Objectives While several monocyclic aromatic hydrocarbons are classified as definite or possible carcinogens to humans, little data exist on their role in prostate cancer (PCa). We examined occupational exposure to benzene, toluene, xylene (BTX) and styrene and PCa risk in a population-based case–control study in Montreal, Canada.
Methods Cases aged ≤75 years diagnosed with PCa in 2005–2009 (n=1920) and population controls frequency-matched on age (n=1989) provided detailed work histories. Experts evaluated the certainty, frequency and concentration of exposure to monocyclic aromatic hydrocarbons in each job lasting ≥2 years. Logistic regression estimated OR and 95% CIs for PCa risk, adjusting for potential confounders.
Results Exposures to BTX were highly intercorrelated, except for durations of exposure at substantial levels. Ever exposure to any BTX was associated with overall PCa (OR 1.27, 95% CI 1.05 to 1.53), while the OR for styrene was 1.19. However, increases in risk were largely confined to low-grade tumours, with ORs of 1.33 (95%CI 1.08 to 1.64) and 1.41 (95% CI 0.85 to 2.31) for ever exposure to any BTX and styrene, respectively, and a duration response pattern for any BTX. Risks for low-grade tumours were elevated among men exposed ≥25 years at substantial levels of benzene (OR 2.32) and styrene (OR 2.44). Some cumulative exposure categories showed increased risks but without clear trends.
Conclusion Exposure to any BTX was associated with higher risks of overall PCa. Prolonged exposures at the substantial level to benzene and styrene increased risks of low-grade tumours. These novel findings were independent from PCa screening.
- prostate cancer
Statistics from Altmetric.com
What is already known about this subject?
Evidence about the potential etiological role of exposure to individual monocyclic aromatic hydrocarbons on prostate cancer risk is sparse and inconclusive.
What are the new findings?
Men exposed ≥25 years to benzene, toluene or xylene, or to styrene, especially at the substantial level, had higher risks of low-grade prostate cancer.
How might this impact on policy or clinical practice in the foreseeable future?
In future evaluations of the potential carcinogenic risk to humans of monocyclic aromatic hydrocarbons, prostate cancer should to be considered as an additional target site.
The aetiology of prostate cancer (PCa) remains elusive. Its only confirmed risk factors are age, first-degree family history and sub-Saharan ancestry; genetic factors explain a portion of familial cancers.1 The descriptive epidemiology of PCa suggests that the environment is at play.2 Factors with suggestive evidence include obesity, physical inactivity, alcohol, diet, night work and certain pesticides.3 Very little data exist on the potential aetiological role of occupational agents, including monocyclic aromatic hydrocarbons (MAHs), which are suspected or recognised carcinogens.
An estimated 375 000 Canadians are exposed at work to benzene, mainly in automotive repair, taxi service and in the printing industry, and 89 000 to styrene, in automotive repair and maintenance or plastic products manufacturing.4 No population estimates are currently available for toluene and xylene. Industrial production of benzene, toluene and xylene (BTX) is mainly derived from petroleum, and these are used as intermediates in the chemical industry and as solvents in paints, inks, dyes, thinners, adhesives and coatings. Exposure also occurs during the manufacturing of polymers, rubber products, shoes and synthetic fibres, in printing, painting and in leather tanning.5–7 Service station attendants and drivers may also be exposed to BTX, used as additives in gasoline.7 Styrene is produced from benzene, with uses include the production of plastics, synthetic rubber and reinforced plastics products such as watercraft and automobile parts.8
The present study investigates associations between lifelong occupational exposure to MAHs and PCa incidence.
The Prostate Cancer & Environment Study (PROtEuS) is a population-based case–control study conducted in Montreal, Canada, assessing the role of environmental factors in PCa development. Eligible subjects were men, aged <76 years at diagnosis or selection, residing in the Montreal area and registered on Quebec’s permanent electoral list. Cases were all patients newly diagnosed with primary histologically confirmed PCa across French-language Montreal hospitals in 2005–2009. These hospitals represented over 80% of all PCa cases diagnosed in the area during the study period according to registry information. The degree of PCa aggressiveness, defined by Gleason’s grade, was extracted from pathology reports. Concurrently, controls were randomly selected from the electoral list of French-speaking men and frequency-matched to cases (±5 years). Study participants (1933 cases and 1994 controls) represented 79% of eligible cases and 56% of eligible controls. This study was approved by the ethics committees of all participating institutions. All participants provided written informed consent.
Trained interviewers conducted face-to-face interviews. Proxy respondents, usually the spouse, were used for 3% of cases and 4% of controls. Information was collected on sociodemographics, anthropometric characteristics, medical history and lifestyle factors.
