Workplace
Original article
1
Department of Public Health
Sciences, Queen’s University,
Kingston, Ontario, Canada
2
University of Montréal
Hospital Research Centre
(CRCHUM), Health Innovation
and Evaluation Hub, Montréal,
Québec, Canada
3
Department of Social and
Preventive Medicine, University
of Montréal, Montréal, Québec,
Canada
Correspondence to
Dr Vikki Ho;
​vikki.​ho@u​ montreal.​ca
Received 30 December 2017
Revised 18 April 2018
Accepted 28 April 2018
To cite: Warden H,
Richardson H,
Richardson L, et al.
Occup Environ Med Epub
ahead of print: [please include
Day Month Year]. doi:10.1136/
oemed-2017-104987
Associations between occupational exposure to
benzene, toluene and xylene and risk of lung cancer
in Montréal
Hunter Warden,1 Harriet Richardson,1 Lesley Richardson,2 Jack Siemiatycki,2,3
Vikki Ho2,3
Abstract
Background  Benzene, toluene and xylene (BTX) are
aromatic hydrocarbons with inconclusive evidence of
lung carcinogenicity. The aim of this research was to
assess the associations between occupational exposures
to BTX agents and lung cancer.
Methods In a population-based case-control study
of lung cancer, occupational histories were obtained
and exposures were assessed by experts. Unconditional
multivariate logistic regression was used to estimate ORs
and 95% CIs, among men, between various metrics of
occupational exposure to BTX and lung cancer, while
adjusting for established and possible risk factors.
Results Considerable overlap was found between
occupational exposure to BTX, where the majority of
exposed participants were exposed to all three chemicals.
Lung cancer was associated with exposure to benzene
(OR=1.35; 95% CI 0.99 to 1.84), toluene (OR=1.31;
95% CI 0.99 to 1.74) and xylene (OR=1.44; 95% CI
1.03 to 2.01). While these results were adjusted for
smoking and other recognised and possible lung cancer
risk factors, they were not mutually adjusted among the
three BTX agents.
Conclusions  Our study provides suggestive evidence
that occupational exposure to one or more of the BTX
agents may be associated with lung cancer.
Introduction
Lung cancer is the leading cause of cancer death
worldwide, accounting for 1.69 million deaths in
2015.1 Though male lung cancer incidence has
plateaued and decreased in the Western world over
the last 25 years, increases are still expected in other
parts of the world.1 Cigarette smoking is by far the
strongest modifiable risk factor for lung cancer.
However, lung cancer is a multifactorial disease and
other determinants can significantly contribute to
the burden of lung cancer even among smokers; it
is estimated that 5%–20% of male lung cancer cases
may be attributable to occupational exposures.2–4
Benzene is an aromatic hydrocarbon that is used
in a wide variety of industries as a solvent and
reagent. It is particularly important in the produc-
tion of plastics, dyes, adhesives, pharmaceuticals
and pesticides. Benzene has long been recognised as
leukaemogenic and is classified as a Group 1 carcin-
ogen by the International Agency for Research on
Cancer (IARC).5 There has been some question as
to whether benzene exposure increases the risk of
Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987
Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
Key messages
What is already known about this subject?
►► Benzene, toluene and xylene (BTX) are aromatic
hydrocarbons with inconclusive evidence of
lung carcinogenicity.
►► Previous research has often suffered from
residual confounding by smoking and
occupational coexposures and few conclusions
can be drawn from the literature.
What are the new findings?
►► The aim of this research was to assess the
associations between occupational exposures to
BTX agents and lung cancer.
►► Our study provides suggestive evidence that
occupational exposure to one or more of the
BTX agents may be associated with lung cancer
in men.
How might this impact on policy or clinical
practice in the foreseeable future?
►► Lung cancer is a significant contributor to the
burden of cancer worldwide and BTX agents are
chemicals found in residential, commercial and
industrial settings.
►► More studies are warranted to support the
findings of our research in order to better
inform future occupational and public health
policies.
lung cancer, which could be a much more signifi-
cant contribution to the national burden of cancer
than its impact as a leukaemogen; but despite the
many results reported on the topic, the evidence is
inconclusive.6–21 The majority of studies concerned
cohorts of petroleum and chemical workers, most
of which found no association.12–21 A few of the
studies, including some with quite extensive expo-
sure assessment protocols, did find increased risk
of lung cancer among workers with benzene expo-
sure.6 7 10 The retrospective cohort studies that
probed this association suffered from the typical
limitations of this design, most importantly in this
case, the possible confounding by factors unavail-
able to the investigators, notably smoking. The
question of benzene’s lung carcinogenicity remains
an open and important one to resolve.
1
 Workplace
The recognition of benzene’s toxicity and carcinogenicity
spurred attempts to seek safer alternatives in many indus-
tries. Toluene and xylene, which are chemical homologues of
benzene, have increasingly been used as substitutes in some
industrial applications. As use of toluene and xylene increases,
it is important to assess their possible relationships with cancer.
Few studies have examined the association between toluene and
xylene exposure and risk of lung cancer;8 13 22–25 findings have
been inconsistent and are incomplete.
In the course of a large lung cancer case-control study
conducted in Montréal, we had the opportunity to assess associ-
ations between occupational exposure to these three agents and
risk of lung cancer. When referring to all three agents collec-
tively, for simplicity we will refer to them as benzene, toluene
and xylene (BTX).
Materials and methods
Study population
A detailed description of the study design and methodology
can be found elsewhere.26–28 Briefly, cases of lung cancer,
