
cancer and noncancer endpoints.

2.4. HAZARD CHARACTERIZATION SUMMARY

This document reconfirms that benzene is a known human carcinogen by all routes of
exposure (U.S. EPA, 1979, 1985). This finding is supported by evidence from three different
areas: human epidemiologic studies, animal data, and improvement in understanding of
mechanisms of action, and numerous studies of dermal absorption in humans and animals. Human
epidemiologic studies of highly exposed occupational cohorts have demonstrated unequivocally
that exposure to benzene can cause acute nonlymphocytic leukemia and other blood disorders,
that is, preleukemia and aplastic anemia (Aksoy, 1976, 1977; Aksoy et al., 1974; Infante et al.,
1977; Rinsky et al., 1981, 1987; Vigliani and Saita, 1964; Hunting et al., 1995; IARC, 1982;
ATSDR, 1997). It is also likely that exposure is associated with a higher risk of chronic
lymphocytic leukemia and possibly multiple myeloma (DeCouflé et al., 1983), although the
evidence for the latter has diminished with recent studies (Hayes et al., 1996, 1997). In
experimental animal species, benzene exposure (both inhalation and oral routes) has been found
to cause cancer in multiple target organ sites such as oral and nasal cavities, liver, forestomach,
preputial gland, lung, ovary, and mammary gland (Section 2.2). It is likely that these responses
are due to interactions of the metabolites of benzene (Section 2.3.1). Recent evidence suggests
that there are likely multiple mechanistic pathways leading to cancer, and in particular
leukemogenesis, from exposure to benzene (Section 3.2).

28
 Myeloid
leukemia

Figure 3. Schematic illustrating various options for modulation of bone marrow cell
populations by benzene metabolites that could result in the induction of aplastic anemia,
leukemia, and immunotoxicity.

Source: Trush et al., 1996.

29
 Additionally, changes in blood and bone marrow consistent with hematotoxicity are
recognized in humans and experimental animals. Clinical outcomes observed are leukopenia,
thrombocytopenia, anemia, and aplastic anemia (ATSDR, 1997). Benzene induces peripheral
blood abnormalities and disrupts hematopoiesis at separate compartments of blood cell formation
(i.e., white, platelet, and red) (ATSDR, 1997). Granulocytic and erythropoietic progenitor cells
are significantly depressed. Chromosomal breakage and loss are increased in mice from exposure
to benzene or its metabolites, which consist of a mixture of phenol and hydroquinone (Section
2.3.2).
The metabolic studies summarized herein suggest that in both laboratory animals and
humans, benzene metabolism exhibits dose-dependent behavior, with the proportion of the
metabolites formed changing considerably depending on the dose of benzene administered.
Benzene metabolism also has been reported to be modulated by coexposure or prior exposure to
other organic chemicals (Medinsky et al., 1994).
Benzene affects bone marrow cells in several different ways. Based on our current
understanding, these effects are produced by the interactive effects of multiple metabolites.
Genotoxic effects are a critical component of the leukemogenic properties of benzene. As more
information becomes available about the epigenetic effects of benzene and the role these effects
play in the leukemogenic process in general, it is likely that these will be shown to have an
important role. Evidence supports the hypothesis that more than one toxic effect contributes to
the leukemogenic process, especially because benzene metabolic products may be able to cause
general disruption of protein functions in bone marrow cells. Protein damage is likely to result in
pleiotropic effects, including general toxicity, alteration of growth factor responses, and DNA
damage. Therefore, the overall picture of benzene-induced leukemogenesis is an increased rate of
genetic damage to hematopoietic cells that occurs in the context of disrupted bone marrow
biology. This situation could encourage not only the production of cells with key genetic
changes, but also the selection and expansion of such cells because of the abnormal marrow.
However, data are not sufficient at this time to state precisely which of the various documented
effects, genotoxic or otherwise, are the critical ones for benzene-induced leukemogenicity.

3. DOSE-RESPONSE ASSESSMENT AND CHARACTERIZATION

In the earlier EPA benzene risk assessment document (U.S. EPA, 1985), the lifetime
leukemia risk due to 1 ppm of benzene in air was estimated to be 2.6 × 10-2. This is the geometric
mean of risk estimates that were calculated on the basis of data from one study on Pliofilm
workers (Rinsky et al., 1981) and two studies of chemical workers (Wong et al., 1983; Ott et al.,

30
1978). On the basis of Rinsky et al.’s (1981) data alone, the risk due to 1 ppm of benzene in air
was estimated to be 4.1 × 10-2 when the relative risk model was used, and
1.8 × 10-2 when the additive risk model was used.
Subsequently, several risk assessments on the basis of Rinsky et al.’s (1981) cohort have
become available (Brett et al., 1989; Crump, 1992; Paxton et al., 1994). More than 100
individual risk estimates using varying assumptions and/or models have been presented, with
outcomes ranging more than 6 orders of magnitude at 1 ppb exposure.
Two dose-response models, a relative and an absolute risk model, were used to calculate
benzene risk estimates using epidemiologic data in the 1985 EPA document. In fitting the dose-
response models, person-years of observation are divided into subgroups according to the
benzene dose (ppm-year). Let Oi be the number of leukemia deaths observed in group I, Ei the
expected number of leukemia deaths in the ith group based on the mortality rates in a comparison
population, di the average benzene dose in the ith group, and Yi the number of person-years in the
ith group. The relative risk model is of the form

E(Oi) = aEi(1+bdi)

and the absolute risk model is of the form

E(Oi) = Ei + (a+bdi)Yi

where E(Oi) is the expected number of leukemia deaths in the ith dose group under the respective
model. The parameters a and b are estimated from cohort data under the assumption that the
number of observed leukemia deaths, Oi, is a Poisson random variable with the expected value
given by one of the two models above. The parameter b represents the potential of benzene to
induce leukemia per unit dose (ppm-year). Once an estimate of the parameter b is obtained, it
was translated into a unit risk (i.e., lifetime risk per unit of ambient air exposure in ppm or µg/m 3)
by a straightforward mathematical manipulation that depends on whether the model is absolute or
relative risk.
The unit risk estimate of 2.6E-2 per ppm was based on the report by Crump and Allen
(1984). Because of the lack of information on exact exposure conditions for individual members
of the cohort, cumulative dose (ppm-years) was used by Crump and Allen to construct dose-
response models. Clearly, the use of cumulative dose is less desirable than the use of actual
concentration (ppm). Its impact on risk estimates, however, is difficult to assess without knowing
the exact exposure concentrations for individuals in the cohort.

31
3.1. DESCRIPTION OF DIFFERENT RISK ASSESSMENTS
Differences between these risk estimates largely derive from differences in the
determination of the exposure estimates used in the dose-response modeling. Rinsky et al. (1981,
1987), Crump and Allen (1984), and Paustenbach et al. (1992, 1993) chiefly center on the levels
that existed in the plants where the Pliofilm workers were employed before 1946. Paustenbach et
al. (1992, 1993) assumed that the samples taken after 1946 underestimated actual levels chiefly
because inadequate measuring devices were used; they asserted that these devices consistently
underestimated exposure by as much as 50%. It was further assumed that the working week was
on the average 51 h, not the 40 h usually assumed. Other assumptions are also given to justify
their high exposure estimates (Paustenbach et al., 1992, 1993).
Much controversy exists concerning the levels of benzene that permeated the workplace
during the early employment years of the Pliofilm workers. It has not been established what those
levels were from the late 1930s until 1946. Actual measurements do not exist before 1946, when
most of the Pliofilm workers were employed, including most of the leukemia victims. After 1946
and into the 1960s, few measurements of actual benzene exposure were taken, and in many
instances they were taken in areas where it was known that high levels of benzene would be
found. Rinsky et al. (1981) maintains that the average exposure to the workers were “within the
limits considered permissible at the time of exposure.” Rinsky agrees that peak exposure to high
levels of benzene probably did occur but, unfortunately, there is no information regarding when
these peak exposures occurred and how large they were for individual members of the cohort.
Several leukemia victims were exposed to as much as 40 ppm 8-h TWA during their early years
with the company. It is believed that actual levels were probably within the range of 35 to 100
ppm during those early years. These levels tended to drop in time as efforts to improve air quality
in the plants were implemented.
Both Brett et al. (1989) and Paxton et al. (1994) assumed that rate ratio (RR) is related to
exposure (ppm-year) by RR(d) = exp(b*d), where d is exposure in ppm-year and b is a parameter
to be estimated (the two assessments differ in the way the parameter b was estimated). Only risk
estimates due to occupational exposure (i.e., 8/h day, 5 days/week, 50 weeks/year) were
presented. To calculate the lifetime risk due to continuous exposure of 1 ppm (i.e., d=76 ppm-
years), the parameter b is multiplied by a factor of (24/8)×(7/5)×(52/50). The resultant risks at 1
ppb and 1 ppm are given in Table 4.
Crump (1992, 1994) presented 96 dose-response analyses by considering different factors
such as (1) different disease end points, (2) additive or multiplicative models, (3) linear/nonlinear
exposure-response relationships, (4) two-exposure measurements (Crump and Allen [1984] vs.
exposure estimates by Paustenbach eventually published in Paustenbach et al. [1993]), and (5)
cumulative or weighted exposure measurements. The risk estimates range from 8.6 × 10-11 to

32
 Table 4. Risk estimatesa calculated on the basis of Pliofilm workers by various investigators

Exposure and
Source Risk at 1 ppm Risk at 1 ppb model

U.S. EPA 1.8E-2 1.8E-5 Crump and Allen,

1985a (7.5E-3, 3.4E-2) (7.5E-6, 3.4E-5) additive risk

4.1E-2 4.1E-5 Crump and Allen,

(1.3E-2, 8.8E-2) (1.3E-5, 8.8E-5) relative risk

Brett et al., 4.0E-3 (1.0E-3, 1.2E-2) to 3.6E-6 (9.5E-7, 6.9E-6) to Crump and Allen,
1989 2.5E-2 (2.5E-3, 9.9E-2) 1.1E-5 (2.2E-6, 1.9E-5) conditional logistic

2.2E-1 (1.2E-2, 1.0) to 2.4E-5 (6.9E-6, 4.2E-5) to Rinsky,

8.4E-1 (1.5E-2, 1.0) 3.4E-5 (8.2E-6, 5.9E-5) conditional logistic

Paxton, 2.2E-3 1.9E-6 Crump and Allen,

1992 (3.8E-5, 4.9E-3) (3.7E-8, 3.7E-6) proportional hazard

4.6E-3 3.5E-6 Paustenbach,

(1.3E-3, 9.0E-3) (1.2E-6, 5.8E-6) proportional hazard

1.8E-2 8.9E-6 Rinsky,

(3.0E-3, 5.5E-2) (2.5E-6, 1.5E-5) proportional hazard

Crump, 1.1E-2 (2.2E-3, 2.0E-2) to 1.1E-5 (2.2E-6, 2.0E-5) to Crump and Allen,
1992; 1994 2.5E-2 (6.0E-3, 1.3E-1) 2.5E-5 (6.0E-6, 1.3E-4) linear

5.4E-3 to 2.5E-2 4.5E-6 to 2.6E-5 Crump and Allen,

nonlinear

7.1E-3 (2.0E-3, 1.2E-2) to 7.2E-6 (2.0E-6, 1.2E-5) to Paustenbach,

1.5E-2 (3.8E-3, 2.6E-2) 1.6E-5 (3.8E-6, 2.6E-5) linear

8.6 × 10-5 to 6.5 × 10-3 8.6 × 10-11 to 5.6 × 10-6 Paustenbach,

nonlinear

a
95% confidence intervals (CI) are provided in parentheses if they can be reconstructed from the original report.
For the EPA risk estimates, the same arithmetic operations (e.g., taking geometric means of several risk
estimates) are applied to the derivations of CI as to the point estimates. Therefore, the statistical confidence
statement for these numbers may not be precise.
b
Not available.
c
The author recommended that 95% upper bound be used to derive unit risk because of the instability of the
maximum likelihood estimate (i.e., the linear component was estimated to be 0).

