Citation
Leukemogenic potential of occupational peak exposure to 1.3-Butadiene

Material Information

Title:
Leukemogenic potential of occupational peak exposure to 1.3-Butadiene
Creator:
Konowal, Anatole
Publication Date:
Language:
English
Physical Description:
xi, 137 leaves : ; 28 cm.

Subjects

Subjects / Keywords:
Butadiene -- Threshold limit values ( lcsh )
Leukemia -- Etiology ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (M.S.)--University of Colorado Denver, 2010.
Bibliography:
Includes bibliographical references (leaves 98-137).
General Note:
Department of Geography and Environmental Sciences
Statement of Responsibility:
by Anatole Konowal.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
671918274 ( OCLC )
ocn671918274

Full Text
LEUKEMOGENIC POTENTIAL OF OCCUPATIONAL
PEAK EXPOSURES TO 1,3-BUTADIENE
by
Anatole Konowal
B.S., University of Illinois at Urbana-Champaign, 1995
A thesis submitted to the
University of Colorado Denver
in partial fulfillment
of the requirements for the degree of
Master of Science
Environmental Sciences
2010


This thesis for the Master of Science
degree by
Anatole Konowal
has been approved
by
John W. Wykof*
(ol^L^/lQ
Date
Richard L. DeGrandchamp


Konowal, Anatole (M.S., Environmental Sciences)
Leukemogenic Potential of Occupational Peak Exposures to 1,3-Butadiene
Thesis directed by Associate Professor John W. Wyckoff
ABSTRACT
The United States Environmental Protection Agency (USEPA), the National
Toxicology Program (NTP) of the United States Department of Health and Human
Services (USDHHS), and the International Agency for Research on Cancer (IARC)
have all concluded that 1,3-butadiene (BD) is carcinogenic in humans. BD is a used
predominantly in the production of synthetic rubber for car and truck tires. The
general public is exposed to BD through inhalation of air mixed with auto exhaust,
gasoline fumes and cigarette smoke, however, the majority of BD exposure occurs in
workers involved in the production of synthetic rubber, resins, and plastics.
In 2002, USEPA released their latest health assessment for BD. A 1995
epidemiological study evaluating leukemia mortality in a large cohort (UAB cohort)
of North American styrene-butadiene rubber (SBR) production workers was selected
by USEPA for the derivation of an inhalation BD unit cancer risk value. This study
demonstrated a positive association, based on the cumulative exposure metric,
between working in the SBR industry and increased leukemia mortality.
An up-to-date alternative to the USEPA BD risk assessment was prepared on behalf
of the Olefins Panel of the American Chemistry Council in 2006. The alternative BD
risk assessment utilized the most recent UAB cohort data (2004) and included an


analysis of peak exposures (exposures of high intensity typically experienced for a
short duration of time) for derivation of a BD unit cancer risk value. It was
determined that peak BD exposures correlated well, and were most predictive of the
observed leukemia in the UAB cohort.
The aim of this project was to evaluate the potential contributions of peak exposures
to leukemogenesis, in support of the findings of the 2006 carcinogenic BD
assessment. A scientific literature search was conducted to identify studies providing
evidence for the predictive value of peak exposure in leukemogenesis. Findings
revealed associations between leukemia and occupational peak exposures to
formaldehyde and benzene. Additionally, intermittent scheduling, and not cumulative
dose administered, of topoisomerase-II inhibitors for the treatment of cancer, has
been implicated in development of secondary acute myeloid leukemia (s-AML).
This abstract accurately represents the content of the candidates thesis. I recommend
its publication.


DEDICATION
I dedicate this thesis to my wife, Aimee, and my daughter, Yelena, for their consistent
support, understanding, and patience while I completed not only my thesis but also
my coursework.


ACKNOWLEDGEMENT
I would like to give thanks to my committee members for their participation and
insights. A special thanks to committee member, David W. Pyatt, for providing me
with this opportunity.


TABLE OF CONTENTS
Figures.....................................................................x
Tables......................................................................xi
Chapter
1. Introduction...........................................................1
1.1 1,3-Butadiene (BD) Characteristics, Uses, and Health Effects...........1
1.2 2002 USEPA Carcinogenic BD Risk Assessment.............................2
1.3 2006 Olefins Panel Alternative BD Carcinogenic Risk Assessment.........3
1.4 UAB Cohort Data Evaluation.............................................5
1.5 Peak Exposure Definition...............................................9
1.6 Peak Exposures and the UAB Cohort.....................................10
1.7 Purpose of the Study..................................................12
2. Exposure Assessment and Selection of Exposure Metrics.................14
2.1 Exposure Assessment...................................................14
2.2 Exposure Metrics......................................................14
2.2.1 Cumulative Exposure...................................................15
2.2.2 Peak Exposure.........................................................16
2.3 OSHA Workplace Exposure Limits........................................17
3. Leukemia..............................................................19
4. Formaldehyde (FA).....................................................20
4.1 FA Characteristics, Uses, and Health Effects..........................20
4.2 FA Carcinogenicity....................................................22
4.2.1 Nasopharyngeal Cancer.................................................22
4.2.2 Lymphohematopoietic Carcinogenicity...................................23
4.3 Cytogenetic and Carcinogenic Studies on FA............................23
vii


4.3.1 Evidence for FA-Induced Leukemic Induction..........................23
4.3.2 Evidence against FA-Induced Leukemic Induction......................25
4.4 Industrial Worker Cohort Data.......................................27
4.4.1 Epidemiological Evidence for Peak Exposures and FA-Induced
Leukemia..........................................................27
4.4.2 Epidemiological Studies Demonstrating No Association between FA and
Leukemia..........................................................30
4.5 Professional Worker Cohort Data.....................................32
4.6 Peak FA Exposures and Leukemia, a Recent Meta-Analysis..............35
5. Benzene (BZ)........................................................37
5.1 BZ Characteristics and Uses.........................................37
5.2 BZ Health Effects...................................................38
5.2.1 Acute Exposure Effects..............................................38
5.2.2 Hematotoxic Effects.................................................39
5.2.3 BZ Carcinogenicity..................................................40
5.3 Animal Studies and BZ Carcinogenesis................................41
5.4 BZ Biotransformation and Toxicity...................................42
5.4.1 First Metabolic Pathway.............................................42
5.4.2 Second Metabolic Pathway............................................43
5.5 BZ Mechanisms of Action.............................................44
5.6 Cumulative BZ Exposure and Leukemia-Epidemiological Evidence........45
5.6.1 Pliofilm Cohort.....................................................46
5.6.2 NCI/CAPM Cohort.....................................................47
5.6.3 Italian Shoe Worker Cohort..........................................48
5.7 Epidemiological Evidence for Peak Exposures and
BZ-Induced Leukemia...............................................49
5.8 OSHA BZ Exposure Limits.............................................56
viii


5.8.1 BZ STEL Reduction.....................................................58
5.9 Possible Mechanism for BZ-Induced Leukemogenesis....................64
5.9.1 Intermittent BZ Exposure and the Cell Cycle............................64
6. Topoisomerase-II Inhibitors (Epipodophyllotoxins).....................67
6.1 Induction of Secondary Leukemias (s-AML) Following
Chemotherapy........................................................67
6.1.1 Alkylating Agent Chemotherapy and s-AML Induction.....................68
6.1.2 Epipodophyllotoxin Chemotherapy and s-AML Induction...................69
6.2 Epipodophyllotoxin Treatment Scheduling and s-AML Induction...........72
7. Evaluation of Other Chemicals for Peak Exposure and
Leukemogenesis......................................................82
7.1 BD Reactive Epoxide Formation and Carcinogenesis......................82
7.2 Ethylene Oxide (EtO)..................................................84
7.2.1 Characteristics, Uses, and Health Effects.............................84
7.2.2 Cytogenetic Evidence for EtO Carcinogenesis...........................85
7.2.3 Epidemiological Evidence for EtO-Induced Leukemogenesis...............85
7.3 Vinyl Chloride (VC)...................................................87
7.3.1 Characteristics, Uses, and Health Effects.............................87
7.3.2 VC Carcinogenicity (Liver Angiosarcoma and Leukemia)..................88
7.4 Chloroprene...........................................................89
7.4.1 Chloroprene Characteristics, Uses, and Health Effects.................89
7.4.2 Chloroprene Carcinogenicity (Leukemia)................................89
7.5 Isoprene..............................................................90
7.5.1 Isoprene Characteristics, Uses, and Health Effects....................90
7.5.2 Final Thoughts on Other Epoxide-Forming Chemicals Evaluated...........91
8. Conclusion............................................................92
References...................................................................98
IX


LIST OF FIGURES
Figure
4.1 Three Potential Models for the Induction of Stem Cell Leukemogenesis
Following Formaldehyde Exposure (Adapted from Zhang et al. (56),
Elsevier B.V., publisher)..........................................25
5.1 The Major Metabolic Pathways Responsible for the Generation of
Genotoxic and Myelotoxic Benzene Intermediates (Adapted
from Kim et al. (192), American Association for Cancer
Research, publisher)..............................................43
6.1 Possible Mechanisms for (1.) Cytotoxic Cell Death or (2.) s-AML Induction
Following Topoisomerase-II Inhibitor Treatment (Adapted from
Smith et al. (268), Wiley-Liss, Inc., publisher)...................72
7.1 1,3-Butadiene (BD) Reactive Metabolite Formation, (Adapted from Albertini
et al. (338), Elsevier Science, publisher)........................83
x


LIST OF TABLES
Table
1.1 Relative rates (RR) for leukemia mortality for the UAB cohort following
exposure (cumulative or peak) to 1,3-butadiene (BD)....................7
1.2 p-values for Cox regression analysis of estimated rate ratios (RR) for the
UAB cohort for the association between leukemia mortality and
1,3-butadiene (BD) exposure, for selected exposure
metrics and models.....................................................8
1.3 /^-values for Cox regression analysis of estimated rate ratios (RR) for the
UAB cohort for the association between myeloid or lymphoid
neoplasms and 1,3-butadiene (BD) exposure, for selected
exposure metrics and models............................................8
4.1 P trend values for associations between leukemia mortality and quantitative
formaldehyde exposure estimates.......................................31
5.1 Standardized mortality ratios (SMR) for leukemia in Italian shoe factory
workers following exposure (cumulative or peak)
to benzene (BZ).......................................................51
5.2 Standardized mortality ratios (SMR) for leukemia and acute non-lymphatic
leukemia in U.S (Illinois) chemical plant workers following exposure
(cumulative or peak) to benzene (BZ)..................................54
6.1 Risk of s-AML following epipodophyllotoxin therapy administered on
various schedules.....................................................79
xi


1. Introduction
1.1 1,3-Butadiene (BD) Characteristics, Uses, and Health Effects
1,3-butadiene (BD), C4H6, is a clear, colorless gas with a mild gasoline-like odor used
in the manufacturing of synthetic rubber for car and truck tires (approximately 60%
of total use) and in the formulation of plastics such as acrylics. Releases of BD occur
from auto exhaust, cigarette smoke, wood, plastic, and rubber burning, as well as
industrial sources. Occupational exposure to BD could be significant in those workers
involved in the production of synthetic rubber, resins, and plastics, and the general
public is exposed to BD daily via inhalation of urban and suburban air, particularly
from air mixed with gasoline fumes, auto exhaust, and cigarette smoke (including
second-hand smoke). Exposure to BD could occur through ingestion of contaminated
drinking water or food, but this contribution is expected to be low compared to
inhalation. In air, the half-life of BD is six hours and evaporation is likely to
eliminate most BD that may have been spilled into water or released into soil (1).
The toxicity of BD has been well characterized from toxicological studies in animals
and human epidemiological studies. Due to the physical nature of BD, far more is
known regarding inhalation exposures, and significant effects from dermal and
ingestion exposure routes have not been identified (1). Short-term, high-dose BD
exposure (10,000 ppm) in humans may cause irritation of the eyes, nose, throat,
nausea, drops in blood pressure and pulse rate, and neurological effects, including
vertigo, headaches, and blurred vision (1,2). Animal studies have demonstrated
reproductive and developmental effects, but no similar studies were located
evaluating these effects in humans (1). Non-cancer effects due to chronic exposure in
humans may lead to increased cardiovascular disease, including rheumatic and
arteriosclerotic heart diseases (1). Animal studies of chronic BD exposure have
1


reported increased respiratory and liver diseases in addition to blood and
cardiovascular diseases (1). The United States Environmental Protection Agency
(USEPA) has classified BD as carcinogenic in humans, as have the National
Toxicology Program (NTP) of the United States Department of Health and Human
Services (USDHHS) and the International Agency for Research on Cancer (IARC)
(3-5). Occupational epidemiological studies have reported increased risk from
bladder, stomach, respiratory, and lymphohematopoietic malignancies, particularly
leukemia, from inhalation BD exposure (1, 2, 6, 7).
1.2 2002 USEPA Carcinogenic BD Risk Assessment
In 2002, USEPA completed its most recent health assessment for 1,3-butadiene (8).
Results from USEPAs health assessment for BD, described on USEPAs Integrated
Risk Information System (IRIS), include derivation of an inhalation unit risk value
for BD as part of the carcinogenic toxicity assessment (2). A 1995 epidemiological
study on leukemia mortality among North American styrene-butadiene rubber (SBR)
production workers by Delzell et al. at the University of Alabama at Birmingham
(UAB cohort), provided cumulative BD exposure estimates along with dose-response
modeling for the USEPAs cancer dose-response assessment (9). In their study,
Delzell et al. demonstrated a positive association, based on the cumulative exposure
metric, between work in the SBR industry and leukemia (6).
USEPA reviewed several epidemiological studies evaluating carcinogenic risk from
BD exposure, and deemed the epidemiological data from the Delzell et al. UAB
cohort study (1995) to be of high-quality and containing the strongest set of data
available for their assessment. USEPA utilized the BD exposure estimates and dose-
modeling (7), calculated by Delzell et al. (1995), for each job in each work area for
2


each study year to determine a cumulative exposure estimate for each worker. With a
maximum follow-up period of 49 years (January 1 1943 to January 1 1992), the study
was deemed to have sufficient power by USEPA to identify leukemia mortality
excesses in the UAB cohort (15,649 male SBR production workers employed for at
least one year at eight plants located in the United States and Canada). In 1998,
Health Canada obtained the UAB cohort data from Delzell et al. (1995) and
conducted their own analyses, which according to USEPA were similar to the
analyses conducted by Delzell et al. (Poisson regression analysis of linear, log-linear,
square root, and power mathematical models) (10). Health Canada, through Poisson
regression analysis of a linear extrapolation mathematical model, determined the unit
cancer risk value for BD to be 0.08 parts per million'1 (ppm1). Regression covariates
identified for the Health Canada (1998)/USEPA (2002) BD risk assessment that were
adjusted for their effects included: age, calendar period, years since hire, race, and
cumulative styrene exposure.
1.3 2006 Olefins Panel Alternative BD Carcinogenic Risk Assessment
In March 2006, a carcinogenic risk assessment for inhalation exposure to BD was
prepared on behalf of the Olefins Panel of the American Chemistry Council to serve
as an up-to-date alternative to the USEPA risk assessment (2002) (11). The authors
used an update of the Delzell et al. (1995) study to derive a unit cancer risk value for
BD, which they calculated to be approximately 900 times lower than the Health
Canada/USEPA unit cancer risk value. The unit cancer risk value (0.000089 ppm-1)
derived by Albertini et al. utilized the most recent UAB cohort data (2004) and
included seven additional years of follow-up as well as revised exposure estimates
that were developed in 2000 (12, 13). The 2000 revised exposure estimates for BD
were approximately one order of magnitude higher than the exposure estimates
3


developed in 1995 (14). The 2004 UAB cohort was comprised of 16,579 SBR
production workers, and consisted of 81 decedents with leukemia, as compared to 58
decedents with leukemia in the 1995 UAB cohort data set (15). The 2006 BD risk
assessment by Albertini et al. identified age and the cumulative number of BD peaks
as statistically significant covariates, and both covariates were utilized in Poisson
regression for linear extrapolation of cancer potency. It should be noted that linear
extrapolation is identified by USEPA as their default approach for identification of
direct acting genotoxic carcinogens, of which BD is assumed to be, as evidenced by
rodent studies describing reactive BD intermediate-DNA interactions, and
particularly interactions with the BD intermediate diepoxybutane (DEB).
Following the recommendations of the USEPA Scientific Advisory Board (SAB)
(1998), in response to the release of USEPAs 1998 BD risk assessment, to utilize the
most recent data and give consideration to peak BD exposures, (16) Albertini et al.
derived their BD unit cancer risk value in a relatively similar manner to that of Health
Canada/USEPA. Both assessments relied on the UAB cohort data set, utilized
relative risk as risk measure, used ppm-years as dose measure for cumulative
exposure, and derived a BD unit cancer risk value through Poisson regression
analysis of a linear mathematical model. Albertini et al. report in their 2006
assessment that small to moderate differences in the derived BD unit cancer risk
values may be attributed to use of updated exposure and response data for the UAB
cohort (2004 vs. 1995 UAB cohort), endpoint differences (leukemia death was
primary or contributing cause vs. leukemia was primary cause only), conversion from
relative risk to extra risk (use of leukemia mortality rates for 70-year lifetime vs.
leukemia incidence rates for an 85-year lifetime), and use of an additional adjustment
factor (no uncertainty factor vs. an uncertainty factor of 2 to adjust for increased
4


sensitivity in females). Albertini et al. state that the most significant difference in the
BD unit cancer risk value between the two assessments is due to the use of
cumulative number of peak BD exposures as a co variate. Inclusion of peak BD
exposures, defined as any exposure greater than 100 ppm in concentration for any
duration of time (14), and as noted by Albertini et al., appears to be the characteristic
most correlated to and predictive of observed leukemia.
1.4 UAB Cohort Data Evaluation
Graff et al. (2005) evaluated the 2004 UAB cohort, via Poisson regression analysis,
and presented data that demonstrated both a positive association and a dose-response
relation, as the total number of BD peaks increased, for peak BD exposures and
leukemia, when adjusted for age and years since hire (17). While a positive
association for cumulative exposure and leukemia was also observed when adjusted
for age and years since hire, the dose-response trend appears to be weaker, as
compared to BD peaks (Table 1.1). When the data was adjusted for potential
confounders, the positive association for BD peaks and leukemia became inconsistent
and the dose-response trend was weakened. Cumulative exposure and leukemia
demonstrated a weakened positive association with adjustment for potential
confounders, but still appears somewhat stronger than the association between BD
peaks and leukemia. According to Albertini et al., demonstration of association
proved to be difficult due to high levels of correlation among the potential
confounders, but Graff et al. did demonstrate that higher intensity BD exposures may
be more pertinent to leukemia mortality than lower intensity BD exposures (11).
The 2004 update of the UAB cohort, prepared by Cheng et al. (2007), was evaluated
via Cox regression analysis to determine if any associations existed between leukemia
5


and the following exposure metrics: cumulative BD exposure, BD peak exposures,
and BD average intensity (15). Cheng at al. demonstrated significant associations
between leukemia for all three exposure metrics when data was adjusted only for age:
however, only BD peak exposures showed a significant association when the data
was adjusted for multiple covariates (Table 1.2). Cheng et al. further analyzed the
updated UAB cohort for potential association between leukemia mortality (combining
lymphoid and myeloid subtypes) and the same three exposure metrics (Table 1.3).
The authors showed a significant association between myeloid neoplasms and BD
peak exposures. In contrast, a significant association for cumulative BD exposure
was observed for lymphoid neoplasms. Both of these significant associations were
reported when the data was age-adjusted only. Once adjustments were made for
multiple covariates, the significant associations disappeared. As with the Graff et al.
(2005) study, a high level of correlation among the covariates may have reduced the
association between peak BD exposures and myeloid neoplasms.
6