between the ages of 35 and 75, were identified in the 18
largest hospitals in the greater Montréal area. Eligible subjects
were identified from 1996 to 2001, with 1236 cases (471
females and 765 males; response rate=86%) and 1512 popu-
lation-based controls (613 females and 899 males; response
rate=70%) successfully recruited into the study. Controls
were selected from electoral lists and frequency-matched to
cases on age and sex. The present investigation was restricted
to male subjects because there were too few females with occu-
pational exposure to BTX to warrant their inclusion. There
was further exclusion of 38 males because of missing smoking
information. The final study sample comprised 733 male cases
and 894 male controls.
Procedures and exposure assessment
Participating subjects were interviewed in person and infor-
mation on sociodemographic characteristics and lifestyle
behaviours was collected as well as detailed job histories. For
participants who had died or were too ill to be interviewed, a
proxy respondent was used when possible. For each job held by
a study participant, the interviewer collected information on the
nature of the company or industry, the job responsibilities and
tasks, equipment or chemicals used as well as the safety measures
associated with each job. Using this information, a team of chem-
ists and industrial hygienists assessed potential exposure to 294
occupational agents in each job held, including BTX.27 28 If the
team of chemists and industrial hygienists, working by consensus
and blinded to case/control status, believed that a subject was
exposed to an agent, then they indicated this and assigned three
dimensions to that assessment: confidence in the assessment,
intensity of exposure and frequency of exposure. The confidence
score represented their confidence that the exposure had actually
occurred and was categorised as possible, probable or definite.
Intensity of exposure was standardised into low, medium and
high categories with reference to benchmark occupational situa-
tions that our team defined as representing low, medium or high
exposure situations. These variables are ordinal. Frequency of
exposure was measured by the number of hours spent at a given
exposure intensity for a typical 40 hours workweek. In addition
to the experts’ judgments of these three dimensions of exposure,
we had information on the duration of exposure in years from
the subject’s job history.
2
Statistical analyses
Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
The purpose was to estimate ORs and 95% CIs for the asso-
ciations between each BTX exposure and risk of lung cancer,
adjusting for relevant potential confounders. Each BTX expo-
sure was examined separately using unconditional logistic
regression. Several models were used, using different parameters
of the exposure variable and of the list of covariates.
Parameterisation of the BTX variables
The basic estimate of association was between subjects ever
exposed to each BTX agent compared with those who were not
exposed to any BTX agent. Participants were considered ever
exposed to a BTX agent if they had been exposed for one or
more years at a confidence score of probable or definite and
any category of intensity and frequency over the course of their
work history. Additionally, we conducted analyses with the ever
exposed variable categorised according to three metrics of expo-
sure: duration (in subcategories labelled  ≤10 years and  >10
years), average intensity (in three subcategories labelled low,
medium and high) and a measure that combines duration, inten-
sity and frequency (in two subcategories labelled non-substan-
tial and substantial). Substantial exposure required that subjects
were assigned medium or high categories of intensity for more
than 2 hours per typical workweek and duration of exposure
was ≥5 years. Also conducted was a risk analysis of a composite
variable, which defined BTX exposure as exposure to any of the
three agents (results in online supplementary table S1). Across
the separate analyses, the unexposed referent was the same and
required that persons be unexposed to all three BTX agents.
Additionally, participants who were categorised as ‘possibly’
exposed to the BTX agent under analysis were excluded, in
order to limit exposure misclassification.
Covariates
An initial analysis was conducted with only age and smoking
included as covariates. Following Leffondré et al,29 we used an
aggregate measure to synthesise the smoking history, comprising
smoking status (yes/no), cumulative amount smoked (log-trans-
formed pack-years) and a categorical representation of time
since cessation (0–2, 3–5, 6–10, 11–15, 16+ years). This aggre-
gate measure, known as the cumulative smoking index (CSI),
facilitates a precise and parsimonious model and has demon-
strated good model fit with our dataset.30 A second estimate
of the OR was conducted with a selection of non-occupational
and occupational covariates. The second analysis included, in
addition to age and the CSI, the following covariates irrespec-
tive of their correlations with the exposure or outcome: years
of schooling, median household income of the census tract of
residence, respondent status (proxy or self) and ethnolinguistic
group (francophone, anglophone, other). Further, we carried
out a backward deletion approach (p value cut-off of 0.20) to
select covariates from the following variables: alcohol consump-
tion and IARC Group 1 and Group 2A occupational lung carcin-
ogens including diesel engine emissions, crystalline silica, coke
dust, coal dust, pitch and tar, soot, cadmium, chromium, nickel,
aluminium, polycyclic aromatic hydrocarbons, wood dust,
asbestos and welding fumes.
Stratification by smoking
In addition to the exploration of the main effects of BTX expo-
sure on lung cancer, analogous analyses were conducted in
each of the BTX-lung cancer relationships within two strata
of smoking history: non-smokers/low-intensity smokers, and
Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987
 Workplace
moderate/high-intensity smokers. Lifetime low-intensity smokers

Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
Table 1  Selected descriptive characteristics of study population
were defined as those who occupied the lowest 25th percentile
or lower of CSI values, while moderate/high-intensity smokers Characteristics Categories N controls=894 N cases=733
were defined as those subjects with CSI values greater than the Age (mean±SD)* – 65.0±7.6 64.2±7.8
25th percentile. Smoking interaction models retained all of the Age group (%) ≤55 years 11.9 13.4
covariates listed within the main effects models, including the 56–65 years 32.1 37.5
CSI variable to control for residual confounding. 66–75 years 56.0 49.1
Ethnolinguistic group Francophone 64.4 77.6
Histological types (%) Anglophone 29.2 17.7
Polytomous logistic regression was used to estimate ORs for Other 6.4 4.6
the associations between each BTX exposure and three main Education (%) Elementary 35.4 44.5
histological types of lung cancer: adenocarcinoma, squamous Secondary 41.6 43.0
cell carcinoma and small cell carcinoma. To test for heteroge- Post-secondary 23.0 12.5
neity across histological types, we used the likelihood ratio test Median family income $ – 35 187±14 097 32 960±14 976
(α=0.05) to compare model fit of polytomous logistic regression CAD (mean±SD)
models that were unconstrained (i.e. where effects were allowed Median family income < $30 000 44.7 51.2
to vary across histological types for all variables) and constrained (%) $30 000–45 000 31.9 30.3
(i.e. where BTX effects were held constant but the effects of >$45 000 23.4 18.6
covariates were allowed to vary). This method has been previ- Smoking (%) Never 17.7 2.5
ously described.31
Current 29.2 67.3
Former 53.1 30.2
Sensitivity analyses Quit smoking (%) 2–<5 years ago 2.80 4.4
The main analyses were repeated in self-responding participants 5–10 years ago 6.60 5.9
only. The main analyses were repeated for benzene and for >10 years ago 43.7 19.9
toluene, but the exposed groups were redefined as those only Cigarette years – 828±794 1530±888
exposed to benzene and toluene, respectively. This sensitivity (mean±SD)†
analysis was not conducted for xylene as there was no one in the Respondent status (%) Self 90.3 60.2
study population that experienced exposure to xylene and not Proxy 9.7 39.8
to benzene or toluene. Results of these sensitivity analyses are *Age in years as determined by time of recruitment into study for controls and age
presented in online supplementary tables S2 and S3. at diagnosis for cases.
All analyses were performed with SAS V.9.4 (Cary, North †Among ever smokers, average number of cigarettes per day multiplied by duration
Carolina, USA). of smoking in years.