33
2.6 × 10-5 at 1 ppb of benzene air concentration and 8.6 × 10-5 to 2.5 × 10-2 at 1 ppm of benzene
air concentration. The largest deviation from the EPA risk number (U.S. EPA, 1985) was
obtained when a nonlinear model and Paustenbach et al. (1993) exposure estimates were used.
When a linear model was used, risk estimates ranged from 7.1 × 10-3 to 2.5 × 10-2 at 1 ppm,
regardless of which exposure measurements were used. When a linear model and Crump and
Allen (1984) exposure measurements were used, the risk at 1 ppm ranged from 1.1 × 10-2 to
2.5 × 10-2. These are close to the 1985 EPA risk estimates. As previously stated, the use of the
updated Rinsky et al. (1987) cohort would not significantly alter risk estimates if the same
exposure-response model and exposure estimates were used. The single factor that affects the
risk estimate most is the assumption of nonlinearity. If low-dose linearity is assumed,
consideration of other factors (e.g., new exposure estimates) will result in no more than a fivefold
difference from the existing EPA risk number.
A need exists to further support these conclusions based on additional research on
biological mechanisms of benzene-induced hematopoiesis and leukemia rather than on statistical
modeling uncertainties alone.

3.2. SHAPE OF THE DOSE-RESPONSE FUNCTION AT LOW DOSES

Too many questions remain about the mode of action for benzene-induced leukemia for
the shape of the dose-response function to be known with certainty. While much progress has
been made in the past few years and a reasonable hypothesis can be generated for the mechanism
of benzene-induced leukemia, it remains simply a hypothesis. Arguments for and against the
dose-response curve being nonlinear at low doses are presented in summary form in Table 5.
Analysis of the Rinsky et al. (1987) data shows that at doses less than 40 ppm-years, the
SMR for leukemia was 1.1 and is not significantly elevated. This has prompted some
investigators to suggest that benzene has a threshold for leukemia induction of about 40 ppm-
years. However, this analysis of leukemia dose-response is based on only nine cases of leukemia,
limiting its value for dose-response analysis. In addition, only six of these cases were AML.
Further, Rinsky et al. (1987) showed a clearly increased SMR for multiple myeloma at doses
below 40 ppm-years, and in a larger Chinese study, involving more than 30 cases, leukemogenic
effects of benzene were observed at exposures well below 200 ppm-years (Yin et al., 1989).
These observations suggest, as expected, that it is difficult to determine the shape of the dose-
response function based on occupational exposure studies alone.

34
 Table 5. Arguments for and against benzene-induced leukemia linearity at low doses

Pro Con
Micronucleus assay is relatively insensitive and may not Micronucleus induction by benzene and its metabolites in
show effects at low doses. mouse bone marrow and in human cells in vitro is nonlinear.

Induction in aneuploidy of other chromosomes (e.g., 7) The induction of aneuploidy of chromosome 9 is nonlinear and
occurs at lower doses, and effect of benzene on is significant only at high levels of exposure (>31 ppm in air)
hyperdiploidy of chromosomes 7, 8, and 9 shows a (Zhang et al., 1996a).
significant linear trend.

Data obtained using accelerator mass spectrometry shows DNA adduct formation is observed by P32-postlabeling only at
that the formation of DNA adducts in mouse bone high doses.
marrow is linear to very low doses.

Errors during repair may cause point mutations. Oxidative DNA damage may contribute to benzene
genotoxicity (Kolachana et al., 1993) but has a high rate of
repair.

Hematotoxicity may increase risk of malignancy but has Hematotoxicity is required for leukemia induction, and this
not been shown to be a prerequisite. will have a threshold.

There is a high background of exposure to benzene and If aneuploidy is critical, then leukemia induction is likely to
its metabolites. Additional environmental exposure will have a threshold. (Numerous molecules of benzene
simply add to this and be linear. There are also metabolites will be required to disrupt microtubules.)
numerous mechanisms of aneuploidy induction, and
aneuploidy is not the only mechanism of suppressor gene
loss and oncogeny activation.

There is a high background exposure to benzene and its The cells in the bone marrow have numerous defense
metabolites, so additional exposure could escape mechanisms.
defenses.

As indicated previously, benzene is not a classic carcinogen; that is, its metabolites are not
genotoxic in simple mutation assays. It most likely produces leukemia by chromosomal damage
rather than simple point mutations. An argument can be made for nonlinearity on the basis that
the induction of chromosome damage by benzene and its metabolites is nonlinear and
in some instances shows a threshold. However, it should be pointed out that the micronucleus
assay of chromosomal damage is relatively insensitive and may not show effects at low doses,
even though some chromosomal damage is occurring. A recent report by Zhang et al. (1996a)
showed that the induction of aneuploidy of chromosome 9 as measured by FISH in interphase
lymphocytes from benzene-exposed workers is significantly elevated only at high levels of
exposure (>31 ppm in air). However, as yet unpublished studies have shown that the induction of
aneuploidy of other chromosomes (e.g., chromosome 7) occurs at lower doses and that the effect

35
of benzene on hyperdiploidy of chromosomes 7, 8, and 9 shows a significant linear trend (Zhang
et al., 1996b).
As discussed earlier, bone marrow DNA adducts as detected by P32 postlabeling after in
vivo exposure to benzene correspond with adducts formed by in vitro treatment with
hydroquinone or 1,2,4-benzentriol (Pathak et al., 1995).
It also has been demonstrated that oxidative DNA damage may contribute to benzene
genotoxicity and thus benzene-induced leukemia (Kolachana et al., 1993; Lagorio et al., 1994).
Because oxidative damage has a high rate of repair and studies in benzene-exposed mice and
human cells in vitro showed that the oxidative DNA damage was rapidly repaired, it could be
argued that this high level of repair will produce a threshold or nonlinearity at low doses.
However, it is errors during this repair process that cause point mutations from oxidative DNA
damage. Further, because there is already a considerable background level of oxidative damage
(Ames and Shigenaga, 1992), additional damage caused by benzene exposure may induce a linear
increase in point mutations.
It also could be proposed that hematotoxicity is required for leukemia induction. Because
hematotoxicity is likely to have a threshold, it is therefore possible that benzene-induced leukemia
will have a threshold and be nonlinear at low doses. In theory, hematotoxicity may increase the
risk of benzene-induced leukemia, because it could cause quiescent stem cells to enter the cycling
feeder cell stage, thereby expressing any genetic damage. However, there is no evidence that
hematotoxicity is a prerequisite for leukemia induction. Cases of leukemia following benzene
exposure without previous hematotoxicity have been reported, but the thoroughness of
monitoring for hematological effects is always a question. Benzene recently also has been shown
to have hematological effects below 10 ppm (Ward et al., 1996), and thus the relevance of a
threshold for hematotoxicity has decreased in most investigators’ estimation.
Irons, Subrahmanyam, Eastmond, and their co-workers have argued that the induction of
aneuploidy is a component of leukemia induction by benzene (Irons and Neptune, 1980;
Subrahmanyam et al., 1991; Eastmond, 1993). If this is true, then it could be argued that
leukemia induction has a threshold because numerous molecules of benzene metabolites would be
required to disrupt microtubules and cause aneuploidy. However, it should be pointed out that
there is a high level of background exposure to benzene and its metabolites. Benzene and its
metabolites are present in our diet and in cigarette smoke. Additional environmental exposure
will simply add to this background. Indeed, McDonald and co-workers have shown that proteins
in both the blood and bone marrow of humans and animals contain high levels of benzene
metabolite adducts and that the exposure of animals to benzene causes a linear increase in 1,4-
benzoquinone adducts on top of this background (McDonald et al., 1993, 1994). This additional
benzene exposure from the environment is likely to have a linear additional effect on the

36
background. Further, there are numerous mechanisms of aneuploidy induction that do not
necessarily involve binding to microtubules, and aneuploidy is not the only genetic mechanism of
suppressor gene loss and oncogeny activation. Care must therefore be exercised in claiming that
benzene is nonlinear on the basis of aneuploidy involvement.
Another theoretical argument is that the cells in the bone marrow have numerous defense
mechanisms to cope with toxic benzene metabolites. However, as discussed above, there is a high
background exposure and so additional exposures could actually escape defenses. Indeed, we
have calculated that there are approximately 10,000 benzene molecules per bone marrow cell
following normal environmental background exposures to benzene. The addition of further
molecules from environmental or occupational exposures will simply add to this and may easily
overwhelm or escape defense mechanisms.
Even if there are threshold levels at which each individual experiences increased leukemia
risk, population variability will almost certainly dictate that there is no one threshold dose that
applies across the population of people exposed to benzene. The data on susceptibility factors for
benzene toxicity and leukemogenicity are growing and will likely shed some light on population
variability in sensitivity to benzene’s adverse effects.