A semistructured questionnaire elicited a work history covering all jobs held ≥1 year during lifetime. For jobs lasting ≥2 years, subjects provided a detailed description covering workplace characteristics, tasks, products and equipment used and protective measures. Thirty specialised questionnaires were also used for complex occupations.
Occupational exposure assessment
Occupations and industries were coded according to Canadian classifications.9 10 Using expert-based assessment,11 a team of trained chemists-hygienists, blind to subjects’ disease status, reviewed each job description to assign exposures to 345 chemical agents. These included BTX and styrene and a generic category for ‘any’ MAH that also encompassed other specific compounds or exposures occurring as mixtures.
For each agent and for each job, experts assigned three semiquantitative indicators: the reliability, that is, the degree of confidence that the exposure occurred (possible, probable and certain), the relative concentration level (low, medium and high, with low representing a level above the background environmental level) and the frequency of exposure in a usual work-week (<5%, 5%–30%, 30%–90% and 90%–100%). Aside from the literature and their experience, experts consulted job-exposure profiles developed to provide guidelines on exposure patterns typically encountered in a given occupation and derived from exposure data of some 20 000 jobs coded in previous population-based studies conducted by our group in Montreal.12 All jobs were coded by two different chemists with final ratings representing a consensus.
Associations between occupational exposure to BTX and styrene, to any BTX or to any MAH and PCa risk were assessed using unconditional logistic regression. ORs for low-grade (Gleason scores <7 or [3+4]) and high-grade (Gleason scores >7 or [4+3]) PCa13 were estimated in polytomous models.
Based on an average latency time of 5 years estimated in the US population,14 we only considered probable or definite exposures occurring >5 years before the index date (diagnosis or interview). Subjects never exposed, or only exposed within the 5 years preceding the index date, constituted the ‘unexposed’ reference group.
Different exposure metrics were built: ever exposure, exposure duration to any and to substantial level (defined as exposure at medium or high concentration for more than 5% of work week) and cumulative exposure, summing the products of the concentration, frequency and duration across all jobs.
To model cumulative exposure, we transformed the ordinal frequency and concentration ratings into quantitative scores. A value was attributed to each frequency interval (eg, 5%–30% of the week) based on the number of hours/week of exposure to a given MAH in the job-exposure profiles database, averaged over a baseline working time of 40 hours. Using the median values of exposed jobs within each frequency interval, the resulting quantitative frequency scores were 2.5%, 12.5%, 50% and 100% for benzene and xylene, 2.3%, 12.5%, 50% and 100% for toluene, 2.5%, 10%, 50% and 100% for styrene and 1.9%, 12.5%, 52.3% and 100% for any MAH. To better reflect the relative weighting of concentration levels, the ordinal values of 1, 2 and 3 were transformed as 1, 5 and 25, representing an overall estimate of the relative scale between the semiquantitative concentration levels assigned by the Montreal experts across the range of agents.15 Duration categories were defined as never, <10 years, 10–25 years and ≥25 years. Cumulative exposure scores were categorised using tertiles of the distribution among exposed controls.
Regression models included age, ancestry, first-degree family history of PCa, household income, education, body mass index, type 2 diabetes, alcohol intake, smoking and occupational physical activity. Missing observations were set to zero, and an extra category (categorical covariates) or a dummy (1/0) variable (continuous covariates) was added to indicate missing data,16 which represented less than 3% of all responses for each covariate. Tests for trend were conducted by treating ordinal exposure categories as continuous variables in a logistic model (Wald test).
Sensitivity analyses were performed (1) assigning alternative weights of 1, 4 and 9 to concentration levels when calculating cumulative exposures 17; (2) restricting to subjects screened for PCa (prostate-specific antigen and/or digital rectal examination) within 2 years of the index date, thereby reducing the likelihood of undetected PCa among controls and of confounding by screening on the associations observed; (3) applying 10-year or 15-year lag times; (4) using a minimum Gleason score of 8 to define high-grade PCa; and (5) stratifying between subjects aged ≤65 versus >65 years at index date.
All analyses were performed using SAS software (V.9.3).
Characteristics of the study population
Overall, 1929 cases (including 538 high-grade PCa) and 1989 controls were included in the analyses. Cases were slightly younger than controls (table 1) and had a lower level of education yet a similar household income. As expected, cases were more likely than controls to have a family history of PCa, to be of sub-Saharan ancestry and to have recently been screened for PCa. Cases and controls had similar patterns of fruit and vegetables consumption, workplace physical activity, alcohol consumption and body mass index, while cases had slightly lower smoking rates and had a lower prevalence of diabetes than controls. Over their lifetime, subjects had held four jobs, on average (median: 4, IQR: 3–5) with a similar distribution in cases and controls.