Table 1.1 Relative rates (RR) for leukemia mortality for the UAB cohort following exposure
(cumulative or peak) to 1,3-butadiene (BD)d
Exposure metric # of leukemia Single category analysis Multiple category analysis
decedents/81a RR (95% CI)b RR (95% CI)c
Cumulative exposure (BD ppm-years) Model lc
0 10 1.0 1.0
>0-33.6 17 1.4 (0.7-3.1) 1.4 (0.5-3.9)
33.7-184.6 18 1.2 (0.6-2.7) 0.9 (0.3-2.6)
184.7-424.9 18 2.9(1.4-6.4) 2.1 (0.7-6.2)
>424.9 18 3.7 (1.7-8.0) 3.0 (1.0-9.2)
Peak exposure (# BD total peaks > 100 ppm) Model 2C
0 10 1.0 1.0
>0-254.6 17 2.1 (0.9-4.5) 1.3 (0.4-3.8)
254.7-965.3 18 2.7 (1.2-5.8) 1.1 (0.4-3.5)
965.4-3136.4 18 3.3 (1.5-7.1) 1.2 (0.4-3.9)
>3136.4 18 4.9 (2.2-10.6) 1.6 (0.5-5.6)
a81 decedents with leukemia from six of eight North American study plants (UAB cohort); 16,579
synthetic rubber industry workers followed from 1943 to 1998.
bAdjusted for age and years since hire; included one category of interest (BD ppm-years or # BD total
peaks > 100 ppm); other categories evaluated but not shown here: styrene and
dimethyldithiocarbamate (DMDTC) exposures; Poisson regression analysis used to examine leukemia
RRs.
c Adjusted for potential confounders in addition to age and years since hire; Model 1 included BD
ppm-years, styrene ppm-years, and DMDTC mg-years/cm; Model 2 included BD peaks, styrene peaks,
and DMDTC mg-years/cm.
dAdapted from Graff et al. (2005),Table 3 (17).
7


Table 1.2 p-values for Cox regression analysis of estimated rate ratios (RR) for the
UAB cohort3 for the association between leukemia mortality and 1,3-butadiene (BD)
exposure, for selected exposure metrics and models'1
Exposure metric Model lb (p-value) Model 2C (p-value)
Cumulative exposure 0.01 0.16
(BD ppm-years)
BD peak exposure <0.01 <0.01
(# BD peaks >100 ppm)
BD average intensity 0.02 0.17
(PPm)________________________________________________________________________
a81 decedents with leukemia from six of eight North American study plants (UAB
cohort); 16,579 synthetic rubber industry workers followed from 1944 to 1998.
bModel 1, adjusted for age only.
c Model 2, adjusted for multiple covariates: age, year of birth, race, DMDTC, years
since hire, and plant.
dAdapted from Cheng et al. (2007), Table 2(15).
Table 1.3 p- values for Cox regression analysis of estimated rate ratios (RR) for the
UAB cohort3 for the association between myeloid or lymphoid neoplasms and 1,3-
butadiene (BD) exposure, for selected exposure metrics and models'1
Exposure metric Model lb (p-value) Model 2C (p-value)
Myeloid neoplasms
Cumulative exposure 0.39
(BD ppm-years)
BD peak exposure 0.04
(# BD peaks >100 ppm)
BD average intensity 0.10
(ppm)
0.64
0.21
0.12
8


Table 1.3 (Cont.)
Exposure metric Model lb (/?-value) Model 2C (/7-value)
Lymphoid neoplasms
Cumulative exposure 0.02 0.10
(BD ppm-years)
BD peak exposure 0.28 0.36
(# BD peaks >100 ppm)
BD average intensity 0.65 0.85
(PPm)_______________________________________________________________________
a81 decedents with leukemia from six of eight North American study plants (UAB
cohort); 16,579 synthetic rubber industry workers followed from 1944 to 1998.
bModel 1, adjusted for age only.
c Model 2, adjusted for multiple covariates: age, year of birth, race, DMDTC, years
since hire, and plant.
dAdapted from Cheng et al. (2007), Table 5(15).
1.5 Peak Exposure Definition
In the next section, Exposure Assessment and Selection of Exposure Metrics, the
peak exposure metric and difficulties defining peaks will be addressed further. Below
is the definition for a peak exposure, as reported by the authors responsible for
evaluation of BD exposure estimations for the UAB cohort. According to Albertini et
al. in their 2006 BD risk assessment, workplace exposures to BD were assumed to be
much higher during World War II and the subsequent post-war expansion period than
from current occupational BD exposures. This position was also adopted by the
Working Group of IARC in their 2008 evaluation of BD. 368,172 person-years were
accrued in the 1995 UAB cohort with a follow-up period of 49 years, beginning in
1943, and although no definitive BD exposure-response data is available, the IARC
Working Group states that it is quite likely that many of the SBR production workers
experienced short but intense exposures to BD during this time. As described by
IARC, these short but intense exposures to BD resemble exposures defined as peak
exposures by Delzell et al. in their 2001 BD exposure estimation revision paper.
9


Sielken et al. (2007) define peak exposures, based on criteria set forth by Delzell et al.
(2001) and Macaluso et al. (2001) as follows: A BD peak occurred whenever the
BD exposure concentration during a day went from below 100 ppm to above 100 ppm
and then returned to below lOOppm. Exposure to BD levels above 100 ppm was
rather common, in large part intermittent, frequently of short duration (several
seconds to several minutes), and not uncommonly to levels of several hundred ppm
(14).
1.6 Peak Exposures and the UAB Cohort
Within the SBR production facilities, polymerization workers, laboratory workers,
and maintenance laborers were all believed to have experienced highly intense and
short duration exposures of BD during the early years of the UAB cohort data set.
Albertini et al. write that many of these workers had 8-hour time-weighted average
(TWA) exposures exceeding the OSHA permissible exposure limit (PEL) of 1 ppm.
For the overall UAB cohort (1995 data), the observed standardized mortality ratio
(SMR) for all cancers was 93, with a 95 % confidence interval (Cl) = 87-99. The
SMR for leukemia was elevated at 131 (95% 0=97-174), but not statistically
significant. When Delzell et al. determined the SMR for the ever-hourly workers, a
statistically significant excess SMR=143 (95% 0=104-191) was observed. Ever
hourly workers who had >10 years of employment and > 20 years since hire, had an
elevated SMR of 224 (95% 0=149-323). Based on job types, the greatest excesses
of leukemia mortality were observed in the polymerization workers (SMR=251, 95%
0=140-414), laboratory workers (SMR=431, 95% 0=207-793), and maintenance
laborers (SMR=265, 95% 0=141-323), workers likely subjected to peak BD
exposures.
10


The 2006 carcinogenic BD assessment by Albertini et al. essentially utilized and
analyzed the same UAB cohort data set as did USEPA for their BD risk assessment
except for one major difference, the inclusion of cumulative number of BD peaks.
Taking into consideration the recommendations from USEPAs SAB following their
1998 BD risk assessment, Albertini et al. incorporated BD peaks into their analyses.
Based on these additional analyses, Albertini et al. demonstrated that leukemia
mortality correlated well with peak BD exposures. Albertini et al. calculated that
excess relative risks due to leukemia are reduced substantially if an adjustment is
made to factor out the cumulative number of BD exposures above 100 pm (peak
exposures).
Selection of an appropriate exposure metric for leukemogenic assessment of
occupational exposures is dependent on understanding the hazards mode or
mechanism of action; however, often this process is unknown or poorly understood
for occupational diseases, as is the case with BD-induced leukemogenesis (18, 19).
USEPA chose to derive their BD unit cancer risk value without the inclusion of peaks
as a significant covariate despite the SABs 1998 recommendations, which took into
consideration the realistic assumption that the publics exposure to BD will likely
never draw near doses associated with peak exposures, particularly doses experienced
by BD monomer and SBR production workers post-World War II. It is possible that
different exposure metrics may group populations of workers differently, leading to
possible misclassification of exposures or study subjects. Therefore, epidemiological
studies should assess multiple exposure metrics to increase the probability of
determining which one is most appropriate (18). Highly concentrated exposures of
short duration can create increased dose rates on target tissues and organs, thus
potentially altering metabolism, overloading protective and repair mechanisms, and
11


elevating tissue responses (20, 21). Peak exposures may also result in different and
possibly more harmful tissue and organ effects than the same administered dose
might, if given with less intensity over a longer time period (21). While there is often
difficulty in defining exactly what a peak exposure is, both the Occupational Health
and Safety Administration (OSHA) and the American Conference of Governmental
Industrial Hygienists (ACGIH), have recognized the potential dangers of peak
exposures. This recognition, based on support from animal and human
epidemiological studies, prompted development of workplace limits by OSHA for
short-term, high intensity exposures known as short-term exposure limits (STEL),
which will be discussed in subsequent sections.
1.7 Purpose of the Study
Albertini et al. (2006), following specific recommendations of USEPAs SAB, have
clearly demonstrated the critical role peaks play in their BD risk assessment. The
Albertini et al. risk assessment used essentially the same data set and analyses for
derivation of the BD inhalation unit cancer risk value as USEPA, but included an
additional analysis for peaks. The purpose of this study was to evaluate the potential
contributions of occupational peak exposures to leukemogenesis in support of the
findings (from the 2006 Albertini et al. BD risk assessment) that peak BD exposures
may be more predictive of increased leukemia mortality than other exposure metrics
commonly utilized in risk assessment. A scientific literature search was conducted to
identify studies providing epidemiological and toxicological evidence for the
predictive value of peak exposure in leukemogenesis. Although the scientific
literature regarding occupational peak exposures to known and suspected carcinogens
is sparse, some data exists that is relevant and will herein be reviewed. Findings
revealed associations between leukemia and occupational peak exposures to
12


formaldehyde and benzene. Additionally, intermittent scheduling, which resembles
the peak exposure metric (high intensity, short-term bursts), of topoisomerase-II
inhibitors (specifically the epipodophyllotoxin class) for the treatment of cancer, has
been implicated in development of secondary acute myeloid leukemia (s-AML), and
was therefore also included as part of this study.
13


2. Exposure Assessment and Selection of Exposure Metrics
2.1 Exposure Assessment
In occupational epidemiology, exposure may be defined as contact between an
individual and a substance, by one of three main routes: inhalation, dermal, and oral
ingestion (25). Before injury to target tissue or a target organ may occur, a critical
exposure concentration must be attained (26). Exposure assessment, an integral part
of the epidemiological study, involves understanding the pathways of exposure, the
magnitude or intensity of the exposure, along with the time course of the exposure
(25). These must be adequately characterized prior to understanding the risk derived
from the exposure-response relationship, which oftentimes results in subtle effects
(25). By using the most appropriate exposure metric, misclassification of exposure,
which has been shown to weaken a studys ability to find an association between a
substance and a disease outcome, will hopefully be limited (27). According to
Semple et al. (2005), choosing a suitable metric to quantify the exposure should be
related to the biological mechanism of action for the substance of interest, particularly
if a strong hypothesis for the mechanism of action exists (25, 26). If no strong
mechanistic data is available, as is the case with BD, the use of exploratory or
alternative exposure metrics should be considered for exposure assessment (28).
2.2 Exposure Metrics
The four most common exposure metrics typically employed in occupational
epidemiology studies: 1. cumulative exposure; 2. average exposure or intensity; 3.
duration of exposure; and 4. peak exposure. Cumulative exposure, the product of
average exposure intensity (concentration) for a given job and exposure duration is
usually expressed in units of ppm-years or some derivative. Cumulative exposure has
been demonstrated to correlate well with disease risk for many exposure-response
14


relationships (29), and is therefore the most common exposure metric utilized (25).
Average exposure, the arithmetic or geometric mean of past exposure concentrations,
is expressed in units of ppm (26). Duration of exposure, a measure of time spanning
from time of first exposure to the onset of disease, is expressed in units of years (26).
As previously discussed, peak exposure is a measure of short-term, high intensity
exposure, and may be expressed in various ways, including frequency (number of
peaks) or intensity of peaks (ppm).
2.2.1 Cumulative Exposure
For acute health effects, an exposure metric related to a short duration of time, such
as peak exposure or simply concentration, should be utilized, while investigation of a
chronic disease would likely involve assessment of the exposure over a longer
duration of time, in which case, cumulative exposure may be the most appropriate
metric (25). In most epidemiological studies, exposures are evaluated using several
different metrics. Most models of cancer epidemiology assume there is a
proportional (linear) relationship between cumulative exposure and risk (26),
meaning cancer risk increases as the cumulative dose increases. In this theoretical
model, each increase in dose will result in an increased risk by a constant quantity
(26). However, if the exposure-response relationship is not a cumulative process, the
association between disease and risk may not be identified (26).
Use of the cumulative exposure metric, frequently gives little or no information
regarding either the intensity or duration of the exposure (25). For example, a subject
may have been exposed at 300 ppm average intensity for two years and suffered ill-
health effects, while another subject may have been exposed to 30 ppm for 20 years
and not suffered any ill-health effects. Both subjects experienced the same
15


cumulative exposure of 600 ppm-years, but only the subject who experienced the
higher dose exposure over a shorter duration of time (300 ppm for two years) fell ill.
Checkoway and Rice (1992), demonstrated how utilization of the cumulative
exposure metric for a case-control study on silicosis (a disease considered linked to
cumulative exposures), did not accurately identify the silicosis risk (30). In their
study, Checkoway and Rice observed that relative risk estimates increased from 1.18
(95% 0=0.97-1.45) to 1.51 (95% 0=1.05-2.17) when the peak exposure metric was
utilized instead of cumulative exposure. According to Semple et al. (2005), who also
examined this data set, this stronger association seen with peaks was likely due to
biologically relevant factors related to excursions above a certain exposure level,
rather than the cumulative dose of the substance. As discussed earlier, the exposure
metric should be chosen based on a biological mechanism of action, if that is known.
Evaluation of cumulative exposure alone did not demonstrate a significant excess of
silicosis risk, as a nonlinear rate of bodily insult (such that may be observed with high
dose short duration exposures) seems to occur with silicosis, thus making peaks a
better predictor of silicosis risk (30).
2.2.2 Peak Exposure
Another common choice for describing exposure is peaks. However, use of this
metric has proven difficult because of the variability observed between studies in
terms of how peaks are defined (ie. frequency or intensity of peaks). Additionally,
peak exposure data is often not available for epidemiological cohort studies
examining past exposures (25). Another difficulty assessing peak exposures is
determining at what exposure level does a peak exposure begin (ie. what constitutes a
peak exposure). Particularly, at what level might a peak become biologically or
toxicologically relevant (20, 25). Semple et al. (2005) state that as there is often
16


difficulty determining where threshold-level health effects begin to occur, thus
making it difficult to determine/estimate at which point an exposure becomes a peak
exposure. Checkoway and Rice (1992), in an effort to clarify what constitutes a peak
exposure provide two definitions of peak exposures: absolute and relative (30). An
absolute peak is an exposure experienced above an enforced exposure limit (such as
PEL), and a relative peak is defined as the highest exposure intensity experienced
over a specified time period. Semple et al. (2005) also state that the duration of a
peak exposure would likely impact any health effects observed.
Another challenge using the peak exposure metric, is the high level of correlation
often observed between cumulative exposures and peak exposures (26). Any peak
exposure experienced by a worker will be incorporated into the cumulative exposure
calculation, such that jobs with high peak exposures are likely to have a high
cumulative exposure as well. With regard to the high correlation between peaks and
cumulative exposure, it is also important to consider the likelihood that peak
exposures and cumulative exposures may induce disease through differing biological
mechanisms, and this may provide difficulty in actually determining which
mechanism may be responsible for increased risk. (26).
2.3 OSH A Workplace Exposure Limits
The Occupational Health and Safety Administration (OSHA) has established legal
limits for the amount of exposure workers may experience in a typical 8-hour
workday. The permissible exposure limit (PEL) is based on an 8-hour time weighted
average (TWA), which is the average exposure over an 8-hour time period. Workers
may experience high exposures during the 8-hour work shift, but their daily
cumulative exposure average must be lower than the PEL at the end of the workday.
17


A worker may experience peak exposures throughout the 8-hour work shift, but the
average exposure should not exceed the PEL. However, if the biological mechanism
of action occurs as a result of peak exposures, then a PEL may not protect the
workers health. As previously discussed, OSH A has also established short-term
exposure limits (STELs), which are designed to protect workers from acute health
effects.
Regulatory exposure limits such as the PEL and STEL are intended to protect
workers health; however, if these workers fall ill despite following exposure
regulations, occupational epidemiological studies could be utilized to provide
feedback regarding the effectiveness of these regulatory limits.
18


3. Leukemia
Leukemia may be defined, as a cancer of the blood or the bone marrow, in which
there is an abnormal or uncontrolled increase of blood cells. There are four main
subtypes of leukemia that are classified according to disease progression or cell of
origin: acute myeloid leukemia (AML), acute lymphocytic leukemia (ALL), chronic
myeloid leukemia (CML), and chronic lymphocytic leukemia (CLL). In acute forms
of leukemia, immature cells, known as blast cells (lymphoblastic or myeloblastic)
undergo rapid expansion, leading to decreased production of normal cells. In turn,
this decrease in normal hematopoiesis hinders the bodys ability to fight infections
and reduces red blood cell and platelet counts. Without treatment, acute leukemias
progress rapidly and are almost invariably fatal. In contrast, chronic leukemias
involve an uncontrolled increase of abnormal mature blood and typically progress
slowly over a period of several months to years. The incidence of most types of
leukemia increases dramatically with age. However, some types of leukemia, such as
ALL, occur more frequently in children. Regarding the cell of origin, leukemias can
occur in either myeloid or lymphoid cell lineages. Myeloid progenitor cells develop
into myeloblasts, that in turn may differentiate into red blood cells, platelets, and
other white blood cells, including macrophages, neutrophils, basophils, and
eosinophils. Lymphoid progenitor cells mature into lymphoblasts that further
differentiate into B and T cell lymphocytes or natural killer (NK) cells (31). Known
or suspected occupational leukemogens (benzene and formaldehyde) as well as
exposure to ionizing radiation are most often associated with AML, while BD has
been associated with CLL in addition to AML (32). All epidemiological studies
presented here classify leukemia subtypes utilizing the World Health Organizations
(WHO) International Statistical Classification of Diseases (ICD), although there have
been significant revisions of this code through time.
19