3.3. DOSE-RESPONSE CHARACTERIZATION

The major result of this update is a reaffirmation of the benzene interim unit risk estimates
derived in EPA’s 1985 interim risk assessment (U.S. EPA, 1985), which established the
probability of humans developing cancer from exposure to 1 ppm of benzene. Review of the
1985 interim risk assessment required addressing two main concerns. The first was use of the
updated epidemiologic data from Rinsky et al.’s (1987) cohort of Pliofilm workers and selection
of appropriate estimates of their exposure to benzene for the derivation of the unit risk estimate.
The second major concern was continued application of the low-dose linearity concept to the
model used to generate estimates of unit risk. It was concluded that at present there is insufficient
evidence to reject this concept.
Use of the update of Rinsky et al.’s (1987) cohort could have only a limited impact on the
EPA (1985) interim risk estimates if the same exposure-response (linear) model and the Crump
and Allen (1984) exposure measurements were used. When the higher estimated exposure
measurements by Paustenbach et al. (1993) were substituted for those of Crump and Allen, the
corresponding risk estimates were reduced by only a factor of, at most, 2. None of the
approaches for estimating exposure has greater scientific support than any other because there
was no ambient air benzene exposure data for the Pliofilm workers prior to 1946. Thus, there is
no clear basis for choosing a single best estimate. Rather, these sets of risk estimates reflect both

37
the inherent uncertainties in the applied model and the limitations of the exposure characterization
and response information in the epidemiologic data.
Without conducting extensive analyses of the raw data in Rinsky et al. (1987), only
theoretical analyses of the impact of various exposure assumptions and presumed etiologic
mechanisms on risk estimates for benzene are possible. There are two approaches for estimating
this impact. One is assuming a biological mechanism of benzene-induced leukemia (e.g.,
assuming that benzene-induced leukemia involves a sequence of genetic and epigenetic changes,
and that some of these steps are effected by benzene exposure), and the other is assuming a linear
model, like that used in the EPA (1985) assessment. In the linear model, use of cumulative
exposure would have less impact on the resultant risk estimate if the concentration were roughly
constant during the work history of the cohort. However, if exposure concentrations in early
work history were higher than in the later years, the unit risk could be overestimated. A
theoretical discussion about this subject can be found in Hoel (1984), which concludes that true
lifetime cancer risk can be significantly under- or overestimated when using epidemiological data
with short exposure.
While the risk estimates would be significantly different if a nonlinear exposure response
model were found to be more plausible, characterizing the shape (i.e., the nonlinearity) of the
exposure-response curve still would require a better understanding of the biological mechanisms
of benzene-induced leukemia. Some recent evidence suggests the possibility that the low-dose
curve could be supralinear because the formation of toxic metabolites plateaus above 25 ppm
benzene in air (Rothman et al., 1996b). This pattern is similar to that seen in laboratory animals
(Sabourin et al., 1989), where the effect per unit dose of benzene is less at high doses than at low
doses. Thus, it is possible that the unit risk is underestimated if linearity is assumed at low doses.
Arguments made in favor of benzene-induced leukemia being nonlinear at low doses can be
matched by arguments opposing this viewpoint. Currently, there is insufficient evidence either to
reject a linear dose-response curve for benzene in the low-dose region or to demonstrate that
benzene is, in fact, nonlinear in its effects. Even if the dose-response relationship were nonlinear,
the shape remains to be determined. Because of current lack of knowledge, continuing the
Agency’s previous approach of using a model with low-dose linearity is recommended. Of the
various approaches employing a linear assumption, the risk at 1 ppm ranges from 7.1 × 10-3 to 2.5
× 10-2 (Table 4).
Based on the Rinsky et al. (1987) study, the risk of leukemia is significantly elevated
(SMR = 1,186; 95% C.I. = 133-4,285) at a dose of 200 to 400 ppm-years (e.g., a person exposed
to a level of 5 to 10 ppm for 40 years). This assumes that exposure occurred for only 8 h each
day. However, Rinsky et al.’s (1987) data suggest that a rise in the SMR may begin at levels
under 40 ppm-years, although the trend does not attain statistical significance until a dose of 200

38
to 400 ppm-years is reached. We, therefore, are less confident that the risk begins to rise below
40 ppm-years (1 ppm for 40 years) than we are of increased risk at doses above 200 ppm-years.
Failure to find elevated risk at exposures to benzene below 40 ppm-years may, however, be a
matter of the lack of power to detect risk as significant below this level rather than an absence of
increased risk. Wong (1995), in a separate analysis of the risk of only AML in the Rinsky et al.
(1981) cohort, calculated an SMR of 0.91 (1 observed, 1.09 expected) in the exposure category
under 200 ppm-years. However, because AML is a subtype within the leukemia category, the
sensitivity for detecting a significant risk at that level of exposure is much lower.
On the other hand, several recent studies appear to support the possibility of elevated risk
at low exposures. Recent data from the Chinese cohort (Yin et al., 1989) suggest that the risk of
AML might be well below 200 ppm-years, although the data analysis still is incomplete and, as
discussed above, exposure estimates may have been biased toward unrealistically low values. Out
of 30 identified leukemia cases reported in that study, 11 reported cumulative exposure of under
200 ppm-years, and of these 11, 7 were subject to average levels of under 5 ppm-years during the
time that they were exposed. In fact, Hayes et al. (1996) added 12 leukemias to this total in an
update of the Yin et al. (1989) study. Interestingly, dosimetry data were calculated on selected
causes of death in this same cohort. Hayes et al. (1996) reported that excess risks of death from
hematopoietic malignancies were found at the level of 10 ppm-years cumulative exposure (9
observed vs. 3.6 expected). Unfortunately, length of employment was not provided in the
Chinese cohort. In addition, Bond et al. (1986a) reported five cases of myelogenous leukemia,
four with cumulative doses of benzene exposure between 1.5 ppm-years and 54 ppm-years. Their
average yearly exposure ranged from 1.0 ppm to 18 ppm. The Wong study (1987) reported that
six of seven cases of leukemia had cumulative benzene exposures of between 0.6 ppm-years and
113.4 ppm-years. Their average yearly exposure ranged from 0.5 ppm to 7.6 ppm. The seventh
case had no measured cumulative dose. Although it is possible that peak exposures could have
occurred at any time during employment, such information is unavailable. However, although the
authors of these studies developed dose-response data for some members of their respective
cohorts, the results of their studies cannot be used in unit risk calculations. In addition to the
questionable exposure estimates, it is clear that these workers were subject to concomitant
exposures to other toxics that also were present in the workplace and that might affect the risk of
cancer. Furthermore, methodological problems are present in these studies.
Based on observations from Rinsky et al. (1981, 1987) and recent studies, the Agency is
fairly confident that exposure to benzene increases the risk of leukemia at the level of 40 ppm-
years of cumulative exposure. However, below 40 ppm-years, the shape of the dose-response
curve cannot be determined on the basis of the current epidemiologic data. Benzene exposure at
40 ppm-years of occupational exposure (8 h/day, 5 days/week, 50 weeks/year) would be

39
equivalent to a lifetime (76 years) environmental exposure of 120 ppb. Hence, 120 ppb would be
a reasonable point of departure (POD) below which the shape of the dose-response curve is
uncertain. To put this POD of 120 ppb into perspective, it is necessary to examine information on
ambient benzene concentrations. Environmental surveys completed around the United States
have provided a variety of information on monitored levels of benzene, using both ambient
measurements and personal exposure measurements. (ATSDR [1997] provides a convenient
summary of much of the data.) Ambient measurements have been made both outdoors and
indoors. Shah and Singh (1988) report that the Volatile Organic Compound National Ambient
Database (1975-1985) contains the following daily median benzene air concentrations: workplace
air (2.1 ppb), indoor air (1.8 ppb), urban ambient (1.8 ppb), suburban ambient (1.8 ppb), rural
ambient (0.47 ppb), and remote (0.16 ppb). The EPA (1987) reports data from 44 sites in 39
cities of the United States, taken during the 6 to 9 a.m. “morning rush hour” periods during June-
September of 1984, 1985, and 1986. The median concentrations at these sites ranged from 4.8 to
35 ppb, with the authors noting that mobile sources (motor vehicles) were the major source of
ambient benzene in these samples. In industrialized areas, Pellizzari (1982) reports outdoor levels
of 0.13 to 5 ppb in Iberville Parish, LA, and Cohen et al. (1989) report median outdoor levels in
the Kanawha Valley region of West Virginia as 0.78 ppb. Cohen et al. (1989) also report that
mean indoor levels in the study were 2.1 ppb (median = 0.64 ppb, maximum = 14.9 ppb).
The EPA’s Total Exposure Assessment Methodology (TEAM) studies showed
consistently that personal exposures to benzene were higher than ambient indoor levels, and that
indoor levels, in turn, were higher than outdoor levels. Wallace (1989) reported that the overall
mean personal benzene exposure (smokers and nonsmokers) from the TEAM data was 4.7 ppb,
compared with an overall mean outdoor ambient level of 1.9 ppb. Median levels of benzene
indoors were broken out by those homes without smokers (mean = 2.2 ppb) and those where one
or more smokers were present (mean = 3.3 ppb). The TEAM authors frequently suggested
smoking as a source for indoor benzene concentrations (Wallace, 1987; Wallace et al., 1989).
Brunnemann et al. (1989) reported that indoor air samples at a smoke-filled bar ranged from 8.1
to 11.3 ppb of benzene. Wester et al. (1986) noted that benzene in the breath of smokers was
higher than that of nonsmokers, and that both were higher than the concentrations in outdoor
ambient air.
Other measurements of benzene concentrations include transient levels of benzene
approaching 1 ppm outside a vehicle while refueling (Bond et al., 1986b). Within a parking
garage, Flachsbart (1992) found a maximum level of 21 ppb. The estimated maximum level in a
basement during the Love Canal situation was about 160 ppb, with levels of about 60 ppb
estimated around an uncontrolled hazardous waste site (Bennett, 1987; Pellizzari, 1982). The

40
maximum single personal monitoring sample, representing one night’s exposure, during the 1981
New Jersey TEAM study was 159.6 ppb.
Using 120 ppb as the POD, the margin of exposure (MOE) can be calculated for several
of these levels. Since the 120 ppb is a 76-year lifetime value, this should be compared with
average levels in whichever exposure scenario is used. For example, if one assumes that 4.7 ppb
is the long-term average exposure for the general population, the MOE would be 120/4.7, or
about 26. If one assumes that the ambient indoor levels of 2.2 ppb cited above represent actual
exposures to nonsmokers, their MOE would be 120/2.2, or 55. If one were to construct a
hypothetical scenario where a person spent their entire life in a smoke-filled bar, the MOE would
drop to a range of 10 to 15.
The purpose of the margin of exposure discussion is to provide the risk manager with a
public health perspective on the adequacy of the difference between an environmental exposure of
interest and human equivalent exposures at the point of departure. Since the mode of action of
benzene exposure-induced leukemia is not well enough understood to support a nonlinear dose-
response analysis approach as is discussed in sections 2.3. and 3., the MOE serves as a default
approach. There are several factors which are not easily accounted for in the benzene data base
and which may impact on interpretation of the MOE and an acceptable risk to the population, i.e.,
the slope of the dose-response curve at or below the calculated POD, the nature and magnitude of
temporal exposure scenarios, extent of human variability and sensitivity within the general
population, use of a precursor effect other than leukemogenesis and the role of benzene and/or its
metabolites in leukemogenesis. These issues must be discussed with the decision-maker if he/she
deems the MOE approach to be a useful alternative to the linear approach which has been
recommended in this document.