Exposure to MAHs in the study population
A total of 1207 (31%) subjects had at least one 2-year job exposed to any MAH more than 5 years before the index date, with a mean lifetime exposure duration of 20.2 years (SD: 13.3 years). Exposed subjects were slightly older, less educated, had a lower household income and a higher tobacco consumption compared with those unexposed.
Over 90% of jobs exposed to any MAH were held by firefighters, shoemakers, motor vehicle and aircraft mechanics, marine craft fabricating, assembling and repairing workers and printers (table 2). The prevalence of exposure to individual MAHs varied between occupational groups, so did exposure dimensions. Firefighters had a high (≥89%) prevalence of exposure to all MAHs, whereas aircraft mechanics, boilermakers and machinists had a very low prevalence of exposure to individual MAHs, reflecting that exposure was mainly from mixtures that could not be assessed individually. Shoemakers, printers and mixing and blending occupations were exposed to BTX but not to styrene. Aside from firefighters and auto mechanics, moulding occupations in the rubber and plastic industries had a relatively high prevalence of exposure to styrene (18%). Corresponding figures according to industries are shown in online supplementary table S1.
Supplementary file 1
The proportions of subjects ever exposed to individual BTX or styrene were 11.2% (n=429), 11.8% (n=456), 9.7% (n=376) and 2.0% (n=78), respectively, with an average exposure duration ranging from 18.0 years for benzene to 22.3 years for styrene. Exposure durations and cumulative exposures within subjects were highly correlated between BTX, with Pearson correlation coefficients between pairs ranging from 0.76 to 0.88, reflecting substantial collinearity. Consequently, for many results, greater attention is given to BTX as a group than to individual compounds. Weaker correlations were found between individual BTX compounds and styrene (Pearson coefficients between 0.28 and 0.43). Similar patterns were observed for duration of exposure at substantial level, except for weaker correlations between benzene and toluene or xylene (Pearson coefficients of 0.46 and 0.35, respectively).
Associations between MAHs and PCa risk
In analyses including all subjects (table 3), we observed increased risks of PCa with ever exposure to any BTX (OR 1.27, 95% CI 1.05 to 1.53). ORs for ever exposure to styrene (1.19) and any MAH (1.09) were marginally elevated. Trends for higher risks were observed with increasing duration of exposure, at any and substantial levels, for any BTX and, suggestively, for benzene.
We found elevated risks in the first and second tertiles of cumulative exposure to BTX, but not in the highest tertile. None of the individual MAHs showed elevated risks with increasing cumulative exposure.
When disease aggressiveness was considered (table 4), risk estimates were generally higher for low-grade than for high-grade cancers, with none of latter reaching statistical significance. Among men with low-grade PCa, ever exposure was associated with ORs of 1.33 (95% CI 1.08 to 1.64) for BTX and 1.41 (95% CI 0.85 to 2.31) for styrene. For low-grade cancers, there was a duration-response pattern (any level) for BTX, but not for styrene, based on few exposed subjects. A duration-response trend was apparent for exposure to benzene at the substantial level; for men exposed ≥25 years, the OR was 2.25 (95% CI 1.20 to 4.21), while the corresponding figure for styrene was 2.44 (95% CI 1.16 to 5.13). Trends of higher risk for low grade PCa with increasing duration of exposure to any BTX, benzene or styrene, at any or at substantial level, were more pronounced in men aged ≤65 years at index date (table 5). Results for cumulative exposure among low-grade cancers corresponded largely to those in the full sample.
Sensitivity analyses using alternative weights for concentration levels, that is, 1, 5, 25 or 1, 4, 9, applying longer lag times (10 or 15 years), restricting subjects to those screened for PCa 2 years earlier or using a more stringent criteria to define aggressive cancers yielded similar results (data not shown).
This paper presents evidence from the largest population-based study to date evaluating the role of occupational exposure to several MAHs in PCa development. In the entire sample, we observed elevated risks of PCa with ever exposure to BTX and according to duration. Marginal increases were found for exposure to styrene or MAHs. ORs were elevated in the lowest or middle tertiles of cumulative exposure for all of the agents, but there were no dose–response patterns.