4. Formaldehyde (FA)
4.1 FA Characteristics, Uses, and Health Effects
Formaldehyde, CH2O, also known as methylene oxide or methanal, is a highly
reactive gas at room temperature. It has a strong and distinct odor and is derived
from both natural and anthropogenic sources. Formaldehyde is found in the
atmosphere, as a constituent of smog, or as a product of automobile exhaust, tobacco
smoke, and wood burning. It may be used as a food preservative and can be found in
household cleaners, cosmetics, and glues or adhesives. Recently, measurable levels
of formaldehyde were found in temporary housing trailers provided by the Federal
Emergency Management Agency (FEMA) for victims of Hurricane Katrina, as it off-
gassed from the composite wood products utilized in trailer construction (33).
Additionally, formaldehyde is produced naturally in the human body, where it is
metabolized routinely by all living cells and acts as an intermediate in the one carbon
pool, as the carbon atom is incorporated into macromolecules (34). Endogenous
concentrations of formaldehyde typically seen in human blood are 2-3 ppm (35).
Because of endogenous formaldehyde production from one carbon metabolism, most
mammals have developed highly efficient enzymatic systems for handling
formaldehyde. In humans, formaldehyde dehydrogenase typically metabolizes ~1.5
ounces of formaldehyde daily (35). Formaldehyde does not accumulate in the
environment or bioaccumulate in plants and animals, and is rapidly broken down in
both water and air (3).
Industrially, formaldehyde is used in the production of urea-formaldehyde resins,
plywood, paper, and fertilizer. Urea-formaldehyde resins, utilized in the production
of housing materials, including particle board, foam insulation, furniture, and carpet,
comprise 23% of the U.S. markets industrial formaldehyde use (3). Formaldehyde is
20


also important in the medical field, where it is utilized for tissue specimen
preservation, as a component in embalming fluids and as an anti-microbial agent.
Due to its expansive use in industry and production, OSHA estimated in 1995 that
2.1 million workers in the U.S. were occupationally exposed to formaldehyde through
inhalation or dermal contact. The majority of individuals exposed are industrial
workers, with a smaller subset including medical lab technicians, morticians,
embalmers, anatomists, and pathologists. OSHA has established regulations for
occupational formaldehyde exposures: a permissible exposure limit (PEL) of 0.75
ppm as an 8-hour time-weighted average (TWA) and a short-term exposure limit
(STEL) of 2 ppm (36). The general population is likely exposed to formaldehyde
through inhalation of indoor air or environmental tobacco smoke. Formaldehyde
releases from recently installed or new building materials or furnishings account for
much of the background exposure (~ 10-20 ppb) in the general population (3).
Human studies have long demonstrated that acute inhalation or dermal exposure to
formaldehyde leads to irritation of the eyes, nose, throat, and skin, respectively (37-
40). Study subjects acutely exposed to formaldehyde at concentrations of 2-3 ppm
(41-49), reported nose and eye irritation beginning just two minutes after exposure in
(46). Formaldehyde, a skin sensitizer, is the eighth most common allergen (50), and
may cause allergic contact dermatitis through the T-cell mediated Type IV-
hypersensitivity reaction (22). Aqueous solutions containing 1-2% formaldehyde
have also been shown to induce positive allergic responses via patch testing in
sensitized individuals with existing skin problems (51-54). According to ATSDR, a
single, high-level exposure to formaldehyde may induce headaches, malaise, and
transient reductions in memory capability, equilibrium, or dexterity (3). Ingestion of
21


formaldehyde may result in severe stomach or esophageal damage (perforation and
ulceration), coinciding with nausea, diarrhea, vomiting, and abdominal pain (3).
4.2 FA Carcinogenicity
4.2.1 Nasopharyngeal Cancer
In 2006, IARC promoted formaldehyde from a group 2A chemical (probable
carcinogen) to a group 1 chemical (carcinogenic to humans) (55). Prior to
reclassification, formaldehyde was considered as only a probable human carcinogen
based on animal data and limited data on induced human carcinogenicity (56). The
new group 1 classification was based on sufficient epidemiological evidence that
formaldehyde causes nasopharyngeal cancer in humans, and was derived from
identified excesses of nasopharyngeal cancer in six cohort studies (57-62) and seven
case-control studies (63-69). Of the seven case-control studies, five reported elevated
rates of nasopharyngeal cancer following formaldehyde exposure, and a statistically
significant excess of nasopharyngeal cancer was reported in the Hauptmann et al.
(2004) National Cancer Institute (NCI) cohort study, consisting of 25, 619 industrial
workers (57). Significant exposure-response patterns were observed for both peak
and cumulative exposure variables, while no excesses were observed for average
intensity or duration of formaldehyde exposure. The 2006 IARC conclusion that
formaldehyde causes nasopharyngeal cancer has been questioned in at least two
papers (70, 71). These authors argue that promotion of formaldehyde from a Group
2a chemical to a Group 1 chemical was premature, and that a re-evaluation of the NCI
cohort study, the largest and most revealing industrial cohort study, should be
undertaken (56). Formaldehyde exposure and induction of nasopharyngeal cancer is
only of peripheral importance to this project, and was therefore only covered
22


superficially and in a basic manner, as the story is much more complex than described
here.
4.2.2 Lymphohematopoietic Carcinogenicity
Prior to their 2006 evaluation, IARC did not address the potential risk associated risk
between occupational formaldehyde exposure and any lymphatic or hematopoietic
cancers (72) due to a lack of consistency in industrial worker cohort studies.
Although the industrial worker cohorts did not clearly demonstrate a positive
association between formaldehyde exposure and lymphohematopoietic malignancies
from both the lymphoid (Hodgkin lymphoma, non-Hodgkin lymphoma, multiple
myeloma, or lymphatic leukemia) and myeloid cell lineages (myeloid leukemia,
monocytic leukemia, polycythemia vera, or myelofibrosis), such an association was
occasionally reported in studies of anatomists, pathologists, funeral workers, and
embalmers. Accordingly, in 2006, IARC re-examined the leukemogenic potential of
formaldehyde and determined that there was strong but not sufficient evidence for a
causal association between leukemia and occupational exposure to formaldehyde
(55). IARC did not consider the evidence sufficient due to a lack of biological
plausibility for how exogenous formaldehyde exposure could be acting on distant
sites like the bone marrow and inducing leukemia.
4.3 Cytogenetic and Carcinogenic Studies on FA
4.3.1 Evidence for FA-Induced Leukemic Induction
While epidemiological studies show a potential association for leukemia and
formaldehyde exposure, particularly for myeloid leukemia, the biological plausibility
of formaldehyde-induced leukemia remains inconclusive. Formaldehyde has been
shown to be genotoxic in both humans and animals, and is therefore capable of
23


causing damage to DNA and/or chromosomes (56). Exposure to formaldehyde
induces DNA-protein cross-links (DPCs), DNA adduct formation, sister chromatid
exchanges (SCEs), chromosomal aberrations (CA), and micronuclei (MN) (3, 22, 56).
Additionally, cytogenetic studies have inconsistently demonstrated genotoxic damage
in lymphocytes of exposed workers (73, 74), but the relevance of this data is unclear.
Data supporting a disruption in hematopoiesis (ie. a decrease in peripheral blood cell
counts), is limited (75). Human leukemogens are known to reduce red blood cell,
white blood cell, and platelet counts at high concentrations in both humans and
experimental animals (76, 77). Formaldehyde exposure has been consistently
negative in all long-term bioassays for hematopoietic toxicity. At issue is the
potential mechanism(s) that allow for highly reactive formaldehyde to either reach the
bone marrow or damage circulating hematopoietic stem and progenitor cells, known
targets of chemical leukemogenesis.
Recently, Zhang et al. observed significantly elevated leukemia-specific
chromosomal changes in myeloid progenitor cells (monosomy 7 and trisomy 8) and
significant decreases in peripheral blood cell counts of formaldehyde exposed
workers in China, thus providing some biological support for leukemogenic potential
of formaldehyde via damage to hematopoietic stem/early progenitor cells in either the
bone marrow or in circulation (75).
Zhang et al. (2010) proposed three potential mechanisms for formaldehyde-induced
leukemogenesis: 1. formaldehyde acts directly on the bone marrow, affecting early
hematopoietic stem cells/progenitor cells; 2. early hematopoietic stem
cells/progenitor cells in peripheral circulation are acted upon by formaldehyde then
travel back to the bone marrow where leukemic transformation occurs; 3. early
24


pluripotent stem cells are acted upon by formaldehyde in the nasal/oral passages then
travel back to the bone marrow where leukemic transformation occurs (Figure 4.1)
(75). Zhang et al. state that, based on their findings and the findings of earlier
cytogenetic damage to the bone marrow and blood hematopoietic stem
cells/progenitor cells, formaldehyde-induced leukemia occurs either in the bone
marrow and/or in peripheral blood (73, 74, 78-80)
1.
Xenobiolic
l
Nasal/oral passages and lung
| Blood
Bone Marrow
DNA Binding
N
Mutation
Macromoiecuiar Binding
I
Toxicity
Initiated Stem Celt
i
i
Leukemia
Formaldehyde
1
Nasal/oral passages and lung
i
Blood /circulating hematopoietic stem cell)
DNA Binding
Mutation
Macromolecutef Binding
Toxicity
initiated Stem Cel
Bone Marrow
leukemia
3.
Formaldehyde
Nasal'oral passages (primitive oturbotent stem coin
DNA Binding
Mutation
Macrcmolecular Binding
i
Toxioty
initialed Stem Cell
| Blood
Bone Marrow
l
i
Leukemia
Figure 4.1 Three Potential Models for the Induction of Stem Cell Leukemogenesis Following
Formaldehyde Exposure (Adapted from Zhang et al. (56), Elsevier B.V., publisher).
4.3.2 Evidence against FA-Induced Leukemic Induction
A very recent study by Lu et al. (2010) examined DNA adduct formation and
distribution, critical steps necessary for initiation of carcinogenesis, in rats following
25


inhaled formaldehyde exposure (81). Study findings demonstrated that formaldehyde
induced DNA adduct formation occurred only at the site of contact, the respiratory
nasal epithelium, and not at any distant sites. These findings are consistent with the
genotoxic and cytotoxic effects observed in formaldehyde-induced nasal
carcinogenicity observed in rats, according to the authors, but do not lend support for
a leukemogenic mechanism for formaldehyde. Lu et al. did not find any
formaldehyde-induced DNA adducts in either the bone marrow or white blood cell
samples. These findings are also consistent with other similar studies demonstrating
that formaldehyde is metabolized too rapidly for it to travel out of the nasal tissue and
to distant sites such as the bone marrow (82-86). The authors do note, however, that
there are key anatomical and physiological differences between the two species that
certainly might affect mechanisms of formaldehyde-induced leukemic transformation.
Although Zhang et al. have published data in support of biological plausibility for
leukemia induction following occupational formaldehyde exposure, there have been
at least 10 chronic carcinogenicity animal studies (87-97) in addition to the recent
findings of Lu et al. (2010), that have reported no association between formaldehyde
exposure (involving both ingestion and inhalation exposure) and development of
neoplasms, including lymphohematopoietic malignancies, at any site other than the
site of administration (98). Additionally, several long-term carcinogenicity animal
studies (again involving both ingestion and inhalation exposure) in which animals
were given high doses of formaldehyde also reported observing no negative
hematological events, including induction of hematopoietic malignancies (87, 91, 92,
94, 99-101).
26


The study results from Zhang et al. (2010) need to be replicated, as evidence
demonstrating a lack of bone marrow toxicity following formaldehyde exposure
suggests a lack of biological plausibility for formaldehyde leukemogenesis. As
reported by IARC, known human leukemogens, such as benzene and DNA
topoisomerase inhibitors, elicit direct toxicity on bone marrow as well as induce
chromosomal aberrations in peripheral lymphocytes (22). Formaldehyde, if capable
of inducing leukemogenesis, must be relying on a differing mechanism of action than
known leukemogens, which may certainly be possible, given that BD is a probable
human leukemogen (based on epidemiologic evidence) not known to induce bone
marrow toxicity (22).
4.4 Industrial Worker Cohort Data
4.4.1 Epidemiological Evidence for Peak Exposures and FA-Induced Leukemia
The decision by IARC to reconsider the leukemogenic potential (from low or no risk
of lymphohematopoietic cancers, in the 1995 monograph, to strong but not sufficient
evidence for leukemogenicity, in the 2006 monograph) of formaldehyde was, in part,
supported by the NCI cohort study results from the Hauptmann et al. 2003 paper, in
which formaldehyde exposure was characterized by several exposure metrics (peak
exposure, average exposure intensity, cumulative exposure, and duration of exposure)
(102). Member of the NCI cohort were employed at one of ten U.S. industrial plants
before January 1, 1966 and followed through December 31, 1994. Relative risks and
95% confidence intervals were calculated using Poisson regression analysis (stratified
by age, calendar year, sex, race, and adjusted for pay category) utilizing the lowest
non-zero category of exposure as the referent group. Peak exposures were defined as
short-term exposures (<15 minutes) that exceeded the 8-hour time-weighted average
(TWA), and were only estimations, as no measurable data for peak exposures existed.
27


The strongest association observed (Table 4.1) was between peak formaldehyde
exposures and leukemia (Ptrend=0.004), but the association between increasing peak
formaldehyde exposure and myeloid leukemia was also statistically significant (Ptrend
=0.009). Hauptmann et al. reported finding no association between cumulative
exposure and leukemia or myeloid leukemia (Ptrend=0.235 and 0.157, respectively) or
for duration of exposure (Ptrend=0.465 and 0.911), and a weak association for average
exposure intensity (PtrCnd=0.242 and 0.088). Ptrend values described herein only took
into account exposed person-years, but Hauptmann et al. also estimated Ptrend values
for exposed and unexposed person-years combined. Ptrend values for the exposed and
unexposed person-years combined, typically showed lower values as would be
expected, with the additional of unexposed subjects. By using four exposure metrics
to look for associations between formaldehyde exposure and leukemia, the authors
state they decreased their risk of an overall false-negative finding, but may have
increased their risk for finding a false-positive result for a single exposure metric.
However, by evaluating multiple aspects for a possible exposure-response
relationship (ie. increasing relative risks for the various exposure metric categories,
statistically significant excesses for relative risk, and the Ptrend for increasing exposure
metric categories), the authors state that they limited false-positive errors for each of
the exposure metrics. In June 2004, the IARC Working Group concluded that since
the study failed to demonstrate an exposure-response relationship with cumulative
exposure, perhaps use of other exposure metrics would be more relevant when
examining the leukemogenic potential of chemical exposures (22).
An update of the NCI cohort study by Beane Freeman et al. (102) was published in
2009. This industrial worker cohorts follow-up was extended by 10 years, through
December 31, 2004. Study results demonstrated a statistically significant association
28


(Table 4.1) for relative risk with increasing peak formaldehyde exposures for all
lymphohematopoietic malignancies (Ptrend=0.02) and Hodgkin lymphoma
(Ptrend=0.01), but no associations were observed for all leukemia and myeloid
leukemia (Ptrend= 0.12 and Ptrend= 0.13, respectively). No significant associations
were found for either the average intensity (Ptrend>0.50 and Ptrend= 0.43) or cumulative
exposure (Ptrend= 0.12 and P trend^-50) metric. When all person years (exposed and
unexposed combined) were included for mortality analysis, the Ptrend decreased to
0.02 for all leukemia and 0.07 for myeloid leukemia. No heterogeneity of relative
risks was found based on race, gender, or pay category. Additionally, controlling for
11 possible confounding chemicals also did not significantly change the results. With
the inclusion of 10 more years of follow-up, the risks associated with leukemia and
myeloid leukemia decreased somewhat since the last update (2003).
Beane Freeman et al. note that for peak exposure and myeloid leukemia, the highest
relative risks due to formaldehyde exposure occurred prior to 1980, the time period
likely associated with substantial formaldehyde exposures. The levels of
occupational formaldehyde exposure decreased after 1980, and the authors state that
workers in the NCI cohort likely had no significant formaldehyde exposure after
1980. For the highest peak exposure category, cumulative risk, as calculated by
extending the calendar-year of follow-up by one year, remained elevated over the
complete follow-up period (1965-2004) for myeloid leukemia, but began declining in
the mid 1990s. Statistically significant Ptrend values for peak formaldehyde exposure
and myeloid leukemia disease association were attained in 1990 and remained
significant through 2000. The authors speculate these results may be due to an
increase in relative risk estimate precision based on additional person-years accrual,
or perhaps that this may be a result of a fairly short-induction time for myeloid
29


leukemia, whereby highest risks were seen within 25 years since first exposure, as
evidenced by time since first exposure and first high peak experienced. In
conclusion, Beane Freeman et al. state that their findings provide some causal
evidence that peak formaldehyde exposures induce lymphohematopoietic
malignancies, particularly myeloid leukemia, but the authors also caution that these
results may be due to chance.
4.4.2 Epidemiological Studies Demonstrating No Association between FA and
Leukemia
The other two large cohort studies evaluated by IARC for leukemogenic potential of
formaldehyde exposure were the Pinkerton et al. (2004) and Coggon et al. (2003)
studies. The Pinkerton et al. study (of U.S. garment workers) suggested a possible
relationship between formaldehyde exposure and leukemia, particularly for myeloid
leukemia (61). This potential relationship was also reported in the Hauptmann et al
(2003) study. According to Pinkerton et al., area monitoring demonstrated steady
levels of formaldehyde exposure with no evidence for peak or intermittent exposures,
but no area monitoring data were available prior to 1970 (personal exposure levels
were assessed for random workers in 1981 and 1984), and historical exposure
estimates were given a quantitative value only after discussions with workers about
past working conditions (103). The myeloid leukemia risk was elevated for workers
who had their first exposure early in the study period (among workers with 20 or
more years since first exposure), when formaldehyde exposures were potentially
higher. The only exposure metric evaluated in this study was duration of exposure.
As the authors noted, early exposures were possibly substantially higher, and while
it is reasonable to assume that garment workers from early in the study period were
30


exposed to formaldehyde peaks, since historical formaldehyde levels were not
available, peaks could not examined.
The Coggon et al. (2003) cohort study of British chemical workers, the third study
reviewed by IARC, failed to show any association between occupational
formaldehyde exposure and leukemia (60). This study also did not evaluate peak
exposures, nor did it examine mortality specifically due to myeloid leukemia, which
has been shown to have a stronger association with formaldehyde exposures in both
the Hauptmann et al. (2003) and Pinkerton et al. (2004) studies.
Table 4.1 P ,rtd values for associations between leukemia mortality and quantitative formaldehyde exposure estimates*
Study/Cause of Peak exposure Average intensity Cumulative Duration of TWA8 intensity
death (ICDb) (ppm)c (ppm)c exposure (ppm-y, ppm-h)c exposure (y)c (ppm)c
# * # * # * # * # *
Hauptmann et al. (102)d
25,619 U.S. industrial workers, employed before 1/1/1966 and followed through 12/31/1994
Leukemia (204- 207) 0.004 0.001 0.242 0.193 0.235 0.183 0.465 0.214
Lymphatic leukemia (204) 0.559 0.279 0.632 0.495 0.476 0.406 0.684 0.498
Myeloid leukemia (205) 0.009 0.003 0.088 0.086 0.157 0.123 0.911 0.423
Other/unspecified leukemia (207) 0.154 0.277 0.71 (0.697) (0.783) (0.740) 0.292 0.402
Study/Cause of Peak exposure Average intensity Cumulative Duration of TWA8 intensity
death (ICDb) (ppm)c (ppm)c exposure (ppm-y, ppm-h)c exposure (y) (ppm)c
# * # * # * # * # *
Beane Freeman et al. (104)"
25,619 U.S. industrial workers, employed before 1/1/1966 and followed through 12/31/2004
Leukemia (204- 207) 0.12 0.02 >0.50 0.50 0.12 0.08
Lymphatic leukemia (204) >0.50 0.30 >0.50 >0.50 0.46 0.41
Myeloid leukemia (205) 0.13 0.07 0.43 0.40 >0.50 0.44
Other/unspecified leukemia (207) >0.50 0.50 >0.50 >0.50 0.15 0.13
31