4. CHILDREN’S RISK CONSIDERATIONS

The effects from exposure to benzene can be quite different among subpopulations.
Children may have a higher unit body weight exposure because of their heightened activity
patterns which can increase their exposures, as well as different ventilation tidal volumes and
frequencies, factors that influence uptake. This could entail a greater risk of leukemia and other
toxic effects to children if they are exposed to benzene at similar levels as adults. Infants and
children may be more vulnerable to leukemogenesis because their hematopoietic cell populations
are differentiating and undergoing maturation. Many confounding factors may affect the
susceptibility of children to leukemia (e.g., nutritional status, lifestyle, ethnicity, and place of
residence). Furthermore, in children, the predominant type of leukemia is lymphatic, while in

41
adults it is a combination of myeloid and lymphatic. Leukemia formerly classified as a single
disease now has been recognized as several different distinct malignancies that are characterized
by varying patterns in terms of age, race, sex, ethnic group, different secular trends, and different
etiologic factors (Linet, 1985).
Some recent research has shown, with limited consistency, that parental occupational
exposure to benzene plays a role in causing childhood leukemia. Shu et al. (1988) conducted a
case-control study of acute childhood leukemia in Shanghai, China, and found a significant
association between acute nonlymphocytic leukemia (ANNL) and maternal occupational
exposures to benzene during pregnancy (OR = 4.0). These excesses occurred among second- or
laterborn children rather than firstborn children. In addition, Mckinney et al. (1991) conducted a
case-control study to determine whether parental occupational, chemical, and other specific
exposures are risk factors for childhood leukemia. They found a significant association between
childhood leukemia and reported preconceptional exposures of fathers to benzene (OR = 5.81,
95% confidence intervals 1.67 to 26.44) and concluded that the results should be interpreted
cautiously because of the small numbers, overlap with another study, and multiple exposures of
some parents. Furthermore, Buckley et al. (1989) conducted a case-control study of occupational
exposures of parents of 204 children (under 18 years of age) with ANNL. They found a
significant association between ANNL and maternal exposure to pesticides, petroleum products,
and solvents. Among many chemicals, benzene was identified as one of the solvents. These
studies, however, have not provided data to indicate how the occupational exposures might affect
offspring. Some possible mechanisms include a germ-cell mutation prior to conception,
transplacental fetal exposures, exposures through breast milk, or direct exposures postnatally to
benzene from the environment.
Data on children exposed to benzene in the environment are very limited. Weaver et al.
(1996) conducted a pilot study that evaluated the feasibility of using trans, trans-muconic acid as a
biomarker of environmental benzene exposure in urban children. Although the authors concluded
that muconic acid could be used as a biomarker in children for environmental exposure, no studies
have been found that used this biomarker to determine actual benzene exposure in children.
In summary, children may represent a subpopulation at increased risk due to factors that
could increase their susceptibility to effects of benzene exposure (e.g., activity patterns), on key
pharmacokinetic processes (e.g., ventilation rates, metabolism rates, and capacities), or on key
pharmacodynamic processes (e.g., toxicant-target interactions in the immature hematopoietic
system). In addition, parental occupational exposures to benzene have been associated with their
increased risk. However, the data to make quantitative adjustments for these factors do not exist
at this time.

42
 5. FUTURE RESEARCH NEEDS

Data insufficiencies in several areas have been noted, and research in these areas ultimately
should provide a better understanding of how benzene causes cancer, particularly the mechanism
of benzene-induced leukemia. Several classes of data are needed on humans, i.e., more extensive
epidemiologic data with good exposure estimation, to permit verification and validation of the
prediction models. Additional research on exposure metrics that incorporates cumulative
exposures and occurrence of various effects inferring risk to general as well as sensitive
subpopulations is needed. More complete data on the preleukemic hematology of benzene-
exposed persons, such as the abnormal monoclonality and blood cell counts seen in such persons,
would be a significant contribution. Specific measures of early genetic damage in humans with
known exposure to benzene will help define the biological events leading up to the disease by
providing internal markers of its progression. This could be potentially useful in risk prediction
and assist in the identification of the steps leading to leukemia induced by exposure to benzene.
Such information may be forthcoming in the near future from a large cohort of benzene-exposed
workers under study in China. Investigators from the National Cancer Institute in the United
States, the Chinese Academy of Preventive Medicine, and the University of California at Berkeley
are currently developing such biomarker information as well as gathering clinical data on
hematologic abnormalities.
A need exists to further validate toxicokinetic models and to assess metabolic
susceptibility factors in human subjects. The collection of such information is problematic at best
because it requires exposure of human volunteers to a known carcinogen. However, data now
being collected in the Chinese cohort on the urinary metabolites of benzene, as well as in vitro
studies of cell-specific metabolism and toxicity in defined human bone marrow cell populations,
may be of use.
Research is needed to reduce some uncertainties in risk assessment of benzene, including
better understanding of the role of specific metabolites of benzene in toxic effects (cancer,
leukemia, hematotoxicity), the shape of the dose-response curve at low levels of benzene
exposure, and the role of DNA adducts and chromosomal aberrations in the development of
leukemia.
Continued basic research in hematopoiesis and leukemia biology is critical for identifying
the mechanisms of leukemogenesis. There remain important unanswered questions about the cell
population that contains targets for leukemic transformation, such as cell number and rate of
division, quiescence patterns, maturation, regulation, and apoptotic behavior. Future
understanding of the phenotypic consequences of common genetic aberration in MDS and AML
also is needed to assist in identifying the stages of leukemic transformation. There still exist some

43
uncertainties as to the exact classification of leukemias and leukemic subtypes in the
epidemiological studies of benzene exposures.
Current uncertainties limit the ability of modeling to explicitly consider all relevant
mechanisms, such as the formation of several types of genetic aberrations; disruption of
proliferation, differentiation, or apoptotic behaviors through genetic change or epigenetic
chemical interference; and the extremely complex and subtle regulation of hematopoietic
processes under normal feedback systems. Future research would be able to quantitatively
describe benzene pharmacokinetics in humans, relate dose measures to the above
pharmacodynamic mechanisms, and account for observed epidemiologic features of benzene-
induced leukemia, such as patterns of latency and susceptibility. For any mechanistic model of
leukemogenesis to be validated, it must be applied to existing data that relate known human
exposures to the probability of contracting MDS/AML. While there are epidemiologic data for
benzene, estimation of exposure is a complex task with considerable uncertainty. Therefore, a
suggested approach is to first develop a biologically based risk model for AML. It should be
recognized that in modeling benzene-induced leukemia in the general population there is
considerable interindividual variability that may influence risk. Some of the genetic factors
important in metabolic variability are becoming known, but other aspects of susceptibility remain
less well characterized. For example, the factors are unknown that control whether patients who
suffer benzene-induced myelosuppression progress to AML or recover after exposure is reduced
or removed. To what extent susceptibility factors will dictate leukemia risk and to what extent
leukemia is a manifestation of stochastic processes are not known.
There are a number of potential subgroups in the general population that may be at
increased risk from benzene exposure. These include but are not limited to indoor house painters
using oil paints, people with hobbies involving glues and solvents, do-it-yourself automobile
repairers working at home, alcoholic individuals, and smokers. With the exception of alcoholics
and smokers, a potential for benzene exposure exists for these individuals via both inhalation and
dermal routes.
Particular emphasis should be placed on research on those sensitive subpopulations who
are believed to be at increased risks (e.g., infants and children, the elderly). Given the high
proliferative rates and rapid rate of development of organ systems in the fetus and the prevalence
of leukemia in children, research directed at determining if the developing fetus is at increased risk
for cancer and noncarcinogenic effects is warranted. Research is needed to show how growth,
development, and aging affect the risk to humans. In addition, environmental and epidemiological
studies are needed to better determine the environmental benzene exposure levels that sensitive
subpopulations such as pregnant women, infants and children, and the elderly are likely to
encounter. Studies are needed to better understand how the absorption, distribution, metabolism,

44
and elimination of benzene varies with age, gender, race, or ethnicity and how this information
can be modeled to predict risk to sensitive subpopulations.
Current understanding of metabolism of benzene and preliminary findings of genetic
heterogeneity suggest that future research should emphasize subpopulations with particular
sensitivity to benzene toxicity and genetic polymorphism of enzymes involved in benzene
metabolism. Although this document focuses on benzene exposure via inhalation, the major route
of exposure to the general population, the potential contribution of dermal exposure among
individuals who use gasoline and other cleaning solvents containing benzene remains unknown.
Therefore, an additional potentially important area is the role and contribution of dermal
absorption, metabolism, and toxicity of benzene in the presence of other solvents including
gasoline. The effects of dermal absorption are an issue of public health concern because
individuals in occupations such as automobile repair and home painting, or who have hobbies
involving solvents, are likely to have higher than average benzene exposure. Another needed area
of research includes an understanding of how the sensitivity of individuals to benzene poisoning is
affected by various disease conditions or abnormalities of hematopoiesis.

6. REFERENCES

Aksoy, M. (1976) Types of leukemia in chronic benzene poisoning. A study in thirty-four patients. Acta Haemat
55:65-72.

Aksoy, M. (1977) Testimony of Mazaffer Aksoy, M.D., to Occupational Safety and Health Administration, U.S.
Department of Labor, July 13, 1977.

Aksoy, M. (1988) Benzene carcinogenicity. Boca Raton, FL: CRC Press.

Aksoy, M. (1989) Hematotoxicity and carcinogenicity of benzene. Environ Health Perspect 82: 193-197.

Aksoy, M; Dincol K; Akgun T; et al. (1971) Haematological effects of chronic benzene poisoning in 217 workers.
Br J Ind Med 28:296-302.