Results Tables 3–5 summarise the associations between BTX expo-

Table 1 describes selected sociodemographic characteristics of
the 733 cases and 894 controls. Cases were more likely to be
current smokers, less educated and were more likely to have
lived in census tract subdivisions with lower median household
income. Additionally, cases were more likely to be of French
descent and to have required a proxy respondent during the
study interview.
The three BTX exposures can occur in similar jobs. Table 2
summarises the most prevalent jobs defined by 4-digit Canadian
Classification and Dictionary of Occupations in which BTX
exposures occurred within the study population. These expo-
sures were spread over a very large number of occupations;
among the more prevalent ones were motor vehicle mechanics,
timber cutters, carpenters, cabinet and furniture makers and
shoemakers.
There was considerable exposure overlap of these agents
among study subjects. Among the 894 controls, 71% were not
exposed to any BTX, 14% were exposed to all three BTX, 8%
were exposed to toluene only, 5% to benzene only and fewer
than 1% to any other combinations of BTX (see online supple-
mentary table S4 for Pearson correlation coefficients between
BTX agents). While this overlap made it difficult to tease out
effects of the individual agents, there was sufficiently distinct
patterns of exposure among the agents that we attempted to
produce risk estimates for each BTX agent. However, we first
estimated the lung cancer risk associated with exposure to any
of the three. Ever exposure to any BTX agent was associated
with an adjusted OR of 1.29 (95% CI 0.99 to 1.68) relative
to an unexposed referent (see online supplementary table S1).
Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987
sures and lung cancer risk, respectively. The following subsec-
tions describe the associations between BTX agents and lung
cancer as well as the results pertaining to the smoking strat-
ified analyses and histological type analyses. The OR results
tended to be higher in analyses that included all covariates
compared with those that only included age and CSI.
Benzene
In contrast to those unexposed to BTX, ever exposure to
benzene was associated with an OR of 1.35 (95% CI 0.99 to
1.84), while substantial exposure to benzene was associated with
a slightly higher OR, although with much wider CIs. Statisti-
cally significant heterogeneity by histological type was observed
(p=0.005), with the strongest association observed for small cell
carcinoma (OR=1.51; 95% CI 0.88 to 2.56). There was sugges-
tive evidence that the effect of benzene on lung cancer risk is
only present among moderate/high-intensity smokers.
Toluene
Ever exposure to toluene was associated with an increased risk
of lung cancer relative to the unexposed referent with borderline
statistical significance (OR=1.31; 95% CI 0.99 to 1.74). Weaker
associations with lung cancer were observed for >10 years expo-
sure and substantial exposure to toluene. Statistically significant
heterogeneity by histological type was observed (p<0.001), with
the strongest association observed for adenocarcinoma of the
lung (OR=1.56; 95% CI 1.09 to 2.24). There was no indication
that the association between toluene and lung cancer differed by
smoking group.
3
 Workplace

Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
Table 2  Most prevalent occupations with exposure to each of benzene, toluene or xylene
Occupation Benzene (Njobs=496) Column % Toluene (Njobs=659) Column % Xylene (Njobs=371) Column %
Motor vehicle mechanic and repair 13.9 9.7 17.3
Food and beverage service workers 0 10.5 0
Logging and timber-related occupations 9.3 0.5 0
Carpentry, cabinet making and wood furniture making 7.9 9.3 9.7
Shoemaking and repair 4.6 3.6 6.2
Printing press 2.8 2.7 3.2
Painting, paperhangers and decorating 4.0 5.6 8.1
Firefighter 2.4 1.8 1.6
Service station attendant 1.0 0.75 1.3
Other occupations* 54.1 55.5 52.6
*Other occupations included: chemists, janitors, tank cleaners, general labourers, aircraft mechanics, machinists, floor sanders, carpet layers, farmers, tire repairmen and chemical
mixers.

Xylene this interaction was not statistically significant. There were no

Ever exposure to xylene was associated with an increased OR for
lung cancer relative to subjects unexposed to BTX (OR=1.44;
95% CI 1.03 to 2.01). Substantial exposure to xylene was not
associated with an increased risk of lung cancer. Statistically
significant heterogeneity by histological type was observed
(p<0.001), with the strongest association observed for adeno-
carcinoma of the lung (OR=1.80; 95% CI 1.17 to 2.77). The
relationship between xylene exposure and lung cancer did not
differ by smoking group.
Discussion
The results of this Montréal-based case-control study suggest
that males exposed to BTX agents experienced modestly
elevated risks of lung cancer relative to unexposed subjects.
The association between benzene exposure and lung cancer
appeared stronger among moderate/high-intensity smokers, but
indications of interactions between smoking and either toluene
or xylene exposure, on risk of lung cancer. Analyses by histo-
logical type revealed that exposure to toluene and xylene had
stronger associations with adenocarcinoma of the lung than with
other types.
Given the high correlations in exposure patterns among BTX
agents leading to possible collinearities, we refrained from
mutual adjustment of OR estimates. This leaves the door open to
confounding among them and we cannot affirm that the effects
we observed for each agent are really due to that agent and not
to one of the others. An analysis that considered any BTX expo-
sure as a composite variable was suggestive of a positive associa-
tion with lung cancer risk.
A major strength of this study was our adjustment for
smoking based on detailed assessment of participants’ smoking
history, which, for the most part, has not been available to past

Table 3  ORs between various metrics of occupational exposure to benzene and lung cancers, among various study groups, in males, in Montréal
Study sample
Controls Cases Exposure metric N controls* N cases† OR0 (95% CI) OR1 (95% CI)
 All All Unexposed 633 486 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 180 166 1.17 (0.89 to 1.55) 1.35 (0.99 to 1.84)
 All All Exposed>10 years 83 72 1.27 (0.85 to 1.89) 1.44 (0.94 to 2.21)
 All All Substantially exposed 24 21 1.19 (0.60 to 2.39) 1.69 (0.84 to 3.42)
 All Adenocarcinoma Unexposed 633 161 1.0 (ref.) 1.0 (ref.)
 All Adenocarcinoma Ever exposed 216 52 1.12 (0.76 to 1.63) 1.35 (0.90 to  2.02)‡
 All Squamous carcinoma Unexposed 633 166 1.0 (ref.) 1.0 (ref.)
 All Squamous carcinoma Ever exposed 216 64 1.33 (0.92 to 1.91) 1.36 (0.92 to  2.00)‡
 All Small cell carcinoma Unexposed 633 81 1.0 (ref.) 1.0 (ref.)
 All Small cell carcinoma Ever exposed 216 30 1.27 (0.79 to 2.07) 1.51 (0.88 to  2.56)‡
Non-smokers and low smokers
 All All Unexposed 326 57 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 99 15 0.81 (0.43 to 1.51) 0.94 (0.49 to  1.81)§
Moderate and heavy smokers
 All All Unexposed 307 429 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 81 151 1.30 (0.94 to 1.80) 1.50 (1.06 to  2.12)§
OR0: adjusted for age and CSI.
OR1: adjusted for age, CSI, respondent status, ethnolinguistic group, years of education, median household income as well as Groups 1 and 2A occupational carcinogens: diesel
engine emissions, crystalline silica, coke dust, coal dust and welding fumes.
*81 controls were excluded from analysis because exposure to benzene was classified as ‘possible’.
†81 cases were excluded from analysis because exposure to benzene was classified as ‘possible’.
‡P value for heterogeneity across histological type ORs: 0.005.
§P value for interaction: 0.21.
CSI, cumulative smoking index.