Table 4.1 (cont.)
Study/Cause of Peak exposure death (lCDb) (ppm) # * Average intensity (ppm) # * Cumulative exposure (ppm-y, ppm-h) # * Duration of exposure (y) # * TWA8 intensity (ppm) # *
Hauptmann et al. (105)
Case-control study of funeral industry workers (embalmers) who died between 1/1/60 and 1/1/86 from LHM (n=l 68)
Lymphoid origin (200-204) (0.111) (0.523) (0.287) (0.598) (0.912) (0.965) 0.360 0.449 (0.605) (0.766)
Non-lymphoid origin (205, 206, 208,209) (0.944) (0.089) (0.997) 0.096 0.523 0.140 0.348 0.046 0.951 0.256
Myeloid leukemia (205) (0.778) 0.036 (0.722) 0.058 0.966 0.192 0.588 0.020 (0.642) 0.396
Acute myeloid leukemia (205) 0.636 0.035 0.869 0.068 0.940 0.284 0.612 0.063 (0.672) 0.441
Tests of trend based on slope estimates for continuous formaldehyde exposure estimates; two-sided likelihood ratio test at a 5%
statistical significance level; values in parentheses indicate negative slope estimates.
bCodes of the International Classification of Diseases (ICD), 8111 revision.
cppm=parts per million; ppm-y=parts per million-years (evaluated in (102, 104)); ppm-h=parts per million-hours (evaluated in
(105)),
y=years; TWA8=8-hour time-weighted average intensity.
dTests of trend based on relative risk (RR)estimates for increasing exposure metric (ie. peak exposure, average intensity
exposure, etc.); calculated by Poisson regression; 2-year lag interval.
Tests of trend based on odds ratio (OR) estimates for increasing exposure metric (ie. peak exposure, average intensity exposure,
etc.); calculated by unconditional logistic regression; 2-year lag interval.
Exposed subjects only.
'Exposed and non-exposed subjects.
4.5 Professional Worker Cohort Data
The seven epidemiological studies (58, 62, 106-110) of professional workers included
in the assessment of leukemogenic potential of formaldehyde by IARC in 2006,
reported small excesses in leukemia mortality (again, particularly mortality from
myeloid leukemia) in pathologists, anatomists, and embalmers. As in the Pinkerton et
al. (2004) and Coggon et al. (2003) studies, peak exposures were not evaluated and
formaldehyde exposures were never actually assessed. Some of studies these utilized
proportionate mortality analyses (PMR), which could bias differences in total
mortality observed for professionals by high selectivity within this group (70). Also
noted by published reviews of this data set by Bosetti et al. (2008) and Collins and
32


Lineker (2004), was exposure to possible confounders such as exposure to other
chemicals and even viral exposures may have led to the excesses of leukemia
observed (70, 111). IARC downplayed the likelihood of viral exposures as the source
of leukemogenesis, stating that this population of workers would not have likely been
exposed to viruses any more or less than the control population.
IARC also stated that professional workers may have been exposed to other
chemicals, but no known leukemogens were ever shown to be present (22).
Furthermore, exposure to chemicals would have differed between the different
occupations and one would not have expected to see similar rates of leukemia
excesses among all the professional occupations (22). Elevated rates of leukemia
remained fairly consistent throughout the professional worker studies from the mid
1980s and early 1990s, and therefore one would have expected to see similar trends in
industrial worker cohort studies conducted during this same time period. A review by
Collins et al. (2001) evaluated formaldehyde exposure by job type and reported that
the highest peak exposures and the highest average daily exposures were likely
experienced by industrial workers rather than the professional workers (112). This
should indicate a greater risk of leukemia for industrial workers, but as Collins and
Lineker elaborate, in a 2004 review and meta-analysis on formaldehyde exposure and
leukemia, due to a lack of consistency among the two types of cohorts, increased
leukemia rates may be a result of employment as an embalmer, pathologist, or
anatomist, and not necessarily related to formaldehyde exposure (70, 111). Although
Collins and Lineker write that the highest peak exposures were likely experienced by
industrial workers, they also point out that there were large differences observed in
exposure levels across studies they evaluated for their 2001 review and meta-analysis
of formaldehyde and pancreatic cancer (112). Individuals exposed in certain
33


professional settings, such as a laboratory or a funeral parlor, may have experienced
an increased frequency or a higher concentration of peak exposures than industrial
workers, or perhaps concomitant chemical exposure may have led to the elevated
rates of leukemia observed in the professional worker studies. Lastly, individuals
who worked, as anatomists, pathologists, or embalmers may have been more likely to
seek medical treatment or had access to better diagnostic procedures, due in part to
their socioeconomic status, than industrial workers, which may have resulted in
reporting bias (111). Linet and Cartwright (1996) report that higher leukemia
mortality rates have been correlated to higher quality medical care (113).
Hauptmann et al. (2009) followed up mortality surveys in the U.S. funeral industry
for three of the seven studies (58, 62, 109) described above. The purpose of this
follow-up was to evaluate the excesses previously observed for leukemia mortality,
and in particular, myeloid leukemia, in professional cohorts (anatomists, pathologists,
and funeral industry workers). Odds ratios and 95% confidence intervals were
calculated using logistic regression analysis on funeral industry workers (n=168) who
died from any type of lymphohematopoietic malignancies between January 1, 1960
and January 1, 1986, and compared with age-matched controls (n=265). Findings
demonstrated a statistically significant increase in mortality from myeloid leukemia
with increasing peak formaldehyde exposure (Ptrend=0.036) for U.S. embalmers (105).
Significant trends were not observed for either increasing average formaldehyde
intensity (Ptrend=0.058) or for increasing 8-hour time-weighted average intensity
(Ptrend=0.396) (Table 4.1). As observed in both the Hauptmann et al. (2003) and
Beane Freeman et al. (2009) studies of the NCI cohort, increasing cumulative
exposure was not associated with myeloid leukemia (Ptrend=0.192). Although
embalming fluids and products contain formaldehyde as well as other chemicals, the
34


authors contend that none of these other chemicals are known leukemogens. The
authors compared results from their funeral industry study to the NCI industrial
worker cohort study (104) and observed that funeral home embalmers had longer
durations of formaldehyde exposure and higher cumulative formaldehyde exposures,
but lower 8-hour time-weighted average intensity. In the NCI cohort study,
increasing peak formaldehyde exposures were divided into four categories. The
highest peak exposure category had a lower bound of > 4 ppm. 77% of funeral
workers (control subjects) who ever embalmed, experienced peak exposure levels
greater than 4 ppm, while only 25% of industrial workers experienced greater than 4
ppm peak exposure levels. Overall, funeral workers appear to have experienced
higher peak formaldehyde exposures that may, in part, explain the significant myeloid
leukemia excess observed in funeral industry workers as compared to non-significant
excesses observed for industrial workers.
4.6 Peak FA Exposures and Leukemia, a Recent Meta-Analysis
In a meta-analysis conducted by Zhang et al. (2009), the authors reviewed 15
epidemiological studies (58, 60-62, 102, 106, 107, 109, 110, 114-119) of
formaldehyde-exposed workers and leukemia (56). The summary relative risks (RR)
for leukemia in the 15 studies were elevated (RR=1.54, 95% confidence interval (Cl)
= 1.18-2.00), with the strongest risks noted for myeloid leukemia (RR-1.90, 0=1.31-
2.76). Just six (58, 61, 62, 102, 107, 109) of the fifteen studies specified leukemia
subtypes, with myeloid leukemia reported for 51% of cases and lymphoid leukemia
reported for 19% of cases. All other cases of leukemia were unspecified. Acute
myeloid leukemia (AML) was the major subtype (64%) of all myeloid leukemia cases
observed, and Zhang et al. hypothesized that formaldehyde exposure increases the
risk for myeloid leukemia more so than lymphoid leukemia, and that AML is the
35


predominant subtype. While conducting their meta-analysis, the authors chose the
peak exposure metric as the most important exposure measure, designating average
exposure intensity, cumulative exposure, and exposure duration as less significant.
Peak exposure was selected as the first metric utilized for analysis by Zhang et al. due
to their belief that other metrics would be less accurate measures of actual
exposure, since workers may have had periods of very high exposure coinciding with
time periods of little or no formaldehyde exposure.
No definitive mechanism of action has been elucidated by which formaldehyde may
induce leukemia. Formaldehyde has been shown to genotoxic in both humans and
experimental, thus capable of causing DNA and chromosomal damage, and potential
subsequent leukemic transformation. Several hypotheses hint at formaldehyde acting
either on the bone marrow or peripheral blood progenitor cells, although these
hypotheses are controversial. Known leukemogens, such as benzene and
topoisomerase-II inhibitors, induce bone marrow toxicity, but formaldehyde, like BD,
appears to induce leukemogenesis through alternative pathways. There is
epidemiologic evidence, although inconsistent, demonstrating a positive association
between formaldehyde and increased leukemia mortality, particularly myeloid
leukemia. Studies evaluating leukemia in garment workers, industrial workers, and
professional workers (anatomists, pathologists, embalmers) have all been cited as
supporting material for the 2006 IARC overall classification of formaldehyde as
carcinogenic to humans. In these studies, which address formaldehyde exposure and
leukemia risk, peak exposures have been shown to be more predictive of leukemia
risk than other exposure metrics.
36


5. Benzene (BZ)
5.1 BZ Characteristics and Uses
Benzene, C6H6, is a clear and colorless liquid with a solvent-like odor that is highly
flammable and non-corrosive (23,120). Due to its low boiling point and high vapor
pressure, benzene is highly volatile and capable of rapid atmospheric evaporation
(23). Benzene is also moderately water soluble (120). This ubiquitous aromatic
hydrocarbon may be found in air, water, and soil and arises from both natural and
industrial sources (120). Natural sources of benzene include volcanic emissions, oil
seeps, forest fires as well as a variety of fruits and vegetables (120, 121). Benzene is
also present in cigarette smoke, crude oil, and gasoline (1-2% by volume in unleaded
gasoline) (120, 121). Industrial sources that contribute to background benzene levels
include automobile exhaust, fuel evaporation from gasoline service stations, and
industrial emissions (120). Benzenes primary use today is as a starting material or as
an intermediate for the synthesis and production of other chemicals (121). For
example, benzene is used to synthesize ethylbenzene for production of styrene
(Styrofoam and other plastics), cumene for production of phenol and acetone,
cyclohexane for the production of nylon and synthetic fibers, as well as in the
manufacturing of lubricants, detergents, dyes, drugs, and pesticides (120).
Commercial production of benzene from coal originated in the mid 1800s after it was
discovered and isolated from coal tar (120). In 1941, benzene was first produced
from petroleum, and since 1959, petroleum has become the predominant source for
benzene in the United States (121). Benzene was ranked 17th in production volume
for all chemicals produced in the United States in 1994, and in 2002, 12.0 billion
pounds of benzene were produced in the United States (121).
37


5.2 BZ Health Effects
5.2.1 Acute Exposure Effects
The non-smoking general population is primarily exposed to benzene through
inhalation of ambient air and to a lesser extent, through ingestion of contaminated
food and water (120). Due to its highly volatile nature, benzene rapidly evaporates
from the skin following dermal contact (122). The primary source of non-
occupational benzene exposure in the United States is tobacco smoke, accounting for
up to 50% of an individuals cumulative exposure (120). Vapors from products such
as furniture wax, paints, organic solvents, and glues that may contain trace levels of
benzene may also be sources of low level exposures (ppb) (120). Occupational
exposure to benzene may occur during the production and/or use of benzene or
benzene-containing products. Workers employed in manufacturing of rubber tires,
benzene production (coke and coal chemical manufacturing, petrochemicals, and
petroleum refining), storage and transport of benzene and benzene-containing
products (such as gasoline), along with such occupations as firefighters, laboratory
technicians, printers, and shoe makers are all subject to potential exposure (120).
This is particularly true in other countries such as China and Brazil where benzene is
still used as a solvent.
Acute, high dose exposure to benzene has been associated with central nervous
system toxicity (drowsiness, dizziness, loss of consciousness, and delirium (120, 123,
124). Other neurological effects reported in animal studies following inhalation
exposure include loss on involuntary reflexes, narcosis, and deceased electrical
activity in the brain (120, 125-128). Dermal (skin irritations and bums) and ocular
effects (eye irritations) have also been reported (120, 129-134). Oral ingestion of
benzene has resulted in swelling and edema as well as death from depression of the
38


central nervous system (CNS) (120,135). In mice, decreased fetal weight and/or the
development of skeletal variants, have demonstrated fetotoxicity following inhalation
exposure of benzene (120, 136-142). However, there is insufficient evidence that
benzene exposure negatively affects human reproduction (120). Immunological
effects following inhalation exposure affect both the humoral and cellular
components, with decreased levels of both antibodies and leukocytes reported in
workers with occupational exposures and supported by animal studies (120, 143-156).
5.2.2 Hematotoxic Effects
While the acute toxicity of benzene has been shown to primarily affect the CNS, by
far the most important target for chronic benzene toxicity is the blood and bone
marrow. Both human and animal studies have extensively demonstrated the
hematotoxic nature of benzene exposure (157-160). Workers with high dose chronic
exposure to benzene may present with significant reductions in any or all of the major
peripheral blood cell types (erythrocytes, leukocytes, and platelets) (120). These
reductions in peripheral blood cell types are known as cytopenias, specifically
anemia, leukopenia, and thrombocytopenia. Pancytopenia, a reduction in all three
major blood lineages, may result from hypoplasia (the condition of having too few
cells) of the bone marrow (120). Additionally, under extremely high prolonged
exposure conditions, a disease known as aplastic anemia can occur. This disease is
characterized by a profound reduction of blood cellular components in the bone
marrow with accompanying decrease observed in the peripheral blood as well (120).
Aplastic anemia can be fatal (30-50% die within one year of diagnosis), and is rarely
associated with an increased risk for the development of acute myeloid leukemia
(AML) in those individuals who do not die from aplastic anemia (23, 120, 122, 148,
156, 157, 161-168). Although aplastic anemia is associated with hypoplasia of the
39


bone marrow, the onset of AML requires a subsequent uncontrolled proliferation of a
clonal population of myeloid blasts, which can greatly increase bone marrow
cellularity. Findings in animal studies have demonstrated that cell death due to
benzene exposure appears to initiate a compensatory proliferation, or a promoting
effect, of bone marrow cells (hyperplasia). It is conceivable that compensatory
hyperplasia may be a potential mechanism for benzene-induced leukemogenesis (154,
169).
5.2.3 BZ Carcinogenicity
5.2.3.1 Leukemia
The United States Environmental Protection Agency (USEPA), the International
Agency for Research on Cancer (IARC), and the National Toxicological Program
(NTP) of the United States Department of Health and Human Services (USDHHS)
have all reached the same conclusion; benzene is a known human carcinogen based
on sufficient evidence in humans and supporting evidence from animal studies (120,
121, 157, 162, 163). The cancer predominantly associated with inhalation exposure
of benzene in humans is leukemia, and specifically, acute non-lymphocytic
(myelogenous or myelocytic) leukemia. Epidemiological studies on a cohort of
workers occupationally exposed to benzene during the manufacturing of rubber
hydrochloride (trade name Pliofilm) at two Ohio plants provide a significant portion
of data IARC, USEPA, and NTP utilized when concluding on the leukemogenic
potential of benzene (164-166). A series of collaborative studies between the
National Cancer Institute and the Chinese Academy of Preventative Medicine
(NCI/CAPM) also provide evidence, based on epidemiological results from a very
large cohort of Chinese workers occupationally exposed to benzene, for the consensus
conclusion on benzenes lymphohematopoietic carcinogenicity (161, 167, 168).
40


Rinsky et al., the authors responsible for the well-conducted epidemiological studies
on the Pliofilm cohort in Ohio, in their 1987 epidemiological risk assessment on
benzene and leukemia, list the case reports, epidemiological studies, and animal
carcinogenesis bioassay studies (some studies which were published more than eighty
years ago) that contributed to the consensus conclusion of both national and
international scientific bodies on the leukemogenic potential of benzene exposure
(148, 165, 166, 170-182).
5.2.3.2 Non-Hodgkins Lymphoma (NHL) and Multiple Myeloma (MM)
Savitz and Andrews (1997) reviewed 18 community-based and 16 industry-based
epidemiological studies on benzene exposure and suggested that AML may not be the
lone cancer induced by benzene and that benzenes carcinogenic effect is broad, and
is likely associated in total lymphatic and hematopoietic cancers, including differing
leukemia cell types (183). According to the ATSDR, studies have suggested that
possible associations may exist between occupational benzene exposure and non-
Hodgkins lymphoma (NHL) and multiple myeloma (MM), but Savitz and Andrews
argue that most studies have not demonstrated positive associations. Recently,
Alexander et al. (2010) conducted a meta-analysis of 8 cohort and 14 case-control
studies and found no independent association between benzene exposure and NHL
(184).
5.3 Animal Studies and BZ Carcinogenesis
Animal studies (in rats and mice) have demonstrated that oral ingestion of benzene is
associated with tumor formation, including Zymbal-gland carcinoma, malignant
lymphoma, skin and oral cavity carcinomas, lung adenomas, mammary gland
carcinomas and ovarian tumors in female mice, and Harderian-gland adenoma and
41


preputial gland squamous cell carcinomas in male mice (120, 121, 182). Currently,
no reproducible animal model for benzene-induced leukemia exists, but benzene has
demonstrated, as described above, its potential as a multiple site carcinogen in rodent
studies from both inhalation and ingestion of benzene (185). Although a lack of oral
carcinogenicity data exists for humans, the consensus conclusion is based on
sufficient inhalation exposure data in humans coupled with the ingestion and
inhalation data from the animal studies (120).
5.4 BZ Biotransformation and Toxicity
5.4.1 First Metabolic Pathway
The mode-of-action by which benzene elicits its leukemogenic effect has not been
well established (157). Benzene itself has not been implicated as having any
carcinogenic and toxic effects, and animal studies have demonstrated that benzenes
reactive metabolites are responsible for the harmful genotoxic and myelotoxic effects
observed (186-189). Due to the fact that during the biotransformation of
benzene many reactive metabolites are formed, it has been difficult to determine
exactly which metabolic pathway or combination of pathways may be responsible for
leukemogenesis (157). Benzene biotransformation is believed to occur through two
metabolic pathways (Figure 5.1), the first of which involves the hepatic metabolites
phenol, catechol, and hydroquinone and the second pathway, which implicates ring-
opened benzene forms (157). Cytochrome P450 2E1 (CYP2E1) has been identified
as the enzyme responsible for the metabolism of benzene (190), and much of the data
regarding benzene toxicity suggests that hepatic conversion of benzene to phenol may
be the primary event (191).
42


bfmn
Vo'-^yTialC:
BsnMnft Oi>d Oxcpin £.fi**wco*l / \
O
Benzene Dioj Expcxide
CHO x COOH
1.4BBnzoquinone 1,2.4-trihydroicybenzene 1 743eruoqunerie
Figure 5.1 The Major Metabolic Pathways Responsible for the Generation of Genotoxic and
Myelotoxic Benzene Intermediates (Adapted from Kim et al. (192), American Association for Cancer
Research, publisher).
5.4.2 Second Metabolic Pathway
The second proposed metabolic pathway, which results in iron-catalyzed ring-
opening of benzene epoxide (193), a benzene intermediate product, and the
subsequent formation of trans, trans-muconaldehyde (MA), a known hematotoxin
capable of inhibiting DNA, RNA, and protein synthesis (194), is described by
Latriano et al. (1986) to be a quantitatively minor metabolite responsible for an
important pathway leading to the formation of toxic metabolites from benzene (195).
According to USEPA (1998), the myelotoxicity and genotoxicity observed following
benzene exposure are resultants of a synergistic reaction between phenol and
catechol, hydroquinone, or muconaldehyde (157). Additionally, CYP2E1 is required
for biotransformation of benzene, and due to genetic polymorphisms, individuals are
43