Aksoy, M; Erden, S; Dincol, G. (1974) Leukemia in show-workers exposed chronically to benzene. Blood 44:837-
841.

Ames, BN; Shigenaga, MK. (1992) Oxidants are a major contributor to aging. Ann NY Acad Sci 663:85-96.

Anderson, D; Yu, TW; Schmezer, P. (1995) An investigation of the DNA-damaging ability of benzene and its
metabolites in human lymphocytes, using the comet assay. Environ Mol Mutagen 26:305-314.

Andreoli, C; Leopardi, P; Crebelli, R. (1997) Detection of DNA damage in human lymphocytes by alkaline single
cell gel electrophoresis after exposure to benzene metabolites. Mutat Res Fundam Mol Mech Mut 377:95-104.

Andrew, LS; Lee, EW; Witmer, CM; et al. (1977) Effects of toluene on the metabolism, disposition and
hemopoietic toxicity of [3H]-benzene. Biochem Pharmacol 26:293-300.

45
Ashby, JA; de Serres, FJ; Draper, M; et al. (1985) Overview and conclusions of the IPCS collaborative study on in
vitro assay systems. In: Ashby, J; de Serres, FJ; Draper, M; Ishidate, M, Jr; Margolin, BH; Matter, BE; Shelby,
MD, eds. Evaluation of short-term tests for carcinogenesis: report of the International Programme on Chemical
Safety's collaborative study on in vitro assays. Vol. 5. Amsterdam: Elsevier Science Publishers, pp. 117-174.

ATSDR (Agency for Toxic Substances and Disease Registry). (1997) Toxicological profile for benzene. Update.
Public Health Service, U.S. Department of Health and Human Services, Atlanta, GA.

Bainton, DF; Ullyot, JL; Farquhar, MG. (1971) The development of neutrophilic polymorphonuclear leukocytes in
human bone marrow. J Exp Med 134: 907-934.

Barale, R; Marrazzini, A; Betti, C; et al. (1990) Genotoxicity of two metabolites of benzene: phenol and
hydroquinone show strong synergistic effects in vivo. Mutat Res 244:15-20.

Bennett, GF. (1987) Air quality aspects of hazardous waste landfills. Haz Waste Haz Mat 4:119-138.

Bhat, RV; Subrahmanyam, VV; Sadler, A; et al. (1988) Bioactivation of catechol in rat and human bone marrow
cells. Toxicol Appl Pharmacol 94:297-304.

Bodell, WJ; Lévay, G; Pongracz, K. (1993) Investigation of benzene DNA adducts and their detection in human
bone marrow. Environ Health Perspect 99:241-244.

Bois, F; Jackson, E; Pekari, K; et al. (1996) Population toxicokinetics of benzene. Environ Health Perspect 104
(Suppl 6):1405-1411.

Bond, GG; McLaren, EA; Baldwin, CL; et al. (1986a) An update of mortality among chemical workers exposed to
benzene. Br J Ind Med 43:685-691.

Bond, AE; Thompson, VL; Ortman, GC; et al. (1986b) Self service station vehicle refueling exposure study. In:
Proceedings of the 1986 EPA/APCA symposium on measurements of toxic air pollutants. Pittsburgh, PA: Air
Pollution Control Association, 458-466.

Brett, S; Rodricks, J; Chinchilli, V. (1989) Review and update of leukemia risk potentially associated with
occupational exposure to benzene. Environ Health Perspect 82:267-281.

Brodfuehrer, JI; Chapman, DE; Wilke, TJ; et al. (1990) Comparative studies of the in vitro metabolism and
covalent binding of 14C-benzene by liver slices and microsomal fraction of mouse, rat, and human. Drug Metab
Dispos 18:20-27.

Brunnemann, KD; Kagan, MR; Cox, JE; et al. (1989) Determination of benzene, toluene and 1,3-butadiene in
cigarette smoke by GC-MSD. Exp Pathol 37:108-113.

Buckley, JD; Robison, LL; Swotinsky, R, et al. (1989) Occupational exposures of parents of children with acute
nonlymphocytic leukemia: a report from the childrens cancer study group. Cancer Res 49:4030-4037.

Caligiuri, MA; Strout, MP; Gilliland, DG. (1997) Molecular biology of acute myeloid leukemia. Semin Oncol
24:32-44.

Chen, H; Eastmond, DA. (1995a) Synergistic increase in chromosomal breakage within the euchromatin induced
by an interaction of the benzene metabolites phenol and hydroquinone in mice. Carcinogenesis 16:1963-1969.

Chen, HW; Eastmond, DA. (1995b) Topoisomerase inhibition by phenolic metabolites: a potential mechanism for
benzene’s clastogenic effects. Carcinogenesis 16:2301-2307.

46
Chen, H; Rupa, DS; Tomar, R; et al. (1994) Chromosomal loss and breakage in mouse bone marrow and spleen
cells exposed to benzene in vivo. Cancer Res 54:3533-3539.

Clean Air Act Amendments of 1990. Subchapter II (42 U.S.C. 7521-7590).

Cohen, MA; Ryan, PB; Yanagiasawa, Y; et al. (1989) Indoor/outdoor measurements of volatile organic compounds
in Kanawha Valley of West Virginia. J Air Pollut Control Assoc 39:1086-1093.

Cooper, KR; Snyder, R. (1988) Benzene metabolism (Toxicokinetics and the molecular aspects of benzene
toxicity). In: Aksoy, M., ed., Benzene carcinogenicity. Boca Raton, FL, CRC Press, p. 33-58.

Cronkite, EP; Drew, RT; Inoue, T; et al. (1989) Hematotoxicity and carcinogenicity of inhaled benzene. Environ
Health Perspect 82:97-108.

Crump, KS. (1992) Exposure-response analyses of Pliofilm cohort. Work supported by Western States Petroleum
Association. Draft.

Crump, KS. (1994) Risk of benzene-induced leukemia: a sensitivity analysis of the Pliofilm cohort with additional
follow-up and new exposure estimates. J Toxicol Environ Health 42:219-242.

Crump, KS; Allen, BC. (1984) Quantitative estimates of risk of leukemia from occupational exposure to benzene.
Prepared for the Occupational Safety and Health Administration by Science Research Systems, Inc., Ruston, LA.

Dean, BJ. (1978) Genetic toxicology of benzene, toluene, xylenes, and phenols. Mutat Res 47: 75-97.

Dean, BJ. (1985) Recent findings on the genetic toxicology of benzene, toluene, xylenes, and phenols. Mutat Res
154:153-181.

DeCouflé, P; Blatter, WA; Blair, A. (1983) Mortality among chemical workers exposed to benzene and other
agents. Environ Res 30:16-25.

Dimitriadis, E; Guy, RL; Hu, P; Cooper, KR; Snyder, R. (1988) Effects of benzene metabolites in combination on
FE uptake into erythrocytes in mice. Toxicologist 8:72.

Dosemeci, M; Li, GL; Hayes, RB; et al. (1994) Cohort study among workers exposed to benzene in China: II.
Exposure assessment. Am J Ind Med 26:401-411.

Dustin, PJ. (1950) The action of mitotic poisons on normal and pathological blood-cell formation. Le Sang 4:21.

Eastmond, DA. (1993) Induction of micronuclei and aneuploidy by the quinone-forming agents benzene and o-
phenylphenol. Toxicol Lett 67:105-118.

Eastmond, DA; Smith, MT; Irons, RD. (1987) An interaction of benzene metabolites reproduces the myelotoxicity
observed with benzene exposure. Toxicol Appl Pharmacol 91:85-95.

Eastmond, DA; Rupa, DS; Hasegawa, LS. (1994) Detection of hyperdiploidy and chromosome breakage in
interphase human lymphocytes following exposure to the benzene metabolite hydroquinone using multicolor
fluorescence in situ hybridization with DNA probes. Mutat Res 322:9-20.

Farris, GM; Everitt, JI; Irons, RD; et al. (1993) Carcinogenicity of inhaled benzene in CBA mice. Fundam Appl
Toxicol 20(4):503-507.

47
Flachsbart, PG. (1992) Human exposure to motor vehicle air pollutants. In: Motor vehicle air pollution: public
health impacts and control measures. Mage, D; Zali, O, eds. ECOTOX, Service of Dept. of Ecotoxicology, Dept. of
Public Health, World Health Organization, Geneva, Switzerland; pp. 85-113.

Forni, A. (1971) Chromosome studies in workers exposed to benzene or toluene or both. Arch Environ Health
22:373-378.

Forni, A. (1994) Comparison of chromosome aberrations and micronuclei in tting genotoxicity in humans.
Toxicol Lett 72(103, Jun):185-190.

Gad-El-Karim, MM; Sagadopa Ramanujam, VM; Legator, MS. (1986) Correlation between the induction of
micronuclei in bone marrow by benzene exposure and the excretion of metabolites in urine of CD-1 mice. Toxicol
Appl Pharmacol 85:464-477.

Gaido, K; Wierda, D. (1987) Suppression of bone marrow stromal cell function by benzene and hydroquinone is
ameliorated by indomethacin. Toxicol Appl Pharmacol 89:378-390.

Ganousis, LG; Goon, D; Zyglewska, T; et al. (1992) Cell-specific metabolism in mouse bone marrow stroma:
studies of activation and deactivation of benzene metabolites. Mol Pharmacol 42:1118-1125.

Goldstein, BD. (1988) Benzene toxicity. Occup Med 3:541-54.

Goldstein, BD; Snyder, CA; Laskin, S; et al. (1982) Myelogenous leukemia in rodents inhaling benzene. Toxicol
Lett 13:169-173.

Guengerich, FP; Kim, D; Iwasaki, M. (1991) Role of human cytochrome P-450 IIE1 in the oxidation of many low
molecular weight cancer suspects. Chem Res Toxicol 4:168-179.

Guy, R; Dimitriadis, E; Hu, P; et al. (1990) Interactive inhibition of erythroid 59Fe utilization by benzene
metabolites in female mice. Chem Biol Interact 74:55-62.

Guy, RL; Peidi, H; Witz, G; et al. (1991) Depression of iron uptake into erythrocytes in mice by treatment with the
combined benzene metabolites p-benzoquinone, muconaldehyde and hydroquinone. J Appl Toxicol 11:443-446.

Hay, EB. (1961) Exposure to aromatic hydrocarbons in a coke oven by-product plant. J Am Ind Hyg Assn 21:386-
391.

Hayes, RB; Yin, SN; Dosemeci, M; et al. (1996) Mortality among benzene-exposed workers in
China. Environ Health Perspect 104(suppl 6):1349-1352.