4 Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987

 Workplace

Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
Table 4  ORs between various metrics of occupational exposure to toluene and lung cancers, among various study groups, in males, in Montréal
Study sample
Controls Cases Exposure metric N controls* N cases† OR0 (95% CI) OR1 (95% CI)
 All All Unexposed 633 486 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 216 198 1.14 (0.88 to 1.48) 1.31 (0.99 to 1.74)
 All All Exposed>10 years 123 97 0.99 (0.70 to 1.38) 1.12 (0.78 to 1.60)
 All All Substantially exposed 37 32 0.95 (0.69 to 2.26) 1.11 (0.63 to 1.97)
 All Adenocarcinoma Unexposed 633 161 1.0 (ref.) 1.0 (ref.)
 All Adenocarcinoma Ever exposed 216 73 1.27 (0.91 to 1.78) 1.56 (1.09 to  2.24)‡
 All Squamous carcinoma Unexposed 633 166 1.0 (ref.) 1.0 (ref.)
 All Squamous carcinoma Ever exposed 216 73 1.24 (0.88 to 1.75) 1.32 (0.92 to  1.90)‡
 All Small cell carcinoma Unexposed 633 81 1.0 (ref.) 1.0 (ref.)
 All Small cell carcinoma Ever exposed 216 33 1.12 (0.70 to 1.98) 1.32 (0.80 to  2.16)‡
Non-smokers and low smokers
 All All Unexposed 326 57 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 108 20 1.11 (0.63 to 1.96) 1.33 (0.74 to  2.40)§
Moderate and heavy smokers
 All All Unexposed 307 429 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 108 178 1.15 (0.86 to 1.54) 1.35 (0.98 to  1.85)§
OR0: adjusted for age and CSI.
OR1: adjusted for age, CSI, respondent status, ethnolinguistic group, years of education, median household income as well as Groups 1 and 2A occupational carcinogens: diesel
engine emissions, crystalline silica, coke dust, coal dust and welding fumes.
*45 controls were excluded from this analysis because exposure to toluene was classified as ‘possible’.
†49 cases were excluded from this analysis because exposure to toluene was classified as ‘possible’.
‡P value for heterogeneity across histological type ORs: 0.0003.
§P value for interaction: 0.97.
CSI, cumulative smoking index.

researchers investigating BTX-lung cancer associations. We
used various measures, including smoking status, duration of
smoking, time since cessation and smoking intensity, incorpo-
rated into one parsimonious measure. Nevertheless, some degree
of residual confounding is still plausible. Further, we explored
our relationships of interest within two smoking strata. A limita-
tion of these stratified analyses was the inclusion of non-smokers
and low-intensity smokers into one stratum, which likely
introduced misclassification. However, there were very few

Table 5  ORs between various metrics of occupational exposure to xylene and lung cancers, among various study groups, in males, in Montréal
Study sample
Controls Cases Exposure metric N controls* N cases† OR0 (95% CI) OR1 (95% CI)
 All All Unexposed 633 486 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 135 118 1.20 (0.87 to 1.65) 1.44 (1.03 to 2.01)
 All All Exposed>10 years 66 51 1.19 (0.76 to 1.88) 1.43 (0.89 to 2.31)
 All All Substantially exposed 26 21 0.92 (0.47 to 1.79) 1.14 (0.58 to 2.27)
 All Adenocarcinoma Unexposed 633 161 1.0 (ref.) 1.0 (ref.)
 All Adenocarcinoma Ever exposed 216 46 1.41 (0.94 to 2.11) 1.80 (1.17 to 2.77)‡
 All Squamous carcinoma Unexposed 633 166 1.0 (ref.) 1.0 (ref.)
 All Squamous carcinoma Ever exposed 216 42 1.25 (0.82 to 1.89) 1.37 (0.88 to 2.13)‡
 All Small cell carcinoma Unexposed 633 81 1.0 (ref.) 1.0 (ref.)
 All Small cell carcinoma Ever exposed 216 19 1.18 (0.67 to 2.08) 1.44 (0.98 to 2.65)‡
Non-smokers and low smokers
 All All Unexposed 326 57 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 76 15 1.09 (0.58 to 2.06) 1.27 (0.65 to 2.45)§
Moderate and heavy smokers
 All All Unexposed 307 429 1.0 (ref.) 1.0 (ref.)
 All All Ever exposed 59 103 1.23 (0.85 to 1.79) 1.54 (1.04 to 2.29)§
OR0: adjusted for age and CSI.
OR1: adjusted for age, CSI, respondent status, ethnolinguistic group, years of education, median household income as well as Groups 1 and 2A occupational carcinogens: diesel
engine emissions, crystalline silica, coke dust, coal dust and welding fumes.
*126 controls were excluded from this analysis because exposure to xylene was classified as ‘possible’.
†129 cases were excluded from this analysis because exposure to xylene was classified as ‘possible’.
‡P value for heterogeneity across histological type ORs: 0.0003.
§P value for interaction: 0.61.
CSI, cumulative smoking index.

Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987 5

 Workplace
non-smoking cases in our study population, which limited our
ability to examine associations among non-smokers only.
Although selection bias due to non-representative participation
is a potential problem in all case-control studies, the high response
rates among cases and controls in our study reduces the possibility
of such bias. Further, the expert assessment of occupational expo-
sures, which were conducted blinded with respect to case-control
status, gives us confidence that any exposure misclassification
would be non-differential. Our methodology, including the self-re-
porting of occupational histories and expert assessment of expo-
sure, has shown good reliability32 and validity.33 34
There were many study participants who had proxy respon-
dents and there was a large difference in the proportion of proxy
respondents used for cases versus controls (40% vs 10%, respec-
tively). This difference may introduce information bias; partic-
ularly since proxy respondents may systematically overestimate
or underestimate occupational exposures. Proxy respondents
have shown reasonable reliability in reporting accurate informa-
tion regarding lifestyle behaviours, but proxy knowledge of the
details of a subject’s tasks and work environment may not be
as complete as the subject might provide.35 This possible differ-
ential in the quality of information obtained from proxies may
have an impact on the ability of the experts to assess exposure.
It is reassuring that results from sensitivity analyses using only
self-responding participants were consistent with ORs observed
using all study participants.
Having analysed three agents and multiple subgroups and
parameterisations, multiple testing is of concern. We certainly
do not consider the CIs or p values to convey strong decisional
information about the plausibility of any associations. Never-
theless, these findings lend a fresh perspective to the BTX-lung
cancer literature, which currently lacks quality studies that have
reliable and validated exposure assessment methodologies and
have adequately adjusted for confounders.
The relationships between BTX exposures and lung cancer
remain inconclusive. The majority of retrospective cohort
studies of benzene-related occupations have reported no associa-
tion or a protective association with lung cancer.12–22 In contrast,
a large prospective Chinese cohort study and two retrospective
cohort studies have reported increased risk of lung cancer among
benzene-exposed workers.6 7 10 Additionally, a population-based
case-control study of lung cancer in Toronto, Ontario observed
a positive association between environmental exposure to vola-
tile organic compounds, including benzene, and lung cancer.11
From a population attributable risk perspective, the importance
of uncovering potential associations between BTX exposures
and lung cancer is high. Lung cancer is a significant contributor
to the burden of cancer worldwide and BTX agents are envi-
ronmentally ubiquitous chemicals found in residential, commer-
cial and industrial settings. In addition to the presence of BTX
in gasoline, toluene and xylene are important solvents used in
painting and rubber production occupations, both of which are
defined by IARC as occupations associated with increased risk
of lung cancer.36 37 However, there is no evidence supporting
the notion that exposure to toluene or xylene is responsible
for the increased risk of lung cancer in these occupations. The
existing body of literature on toluene and xylene exposure and
lung cancer is limited in breadth and quality of evidence. Two
Swedish cohort studies of rotogravure workers have demon-
strated that toluene-exposed workers were at a greater risk of
lung cancer relative to the general Swedish population.23 25 These
two studies used ambient workplace levels of toluene to estimate
individual-level exposure and both studies comprised persons
working in a niche industry with additional chemical workplace
6
exposures. The comparability of these two occupational cohort
Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
studies to the present investigation is limited. Prior research by
Gérin et al reported increased risks of lung cancer and squa-
mous cell carcinoma of the lung among individuals exposed to
high and medium/high levels of cumulative exposure to xylene,
respectively, in a separate Montréal-based case-control study.8
While we also observed increased lung cancer risk among indi-
viduals exposed to xylene, Gérin and colleagues did not observe
elevated lung cancer risk among subjects exposed to benzene and
toluene. Differences in BTX exposure parameterisations and low
statistical power associated with histological type analyses limit
the comparisons that can be made with the present investigation.
The three BTX agents are volatile chemicals, and inhalation is
the primary route of exposure in most settings. The metabolism
of these chemicals is concentrated in the liver, with suggestive
evidence that benzene metabolism may be initiated in the lungs
at low exposure doses.38 It is not known if toluene or xylene
metabolism is initiated in pulmonary tissue. There is some
evidence that metabolites of BTX agents are associated with
biomarkers of genotoxicity in humans,39 40 including oxidative
stress, but whether these genotoxicities manifest in cancer induc-
tion or progression is unknown.
In summary, our results suggest that exposure to BTX agents
was associated with modest increases in lung cancer risk. Risk
estimates were slightly stronger for adenocarcinoma of the
lung compared with other histological types among individuals
exposed to toluene and xylene. Smoking did not clearly modify
the effects of BTX exposures. Considering past research and the
present results, we infer that there is not yet a clear pattern of
findings regarding the lung carcinogenicity of BTX agents. The
body of evidence is rather sparse and the results inconsistent.
Further research opportunities should be sought where expo-
sure contrasts are high and exposure assessment can be of high
quality. This could be in context of cohort studies or case-control
studies, depending on circumstances.
Acknowledgements  LR contributed to the design of the studies and assisted
in the development and coordination of the data collection methods as well as
contributing to the preparation of this manuscript. Exposure assessment methods
were expertly developed and implemented by Michel Gérin, Louise Nadon, Ramzan
Lakhani, Denis Bégin and Benoit Latreille. We are grateful for the careful collection of
data by numerous research assistants and interviewers. The authors wish to express
their gratitude for the collaboration of all Montréal hospitals.
Contributors  Each author has directly participated in the planning, execution or
analysis of the study. JS and LR designed the study. HW, HR, LR and VH designed the
study’s analytic strategy and interpreted the results. JS provided expert advices in study
design and data analysis and in the interpretation of study results. HW, HR and VH
drafted the manuscript. All authors critically revised and commented on the manuscript.
Funding  Funding for the study was provided by the National Cancer Institute of
Canada, the Fonds de recherche en santé du Québec and the Canadian Institutes for
Health Research. JS’s research team was supported in part by the Canada Research
Chairs programme and the Guzzo-SRC Chair in Environment and Cancer. LR is
supported in part by a salary award (NIH-R01) from the National Institutes of Health.
VH is currently supported by the Cancer Research Society, Fonds de recherche du
Québec—Santé (FRQS) and Ministère de l’Économie, de la Science et de l’Innovation
du Québec (MESI).
Competing interests  None declared.
Patient consent  Obtained.
Ethics approval  Approval was obtained from the Institutional Review Boards of
the Institut Armand-Frappier, McGill University, the Université de Montréal and all
participating hospitals.
Provenance and peer review  Not commissioned; externally peer reviewed.
© Article author(s) (or their employer(s) unless otherwise stated in the text of the
article) 2018. All rights reserved. No commercial use is permitted unless otherwise
expressly granted.
Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987