likely to vary in their expression levels of CYP2E1, thereby affecting concentrations
of toxic benzene metabolites produced that in turn may affect the level of toxicity
observed.
5.5 BZ Mechanisms of Action
Rickert et al. (1979) have shown that catechol and hydroquinone persist in the bone
marrow following benzene exposure (196). Bone marrow contains high
concentrations of peroxidases (197), including myeloperoxidase (MPO), which may
easily bioactivate phenolic metabolites to reactive quinones (198). MPO activation of
reactive quinones, such as hydroquinone, is known to induce DNA adduct formation
(199-201). DNA adduct formation is one of several proposed mechanisms for
benzene genotoxicity. Other proposed mechanisms for benzene-induced
leukemogenesis include, oxidative DNA damage, inhibition of topoisomerase II
leading to clastogenic chromosomal effects as seen with epipodophyllotoxins, and
aneugenic effects resulting in mitotic damage and chromosome malsegregation
during cytokinesis (202-204).
Whysner et al., in their 2004 review, conclude that benzene acts most likely through a
clastogenic manner based on the evidence they compiled through examination of
genotoxicity studies on benzene metabolites in humans and in human-derived cells
(205). Since topoisomerase II-inhibitors, such as etoposide, as described earlier,
induce secondary or therapy-related AML, it certainly is possible that benzene
metabolites may act in a similar manner due to the induction of AML (recognized as
the predominant leukemia subtype) following benzene exposure. Damage incurred
by topoisomerase II, histone proteins, tubulin, and other DNA-associated proteins by
some combination of benzenes reactive metabolites potentially leads to DNA strand
breaks, chromosomal translocations, mitotic recombination, and aneuploidy (157). If
44


these types of insults occur in lymphohematopoietic stem cells or early progenitor
cells in the bone marrow, the potential for emergence and growth of a leukemic clone
is enhanced due to such molecular activities as gene fusion, suppressor gene
inactivation, proto-oncogene activation as a result of the DNA, and/or chromosomal
damage induced by benzene exposure (157).
5.6 Cumulative BZ Exposure and Leukemia-Epidemiological Evidence
The epidemiological studies presented earlier, from which the consensus conclusion
on benzene leukemogenicity was reached, provide sufficient causal evidence that
occupational exposure to benzene leads to increased mortality from leukemia, and
particularly AML. The Pliofilm (Ohio) cohort study, consisting of 1,165 white males
employed between 1940 and 1965 and vital status ascertained through 1981, is
generally considered as providing the strongest data set for the leukemogenic
potential of benzene because of limited exposure to confounding chemicals and a
large range of exposure estimates (120, 165). According to USEPA IRIS
carcinogenic benzene assessment (122), the Rinsky data (increasing cumulative
exposure associated with increasing risk of leukemia) was utilized for derivation of
unit cancer risk estimates by Crump (206, 207). The NCI/CAPM Chinese worker
cohort studies are also commonly cited as providing suggestive evidence for excess
mortality from leukemia due to benzene exposure, as these studies evaluated the
incidence rates for very large cohorts of benzene-exposed (74,828) and non-exposed
(35,805) workers employed in 672 factories in 12 cities in China, thus allowing for
the detection of significant leukemia excesses from low levels of benzene exposure
(120). Evidence compiled from both series of studies indicates that risk for leukemia
rises as the dose of benzene increases, implicating cumulative exposure responsible
for benzene-induced leukemia. Additional suggestive evidence demonstrating that the
45


risk of leukemia rises with increasing benzene exposure is provided in another series
of studies that evaluated SMR values for approximately 1,700 male and female Italian
shoe factory workers (208, 209). Although cumulative exposure appears to be the
dominant exposure metric associated with elevated risk of leukemia, peak exposures
were not evaluated, and close examination of the data reveals that while the Pliofilm
studies of Rinksy et al. (1987) strongly support this argument, the evidence is not
quite as strong when examining the data for a much larger study, comprised of the
Chinese workers cohort from the NCI/CAPM collaborative studies (Hayes et al.,
1997).
5.6.1 Pliofilm Cohort
The Rinsky et al. (1981) epidemiological risk assessment utilized for derivation of
unit cancer risk estimates concludes that cumulative benzene exposure may be the
strongest single predictor for death from leukemia. SMRs for leukemia mortality
increased from 109, to 322, to 1186, and to 6637 as cumulative benzene exposure
increased from less than 40 ppm-years, to 40-199 ppm years, to 200-399 ppm-years,
and 400 or greater ppm-years. The authors also evaluated duration of exposure and
the rate of exposure (cumulative exposure divided by the rate of exposure), but found
no significant contributions from either of these two exposure metrics in regards to
predicting death from leukemia. Rinsky et al. mention that incidences involving high
exposures (peaks), such as exposures occurring during spills, were not evaluated and
therefore not reflected in exposure calculations.
In 2002, Rinsky et al. updated their Pliofilm cohort study (164). The updated 2002
cohort consisted of 1,291 workers followed for an additional 15 years (through 1996).
While the SMR for leukemia decreased from 3.37 (95%CI=1.54-6.41) to 2.56 (95%
46


0=1.43-4.22), cumulative exposure was still significantly associated with leukemia
mortality. Rinsky et al. write that they evaluated peak benzene exposure, which they
define as the highest level of exposure experienced for any amount of time, however
no results were presented in their updated study. The authors suggest that low
cumulative exposure to benzene may lead multiple myeloma, as evidenced by non-
significantly elevated risks, and high cumulative exposure leads to leukemia (164,
210). Although peak exposure data was not presented in the study and a strong
exposure-response effect was still noted, it certainly may be possible that workers
experienced peaks at some point during their employment, as a part of either low or
high cumulative exposure, and these intermittent exposures may have contributed to
the excess risks observed for leukemia.
5.6.2 NCI/CAPM Cohort
In contrast to the convincing findings of the Rinksy et al. Pliofilm cohort studies, the
1997 Hayes et al. study examining a large cohort of Chinese workers (NCI/CAPM),
demonstrated a weak, albeit significant association between cumulative exposure and
leukemia (161). These findings, while demonstrating a modest dose-response effect,
suggest that cumulative exposure of less than 40 ppm-years may cause leukemia. In
the Rinsky et al. (1987) study, significant excesses for risk of death from leukemia
begin at cumulative exposures greater than 200 ppm-years, and this risk was shown to
rise exponentially with increasing cumulative dose (165). Hayes et al. observed a
weak significant trend (p=0.04) for increasing cumulative exposure and relative risk
for leukemia and a non-significant trend (p=0.06) for increasing cumulative exposure
and relative risk for ANLL. Analysis by duration of exposure for increasing exposure
duration showed no significant trend for leukemia or ANLL, and analysis by average
exposure demonstrated a weak association with increasing average exposure for both
47


leukemia and ANLL. Additionally, leukemia deaths were generally associated with
benzene exposure greater than 20 ppm in the Rinsky et al. study (1987), while levels
associated with leukemia mortality as observed by Hayes et al., were much lower,
with exposure levels below 10 ppm. As with the Rinksy et al. study (1987), peak
exposures and their leukemogenic potential were not investigated by Hayes et al.
5.6.3 Italian Shoe Worker Cohort
A cohort of 1,008 men and 1,005 women, employed at a shoe manufacturing plant in
Florence, Italy between 1950 and 1984 (whose vital status was ascertained through
December 31, 1984), was evaluated by Paci et al. (1989) for leukemia mortality
following benzene exposure (209). In Italy, a law limiting the amount of benzene in
glue (no more than 2% benzene by weight) used for shoe production was passed in
1963. Prior to passage of this law, the three most common shoe glues contained an
average of 81% benzene by weight (208). Paci et al. observed increased risk only
among workers first employed prior to 1963. Paci et al. also showed a clear absence
of any dose-response relationship based on duration of employment. An SMR of 769
was observed for male workers with less than one year of employment. For male
workers with of 1-5 years of employment the SMR was 555, and an SMR of 416 was
observed for male workers with greater than five years of employment. Paci et al.
state that duration of work is considered an indicator of dose, and this result may be
due to poor exposure classification of dose, since no information was available
regarding airborne benzene concentrations in the plant. The authors, however, give
another explanation for the lack of a dose-response and risk relationship, writing that
perhaps the excess risk observed may have been the result of high exposures over a
shorter exposure interval.
48


5.7 Epidemiological Evidence for Peak Exposures and BZ-Induced Leukemia
Since no quantifiable information on airborne benzene exposure for the shoe
manufacturing plant in Italy was provided by the Paci et al. (1989) study, a new study
was undertaken by Seniori Costantini et al. (2003), in which a quantitative exposure
assessment of benzene exposure was completed along with the addition of 15 years of
follow-up (through December 1999) for determination of any association between
benzene exposure and risk of leukemia for the Italian shoe workers cohort (208).
Study objectives included reconstruction of work histories, creation of time-specific
job-exposure matrix with benzene exposure estimates, estimation of benzene
exposure for all workers in the cohort, and leukemia risk estimates in relation to
benzene exposure. Cumulative exposure (sum of the products of job-specific benzene
concentrations and the duration of employment) was calculated and peak exposure
(highest intensity of benzene exposure regardless of job exposure duration) was
estimated for all workers in the cohort (Table 5.1). Seniori Costantini et al. divided
cumulative exposures into four categories, <40, 40-99, 100-199, and >200 ppm-years.
These categories were chosen so as to compare results to the 1987 Rinsky et al. study.
The SMRs for leukemia in men and women combined were elevated for all but the
<40 ppm-years category, the lowest exposure category, and the results were similar
when SMRs were calculated for men only. The SMRs were significantly elevated in
the highest exposure category, with an SMR of 7.0 (95% 0=1.0-18.0) observed for
male workers in this category. Seniori Costantini et al. write that their results are
entirely consistent with the 1987 Rinsky et al. study, in that increasing risk of
leukemia is associated with increasing cumulative benzene exposure.
Increasing peak exposure findings for male and female workers combined and male
workers only, were also associated with increasing leukemia risk. For their study,
49


Seniori Costantini et al. examined three peak exposure categories <1 ppm, 1-29 ppm,
and >30 ppm. Elevated risk was associated with high peak exposure, particularly
with benzene exposures greater than 30 ppm. Although Seniori Costantini et al. did
find suggestive evidence of an association with increasing leukemia risk and peak
exposures, just as they did with cumulative exposure, they state having difficulty in
interpretation of the peak exposure results because the two exposure metrics, peak
and cumulative exposure, are highly correlated. The authors also state that for
reasons of consistency with previous studies and biologic plausibility, they accept
the cumulative exposure data as more relevant. Seniori Costantini et al. chose not
only to evaluate peak exposures, but to also include these results, which demonstrated
an association between peak exposures and leukemia in their study, and reasoning
such as maintaining consistency with other similar studies, seems to disregard the
significance of the peak benzene exposure results. With regard to biologic
plausibility, numerous animal studies were conducted prior to publication of this
study that examined intermittent benzene exposure and demonstrated loss of bone
marrow cellular proliferation and the occurrence of chromosomal damage, both
hallmarks of aplastic anemia and leukemia in humans (211-221). Data provided by
these toxicological animal studies provides suggestive evidence for an association
between intermittent exposure and leukemia, some of which was utilized by OSHA in
the writing of their final ruling on occupational exposure to benzene in 1987.
50


Table 5.1 Standardized mortality ratios (SMR) for leukemia in Italian shoe factory workers
following exposure (cumulative or peak) to benzene (BZ)a (ppm=parts per million)_____
Exposure Male and female workers combined Male workers only
metric SMR (95% CI)b obs. exp. SMR (95% CI)b obs. exp.
Cumulative BZ exposure <40 ppm- 1.3 (0.3-3.7) 3 2.36 1.4 (0.2-5.0) 2 1.45
years 40-99 ppm- 4.1 (0.5-14.7) 2 0.49 3.7 (0.1-20.6) 1 0.27
years 100-199 ppm- 2.5 (0.3-9.1) 2 0.79 3.0 (0.4-10.9) 2 0.66
years >200 ppm- 5.1 (1.4-13.0) 4 0.79 7.0(1.9-18.0) 4 0.57
years Peak BZ exposure <1 ppm 0.8 (0.02-4.7) 1 ' 1.2 1.2 (0.03-6.7) 1 0.8
1-29 ppm 2.6 (0.7-6.8) 4 1.5 2.5 (0.3-9.1) 2 0.8
>30 ppm 3.5 (1.3-7.6) 6 1.7 4.5 (1.1-9.9) 6 1.3
Cohort of 1,687 Italian shoe factory workers followed from 1/1/50 to 12/31/99; Seniori Costantini
et al. (2003) (208).
b95% 0=95% confidence interval; obs.=observed cases and exp.=expected cases; expected cases
calculated by use of national (1950-1969) and regional leukemia rates (1970-1993) specific for 5-
year calendar periods and 5-year age classes.
Increasing cumulative exposure to benzene is predominantly associated with
increasing risk of leukemia, according to the epidemiological results of several cohort
studies reviewed and cited by national and international scientific bodies (USEPA,
NTP, and IARC) that have reached a consensus conclusion on the carcinogenicity of
benzene. Although the cumulative exposure metric is considered the strongest
predictor of benzenes leukemogenic potential following occupational exposure, the
authors of these studies have not specifically evaluated peak exposure as an
independent exposure metric, instead any peak exposures experienced may be
assumed incorporated into the cumulative exposure dose. Seniori Costantini et al.
(2003) did evaluate peak benzene exposures in their cohort study of Italian shoe
51


workers, but chose not to investigate further the positive association they
demonstrated. An epidemiological study conducted by Ireland et al. (1997) (222) and
a subsequent update by Collins et al. (2003) (223), provide evidence for a predictive
value for peak exposures in leukemogenesis while showing limited evidence for the
cumulative exposure relationship (120).
The Ireland et al. (1997) study examines cancer mortality and low-level benzene
exposure among 4,172 male workers employed between 1940 and 1977, vital status
ascertained through December 31, 1991, at a Monsanto plant in Sauget, IL (222).
Previous studies reported no increases or only slightly elevated rates for leukemia or
AML following low-level benzene exposures (177, 224-226). Ireland et al. estimated
peak exposures for all jobs, and specifically the number of days for which a worker
experienced a 15-minute exposure greater than 100 ppm. Peak exposure categories
were divided into four categories: no days >100 ppm, < 7 days over 100 ppm, 7-40
days over 100 ppm, and > 40 days over 100 ppm. Cumulative exposure was also
assessed, and categories were divided into three groups: < 12 ppm-months, 12-72
ppm-months, and > 72 ppm-months. Ireland et al. reported that just 1% of their study
cohort accumulated over 600 ppm-months of cumulative benzene exposure as
compared to 50 % of the Pliofilm cohort from the Rinsky et al. (1981) study,
implying that cumulative benzene exposure in their study was low-level. Study
results revealed imprecise, but elevated rates of leukemia with increasing cumulative
exposure with similar results observed for the peak exposure categories (data not
shown in the study). Leukemia SMR for non-exposed workers was 1.1 (95% 0=0.4-
2.6), < 12 ppm-months SMR was 2.5 (95% CI=0.3-8.9), 12-72 ppm-months SMR
was 0.0 (95% 0=0.0-5.4), and > 72 ppm-months SMR was 4.6 (95% 0=0.9-13.4).
As observed in the Costantini et al. (2003) study, an association between peak
52


exposure and leukemia appears to exist, albeit an imprecise association. The authors
state that their leukemia mortality rates may have been limited due to exposure
underestimation, as exposure estimation for the 1940s and 1950 was difficult with
little information provided. The study may also have been limited by the small
number of deaths observed, but four of the five leukemia deaths in production
workers occurred in workers exposed prior to 1950, a time in which benzene
exposures could actually have been much higher than the estimations created by the
authors. Ireland et al. concluded that their study provided some evidence that low
level benzene exposure, as compared to the Pliofilm cohorts benzene exposure
levels, increases the risk of leukemia, regardless of whether it was cumulative
exposure or peak exposure as the exposure metric evaluated.
Collins et al. (2003) updated cancer mortality rates for this cohort (Monsanto) of
4,172 male workers with an additional seven years of follow-up (through December
31, 1997) and introduced 245 female workers into the cohort (223). Cumulative
benzene exposure was examined using the same exposure categories (< 1 ppm-year,
1-6 ppm-years, > 6 ppm-years) as in the previous study. Peak exposure, also the
same as in the previous study, estimated the number of days with 15-minute
exposures greater than 100 ppm, and was divided into four exposure categories (no
days over 100 ppm, < 7 days over 100 pm, 7-40 days over 100 ppm, and > 40 days
over 100 ppm). Collins et al. found little evidence for increasing risk of leukemia
with increasing cumulative benzene exposure (Table 5.2). The authors reported that
there were more leukemia deaths observed than expected for workers in the highest
peak exposure category, in workers with 40 or more days with peak exposures over
100 ppm. A similar result was also observed for multiple myeloma (SMR=4.0, 95%
CI=0.8-11.7) and for AML (SMR=4.1, 95% 0=0.5-14.9) and workers in the highest
53


peak exposure category respectively. Collins et al. conclude that the highest risks
observed were in those workers that experienced high, intermittent benzene exposures
greater than 100 ppm, but that imprecision and exposure misclassification, as with the
Ireland et al. (1997) study, may have limited their results.
Table 5.2 Standardized mortality ratios (SMR) for leukemia and acute non-lymphatic leukemia in
U.S (Illinois) chemical plant workers following exposure (cumulative or peak) to benzene (BZ)a
(ppm=parts per million)__________________________________________________________________
Exposure For leukemia (ICD 204-207)b For acute non-lymphatic leukemia
metric SMR (95% CI)b obs. exp. (ICD 205.0-206.0)b SMR (95% CI)b obs. exp.
Cumulative BZ exposure No exposure 1.0 (0.5-1.8) 10 10.4 0.8 (0.1-2.8) 2 2.6
<1 ppm-years 0.7 (0.1-2.5) 2 2.9 1.4 (0.1-5.1) 1 0.7
1-6 ppm- 1.4 (0.4-3.6) 4 2.9 2.7 (0.3-9.9) 2 0.7
years >6 ppm-years 0.7 (0.6-3.8) 6 3.5 2.2 (0.3-8.1) 2 0.9
Peak BZ exposure11 No exposure 1.0 (0.5-1.6) 15 15.6 1.3 (0.4-3.0) 5 3.9
<7 days 1.2 (0.0-6.8) 1 0.8 0.0 (0.0-10.2) 0 0.2
7-40 days 0.7 (0.0-4.0) 1 1.4 0.0 (0.0-10.4) 0 0.4
>40 days 2.7 (0.8-6.4) 5 1.8 4.1 (0.5-14.9) 2 0.5
aCohort of 4,417 Monsanto plant (Sauget, Illinois) workers followed from hire date to 12/31/1997;
Collins et al. (2003) (223).
bCodes of the International Classification of Diseases (ICD), 8th revision; 95% CI=95%
confidence interval; obs.=observed and exp.=expected cases; SMRs calculated by Fisher exact
method; compared to death rates for Illinois population.
cNumber of days with peak exposures over 100 ppm.
Much of the benzene carcinogenic risk assessment data assumes that a cumulative
relationship between risk and exposure exists, and that the concentration of benzene
exposure is independent of this relationship (165, 223, 227-230). According to
Collins et al. (2003), other alternative theories have been proposed regarding
benzenes mode of toxicity. The first theory involves possible existence of a
54