Hayes, RB; Yin, SN; Dosemeci, MS; et al. (1997) Benzene and the dose-related incidence of hematologic
neoplasms in China. J Nat Cancer Inst 89:1065-1071.

Heim, S. (1996) Clonal chromosome abnormalities in neoplastic cells: evidence of genetic instability? Cancer Surv
28:247-260.

Henderson, RF; Sabourin, PJ; Medinsky, MA; et al. (1992) Benzene dosimetry in experimental animals: relevance
for risk assessment. In: D’Amato, REA, ed. Relevance of animal studies to the evaluation of human cancer risk.
New York: Wiley-Liss, pp. 93-105.

Hoel, D. (1984) The impact of occupational exposure patterns on quantitative risk estimation. In: Hoel, D; Merrill,
R; Perera, F, eds. Risk quantitation and regulatory policy. Banbury Report 19, Cold Spring Harbor Laboratory,
Cold Spring Harbor, NY, pp. 105-118.

48
Hogstedt, B; Holmen, A; Karlsson, A; et al. (1991) Gasoline pump mechanics had increased frequencies and sizes
of micronuclei in lymphocytes stimulated by pokeweed mitogen. Mutat Res 263:51-55.

Hu, P; Guy, R; Witz, G; Goldstein, B; et al. (1990) Suppression of erythropoiesis in mice by combined treatment
with the benzene metabolites p-benzoquinone (BQ), muconaldehyde (MUC) and hydroquinone (HQ). Toxicologist
10:58.

Huff, JE; Haseman, JK; DeMarini, DM; et al. (1989) Multiple-site carcinogenicity of benzene in Fischer 344 rats
and B6C3F1 mice. Environ Health Perspect 82:125-163.

Hunting, KL; Longbottom, H; Kalavar, SS; et al. (1995) Haematopoietic cancer mortality among vehicle
mechanics. Occup Environ Med 52:673-678.

IARC (International Agency for Research on Cancer). (1982) IARC monographs on the evaluation of carcinogenic
risks of chemicals to humans: some industrial chemicals and dyestuffs. Vol. 29, Lyon, France: IARC, pp. 93-148.

Infante, PF; Rinsky, RA; Wagoner, JK; et al. (1977) Leukemia in benzene workers. Lancet 2:76-78.

Inoue, I; Seiji, K; Watanbe, TH; et al. (1988a) Mutual metabolic suppression between benzene and toluene in man.
Arch Occup Environ Health 60:15-20.

Inoue, O; Seiji, K; Kasahara, M; et al. (1988b) Determination of catechol and quinol in urine of workers exposed
to benzene. Br J Ind Med 45:487-492.

IPCS (International Programme on Chemical Safety). (1993) Benzene: Environmental health criteria 150. Geneva:
World Health Organization.

Irons, RD; Dent, JG; Baker, TS; et al. (1980) Benzene is metabolized and covalently bound in bone marrow in situ.
Chem Biol Interact 30:241-245.

Irons, RD; Neptune, DA. (1980) Effects of the principal hydroxy-metabolites of benzene on microtubule
polymerization. Arch Toxicol 45:297-305.

Irons, RD; Sawahata, T. (1985) Phenols, catechols, and quinones. In: Anders, MW. ed. Bioactivation of foreign
compounds. New York: Academic Press, pp. 259-281.

Irons, RD; Stillman, WS. (1993) Cell proliferation and differentiation in chemical leukemogenesis. Stem Cells
11:235-242.

Irons, RD; Stillman, WS. (1996) The process of leukemogenesis. Environ Health Perspect 104(suppl 6):1239-1246.

Irons, RD; Stillman, WS; Colagiovanni, DB; et al. (1992) Synergistic action of the benzene metabolite
hydroquinone on myelopoietic stimulating activity of granulocyte/macrophage colony-stimulating factor in vitro.
Proc Natl Acad Sci 89:3691-3695.

Johansson, B; Mertens, F. (1988) The breakpoints of spontaneous chromosomes aberrations in lymphocytes from
patients with non-Hodgkin’s lymphoma are nonrandomly distributed. Hereditas 108:243-249.

Johansson, I; Ingel, MA; Sundberg, M. (1988) Benzene metabolism by ethanol-, acetone-, and benzene-inducible
cytochrome P-450 (IIE1) in rat and rabbit liver microsomes. Cancer Res 48:5387-5390.

Karp, JE; Smith, MA. (1997) The molecular pathogenesis of treatment-induced (secondary) leukemias:
foundations for treatment and prevention. Semin Oncol 24:103-113.

49
Kenyon, EM; Medinsky, MA. (1995) Incorporation of heterogeneous enzymes distribution into a physiological
model for benzene, phenol and hydroquinone. Toxicologist 15:49.

Kipen, HM; Cody, RP; Crump, KS; et al. (1988) Hematologic effects of benzene: A thirty-five year longitudinal
study of rubber workers. Toxicol Ind Health 4:411-430.

Kipen, HM; Cody, RP; Goldstein, BD. (1989) Use of longitudinal analysis of peripheral blood counts to validate
historical reconstructions of benzene exposure. Environ Health Perspect 82:199-206.

Kline, SA; Xiang, Q; Goldstein, BD; et al. (1993) Reaction of (E, E)-muconaldehyde and its aldehydic metabolites,
(E,E)-6-oxohexadienoic acid and (E,E)-6-hydroxyhexa-2,4-dienal, with glutathione. Chem Res Toxicol 6:578-583.

Kolachana, P; Subrahmanyam, VV; Meyer, KB; et al. (1993) Benzene and its phenolic metabolites produce
oxidative DNA damage in HL60 cells in vitro and in the bone marrow in vivo. Cancer Res 53:1023-1026.

Koop, DR; Laethem, CL; Schnier, GC. (1989) Identification of ethanol-inducible P450 isozyme 3a (P450IIE1) as a
benzene and phenol hydroxylase. Toxicol Appl Pharmocol 98:278-288.

Kracke, RR; Parker, FP. (1934) The etiology of granulopenia (agranulocytosis) with particular references to drugs
containing the benzene ring. Am J Clin Pathol 4:451-469.

Lagorio, S; Tagesson, C; Forastiere, F; et al. (1994) Exposure to benzene and urinary concentrations of 8-
hydroxydeoxyguanosine, a biological marker of oxidative damage to DNA. Occup Environ Med 51:739-743.

Latriano, L; Goldstein, BD; Witz, G. (1986) Formation of muconaldehyde, an open ring metabolite of benzene, in
mouse liver microsomes: an additional pathway of toxic metabolites. Proc Natl Acad Sci USA 83:8356-8360.

Lee, SST; Buters, JTM; Pineau, T; et al. (1996) Role of CYP2E1 in the hepatotoxicity of acetaminophen. J Biol
Chem 15:3012-3022.

Legathe, A; Hoener, BA; Tozer, TN. (1994) Pharmacokinetic interaction between benzene metabolites, phenol and
hydroquinone, in B6C3F1 mice. Toxicol Appl Pharmacol 124:131-138.

Lévay, G; Pathak, DN; Bodell, WJ. (1996) Detection of DNA adducts in the white blood cells of B6C3F1 mice
treated with benzene. Carcinogenesis 17:151-153.

Lévay, G; Ross, D; Bodell, WJ. (1993) Peroxidase activation of hydroquinone results in the formation of DNA
adducts in HL-60 cells, mouse bone marrow macrophages and human bone marrow. Carcinogenesis 14:2329-2334.

Linet, MS. (1985) The leukemias: epidemiologic aspects. New York: Oxford University Press.

Low, LK; Meeks, J; Norris, KJ; et al. (1989) Pharmacokinetics and metabolism of benzene in Zymbal gland and
other key target tissues after oral administration in rats. Environ Health Perspect 82:215-222.

Low, LK; Lambert, C; Meeks, J; et al. (1995) Tissue-specific metabolism of benzene in Zymbal gland and other
solid tumor target tissues in rats. J Am Coll Toxicol 14:40-60.

Maltoni, C; Conti, B; Cotti, G. (1983) Benzene: a multipotential carcinogen. Results of the long-term bioassays
performed at the Bologna Institute of Oncology. Am J Ind Med 4:589-630.

Maltoni, C; Conti, B; Perino, G; et al. (1988) Further evidence of benzene carcinogenicity. Results on Wistar rats
and Swiss mice treated by ingestion. Ann NY Acad Sci 534:412-426.

50
Marrazzini, A; Chelotti, L; Barrai, I; et al. (1994) In vivo genotoxic interactions among three phenolic benzene
metabolites. Mutat Res 341:29-46.

McDonald, TA; Waidyanatha, S; Rappaport, SM. (1993) Production of benzoquinone adducts with hemoglobin
and bone-marrow proteins following administration of [13C6]benzene to rats. Carcinogenesis 14:1921-1925.

McDonald, TA; Yeowell, OCK; Rappaport, SM. (1994) Comparison of protein adducts of benzene oxide and
benzoquinone in the blood and bone marrow of rats and mice exposed to [14C/13C6]benzene. Cancer Res 54:4907-
4914.

McKelvey-Martin, VJ; Green, MHL; Schmezer, P; et al. (1993) The single cell gel electrophoresis assay (comet
assay): a European review. Mutat Res 288:47-63.

McKinney PA; Alexander, FE; Cartwright, RA; et al. (1991) Parental occupations of children with leukemia in
west Cumbria, north Humberside, and Gateshead, Br Med J 302:681-686.

Medinsky, MA; Schlosser, PM; Bond, JA. (1994) Critical issues in benzene toxicity and metabolism: the effects of
interactions with other chemicals on risk assessment. Environ Health Perspect 102 (suppl 9):119-124.

Medinsky, MA; Kenyon, EM; Seaton, MJ; et al. (1996) Mechanistic considerations in benzene physiological model
development. Environ Health Perspect 104(suppl 6):1399-1404.

Moszczynski, P; Lisiewicz, J. (1982) T and B cells and occupational exposure to benzene and its homologues (with
regard to other blood cells). Rev Esp Oncol 29:49-55.

Mullin, AH; Rando, R; Esmundo, F; et al. (1995) Inhalation of benzene leads to an increase in the mutant
frequencies of a lacI transgene in lung and spleen tissues of mice. Mutat Res 327:121-129.

National Toxicology Program, Public Health Service, U.S. Department of Health and Human Services. (1986) NTP
technical report on the toxicology and carcinogenesis studies of benzene (CAS No. 71-43-2) in F344/N rats and
B6C3F1 mice (gavage studies). NTP TR 289. Research Triangle Park, NC: National Institutes of Health.