References
1 International Agency for Research on Cancer. Lung cancer estimated incidence,
mortality and prevalence worldwide, 2012.
2 Purdue MP, Hutchings SJ, Rushton L, et al. The proportion of cancer attributable to
occupational exposures. Ann Epidemiol 2015;25:188–92.
3 Driscoll T, Nelson DI, Steenland K, et al. The global burden of disease due to
occupational carcinogens. Am J Ind Med 2005;48:419–31.
4 Parkin DM. 14. Cancers attributable to occupational exposures in the UK in 2010. Br J
Cancer 2011;105 (Suppl 2):S70–S72.
5 IARC Monographs. Overall evaluations of carcinogenicity: an updating of IARC
Monographs volumes 1 to 42, 1987.
6 Linet MS, Yin SN, Gilbert ES, et al. A retrospective cohort study of cause-specific
mortality and incidence of hematopoietic malignancies in Chinese benzene-exposed
workers. Int J Cancer 2015;137:2184–97.
7 Sorahan T, Kinlen LJ, Doll R. Cancer risks in a historical UK cohort of benzene exposed
workers. Occup Environ Med 2005;62:231–6.
8 Gérin M, Siemiatycki J, Désy M, et al. Associations between several sites of cancer
and occupational exposure to benzene, toluene, xylene, and styrene: results of a case-
control study in Montreal. Am J Ind Med 1998;34:144–56.
9 Mattei F, Liverani S, Guida F, et al. Multidimensional analysis of the effect of
occupational exposure to organic solvents on lung cancer risk: the ICARE study.
Occup Environ Med 2016;73:368–77.
10 Collins JJ, Ireland B, Buckley CF, et al. Lymphohaematopoeitic cancer mortality among
workers with benzene exposure. Occup Environ Med 2003;60:676–9.
11 Villeneuve PJ, Jerrett M, Brenner D, et al. A case-control study of long-term exposure
to ambient volatile organic compounds and lung cancer in Toronto, Ontario, Canada.
Am J Epidemiol 2014;179:443–51.
12 Satin KP, Wong O, Yuan LA, et al. A 50-year mortality follow-up of a large cohort of oil
refinery workers in Texas. J Occup Environ Med 1996;38:492–506.
13 Anttila A, Pukkala E, Riala R, et al. Cancer incidence among Finnish workers exposed
to aromatic hydrocarbons. Int Arch Occup Environ Health 1998;71:187–93.
14 Consonni D, Pesatori AC, Tironi A, et al. Mortality study in an Italian oil refinery:
Extension of the follow-up. Am J Ind Med 1999;35:287–94.
15 Wong O, Harris F, Rosamilia K, et al. Updated mortality study of workers at a
petroleum refinery in Torrance, California, 1959 to 1997. J Occup Environ Med
2001;43:1089–102.
16 Sorahan T, Nichols L, Harrington JM. Mortality of United Kingdom oil refinery and
petroleum distribution workers, 1951-1998. Occup Med 2002;52:333–9.
17 Lewis RJ, et al. Mortality and cancer morbidity in a cohort of Canadian petroleum
workers. Occup Environ Med 2003;60:918–28.
18 Tsai SP, Wendt JK, Cardarelli KM, et al. A mortality and morbidity study of refinery and
petrochemical employees in Louisiana. Occup Environ Med 2003;60:627–33.
19 Bloemen LJ, et al. Lymphohaematopoietic cancer risk among chemical workers
exposed to benzene. Occup Environ Med 2004;61:270–4.
20 Huebner W, Wocjik N, Rosamilia K, et al. Mortality updates (1970-1997) of two
refinery/petrochemical plant cohorts at Baton Rouge, Louisiana, and Baytown, Texas.
Occup Env Med 2004;46:1229–45.
Warden H, et al. Occup Environ Med 2018;0:1–7. doi:10.1136/oemed-2017-104987
Workplace
21 Gun RT, Pratt N, Ryan P, et al. Update of mortality and cancer incidence in the
Occup Environ Med: first published as 10.1136/oemed-2017-104987 on 15 May 2018. Downloaded from http://oem.bmj.com/ on 18 May 2018 by guest. Protected by copyright.
Australian petroleum industry cohort. Occup Environ Med 2006;63:476–81.
22 Svensson BG, Nise G, Englander V, et al. Deaths and tumours among rotogravure
printers exposed to toluene. Br J Ind Med 1990;47:372–9.
23 Lehman EJ, Hein MJ. Mortality of workers employed in shoe manufacturing: an
update. Am J Ind Med 2006;49:535–46.
24 Malker HS, Gemne G. A register-epidemiology study on cancer among Swedish
printing industry workers. Arch Environ Health 1987;42:73–82.
25 Lynge E, Anttila A, Hemminki K. Organic solvents and cancer. Cancer Causes Control
1997;8:406–19.
26 Gérin M, Siemiatycki J, Kemper H, et al. Obtaining occupational exposure histories in
epidemiologic case-control studies. J Occup Med 1985;27:420–6.
27 Siemiatycki J. Risk factors for cancer in the workplace. Boca Raton: CRC Press, 1991.
28 Ramanakumar AV, Parent ME, Menzies D, et al. Risk of lung cancer following
nonmalignant respiratory conditions: evidence from two case-control studies in
Montreal, Canada. Lung Cancer 2006;53:5–12.
29 Leffondré K, Abrahamowicz M, Siemiatycki J, et al. Modeling smoking history: a
comparison of different approaches. Am J Epidemiol 2002;156:813–23.
30 Leffondré K, Abrahamowicz M, Xiao Y, et al. Modelling smoking history
using a comprehensive smoking index: application to lung cancer. Stat Med
2006;25:4132–46.
31 Glynn RJ, Rosner B. Methods to evaluate risks for composite end points and their
individual components. J Clin Epidemiol 2004;57:113–22.
32 Baumgarten M, Siemiatycki J, Gibbs GW. Validity of work histories obtained by
interview for epidemiologic purposes. Am J Epidemiol 1983;118:583–91.
33 Siemiatycki J, Fritschi L, Nadon L, et al. Reliability of an expert rating procedure for
retrospective assessment of occupational exposures in community-based case-control
studies. Am J Ind Med 1997;31:280–6.
34 Fritschi L, Nadon L, Benke G, et al. Validation of expert assessment of occupational
exposures. Am J Ind Med 2003;43:519–22.
35 Boyle CA, Brann EA. Proxy respondents and the validity of occupational and other
exposure data. The Selected Cancers Cooperative Study Group. Am J Epidemiol
1992;136:712–21.
36 International Agency for Research on Cancer. Rubber-manufacturing industry Vol.
100F, Monograph, 2012.
37 International Agency for Research on Cancer. Occupational exposure as a painter.
Vol. 100 F, IARC Monographs on the Evaluation of Carcinogenic Risks to Humans,
2012.
38 Powley MW, Carlson GP. Cytochromes P450 involved with benzene metabolism in
hepatic and pulmonary microsomes. J Biochem Mol Toxicol 2000;14:303–9.
39 Xiong F, Li Q, Zhou B, et al. Oxidative Stress and Genotoxicity of Long-Term
Occupational Exposure to Low Levels of BTEX in Gas Station Workers. Int J Environ
Res Public Health 2016;13:1212.
40 Edokpolo B, Yu QJ, Connell D. Health risk assessment of ambient air concentrations of
benzene, toluene and xylene (BTX) in service station environments. Int J Environ Res
Public Health 2014;11:6354–74.
7