threshold concentration for benzene, and concentrations below this threshold do not
induce bone marrow toxicity or leukemogenicity (216, 223, 231, 232). A second
theory emphasizes the importance of intermittent exposure and benzenes myelotoxic
effects (216, 223, 231, 232). Following interpretation of their results, Collins et al.
(2003) report that a high number of peak exposures to benzene appeared to be a
better predictor of AML risk than cumulative exposure. The authors further state
that if benzene concentration is important in the determination of increasing leukemia
risk, use of cumulative exposure as the only exposure metric may obscure the true
risk of benzene exposure. Collins et al. (2003) mention that analysis of the Pliofdm
worker cohort has been re-evaluated, and that rather than examining cumulative
exposure, exposure concentration is instead examined (231). The results from this re-
evaluation demonstrated increased leukemia risk in workers that experienced a
minimum benzene exposure of 20-60 ppm, indicating that a critical concentration
must be reached in order for leukemia risk to rise, and that dose rate and threshold
concentration in workers occupationally exposed to benzene need further study.
Rushton and Romaniuk, in a 1997 case-control study to investigate leukemia risk and
benzene exposure for petroleum workers in the U.K., found no evidence for an
association between increasing risk with increasing cumulative exposure or intensity
of benzene exposure (233). Rushton and Romaniuk, did however, find excesses of
AML and monocytic leukemia in cases of workers with cumulative exposures 4.5-45
ppm-years that ever experienced daily or weekly peak exposure to benzene. An odds
ratio of 1.05 (95%CI=1.00-1.10) was calculated for a conditional logistic regression
model for AML and monocytic leukemia (p=0.04) when years at peak exposure
(weekly, > 3 ppm, 15-60 minutes) was the lone variable. The odds ratio decreased to
non-significant value (p=0.19) when the regression model consisted of both
55


cumulative exposure and years at peak exposure variables. The authors concluded
that there was some suggestion of a relationship between peak exposures and
leukemia, particularly AML and monocytic leukemia, but that further work is needed
to quantify peaks and explore the mechanisms making peak exposure-induced
leukemia possible.
5.8 OSHA BZ Exposure Limits
In 1987, the Occupational Health and Safety Administration (OSHA) amended its
existing standards for benzene for both the permissible exposure limit (PEL) and the
short-term exposure limit (STEL) (23). The rationale for the enactment, as per
OSHA, is to protect workers from benzene exposure, since
.. .employees exposed to benzene face a significant health risk and that this standard
will substantially reduce that risk. The record in this rulemaking demonstrates that
employees occupationally exposed to benzene are at risk of developing pancytopenia,
aplastic anemia, multiple myeloma, leukemia and other blood dyscrasias.
The PEL, calculated as an 8-hour time-weighted average (TWA), was reduced from
10 ppm to 1 ppm. The PEL is enforced to ensure that the cumulative benzene
exposure for workers over an 8-hour shift does not exceed a total of 1 ppm of benzene
in an 8-hour work shift. The STEL was reduced from 25 ppm to 5 ppm, and the
STEL is used to enforce that peak, or highly intense and intermittent benzene
exposure does not exceed 5 ppm over a 15 minute time period. OSHA encountered
opposition not only to the reduction of the PEL to 1 ppm, but also from those who
contended that there was no need for any STEL whatsoever with regard to benzene
exposure. Those opposed to an STEL stated that there was no conclusive scientific
evidence on which to base an STEL, since no decent animal model existed to study
benzene and STELs and exposure estimates were only approximations. Additionally,
56


those opposed to the STEL believed that the new standard (1 ppm PEL) alone would
protect workers from acute health effects and peak exposures (23).
JRB Associates conducted a feasibility analysis for OSHA in 1983 in which they
spent almost one year identifying benzene emission sources, conducting mail and
telephone surveys, visiting plants, and consulting with industry experts to determine
where and how occupational exposure to benzene may occur and how to control for
these exposure sources (23). JRB determined that workers in petroleum refinery,
petrochemical industry, and coke and coal chemical operations were likely to
experience peak exposures to benzene. More specifically, JRB identified the
following activities: sampling, railcar and tank truck loading, tank, gauging,
laboratory quality control, and maintenance as activities likely to expose workers to
short-term peak exposures. OSHA concluded that reducing the STEL was
technologically feasible in order to limit the leukemogenic potential of occupational
benzene exposure. OSHA viewed the PEL to be an ineffective method for control of
highly intense short-term exposures in those workers mentioned earlier because the
peak exposures get lost in the average... An occupational peak exposure of 480 ppm
for 1 minute and no additional exposure to benzene for the remaining 479 minutes in
the 8-hour work day would be equivalent to achieving the 1 ppm PEL. However, if
this type of exposure occurred daily, such as with work involving pipe repair, these
repeating, intermittent exposures to benzene could lead to an increased risk of
leukemia due to insufficient recovery time between acute insults to benzene that are
known to induce pancytopenia and aplastic anemia. Thus, enforcement of an STEL
would protect workers from excursions deemed permissible by the 8-hour TWA PEL.
57


5.8.1 BZ STEL Reduction
The STEL reduction from a 25 ppm ceiling and a 50 ppm peak to 5 ppm average over
a 15 minute time period was promulgated by OSHA following determination that
peak exposures seem to cause more damage than continuous exposure does at an
equal or lower exposure level (23). This determination was based on experimental
animal studies and human epidemiological data, and OSHA states that by controlling
for peak exposures, the PEL of 1 ppm would also be achieved, and that cumulative
dose might also be reduced in some situations. In summary, OSHA presented their
three arguments for the need of an STEL in addition to a reduced PEL: 1) workers
that experienced intermittent peak exposures have demonstrated significantly elevated
levels of leukemia; 2) in animal studies, low level benzene exposure (1 ppm average
exposure) resulted in chromosomal aberrations, and in humans, average exposure
levels that were slightly higher than 1 ppm, induced leukemia and chromosomal
breakage at a significant level; 3) animal data suggests that intermittent benzene
exposure increases rates of chromosomal changes and aplastic anemia more than
continuous exposure to equal or lower concentrations of benzene.
5.8.1.1 Leukemia Mortality Studies Cited for BZ STEL Reduction
Paxman and Rappaport, in their 1990 analysis of OSHAs STEL for benzene,
summarize the three human studies (two leukemia mortality studies and one
cytogenetic study) OSHA cited for the justification of firstly, keeping a STEL in
place for benzene and secondly, for reduction of the STEL from 25 ppm to 5 ppm
(24). The first mortality study cited by OSHA, according to Paxman and Rappaport,
was authored by Divine and Barron (1986) and re-examined a previous study (Divine,
Barron, and Kaplan 1985) on the Texaco Mortality Study (TMS) cohort that consisted
of approximately 19,000 refinery, research, and petrochemical white male workers
58


employed at least five years between 1947 and 1977 (234). The initial 1985 TMS
cohort study showed no statistically significant association for death from any
specific cause of death for the entire cohort (235). There was a small non-significant
elevation in SMR (SMR=118, 95% Cl 87-156) for deaths from leukemia observed for
the entire cohort. The 1986 study examined specific subgroups of workers in the
TMS cohort and observed significantly elevated SMRs for those workers employed
as pipefitter and boilermakers. An SMR of 212 (95% Cl 123-339) was observed for
pipefitters and boilermakers employed longer than one year, and an SMR of 285
(95% Cl 147-498) was observed for the same subgroup for workers employed more
than five years.
Upon closer review of OSHAs Final rule on occupational exposure to benzene
(1987), however, OSHA actually cited an even earlier study of the TMS cohort by
Divine and Barron (1983, unavailable) than the 1986 study described by Paxman and
Rappaport. In the 1983 case-control study cited specifically by OSHA in their
document, significantly elevated rate ratios were demonstrated for utilities workers
(RR=4.6), utilities controlmen (RR=5.1), and pipefitters (RR=2.7) (23). OSHA stated
that these subgroups of workers from the TMS cohort were highly likely to have
experienced short-term intermittent bursts of benzene, as major duties for pipefitters
and utilities workers included repair of broken pipes and leaking seals in streams
containing 5% to 100% benzene. OSHA concluded that the excess in leukemogenic
deaths observed was directly attributable to the mode of benzene exposure
experienced by these subgroups of workers.
The second mortality cited by OSHA was a 1983 study by Wong submitted to the
Chemical Manufactures Association (CMA) (236). 4,602 male chemical workers
59


employed for at least six months between 1946 and 1975 at one of seven plants that
experienced either intermittent exposures of benzene (low <25 ppm, medium 25-100
ppm, and high >100 ppm) or continuous exposures to benzene were compared to
3,074 workers from the same or similar plants that had experienced no occupational
exposure to benzene. Wong observed a significant increase (RR=4.66, p=0.03) in
deaths from leukemia for all benzene exposed workers (intermittent and continuous)
and a statistically significant dose-response was detected when the data was measured
by cumulative exposure. While no significant dose-response relationship was
detected for peak benzene exposures, a relative risk of 3.4 was observed for workers
that experienced peaks below 25 ppm, and workers that experienced peaks above 100
ppm, demonstrated a relative risk of 3.1. The elevated relative risks, while not
significant or consistent for a dose-response relationship, might indicate that high
levels of peak exposures to benzene may not be necessary for induction of leukemia,
and rather, low-level intermittent exposure to benzene may itself be sufficient.
Experimental animal studies cited by OSHA regarding the hematotoxic effects of
intermittent low-level benzene exposure will be examined following review of the
cytogenetic study cited by OSHA in their STEL assessment for benzene.
5.8.1.2 Cytogenetic Studies Cited for BZ STEL Reduction
A cytogenetic study by Picciano (1979) revealed statistically significant increased
rates of chromosomal aberrations in 52 workers exposed to low-levels of benzene
(less than 10 ppm) from one month to 26 years as compared to 42 workers with no
known chemical or radiation exposures (237). Picciano notes that increased rates of
chromosomal aberrations have been associated with increased rates of leukemias
(238). Exposed workers had twice as many chromosomal breaks (deletions) and
three times as many marker chromosomes (rings, dicentrics, translocations, and
60


exchange figures) as compared to non-exposed workers. Piccianos work was a
further analysis of a cytogenetic study conducted by Dow Chemical Company, whose
results were released in 1978. In the 1978 study, DOW stated that for this group of
52 employees, the 8-hour TWA for benzene exposure was between 2-10 ppm and that
peak exposure levels (from 50 ppm to greater than 100 ppm) were determined by 2-3
minute sampling periods, with ceiling concentrations of 25 ppm determined by 15-
minute sampling (23). The Dow study author, Holder, notes that while average
benzene exposures were below 10 ppm with peak exposures exceeding 100 ppm at
times, the frequency of peaking in these workers was not known. According to
OSHA, Holder implied that the chromosomal damage observed was due to peak
exposures, and that chromosomal damage was even elevated at average exposure
levels below 1 ppm. OSHA determined that workers experiencing intermittent peak
benzene exposures were at a higher risk for health effects, particularly development
of leukemia, than those workers who experienced the same dose of benzene under
continuous exposure utilizing the cumulative exposure metric.
5.8.1.3 Toxicological Evidence for BZ STEL Reduction
OSHAs review of toxicological evidence in support of the benzene STEL, included
study results presented by Dr. Richard Irons, who was Senior Scientist with the
Chemical Industry Institute of Toxicology (CUT) from 1977-1988. Irons
demonstrated that intermittent administration of hydroquinone, a benzene metabolite
Irons had previously determined responsible for bone marrow suppression, was more
potent in the induction of bone marrow aplasia (akin to aplastic anemia in humans)
than continuous administration of hydroquinone (23). Mice were either administered
hydroquinone intermittently for three days a week then given four days to recover, or
given hydroquinone continuously for five days. Irons reported that, over a 30 day
61


period, animals treated with intermittent hydroquinone exposure received
approximately 45% of the dose that the animals in the continuous exposure
experiments would have received, yet these intermittently exposed animals showed a
more pronounced increase in bone marrow suppression in addition to increased
mortality six to eight weeks following hydroquinone administration. Based on his
animal findings, Irons suggested that workers need protection from peak exposures to
benzene, and not just through the lowering of cumulative exposure as measured by
the PEL.
Dr. Raymond Tice (1986), who was a staff scientist at Brookhaven National
Laboratory, submitted data to OSHA that suggested bone marrow damage resulting
from intermittent exposure to benzene was more significant than damage observed
under continuous exposure at greater levels of benzene (23). Mice were exposed to
benzene (300 ppm for 13 weeks) for either five days per week or three days per week.
Results demonstrated that the mice with intermittent exposure (three days per week),
had longer bone marrow suppression than those mice with continuous exposure.
Tices findings were similar to those of Irons, whereby intermittent exposure was
contributing to a longer and more severe depression of bone marrow erythropoiesis
than continuous exposure.
Additional studies cited by OSHA demonstrated chromosomal damage in animals
resulting from intermittent, low-level benzene exposure. Tice et al. (1980) observed
that a single short-term (four hour) dose of benzene (3,100 ppm) significantly
increased the frequency of sister chromatid exchanges (SCEs) in cells of the murine
bone marrow (220). An inhibition of bone marrow cellular proliferation was also
reported. Tice et al. (1982) reported induction of SCEs in the murine bone marrow
62


following a four-hour inhalation exposure of 28 ppm benzene (total of 112 ppm)
(219). This result represented the lowest reported cumulative concentration of
benzene that induced cytogenetic damage of the bone marrow until the results of the
Erexson et al. study, another study cited by OSHA, were published. Erexson et al.
(1986) measured cytogenetic damage by means of SCE and micronuclei (MN)
induction in peripheral blood lymphocytes (PBLs) in both mice and rats (218). Mice
were exposed to benzene (0, 10, 100, or 1,000 ppm) for six hours, and rats were also
exposed to benzene for six hours, but at differing concentrations (0, 0.1, 0.3, 1,3, 10,
or 30 ppm). Measurable statistically significant cytogenetic damage occurred in rat
PBLs exposed to benzene concentrations 1 ppm or higher and 10 ppm or higher in
mice PBLs. The final toxicological animal study cited by OSHA, Styles and
Richardson (1984), examined cytogenetic effects on rats exposed to benzene vapor
(concentrations of 1, 10, 100, 1,000 ppm) for six hours (217). Results from this study
demonstrated significant clastogenic effects in rat bone marrow cells at 100 and 1,000
ppm. Collectively, OSHA cited the evidence from these experimental animal studies
as pertinent to justification of the STEL. Low-level intermittent benzene exposure is
capable of exhibiting more pronounced cytogenetic damage and suppression of bone
marrow cellularity in animals than continuous benzene exposure, thus suggesting that
peak exposures certainly may contribute, to some extent, to benzene-induced
leukemogenesis.
63


5.9 Possible Mechanism for BZ-Induced Leukemogenesis
5.9.1 Intermittent BZ Exposure and the Cell Cycle
Short-term, intermittent benzene administration and myelotoxicity in rat bone marrow
cells was studied cytofluorometrically by Irons et al. (1979) (216). Proliferating bone
marrow cells were evaluated following benzene exposure to ascertain if any cell
cycle-specific activity was occurring. Relative to DNA synthesis, the cell cycle may
be divided into four phases: Gl, phase during which no DNA synthesis occurs; S,
phase of DNA synthesis; G2, phase between conclusion of DNA synthesis and
mitosis, during which the cell contains twice the amount of DNA content (diploid)
found in Gl (haploid); and M, mitosis, or physical cell division phase.
Cytofluorometric DNA analysis revealed that benzene cytotoxicity did not affect the
entire cycling cell population, rather only affecting cells in either the G2 or M phases
of the cell cycle. These finding suggest benzene cytotoxicity was arresting the cell
cycle in G2 or M phase, thus affecting the bone marrow cell population undergoing
early differentiation while sparing non-dividing cells. Benzene and its metabolites
had previously been shown to cause such mitotic abnormalities as metaphase arrest
(239, 240). Irons et al. speculate that due to this cell cycle-specific effect, a rapid
reduction in bone marrow cellularity might be occurring, initially with a loss of
differentiating cells, leading to an eventual loss of mature cells as well, as the
progenitor bone marrow cell population decreases.
Irons conducted a further cytofluorometric study examining cell cycle dynamics of
proliferating bone marrow cells in 1981; his findings demonstrated toxicity from
benzene exposure affected proliferating bone marrow cells late in S phase and led to
an accumulation of these proliferating cells in G2/M phase (215). Scientific literature
supports the hypothesis that cytotoxic agents may prevent proliferating cells from
64


proceeding through the cell cycle (241). Under normal conditions, a population of
cells undergoes asynchronous cell division and differentiation, but if a cytotoxic
agent, such as benzene, is able to block progression through the cell cycle at a distinct
phase, the population of dividing cells may become synchronized or partially
synchronized (215). The surviving cell population will then accumulate at the phase
of cell cycle block, as evidenced by Irons findings, and a population of
undifferentiated, or immature cells will arise. Tunek et al. (1981) demonstrated that
high levels of intermittent benzene exposure effectively reduced bone marrow
cellularity by 86-95% in the bone marrow of mice (221). Regeneration of bone
marrow cells occurred following cessation of benzene exposure, and the authors
observed that the population of granulopoietic stem cells in S phase was higher than
in controls, indicating an accelerated proliferation rate.
Dempster and Snyder (1991), however, reported observing no elevation in
granulopoietic stem cell numbers in S-phase of the cell cycle following low-level (10
ppm), short-term benzene exposure (242). Granulopoietic stem cells did increase
their rate of differentiation and produced a higher number of progeny, while the
percentage of this population of granulocytic progenitor cells in S phase remained
unchanged (no proliferation was observed) after benzene exposure. Dempster and
Snyder hypothesize that recovery of the granulocytic progenitor cells was likely a
result of induction of the pluripotent stem cell population post-benzene exposure.
Gill et al. proposed a similar theory in their 1980 study that examined the importance
of pluripotent stem cells and benzene exposure (243). In response to a benzene-
induced depletion of a differentiated granulocytic stem cell population, pluripotent
stem cells, which typically divide slowly, may be recruited into a heightened state of
proliferation that in turn, may render the pluripotent stem cells more vulnerable to
65


injury from benzene exposure, and possibly development of aplastic anemia and
leukemia (243-246).
Benzene is globally recognized as a known human carcinogen, and more specifically
as a leukemogen, based on consistent epidemiological and toxicological evidence.
The hematoxic nature of benzene is also well documented, as benzene is capable of
negatively affecting bone marrow and peripheral blood cellularity. The Pliofilm and
NCI/CAPM cohort study results have clearly demonstrated a cumulative exposure
dose-response relationship between benzene and leukemia mortality, particularly for
AML. While most epidemiological studies have determined that cumulative benzene
exposure is responsible for excess leukemia risk, several studies have shown that
peak exposures to benzene may also contribute to its leukemogenic potential. Some
of these studies have chosen to disregard such findings, either for a desire to maintain
consistency with previously reported results or because of difficulties in the
interpretation of their results due to a high level of correlation between cumulative
and peak exposures. In addition to the epidemiological study findings demonstrating
the significance of peak benzene exposure to leukemogenesis, animal studies cited for
OSHAs 1987 benzene STEL reduction, also provide suggestive evidence for an
association between intermittent benzene exposure and both a loss of bone marrow
cellularity and chromosomal damage. This animal data suggests that intermittent
benzene exposure produces stronger genotoxic and cytotoxic effects than continuous
benzene exposure to equal or lower concentrations of benzene. A mechanism by
which peak benzene exposure may induce leukemogenesis via cell cycle block and
creation of a synchronous population of proliferating bone marrow cells was also
examined.
66