Niculescu, R; Bradford, HN; Colman, RW; et al. (1995) Inhibition of the conversion of pre-interleukins-1 alpha
and 1 beta to mature cytokines by p-benzoquinone, a metabolite of benzene. Chem Biol Interact 98:211-222.

Oberly, TJ; Bewsey, BJ; Probst, GS. (1984) An evaluation of the L5178Y TK+/- mouse lymphoma forward
mutation assay using 42 chemicals. Mutat Res 125:291-306.

Oberly, TJ; Rexroat, MA; Richardson, KK; et al. (1990) An evaluation of the CHO/HGPRT mutation assay
involving suspension cultures and soft agar cloning: results for 33 chemicals. Environ Mol Mutagen 16:260-271.

Occupational Safety and Health Administration. (1987) Final rule on occupational exposure to benzene. Federal
Register 54:34660-34762.

Ott, MG; Townsend, JC; Fishbeck, WA; et al. (1977) Mortality among individuals occupationally exposed to
benzene. Exhibit 154, OSHA benzene hearings, July 9-August 19, 1977.

Ott, MG; Townsend, JC; Fishbeck, WA; et al. (1978) Mortality among individuals occupationally exposed to
benzene. Arch Environ Health 33:3-10.

Parke, DV; Williams, RT. (1953) Studies in detoxification. The metabolism of benzene. (A) The formation of
phenylglucuronide and phenyl sulfuric acid from 14C-benzene. (B) The metabolism of 14C-phenol. Biochem J
55:337-340.

51
Parmley, R. (1988) Mammals. In: Vertebrate blood cells. Rowley, AF; Ratcliffe, NA, eds. Cambridge: Cambridge
University Press, pp. 337-424.

Pathak, DN; Lévay, G; Bodell, WJ. (1995) DNA adduct formation in the bone marrow of B6C3F1 mice treated
with benzene. Carcinogenesis 16:1803-1808.

Paustenbach, DJ; Price, PS; Ollison, W; et al. (1992) Reevaluation of benzene exposure for the Pliofilm
(rubberworker) cohort (1936-1976). J Toxicol Environ Health 36:177-231.

Paustenbach, D; Bass, R; Price, P. (1993) Benzene toxicity and risk assessment, 1972-1992: implications for future
regulation. Environ Health Perspect 101 (Suppl 6):177-200.

Paxton, MB. (1996) Leukemia risk associated with benzene exposure in the Pliofilm cohort. Environ Health
Perspect 104(suppl 6):1431-1436.

Paxton, MB; Chinchilli, V; Brett, SM; et al. (1992, Jan. 27) Reanalysis and update of the leukemogenic risk
associated with occupational benzene exposure in Pliofilm cohort. Environ Corporation, Arlington, VA.

Paxton, MB; Chinchilli, V; Brett, SM; et al. (1994) Leukemia risk associated with benzene exposure in the
Pliofilm cohort. II. Risk estimate. Risk Anal 14:155-161.

Pedersen-Bjergaard, J; Pedersen, M; Roulston, D; et al. (1995) Different genetic pathways in leukemogenesis for
patients presenting with therapy-related myelodysplasia and therapy-related acute myeloid leukemia. Blood
86:3542-3552.

Pellizzari, ED. (1982) Analysis for organic vapor emissions near industrial and chemical waste disposal sites.
Environ Sci Technol 16:781-785.

Plappert, U; Barthel, E; Raddatz, K; et al. (1994) Early effects of benzene exposure in mice. Hematological versus
genotoxic effects. Arch Toxicol 68:284-290.

Raabe, GK; Wong, O. (1996) Leukemia mortality by cell type in petroleum workers with potential exposure to
benzene. Environ Health Perspect 104(suppl 6):1381-1392.

Reddy, MV; Bleicher, WT; Blackburn, GR; et al. (1990) DNA adduction by phenol, hydroquinone, or
benzoquinone in vitro but not in vivo: nuclease P1-enhanced 32P-postlabeling of adducts as labeled nucleoside
bisphosphates, dinucleotides and nucleoside monophosphates. Carcinogenesis 11:1349-1357.

Reddy, MV; Schultz, SC; Blackburn, GR; et al. (1994) Lack of DNA adduct formation in mice treated with
benzene. Mutat Res 325:149-155.

Renz, JF; Kalf, GF. (1991) Role for interleukin-1 (IL-1) in benzene-induced hematotoxicity: inhibition of
conversion of pre-IL-1 alpha to mature cytokine in murine macrophages by hydroquinone and prevention of
benzene-induced hematotoxicity in mice by IL-1 alpha. Blood 78:938-944.

Rickert, DE; Baker, TS; Bus, JS; et al. (1979) Benzene disposition in the rat after exposure by inhalation. Toxicol
Appl Pharmacol 49:417-423.

Rinsky, R; Young, RJ; Smith, AB. (1981) Leukemia in benzene workers. Am J Ind Med 2:217-245.

Rinsky, RA; Smith, AB; Hornung, R; et al. (1987) Benzene and leukemia: an epidemiologic risk assessment. N
Engl J Med 316:1044-1050.

52
Robertson, ML; Eastmond, DA; Smith, MT. (1991) Two benzene metabolites, catechol and hydroquinone, produce
a synergistic induction of micronuclei and toxicity in cultured human lymphocytes. Mutat Res 249:201-209.

Ross, D. (1996) Metabolic basis of benzene toxicity. Eur J Haematol 57:(suppl):111-118.

Ross, D; Siegel, D; Gibson, N; et al. (1990) Activation and deactivation of quinones catalyzed by DT-diaphorase.
Evidence for bioreductive activation of diaziquone (AZQ) in human tumor cells and detoxification of benzene
metabolites in bone marrow stroma. Free Radic Res Commun 8:373-381.

Rothman, N; Haas, R; Hayes, RB; et al. (1995) Benzene induces gene duplicating but not gene inactivating
mutations at the glycophorin A locus in exposed humans. Proc Natl Acad Sci USA 92:4069-4073.

Rothman, N; Smith, MT; Hayes, RB; et al. (1996a) An epidemiologic study of benzene’s early biologic effects in
heavily exposed workers in Shanghai, China. Environ Health Perspect 104(suppl 6):1365-1370.

Rothman, N; Li, GL; Dosemeci, M; et al. (1996b) Hematotoxicity among Chinese workers heavily exposed to
benzene. Am J Ind Med 29:236-246.

Rothman, N; Smith, MT; Hayes, Rb; et al. (1997) Benzene poisoning, a risk factor for hematological malignancy,
is associated with the NQ01 C-609->T mutation and rapid fractional excretion of chlorzoxazone. Cancer Res
57:2839-2842.

Rushton, R; Romaniuk, H. (1997) A case-control study to investigate the risk of leukaemia associated with
exposure to benzene in petroleum marketing and distribution workers in the United Kingdom. Occup Environ Med
54:152-166.

Sabourin, PJ; Bechtold, WE; Birnbaum, LS; et al. (1988) Differences in the metabolism and disposition of inhaled
[3H] benzene by F344/N rats and B6C3F1 mice. Toxicol Appl Pharmacol 94:128-140.

Sabourin, PJ; Bechtold, WE; Griffith, WC; et al. (1989) Effect of exposure concentration, exposure rate, and route
of administration on metabolism of benzene by F344 rats and B6C3F1 mice. Toxicol Appl Pharmacol 99:421-444.

Sabourin, PJ; Medinsky, MA; Bechtold, WE; et al. (1989) Disposition of benzene in animals and man. In:
Mehlman, MA, ed., Advances in Modern Environmental Toxicology, Vol. 16. Benzene: occupational and
environmental hazards - scientific update, Princeton, NJ: Princeton Scientific Publishing Co., Inc., p. 153-176.

Sammett, D; Lee, EW; Kocsis, JJ; Snyder, R. (1979) Partial hepatectomy reduced both the metabolism and toxicity
of benzene. J Toxicol Environ Health 5:785-792.

Sarto, F; Cominato, I; Pinton, AM; et al. (1984) A cytogenetic study on workers exposed to low concentrations of
benzene. Carcinogenesis 5:827-832.

Sasiadek, M. (1992) Nonrandom distribution of breakpoints in the karyotypes of workers occupationally exposed to
benzene. Environ Health Perspect 97:255-257.

Savitz, DA; Andrews, KW. (1996) Risk of myelogenous leukaemia and multiple myeloma in workers exposed to
benzene. Occup Environ Med 53:357.

Sawahata, T; Rickert, DE; Greenlee, WF. (1985) Metabolism of benzene and its metabolites in bone marrow. In:
Irons, RD, ed, Toxicology of the blood and bone marrow. New York: Raven Press, p. 141-148.

Schlosser, PM; Bond, JA; Medinsky, MA. (1993) Benzene and phenol metabolism by mouse and rat liver
microsomes. Carcinogenesis 14(12):2477-2486.

53
Schlosser, PM; Kenyon, EM; Seaton, MJ; et al. (1995) Determinants of benzene metabolism and disposition. CIIT
Activities 15(6):1-9.

Schnatter, AR; Armstrong, TW; Thompson, LS; et al. (1996a) The relationship between low-level benzene
exposure and leukemia in Canadian petroleum distribution workers. Environ Health Perspect 104(suppl 6):1375-
1379.

Schnatter, AR; Nicolich, MJ; Bird, MG. (1996b) Determination of leukemogenic benzene exposure concentrations:
refined analyses of the Pliofilm cohort. Risk Anal 16:833-840.

Seaton, MJ; Schlosser, PM; Bond, JA; et al. (1994) Benzene metabolism by human liver microsomes in relation to
cytochrome P450 2E1 activity. Carcinogenesis 15:1799-1806.

Seaton, MJ; Follansbee, MH; Bond, JA. (1995) Oxidation of 1,2-epoxy-3-butene to 1,2:3,4-diepoxybutane by
cDNA-expressed human cytochromes P450 2E1 and 3A4 and human, mouse and rat liver microsomes.
Carcinogenesis 16:2287-2293.

Shah, JJ; Singh, HB. (1988) Distribution of volatile organic chemicals in outdoor and indoor air. Environ Sci
Technol 22:1381-1388.

Shu, XO; Gao, YT; Brinton, LA; et al. (1988) A population-based case-control study of childhood leukemia in
Shanghai. Cancer 62:635-644.

Smart, R; Zannoni, V. (1984) DT-diaphorase and peroxidase influence the covalent binding of the metabolites of
phenol, the major metabolite of benzene. Mol Pharmacol 26:105-111.