Ever exposure, duration of exposure and cumulative exposure to BTX were highly correlated, and associations could often not be disentangled. One exception was for duration of substantial exposure to benzene, which was less correlated with toluene or xylene. The elevated PCa risk with duration of exposure to benzene, at any and at the substantial level, appeared somewhat stronger and more consistent than for other MAHs and showed evidence of dose–response patterns. Given its heterogeneity in contents, the combined group of MAHs is probably less informative.
As this is, to our knowledge, the first study of exposure to MAHs in PCa risk examining tumour grade, a comparison with prior literature is not possible. The positive associations observed here were largely confined to low-grade cancers. We can speculate on potential explanations. Gleason’s grade describes tumour cell differentiation patterns, with higher grade tumours being more aggressive and having a poorer prognosis.18 It is unknown whether prostate tumours arise well differentiated and then progress to less differentiated forms or if Gleason grade is an early and largely unchanging feature. However, recent findings suggest that grade may be established early in tumour pathogenesis and that Gleason grade progression is uncommon.18 Low-grade PCa has been shown to diverge early from high-grade PCa, and there appears to be no direct progression from low grade to metastatic disease.19 Hypothetically, it may be that low-grade and high-grade PCa would share different risk factors. Prior observations, where obesity and alcohol intake were linked to aggressive PCa more specifically,20 21 support this. According to this, the occupational exposures studied might be more important aetiological factors for low-grade than for high-grade PCa.
A second potential explanation for different results between low-grade and high-grade cancers may relate to the lower statistical precision for high-grade cancers. There was indeed considerable overlap in CIs between the two groups. Third, it could be that a disease detection issue is at play, as PCa screening practices may differ among occupational groups. The positive associations observed for low-grade cancers would be consistent with a greater screening frequency among exposed subjects. However, we found no detectable differences in the reported frequency of PCa screening in the 5 years preceding the index date between control subjects exposed and unexposed to the different MAHs (data not shown). The frequency of PCa screening among cases is not informative as a higher testing frequency might reflect diagnostic efforts.
Some of the increased risks observed in this study appeared to be more pronounced among men ≤65 years of age. Early-onset (≤55 years) PCa is thought to be distinct aetiologically, often including a more significant genetic component.14 Striking differences in the mechanistic landscapes of structural genomic alterations have been observed between early-onset (≤50 years) and elderly onset PCa.22 However, diagnoses before age 50 or 55 years are rare (typically 2% and 10% of new diagnoses, respectively), and there were too few subjects in our sample to study them. A census-based study from Nordic countries reported higher risks of PCa before age 50 years among firefighters and military personnel than for later-onset disease, raising the possibility that occupational exposures might be more relevant for the younger age group.23
Comparison with other studies
While increased risks of PCa have been reported in certain occupational groups, including in this study sample,24 very little evidence has been accrued on the role of specific occupational agents. The strongest evidence yet is probably for specific pesticides.25
Industry-based studies have evaluated the risk of PCa resulting from exposure to individual MAHs,26–29 especially styrene, with inconclusive evidence. Studies of workers in the reinforced plastics industry,30–34 characterised by higher and more specific exposures to styrene,35 found no increase in PCa mortality30 31 33 34 or incidence,32 nor was there increasing PCa risk with higher exposure,30 31 34 cumulative exposure,33 34 exposure duration,32–34 time at first exposure or exposure probability.32 Most studies relied on mortality data, which is less well suited for studying PCa aetiology, and were limited by the paucity of information on lifestyle factors. The lack of detailed administrative records also precluded taking into account the full variability in exposure between workers within a department30 or company.32 Two nested case–control analyses of industrial cohorts investigated the relationships between PCa mortality36 or incidence37 and exposure to specific MAHs, reporting positive associations with exposure to benzene37 or xylene.36 A review focusing on styrene exposure suggested no overall association with PCa.35
To our knowledge, the only other population-based study exploring associations between several MAHs and PCa risk was conducted in Montreal in the 1980s, comprising 449 cases.17 While no association was found with exposure to BTX, exposure to styrene at the high/medium level was associated with increased risk (OR 5.5, 95% CI 1.4 to 21.8, seven exposed cases).
Detailed descriptions were obtained for jobs lasting ≥2 years to limit interview and exposure assessment burden. However, this had a minimal impact as shorter jobs represented <4% of the overall career coverage in this population.
The exposure assessment method provided semiquantitative estimates of the frequency and concentration. Computing cumulative exposure required assigning quantitative scores to better reflect absolute exposure levels for these dimensions. We applied two different arbitrary scales felt to approximate absolute concentration levels across categories. No quantitative measurement data are available to support one scale over the other, but the interpretation of findings was not altered according to the scale applied.