6. Topoisomerase-II Inhibitors (Epipodophyllotoxins)
6.1 Induction of Secondary Leukemias (s-AML) Following Chemotherapy
Therapy-related (t-AML) or secondary acute myeloid leukemia (s-AML) is a
complication that arises from chemotherapeutic cancer treatments. Secondary AML
can develop in both children and adults and there appears to be no increased
susceptibility in children based on age at time of treatment (247). The two classes of
chemotherapeutics associated with the induction of secondary leukemias are the
alkylating agents (melphalan, cyclophosphamide, cisplatin) and topoisomerase-II
inhibitors (epipodophyllotoxins and anthracyclines). Kyle et al. (1970) initially
reported observing s-AML cases in patients being treated with melphalan for multiple
myeloma (248). Since this report, the leukemogenic potential for most of the other
alkylating agents has been realized following primary treatment for hematologic
neoplasms (Hodgkins disease, Non-Hodgkins lymphomas, acute lymphoblastic
leukemia), solid tumors (breast, lung, and testicular cancer), and nonmalignant
diseases (rheumatoid arthritis, psoriasis) (249). Pui et al. (1989) first observed that
topoisomerase II-inhibitors, specifically epipodophyllotoxins, administered to
children with acute lymphoid leukemia (ALL), demonstrated increases in cumulative
risk for s-AML (250). The variables associated with the development of secondary
leukemias following administration of either alkylating agents or epipodophyllotoxins
for the treatment of primary cancers are genetic host factors, concomitant medication,
and schedule of administration (251). Numerous studies have demonstrated that high
cumulative doses of alkylating agents are associated with s-AML (252-258). In a
1987 study of 391 patients treated with alkylating agents for Hodgkins disease, total
cumulative dose was found to be directly proportional to the cumulative risk of s-
AML. The cumulative 10-year risk for patients with a low cumulative dose was
6.4%, intermediate dose was 11.3%, and for high dose was 37.5% (258). However,
67


unlike with alkylating agents, the administration of frequent, intermittent doses of
epipodophyllotoxins, and not the cumulative dose, has been associated with induction
of s-AML (251, 259-269). Hijiya et al. state in their 2009 review that
epipodophyllotoxins provide the best-documented evidence for topoisomerase II-
inhibitor-associated s-AML and that this compelling evidence arises from studies
demonstrating that the scheduling of epipodophyllotoxins is a stronger indicator of
leukemogenesis than cumulative dose (251). Weekly or twice weekly
epipodophyllotoxin administration is characterized by intermittent dosing, often at
high frequency and intensity.
The schedule of administration associated with increased rates of induction of s-AML
resembles the peak exposure dose metric (high dose with intermittent exposure)
possibly responsible for the observed elevated rates of leukemogenesis in individuals
occupationally exposed to 1,3-butadiene, formaldehyde, and benzene. As detailed
below, patients treated with overall low cumulative doses of epipodophyllotoxins
would be expected to have lower risk for secondary leukemias, however, often the
opposite was observed, and these patients actually demonstrated greater risk for s-
AML than patients who had received higher cumulative doses.
6.1.1 Alkylating Agent Chemotherapy and s-AML Induction
Therapy-related leukemia due to alkylating agent chemotherapy is characterized by
partial or complete loss of chromosomes 5 (5q-/-5) and/or 7 (7q-/-7) with an
occurrence rate of 85 to 95% (270, 271). The causality of secondary leukemias
resultant from alkylating agent chemotherapy may also be attributed to cytogenetic
damage created through DNA adduct formation, as the alkylating agent binds to
purine and pyrimidine DNA bases, that in turn, leads to gene mutations (272).
68


Conversely, some alkylating agents may induce covalent cross-linking between the
two strands of DNA, leading to prevention of transcription and replication (273-275).
Alkylating agents target cells in all phases of the cell-cycle, and if damaged DNA is
not corrected by DNA repair mechanisms, impacted cells will either be killed or
undergo mutagenesis (276). The mean latency period for development of s-AML is
5-7 years following therapy (277), with initial disease presentation as pre-leukemic
pancytopenia or myelodysplastic syndrome (MDS), and eventual progression to AML
(278). Certain required genetic events following chromosomal loss may explain the
longer latency period observed as MDS transforms to s-AML (277). The Ml and M2
FAB types (myeloblastic) are commonly associated with alkylating agent-induced s-
AML (251). The incidence rate for patients treated with alkylating agents ranges
from 1 to > 20% and is dependent upon regimen (253, 254, 265, 279-284). In
addition to high cumulative dose, risk factors such as age, radiation therapy, and
splenectomy increase secondary leukemogenic potential (252, 258, 285-289). Certain
host factors may also genetically predispose individuals to alkylating-agent-related s-
AML. These include a primary incidence of genetic disorders such as
neurofibromatosis type 1, Fanconi anemia, defects in the GSTT1 (glutathione
transferase theta 1) gene, and germline p53 mutations, as observed in children with
rhabdomyosarcoma (277, 290-293).
6.1.2 Epipodophyllotoxin Chemotherapy and s-AML Induction
While s-AML due to alkylating agent treatment typically occurs following a pre-
leukemic phase and a latency period of 5-7 years, s-AML following
epipodophyllotoxin (teniposide and etoposide) therapy occurs relatively quickly, with
a latency period of 2-3 years and no pre-leukemic stage (251). Epipodophyllotoxin
therapy-related s-AML is typically of the M4 (myelomonocytic) or M5 FAB subtype
69


(monoblastic), but other subtypes have also been reported (250). The incidence rate
for s-AML following epipodophyllotoxin treatment is approximately 2-12% (268).
Although cumulative dose and s-AML have been shown to be strongly associated
with alkylating agent therapy, intermittent weekly or twice weekly scheduling of
epipodophyllotoxin therapy appears to be a stronger indicator for leukemogenic
potential. According to Hijiya et al. (251), data regarding s-AML following
epipodophyllotoxin therapy and cumulative dose are inconsistent (260, 266, 294), and
observations have been made that prolonged treatment administration with low doses
of epipodophyllotoxins may actually reduce the risk of s-AML (295, 296). Other risk
factors associated with increased incidence of s-AML due to epipodophyllotoxin
treatment include primary incidence of ALL (265, 297) and the contributions of
several host factors, specifically polymorphisms of a CYP3 A (298), glutathione S-
transferase PI (GSTP1) (299), thiopurine methyl transferase (TPMT) (300), and
NAD(P)H:quinone oxireductase (NQOl) (301) genes (251). Additionally,
treatments with supportive medications, particularly L-asparaginase therapy (302,
303) and administration of concomitant medications, including alkylating agents
(259, 294, 304, 305) and anti-metabolites (260, 262) have also been associated with
increased rates of susceptibility to s-AML (251).
6.1.2.1 Epipodophyllotoxin Mechanisms of Action
DNA topoisomerase II is an ATP-dependent enzyme responsible for the relaxation of
supercoiled DNA for the purpose of making DNA available for transcription factors
and other DNA-binding proteins (249). This relaxation is made possible by the
cleavage and subsequent re-ligation of both DNA strands. Normally, these DNA
strand breaks are repaired as homologous chromosomes realign. With exposure to
DNA topoisomerase II inhibitors, the rate of DNA re-ligation is decreased, as the
70


covalent DNA-enzyme molecule forms a stable complex, thus inhibiting normal
DNA repair (265, 306-310). The stable DNA-enzyme complex interrupts normal
homologous chromosomal recombination and induces illegitimate recombination of
non-homologous chromosomes, leading to sister chromatid exchanges (SCEs) and
chromosomal aberrations (CAs), such as balanced translocations that induce
mutational transformations in normal myeloid progenitor cells (260, 311-313).
The epipodophyllotoxins are non-intercalating DNA topoisomerase II inhibitors,
while anthracyclines are intercalative agents (265). Most epipodophyllotoxin-related
leukemogenesis is characterized by breakpoints within the MLL gene cluster
involving translocations to chromosome band 1 lq23, and to a lesser extent,
translocations breakpoints at t(8;21), which joins the ETO and AML1 genes; t(3;21),
which fuses AML1 and EVIJ, EAP, and MDS1 genes; inv(16), which involves
rearrangement of CBFB and MYH11 genes; t(8; 16), which creates MOZ/CBP fusion
gene; t(15;17), which disrupts the PML and RARa genes; and t(9;22), which creates
the BCR/ABL fusion gene (259, 262, 265, 314-324). The translocation breakpoints
described above have all been implicated as cleavage sites for DNA topoisomerase-II
inhibitors during anticancer therapy, and it is believed the interaction between the
epipodophyllotoxin, DNA, and DNA topoisomerase II and the resultant chromosomal
aberrations (see Figure 6.1) induce leukemogenesis (265). Treatment schedules
involving intermittent use with frequent and high doses, such as the weekly or
biweekly administration, may not allow adequate time for proper DNA repair, while
infrequent administration resulting in a high cumulative dose may actually allow
sufficient time in between doses for DNA repair (260).
71


Etoposide/topo-II/DNA complex
Replication block
I
Replication block bypass
I
Non-homologous recombination
1. Essential gene inactivation 2. Leukemic gene activation
i I
Depletion of essential gene product Transforming protein production
CelAeath Leukemic transformation
Figure 6.1 Possible Mechanisms for (I.) Cytotoxic Cell Death or (2.) s-AML Induction Following
Topoisomerase-II Inhibitor Treatment (Adapted from Smith et al. (268), Wiley-Liss, Inc., publisher).
6.2 Epipodophyllotoxin Treatment Scheduling and s-AML Induction
The 1991 study published by Pui et al. (260) examined factors that might influence
the risk of developing s-AML following epipodophyllotoxin anticancer treatment in
734 children with primary ALL (treated between May 1979 and September 1988 at
St. Jude Childrens Research Hospital in Memphis, TN), and was the first of several
studies to conclude that the schedule of epipodophyllotoxin administration, and not
the cumulative dose, was responsible for increased risk of s-AML. The authors
reported that patients receiving weekly or twice weekly doses of teniposide with or
without etoposide were 12 times more likely to contract epipodophyllotoxin-related s-
AML than patients treated with other treatment schedules (teniposide every other
week) (Table 6.1). The overall cumulative risk for developing s-AML within six
years following epipodophyllotoxin treatment was 3.8%, with a 95% confidence
72


interval (Cl) of 2.3%-6.1%, for all patients in complete continuous remission
regardless of treatment schedule.
The risks were elevated when treatment schedule subgroups were examined closer.
Patients undergoing the X-HR treatment regimen (a cumulative dose of 4,620 mg/m2,
teniposide given twice weekly) had a six year cumulative risk of 12.3% (95% 0=5.7-
25.4) and patients undergoing the XI-HR3 treatment regimen (a cumulative dose of
5,100 mg/m2 of teniposide and 9,000 mg/m2 of etoposide, given weekly) had a six
year cumulative risk of 12.4% (95% 0=6.1-24.4). Patients in the treatment group,
XI-HR2, received a cumulative dose of 5,100 mg/m2 of teniposide and 9,000 mg/m2
of etoposide every other week, and had a cumulative risk of 1.6% (95% CI=0.4-6.1).
For several other treatment groups in which epipodophyllotoxins were not given or
given at a frequency different than the weekly or twice weekly administration,
cumulative six year risk values for development of s-AML were extremely low or
negligible [the X-LR1 regimen patient group received no etoposide or teniposide and
reported a cumulative risk of 1.0% (95% CI=0.6-6.3)].
When comparing the XI-HR2 and XI-HR3 subgroup, the authors reported that both
groups received identical forms of chemotherapy and radiotherapy, and the only
differing risk factor was the scheduling of epipodophyllotoxin treatment. After
multivariate analysis in which treatment scheduling (weekly or twice weekly
administration) was adjusted for, three biologic factors (T-cell immunophenotype,
mediastinal mass, and initial central nervous system leukemia) and treatment-related
factors (cumulative dose, radiotherapy, and teniposide therapy) displayed no
significance relating to development of s-AML following epipodophyllotoxin
treatment. With competing covariate adjustments, the frequency of treatment
73


continued to be significant, displaying a relative risk of 6.7 (95% 0=1.5-30.9,
/X0.001).
A study published by Ratain et al. (1987) documented four patients out of 119
patients with advanced non-small-cell carcinoma of the lung developing acute non-
lymphocytic leukemia (ANLL) following treatment with etoposide (259). The
median cumulative dose of etoposide given to those patients who developed t-ANLL
was 6,795 mg/m2 and the median cumulative dose given to non-leukemic patients
was 3,025 mg/m (/?<0.01). The schedule of etoposide administration was weekly for
12 weeks (each dose was 300 mg/m2) and thereafter biweekly administration. The 2
year cumulative risk was calculated to be 15% 11% at two years and 44% 24% at
2.5 years (Table 6.1). The authors concluded that high cumulative doses of etoposide
might have been causative in the development of t-ALL since radiotherapy (three of
the four patients received no radiotherapy) and classical alkylating agents (none given
to any patients) did not appear to be factors. Ratain et al. did report, however, that the
increased incidence of t-ALL observed could possibly have been a result of the
unusual schedule and dose intensity of etoposide used in this study. As this study
was published prior to the pivotal Pui et al. (1991) study that demonstrated that
scheduling of epipodophyllotoxin administration, and not cumulative dose, was the
likely causal factor of s-AML, it appears to have received little attention. Yet, in the
context of this paper, it certainly merits attention at this point, as this study reinforces
the hypothesis that frequent intermittent (peak) exposures increase leukemogenic
potential.
Hijiya et al. (2007) updated the 1991 Pui et al. study for six year cumulative risk for
epipodophyllotoxin-related s-AML to include patients through 1998 (269). The
74


original study reported 21 patients developed s-AML, and the update reported 29
patients with s-AML. Original risk estimates were 12.3% for administration of
teniposide twice weekly (updated risk was 7.1%), 12.4% for administration of
teniposide and etoposide weekly (updated risk was 8.3%), and 0-1.6% (updated risk
was 0-2.0%) for other schedules, including no administration of teniposide or
etoposide (Table 6.1). Although the 6 year cumulative risks decreased from the
original estimates, all values remained significant (p=0.02). Multivariate analysis
again revealed that schedule of administration and not cumulative dose was a
significant risk factor for the development of treatment-related s-AML.
A 1993 paper by Sugita et al. (263) reported an s-AML four year cumulative risk of
18.4% in NHL and Ki-1 lymphoma patients who had received VP-16 (etoposide)
twice weekly (total of 50 weeks) for a cumulative dose of 5,600 mg/m2 on the T-8801
protocol. 5 of 38 NHL/Ki-1 patients developed s-AML, while none of the 46 patients
who underwent etoposide therapy for NHL utilizing the B-8801 protocol (VP-16
schedule was daily x 4 either 0.5 years or 1.5 years depending on cancer stage) for a
cumulative dose of either 5,200 mg/m or 10,000 mg/m (depending on cancer stage)
developed s-AML (Table 6.1). The authors reported that the elevated risk observed
for s-AML was largely dependent on the schedule of etoposide administration since
the T-8801 protocol relied on twice weekly doses with a lower cumulative dose than
that of protocol B-8801.
Winnick et al. (262) examined development of s-AML following epipodophyllotoxin
(VP-16) therapy in 205 children treated with the Dallas/Fort Worth (DFW) protocol
between January 1986 and July 1, 1991 for B-cell lineage ALL. No patients received
alkylating agent therapy or radiotherapy, and the 4 year risk of s-AML was
75


determined to be 5.9% 3.2% (Table 6.1). Scheduling of epipodophyllotoxin
administration was somewhat different from the St. Jude schedule examined by Pui et
al. (1991), and all but one patient received VP-16, not VM-26 (teniposide). VP-16
was given twice weekly every nine weeks until 30 months of complete continuous
remission, unlike in the Pui et al. study, where VM-26 (teniposide) was given twice
weekly for two consecutive weeks over a ten week period during the first year. The
cumulative dose of VP-16 given was 9.9 g/m2. The possibility exists that the
cumulative dose of VP-16 might be responsible, at least partially, for the elevated risk
reported by Winnick et al., although the authors do not mention examining
cumulative dose as a possible risk factor. Additionally, as noted in other studies
reviewed for this paper, concomitant medication may potentiate the s-AML risk of
epipodophyllotoxin-related s-AML. The authors mention that the DFW protocol
utilized methotrexate (MTX), L-asparaginase (L-asp), cytarabine (ara-C), all drugs
with known abilities to interfere with DNA repair and synthesis. These drugs, as well
as cisplatin and certainly the use of alkylating agents, have also commonly been
utilized in other epipodophyllotoxin anticancer protocols regardless of schedule of
epipodophyllotoxin administration.
Pui et al. (1995) concluded that L-asp given to patients immediately preceding
etoposide treatment increased s-AML rates, possibly due to enzymatic activity
increases in systemic exposure to etoposide or its catechol metabolites, may lead to
reductions in cellular ability to repair DNA damage (302). The authors noted that
those patients given L-asp actually had received lower cumulative doses of etoposide
(patients in Study XIII also received high doses of methotrexate preceding etoposide
administration). Patients were enrolled in Study XIII, in which etoposide
administration was every other week or less often, scheduling (Study XI) which
76


originally demonstrated very low or negligible risks overall in the 1991 Pui et al.
study. The two year cumulative risk estimate was 5.4% (95% 0=0-11%) and
significantly higher than the risk estimate (1.1%) for Study XI in the 1991 study.
Hijiya et al. (2007) updated the 1995 Pui et al. study on L-asp and s-AML and
reported a decreased but significant risk estimate of 2.6% (p=0.01) for use of L-asp
immediately preceding etoposide (269). Etoposide administration for this study did
not follow the weekly or twice weekly schedule Pui et al. and others examined, that
led to the conclusion that frequent, intermittent doses of epipodophyllotoxins elevated
risk for s-AML, but this data demonstrated that other factors, particularly other
medications known to interfere with DNA repair, certainly may affect leukemogenic
potential of DNA topoisomerase II inhibitors for cancer treatments.
While several studies published by Smith et al. examining therapy-related AML
following epipodophyllotoxin treatment do not specifically investigate scheduling,
they do provide some peripheral support for the importance of both genetic and
concomitant medication risk factors and are therefore briefly reviewed here (266-
268). The Cancer Therapy Evaluation Program (CTEP) of the National Cancer
Institute (NCI) developed a monitoring plan to assess risk estimates in patients with
pediatric and adult solid tumors treated with epipodophyllotoxins at different
cumulative doses. Smith et al. (1993) (267) identified 12 such NCI-supported
Cooperative Group clinical trials, where 1 trial used a moderate dose of teniposide
(900 mg/m2), and the other 11 trials utilized etoposide at a low dose (< 1,500 mg/m2),
moderate dose (1,500-3,999 mg/m2), and a high cumulative dose (> 4,000 mg/m2). A
six year cumulative risk estimate of 3.2% (95% Cl, 1.2-8.6) was calculated after two
cases of AML, one case of MDS, and one case of MDS progressing to AML were
documented for patients who had received a low cumulative dose of etoposide. In
77


addition to receiving low cumulative doses of VP-16 (600 or 900 mg/m2), these four
patients were also treated with a variety of alkylating agents, including cisplatin and a
relatively high cumulative dose of cyclophosphamide. When Smith et al. compared
the 3.2% six year cumulative s-AML risk probability they observed to other studies
which examined AML following alkylating agent chemotherapy (255, 279, 280, 283,
287, 289, 325), they observed similar rates. From this data, it would appear that use
of a high cumulative dose of alkylating agent chemotherapy was responsible for the
elevated s-AML risk estimates, as evidenced by the development of MDS and
progression from MDS to s-AML, but interestingly, the latency period for the four
patients ranged from 2.0 to 3.3 years, which is typically characteristic of
epipodophyllotoxin-related s-AML. The schedule of VP-16 administration was 100
mg/m2 daily for three days in the low cumulative dose group that documented the
four patients with s-MDS/AML. Of note, the four cases of s-MDS/AML were all
patients treated for rhabdomyosarcoma, and as mentioned earlier, the p53 germline
mutation associated with this disease has been implicated as a host factor that may
genetically predispose an individual to development of s-AML following alkylating
agent therapy.
While the scheduling of VP-16 does not appear to play a factor in s-AML in this
paper, the authors did note that scheduling of epipodophyllotoxin administration and
development of s-AML may be reasonably plausible, as epipodophyllotoxins are cell
cycle-specific drugs, and that considerable differences in antitumor effect were
observed in a Slevin et al. (1989) paper that examined the effect etoposide scheduling
had on etoposide activity in small-cell lung cancer (326). Chen et al. (1996) tested
leukemia cells in vitro to determine if a prolonged low dose schedule of etoposide
(compared to short exposure with high concentration) could produce the desired
78


cytotoxic effects of anticancer therapeutics while decreasing non-homologous site-
specific DNA recombination (296). After 6 days of culture, leukemia cells receiving
the prolonged low dose demonstrated a significantly lower deletion frequency when
compared to the high dose/short exposure to etoposide (p=0.0003).
Table 6,1 Risk of s-AML following epipodophyllotoxin therapy administered on various schedules
# # of s- Cumulative dose Cumulative risk
Study Schedule Patients AML (mg/m2)a of s-AMLb
Ratain et al. (259) Weekly X 12 weeks, then biweekly 119 4 4,382-7,950 15% 11% at 2 years 44% 24% at 2.5 years
Pui et al. (260) Twice weekly 85 6 9,240 12.3% at 6 years (5.5-27.4)
Weekly 84 7 19,200 12.4% at 6 years (5.9-26.0)
Every other week 148 2 19,200 1.6% at 6 years (0.4-6.1)
Amylon et al. (261) Twice weekly 646 14 7,200 7.2% at 5 years (4.3-12.2)
Winnick et al. (262) Twice weekly 203 10 5,100-9,900 5.9% at 4 years (3.2-11.0)
Sugita et al. (263) Twice weekly 38 5 5,600 18.4% at 4 years (7.7-44.2)
Daily X 4 46 0 5,200-10,000 0% at 4 years (< 6.3)
Smith et al. (266) 2 or 3 X weekly 451 8 600-900 2.1% at 4 years (< 3.7) 3.3% at 6 years (<5.9)
5 days/week 1270 4 1,500-3,000 0.4% at 4 years (<1.0) 0.7% at 6 years (<1.6)
5 days/week 570 5 3,000-10,000 1.4% at 4 years (<2.9)
2.2% at 6 years
(<4.6)
79