Smith, MT. (1996) The mechanism of benzene-induced leukemia: a hypothesis and speculations on the causes of
leukemia. Environ Health Perspect 104:1219-1225.

Smith, MT; Yager, JW; Steinmetz, KM; et al. (1989) Peroxidase-dependent metabolism of benzene’s phenolic
metabolites and its potential role in benzene toxicity and carcinogenicity. Environ Health Perspect 82:23-29.

Snyder, CA; Goldstein, BD; Sellakumar, AR; et al. (1980) The inhalation toxicology of benzene: incidence of
hematopoietic neoplasms and hematotoxicity in AKR/J and C57BL/6J mice. Toxicol Appl Pharmacol 54:323-331.

Snyder, CA; Goldstein, BD; Sellakumar, A; et al. (1982) Toxicity of chronic benzene inhalation: CD-1 mice
exposed to 300 ppm. Bull Environ Contam Toxicol 29:385-391.

Snyder, CA; Goldstein, BD; Sellakumar, AR; et al. (1984) Evidence for hematotoxicity and tumorigenesis in rats
exposed to 100 ppm benzene. Am J Ind Med 5:429-434.

Snyder, R; Kalf, GF. (1994) A perspective on benzene leukemogenesis. Crit Rev Toxicol 24:177-209.

Snyder, R; Longacre, SL; Witmer, CM; et al. (1981) Biochemical toxicology of benzene. Rev Biochem Tox 3:123-
154.

Stephens, EA; Taylor, JA; Kaplan, N; et al. (1994) involvement of 11p15 and 3q21q26 in therapy-related myeloid
leukemia (t-ML) in children. Case reports and review of the literature. Cancer Genet Cytogenet 75:11-22.

Subrahmanyam, VV, Sadler, A; Suba, E; et al. (1989) Stimulation of the in-vitro bioactivation and detoxification
of benzene metabolites. Drug Metab Dispos 17:348-350.

54
Subrahmanyam, V; Doane-Setzer, P; Steinmetz, K; et al. (1990) Phenol-induced stimulaion of hydroquinone
bioactivation in mouse bone marrow in vivo: possible implications in benzene myelotoxicity. Toxicology 62:107-116.

Subrahmanyam, VV; Ross, D; Eastmond, DA; et al. (1991) Potential role of free radicals in benzene-induced
myelotoxicity and leukemia. Free Rad Biol Med 11:495-515.

Tompa, A; Major, J; Jakab, MG. (1994) Monitoring of benzene-exposed workers for genotoxic effects of benzene:
improved-working-condition-related decrease in the frequencies of chromosomal aberrations in peripheral blood
lymphocytes. Mutat Res 304:159-165.

Trush, MA; Twerdok, LE; Rembish, SJ; et al. (1996) Analysis of targer cell susceptibility as a basis for the
development of a chemoprotective strategy against benzene-induced hematotoxicities. Environ Health Perspect
104(6):1227-1234.

Tunek, A; Olofsson, T; Berlin, M. (1981) Toxic effects of benzene and benzene metabolites on granulopoietic stem
cells and bone marrow cellularity in mice. Toxicol Appl Pharmacol 59:149-156.

Tuo, JS; Loft, S; Thomsen, MS; et al. (1996) Benzene-induced genotoxicity in mice in vivo detected by the alkaline
comet assay: reduction by CYP2E1 inhibition. Mutat Res 368:213-219.

U.S. Environmental Protection Agency. (1979) Final report on population risk to ambient benzene exposures.
Prepared by the Carcinogen Assessment Group, Research Triangle Park, NC. EPA/450/5-80-004.

U.S. Environmental Protection Agency. (1985, Feb. 15) Interim quantitative cancer unit risk estimates due to
inhalation of benzene. Prepared by the Carcinogen Assessment Group, Office of Research and Development,
Washington, DC. EPA/600/X-85-022.

U.S. Environmental Protection Agency. (1986, Sept. 24) Guidelines for carcinogen risk assessment. Federal
Register 51(185):33992-34003.

U.S. Environmental Protection Agency. (1987) June-September, 6-9 AM, ambient air benzene concentrations in 39
U.S. cities, 1984-1986. Research Triangle Park, NC: U.S. Environmental Protection Agency, Atmospheric
Sciences Research Lab. EPA/600/D-87/160.

U.S. Environmental Protection Agency. (1996, April 23) Proposed guidelines for carcinogen risk assessment.
Federal Register 61(79):17960-18011.

Utterback, DF; Rinsky, RA. (1995) Benzene exposure assessment in rubber hydrochloride workers: A critical
evaluation of previous estimates. Am J Ind Med 27:661-676.

Valentine, JL; Lee, SST; Seaton, MJ; et al. (1996) Reduction of benzene metabolism and toxicity in mice that lack
CYP2E1 expression. Toxicol Appl Pharmacol 141:205-213.

Van den Berghe, H; Louwagie, A; Broeckaert-Van Orshoven, A; et al. (1979) Chromosome analysis in two
unusual malignant blood disorders presumably induced by benzene. Blood 53:558-566.

Vigliani, EC; Forni, A. (1976) Benzene and leukemia. Environ Res 11:122-127.

Vigliani, EC; Saita, G. (1964) Benzene and leukemia. N Engl J Med 271:872-876.

Wallace, LA. (1989) Major sources of benzene exposure. Environ Health Perspect 82:165-169.

Wallace, LA; Pellizzari, ED; Hartwell, TD; et al. (1987) Exposures to benzene and other volatile compounds from
active and passive smoking. Arch Environ Health 42:272-279.

55
Ward, JB; Ammenhauser, MM; Ramanujam, VM; et al. (1992) The mutagenic effects of low level sub-acute
inhalation exposure to benzene in CD-1 mice. Mutat Res 268:49-57.

Ward, E; Hornung, R; Morris, J; et al. (1996) Risk of low red or white cell count related to estimated benzene
exposure in a rubberworker cohort (1940-1975). Am J Ind Med 29:247-257.

Weaver, VM; Davoli, CT; Heller, PJ; et al. (1996) Benzene exposure, assessed by urinary trans, trans-muconic
acid, in urban children with elevated blood lead levels. Environ Health Perspect 104(3):318-323.

Wester, RC; Maibach, HI; Gruenke, LD; et al. (1986) Benzene levels in ambient air and breath of smokers and
nonsmokers in urban and pristine environments. J Toxicol Environ Health 18:567-573.

Witz, G; Rao, GS; Goldstein, BD. (1985) Short-term toxicity of trans, trans-muconaldehyde. Toxicol Appl
Pharmacol 80:511-516.

Witz, G; Maniara, W; Mylavarapu, V; et al. (1990a) Comparative metabolism of benzene and trans,trans-
muconaldehyde to trans,trans-muconic acid in DBA/2N and C57BL/6 mice. Biochem Pharmacol 40:1275-1280.

Witz, G; Gad, SC; Tice, RR; et al. (1990b) Genetic toxicity of the benzene metabolite trans, trans-muconaldehyde
in mammalian and bacterial cells. Mutat Res 240:295-306.

Wong, O. (1987) An industry-wide study of chemical workers occupationally exposed to benzene. Br J Ind Med
44:382-395.

Wong, O. (1995) Risk of acute myeloid leukemia and multiple myeloma in workers exposed to benzene. Occup
Environ Med 52:380-384.

Wong, O; Morgan, RW; Whorton, MD. (1983, Dec. 8) An industry-wide mortality study of chemical workers
occupationally exposed to benzene. Technical report submitted to Chemical Manufacturers Association by
Environmental Health Associates, Berkeley, CA.

Wright, EC. (1995) The pathogenesis of leukemia. In: Radiation toxicology: bone marrow and leukemia. Hendry,
JH; Lord, BI, eds. London: Taylor and Francis, pp. 245-274.

Yager, JW; Eastmond, DA; Robertson, ML; et al. (1990) Characterization of micronuclei induced in human
lymphocytes by benzene metabolites. Cancer Res 50:393-399.

Yin, SN; Li, GL; Tain, FD; et al. (1987) Leukaemia in benzene workers: A retrospective cohort study. Br J Ind
Med 44:124-128.

Yin, SN; Li, GL; Tain, FD; et al. (1989) A retrospective cohort study of leukemia and other cancers in benzene
workers. Environ Health Perspect 82:207-213.

Yin, SN; Linet, MS; Haynes, RB; et al. (1994) Cohort study among workers exposed to benzene in China: I.
General methods and resources. Amer J Ind Med 26:383-400.

Yin, SN; Hayes, RB; Linet, MS; et al. (1996) A cohort study of cancer among benzene-exposed workers in China:
Overall results. Am J Ind Med 29:227-235.

Young, BD; Saha, V. (1996) Chromosome abnormalities in leukaemia: the 11q23 paradigm. Cancer Surv 28:225-
245.

56
Zhang, L; Robertson, ML; Kolachana, P; et al. (1993a) Benzene metabolite, 1,2,4-benzenetriol, induces
micronuclei and oxidative DNA damage in human lymphocytes and HL60 cells. Environ Mol Mutagen 21:339-348.

Zhang, Z; Kline, SA; Kirley, TA; et al. (1993b) Pathways of trans, trans-muconaldehyde metabolism in mouse
liver cytosol: reversibility of monoreductive metabolism and formation of end products. Arch Toxicol 67:461-467.

Zhang, L; Venkatesh, P; Creek, MLR; et al. (1994) Detection of 1,2,4-benzenetriol induced aneuploidy and
microtubule disruption by fluorescence in situ hybridization and immunocytochemistry. Mutat Res 320:315-327.

Zhang, Z; Schafer, F; Schoenfeld, H; et al. (1995) The hematotoxic effects of 6-hydroxy-trans, trans-2,4-
hexadienal, a reaction metabolite of trans, trans-muconaldehyde, in CD-1 mice. Toxicol Appl Pharmacol 132:213-
219.

Zhang, L; Rothman, N; Wang, Y; et al. (1996a) Interphase cytogenetics of workers exposed to benzene. Environ
Health Perspect 104(suppl 6):1325-1329.

Zhang, L; Rothman, N; Wang, Y; et al. (1996b) Aneuploidy of chromosomes 7, 8, and 9 detected by fluorescence
in situ hybridization in workers exposed to benzene. Cancer Res (submitted).

Zhu, H; Li, Y; Trush, MA. (1995) Differences in xenobiotic detoxifying activites between bone marrow stromal
cells from mice and rats: implications for benzene-induced hematotxicity. J Toxicol Environ Health 46:183-201.

57