Although the sample size was relatively large, the number of exposed subjects was more limited for styrene and for analyses based on categories or subgroups, which may have contributed to statistical instability and compromised our ability to detect associations.
While participation rates in our study were comparable or better as compared with other studies entailing extensive inperson data collection, they were more modest (56%) for controls. This could have influenced our results if participation was associated with socioeconomic characteristics that were also associated with MAH exposure. However, according to census tract data, the rates for recent immigration, unemployment, low educational level and low household income were similar in living areas of participants and non-participants, both among cases and controls, indicating that the potential for selection bias in this study is limited.
We had information on multiple suspected potential confounders for PCa, such as socioeconomic and lifestyle factors, and these were included in our models. While coexposure with workplace agents such as other solvents, lubricants or PAHs could have occurred, no occupational agents could be considered as clear a priori confounders in the MAHs–PCa associations, other perhaps than for some specific pesticides. We did not have information on the latter, although their prevalence of exposure is expected to have been very low in our population. Besides, our study covered exposures across a wide range of occupations in the population, rather than being specific to a limited number of occupational settings where potential confounding by coexposures might be stronger.
Our study benefited from an expert-based exposure assessment approach, which is recognised as the reference method for such a study design.38 39 Evaluations of the data collection and exposure assessment methods in previous case–control studies in Montreal have shown that the job histories were valid40 and that the exposure codings were reliable and highly sensitive.41–43
The large sample size, the availability of information on disease aggressiveness and the possibility to account for disease latency are also noteworthy. The study was set in a population with free and universal access to healthcare and which was relatively uniformly and regularly screened for PCa at the time of subjects’ ascertainment, thereby reducing the potential for disease misclassification due to underdetection of PCa among controls. Moreover, we had the ability to restrict analyses to recently screened subjects, which yielded results analogous to those in the main analyses.
The International Agency for Research on Cancer classified benzene as carcinogenic to humans and styrene as possibly carcinogenic to humans, primarily from associations with lymphohaematopoietic cancers.44 Toluene and xylene are not classifiable as to their carcinogenicity. There is strong evidence of genotoxic effects of benzene metabolites on haematopoietic stem cells.45 46 Initiation of PCa has been hypothesised to result from oxidation of the metabolites catechols of estrogens or benzene to quinones, which react with DNA to form depurinating adducts. Apurinic sites produced in DNA from the loss of these depurinating adducts are converted into tumour-initiating mutations by error-prone repair. Higher level of oestrogen–DNA adducts has been found in the urine of men with PCa compared with healthy controls47
In industry-based cohort studies, exposure to styrene has been linked to lymphatic and haematopoietic cancers, but the excess risk was not consistent.8 35 The prostate is hormone dependent and exposure to exogenous hormone modulators may lead to carcinogenesis. An endocrine disruptor activity of styrene had been suspected based on increased prolactin levels observed in exposed female workers, but this was not supported by experimental data.48
Exposures to BTX were highly correlated, hampering the evaluation of separate associations. Our findings provide evidence for an increased risk of low-grade PCa among men occupationally exposed to BTX, with benzene showing the most consistent results, and styrene over prolonged periods, especially at substantial levels. Results based on cumulative exposure were equivocal.
The authors would like to thank all members of the Epidemiology and Biostatistics Unit at INRS-Institut Armand-Frappier who were involved in the conduct of the study. In particular, we want to highlight the work of Louise Nadon, Ramzan Lakhani, Mounia Senhaji Rhazi and Robert Bourbonnais, who carried out the occupational exposure assessment. The contribution of Hughes Richard in the preparation of the databases is also gratefully acknowledged. We thank the urologists from the participating hospitals for their collaboration in patients’ access.
Contributors AB-L conducted the data analysis and prepared the manuscript. M-EP conceived and led the PROtEuS study. J-FS and M-EP contributed to the interpretation of the data and critically revised the manuscript. All authors read and approved the final version of the manuscript.
Funding This study was supported financially through grants from the Canadian Cancer Society (grants no. 13149, 19500, 19864 and 19865), the Cancer Research Society, the Fonds de Recherche du Québec—Santé (FRQS), FRQS-Réseau de recherche en santé environnementale and the Ministère du Développement économique, de l’Innovation et de l’Exportation du Québec. M-EP and J-FS received career awards and a fellowship, respectively, from the FRQS.
Competing interests None declared.
Patient consent Not required.
Ethics approval CER – Institut national de la recherche scientifique.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Our data can be shared upon agreement with the corresponding author. No additional data are available.
Presented at This work was conducted at INRS-Institut Armand-Frappier, Université du Québec, Laval, Québec, Canada.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.