Table 6.1 (Cont.)
# # of s- Cumulative dose Cumulative risk
Study Schedule Patients AML (mg/m2)a of s-AMLb
Hijiya et al. (269) (update of (260)) Twice weekly 85 6 9,240 7.1% at 6 years (1.6-8.3)d
Weekly 84 7 19,200 8.3% at 6 years (2.4-14.2)d
Other 39-155 Unknown 0-19200 0-2.0% at 6 years (<4.4)d
c Calculation for cumulative dose (2:1 conversion factor for teniposide to etoposide) (268).
b95% confidence interval (in parantheses). Only the upper 95% confidence bound is shown if risk is 0.
c95% confidence interval calculated by Smith et al. (268) using equation provided by Smith et al. in
(267).
Standard error (SE) given by Hijiya et al. (269). 95% upper confidence bound calculated by
multiplying SE by 1.96, the 0.975 quantile for a normal distribution.
In an update of their 1993 paper, Smith et al. (1999) re-examined six year cumulative
risk estimates for patients with documented s-MDS/AML following
epipodophyllotoxin treatment through the 12 NCI-supported Cooperative Group
clinical trials (266). 17 patients were identified with s-MDS/AML (8 in the low
etoposide cumulative dose group, 4 in the moderate dose group, and 5 in the high
cumulative dose group). Patients in the low cumulative etoposide dose group
followed a schedule of administration daily for three days, while the patients in both
the moderate and high cumulative dose groups were given etoposide daily for five
days. Risk estimates were as follows: 3.3% for the low cumulative etoposide dose
group (600-900 mg/m2), 0.7% for the moderate etoposide dose group (1,500-3,000
mg/m2), and 2.2% for the high etoposide group (3,000-10,000 mg/m2) (Table 6.1).
This data reinforces the argument that factors other than cumulative dose of etoposide
were responsible for determination of s-AML risk in the 12 NCI-supported clinical
trials. While it appears likely the s-AML is related to concomitant use of other
medications, particularly alkylating agents, and genetic host factors, the authors
80


noted, as they had in the 1993 paper, the importance that scheduling of administration
may have when determining risk for secondary leukemias.
In a 1994 review article by Smith et al. (268) examining s-AML following
epipodophyllotoxin treatment, the authors cite an abstract by Amylon et al. (1992)
(261) comparing risk estimates for s-AML based on various schedules of
administration. The cumulative risk of s-AML in the Amylon et al. study is reported
to be 7.2% (95% Cl, 4.3-12.2) for twice weekly scheduling with a cumulative dose of
7,200 mg/m2 for teniposide (Table 6.1). No other information is provided
(particularly if any other risk factors, ie. cumulative dose, were estimated) and a
literature search was conducted, but this publication was not found.
Intermittent scheduling (weekly or biweekly administration) and not cumulative dose
of topoisomerase-II inhibitors (epipodophyllotoxins class) has been shown to be more
predictive of s-AML induction following primary cancer therapy. Studies have
demonstrated that patients undergoing a weekly or biweekly schedule of
epipodophyllotoxin (etoposide or teniposide) administration, resulting in an overall
low cumulative dose, are at greater risk for s-AML than patients undergoing
treatment with infrequent scheduling that results in an overall higher cumulative dose.
The intermittent scheduling of epipodophyllotoxin treatment of cancer resembles the
peak exposure metric, whereby doses are given at high concentrations for short
periods of time, potentially disallowing sufficient time for adequate DNA repair in
between administered doses. Leukemogenic mechanisms by which
epipodophyllotoxins induce s-AML may be similar to those of benzene, as both are
known to be cytotoxic, cell-cycle phase-specific compounds that elicit genotoxic
effects via mutational transformation of myeloid progenitor cells by SCEs and CAs.
81


7. Evaluation of Other Chemicals for Peak Exposure and Leukemogenesis
7.1 BD Reactive Epoxide Formation and Carcinogenesis
IARC has concluded that BD is a group 1 carcinogen, as sufficient causal evidence
exists in humans (due in part to increased mortality from leukemia) and experimental
animals (4). Based on review of mechanistic and epidemiological data on BD,
Melnick and Kohn (1995) agree that BD is a human carcinogen and state,
additionally, that the mouse provides an appropriate animal model for the study of
human cancers following BD exposure, particularly lymphohematopoietic
malignancies (327). In the mouse, animal studies have demonstrated induction of
lymphomas following BD exposure, while epidemiological studies have shown
increased mortality from leukemia and other lymphohematopoietic neoplasms
following occupational BD exposure (327).
Biotransformation of BD is believed to be a critical mechanism responsible for BD
carcinogenesis. During BD metabolism, reactive epoxides may be produced (Figure
7.1) that have been shown to be carcinogenic in humans as well as all mammalian
species (327). BD is initially oxidized to the monoepoxide l,2-epoxy-3-butene (EB)
via CYP2E1 (328). EB may then be further oxidized to l,2:3,4-diepoxybutane
(DEB), or EB may undergo hydrolysis to form l,2-dihydroxy-3-butene (BD-diol)
which may then be oxidized to 3,4-epoxybutane-1,2-diol (EBD or EB-diol) (329).
The three epoxide metabolites generated (EB, DEB, and EBD) are known alkyating
agents capable of reacting with double stranded DNA and subsequent adduct
formation. Crosslinking of DNA and BD metabolites forms N7-alkylguanine adducts
that have been found in liver DNA of mice and in the urine of humans exposed to BD
(330-335). Adducts formed with EB, DEB, and EBD are mutagenic and may lead to
genetic alterations via blockage of DNA replication or induction of incorrect
82


nucleotide incorporation (4). Unlike studies in rats, BD has been shown to be
genotoxic in the bone marrow of mice, resulting in six to eight-fold increases in
SCEs, micronuclei, and chromosomal aberrations (336, 337).
1,3-Butadiene !BDi
sr
P450
CYP2E1
CYP2A6
P450
CYP2E1
CYP3A4
1,2:3,4-Oiepoxybutane
mm)
GST
Detoxification
product
Detoxification
product
\
\
EH
\ i
O
1.2- Epoxy-3- butene
(EB>
GST
EH
I OH
HO
EH
1.2-Dihydroxy-3-butene
{80-diOl)
P450
HO

OH
o
OH
1 -Hydroxy-2-(N-acety!cysteinyi)-
3-butena (M2|
and
1 -fN-acatyteysteinyri-2-hydroxy-
3-butene
t ,2-D*ydroxy-3,4-epoxybutane
(FB-dioli
OH
'1
SR
SR
J
HO.
v-x
OH
1,2-D (N-acetylcysteiny))-
butane
(Ml)
Figure 7.1 1,3-Butadiene (BD) Reactive Metabolite Formation, (Adapted from Albertini et al.(338),
Elsevier Science, publisher).
Evidence from human and animal studies has demonstrated the lymphohematopoietic
cancer-inducing effects of BD-intermediate epoxide products. Acknowledgement of
the significance of BD-intermediate epoxides and DNA adduct formation as a
83


potential mechanism for carcinogenesis led to the evaluation of other epoxide or
epoxide-forming chemicals (ethylene oxide, vinyl chloride, chloroprene, and
isoprene) for leukemogenic potential following occupational peak exposures. In
general, all of the epidemiological studies reviewed for occupational exposure to
ethylene oxide, vinyl chloride, chloroprene, and isoprene did not specifically examine
peak exposures, and only workers with ethylene oxide exposure demonstrated
limited, inconsistent elevations for leukemia mortality.
7.2 Ethylene Oxide (EtO)
7.2.1 Characteristics, Uses, and Health Effects
Ethylene oxide, C2H4O, a monofunctional epoxide and simple ether, is a colorless,
somewhat sweet-smelling gas at room temperature that is stable in aqueous solutions,
but may be highly explosive when heated (327, 339, 340). In the U.S., ethylene oxide
production ranks among the top 25 for highest volume of chemicals produced (339).
Ethylene oxides major use is as an intermediate in the production of other chemicals,
including ethylene glycol (antifreeze) and polyesters (340). Other uses for ethylene
oxide include use as a fumigant and sterilant, particularly for medical items and
surgical equipment in hospitals and medical and dental clinics as well as the facilities
themselves (339). Additional uses for ethylene oxide include insect control for
agricultural production of nuts and spices (340). Occupational exposure to ethylene
oxide via inhalation is the most common route of exposure, although ingestion of
contaminated food or dermal exposure is also possible but less common (340).
Workers involved in production of ethylene oxide, use of ethylene oxide in the
production of other chemicals, facility sterilizations, and sterilization of medical
equipment and devices typically experience the highest levels of ethylene oxide
exposure (339).
84


7.2.2 Cytogenetic Evidence for EtO Carcinogenesis
Ethylene oxide, like the BD metabolites, is capable of producing N7-alkylguanine
adducts and has been classified as a Group 1 human carcinogen by IARC based on
limited evidence in humans and sufficient evidence in experimental animals (4).
Melnick and Kohn (1995) conclude that ethylene oxide and butadiene both induce
cancer in a similar mechanistic manner involving epoxide-DNA adduct formation
(327). Ethylene oxide has been found to be genotoxic in prokaryotic and lower
eukaryotic organisms as well as in mammals (339), hence its use as a sterilant and
fumigant. Significant increases in SCEs and chromosomal abnormalities have been
demonstrated in peripheral lymphocytes, micronuclei observed in erythrocytes, and
single-strand DNA breaks in peripheral mononuclear blood cells of workers exposed
to ethylene oxide (4, 339, 341-344).
7.2.3 Epidemiological Evidence for EtO-Induced Leukemogenesis
Epidemiological studies have not shown any consistent association for ethylene oxide
exposure and increased risk of leukemia. A series of reports by Hogstedt et al. (1979,
1979, 1986, 1988) evaluating Swedish workers in ethylene oxide production and
sterilization of hospital equipment, demonstrated significant excesses of leukemia
mortality (345-348). It was possible for these workers to have experienced peak
exposures for short time periods during earlier years of employment, but as
mentioned earlier, peak exposure was not evaluated, and efforts were made during the
mid-1970s, according to Hogstedt et al. (1986), to reduce ethylene oxide exposure
levels (345). In the 1986 study, the authors concluded that even low-level
intermittent exposure to ethylene oxide induced leukemia, based on their
epidemiologic data and existing toxicological data that indicated a causal relationship.
85


The Hogstedt findings enhanced interest to determine if ethylene oxide was indeed
leukemogenic and several studies on large cohort size (349-357) were published in
the 1980s and the 1990s attempting to verify the study findings of leukemia excess
(358). None of these studies showed any significant association between leukemia
and ethylene oxide exposure, but the Bisanti et al. (1993) study on a cohort of 1,971
Italian chemical workers did report significant excesses for lymphosarcoma and
reticulosarcoma (SMR=682, 95% CI=186-1745) (352). In the Bisanti et al. study,
The SMR was leukemia was elevated, although not significantly, at SMR=193
(95%CI=23-699). The study authors also reported that 1,334 of the Italian workers
may have been exposed to 27 other toxic gases in addition to ethylene oxide
exposure. The Hogstedt studies, like the Bisanti et al. study, also suffered from
potential concomitant chemical exposures (benzene included), in addition to small
study size and low power. A 1993 study, conducted by Stayner et al. on
approximately 18,000 workers in 13 U.S. facilities involved in medical item
sterilization using ethylene oxide with follow-up through 1987, evaluated cumulative
exposure, duration of exposure, average time-weighted exposure, and maximum time-
weighted exposure metrics (359). Stayner et al. stated it was not possible to evaluate
short-term, peak exposures since this was a retrospective study and no real-time data
existed for each cohort member. No significant trend for leukemia risk was observed
for any of the exposure metrics.
Teta et al. (1999) conducted a meta-analysis for leukemia risk on 10 independent
cohorts from five countries (33,000 workers) and reported an SMR of 1.08 (95% Cl
0.61-1.93) with inclusion of the Hogstedt studies and an SMR=0.95 (95% 0=0.64-
1.35) excluding the Hosgtedt studies, which were excluded on the basis of leukemia
heterogeneity (360). Teta et al. concluded that the association between ethylene
86


oxide and leukemia reported in the Hogstedt studies may be inaccurate due to the
findings of several studies on workers exposed to ethylene oxide in the early years of
the chemical industry that had strong follow-up coupled with the results of their meta-
analysis and tests of heterogeneity which demonstrated no increase in leukemia
mortality. Shore et al. (1993) assessed 10 epidemiological cohorts for evidence of
carcinogenicity due to ethylene oxide exposure (361). SMR was 1.06 (95% Cl 0.73-
1.48) for leukemia and the authors also evaluated trends by intensity or frequency of
exposure, by cumulative exposure, by duration of exposure, and by latency (time
since first exposure). No significant trends were observed for the four exposure
metrics and leukemia. Peak exposure trends were not evaluated. Three more recent
epidemiological study updates also found no evidence for any association between
ethylene oxide exposure and leukemia (362-364).
7.3 Vinyl Chloride (VC)
7.3.1 Characteristics, Uses, and Health Effects
Vinyl chloride, a chlorinated aliphatic hydrocarbon, is a colorless gas with a
somewhat sweet odor at room temperature (365). Vinyl chloride is utilized almost
entirely (95% of vinyl chloride monomer) for the production of polyvinyl chloride
(PVC) and copolymers. PVC is a plastic resin that may be found in furniture,
automobile parts, flooring, medical supplies, and irrigation systems (365). Exposure
to vinyl chloride in the general population may occur through inhalation, ingestion, or
dermal routes, but exposure levels are typically considered very low (366).
Occupational exposure to vinyl chloride involves post-production activities, such as
piping of the monomer for transport, usage, storage, or during polymerization to form
PVC resins (365).
87


7.3.2 VC Carcinogenicity (Liver Angiosarcoma and Leukemia)
IARC has concluded that vinyl chloride is a Group 1 human carcinogen based on
sufficient evidence in both humans and experimental animals (4). Vinyl chloride is
oxidized by CYP2E1 to chloroethylene oxide, an epoxide, which in turn may be
rearranged to form chloroacetaldehyde (367). Both of the vinyl chloride metabolites,
chloroethylene oxide and chloroacetaldehyde, are capable of DNA adduct formation
(368), and subsequent carcinogenicity through genotoxic mechanisms likely similar
to that of BD and ethylene oxide. Animal studies in mice and rats have revealed that
inhalation of vinyl chloride induces angiosarcomas (numerous sites) and mammary
tumors, along with lung tumors in mice and hepatocellular carcinomas in rats (369-
371). The IARC conclusion on the carcinogenicity of vinyl chloride is based on
strong evidence for angiosarcoma of the liver and hepatocellular carcinoma formation
in humans (4, 369, 372-375). Overall, there has not been any association found
between leukemia and vinyl chloride exposure (373, 374, 376-379). A German study
by Weber et al. (1981) (380) and a study from the former Soviet Union by Smulevich
et al. (1988) (381), both showed significant excesses of lymphohematopoietic
malignancies, although disease classification did not distinctly evaluate for leukemia.
In 2000, Kielhom et al. reviewed the epidemiological literature on occupational
exposure to vinyl chloride to determine if tighter exposure standards had reduced the
risk of liver angiosarcoma (382). As part of their assessment, Kielhom et al. also
evaluated lymphohematopoietic neoplasm data and concluded that overall results
from the two studies demonstrating excesses in lymphohematopoietic malignancies
were not significant. A meta-analysis of epidemiological studies evaluating
occupational vinyl chloride exposure conducted by Boffetta et al. (2003)
demonstrated no increase in SMR for leukemia (383). Boffetta et al. also analyzed
and categorized the findings by Weber et al. (1981) and Smulevich (1986), and
88


concluded no increased SMR for any of the three categories, of which one category
was leukemia.
7.4 Chloroprene
7.4.1 Chloroprene Characteristics, Uses, and Health Effects
Chloroprene, or 2-chloro-1,3-butadiene, is a colorless liquid with a strong odor at
room temperature (384). When exposed to fire, chloroprene is highly explosive, and
at room temperature chloroprene is highly reactive and may polymerize or form
peroxides (385). The main use for chloroprene is as a monomer for the production of
polychloroprene (neoprene) elastomer, a synthetic rubber found in footwear, wetsuits,
laptop computer sleeves, flame-resistant cushions, and roof coatings (386).
Chloroprene does not occur naturally in the environment and the main route of
exposure is inhalation, particularly for workers manufacturing either chloroprene or
polychloroprene (384). IARC has concluded that chloroprene is possibly
carcinogenic to humans (Group 2B) based on sufficient evidence in animals and
inadequate evidence in humans (386). Chloroprene biotransformation is believed to
occur via oxidation to form epoxide intermediates, 2-chloro-3,4-epoxybutene-1 and
2-chloro-1,2-epoxybutene-3 (387). According to NTP, evidence suggests that
chloroprene and BD induce all of the same types of tumors except for lymphomas and
ovarian malignancies in mice (384), indicating a potential lack of leukemogenic
potential in humans, as based on findings for BD exposure in mice (327).
7.4.2 Chloroprene Carcinogenicity (Leukemia)
IARCs evaluation of the carcinogenic potential of chloroprene was based on two
studies (388, 389), for which no consistent excess for cancer at any site was observed
(386). In 2007, Marsh et al. conducted a historical cohort study (consisting of
89