Objectives The contribution of occupational exposures to the cancer burden can be estimated using population-attributable fractions, which is of great importance for policy making. This paper reviews occupational carcinogens, and presents the most relevant risk relations to cancer in high-income countries using France as an example, to provide a framework for national estimation of cancer burden attributable to occupational exposure.
Methods Occupational exposures that should be included in cancer burden studies were evaluated using multiple criteria: classified as carcinogenic or probably carcinogenic by the International Agency for Research on Cancer (IARC) Monographs volumes 1–114, being a primary occupational exposure, historical and current presence of the exposure in France and the availability of exposure and risk relation data. Relative risk estimates were obtained from published systematic reviews and from the IARC Monographs.
Results Of the 118 group 1 and 75 group 2A carcinogens, 37 exposures and 73 exposure-cancer site pairs were relevant. Lung cancer was associated with the most occupational carcinogenic exposures (namely, 18), followed by bladder cancer and non-Hodgkin’s lymphoma. Ionising radiation was associated with the highest number of cancer sites (namely, 20), followed by asbestos and working in the rubber manufacturing industry. Asbestos, bis(chloromethyl)ether, nickel and wood dust had the strongest effect on cancer, with relative risks above 5.
Conclusions A large number of occupational exposures continues to impact the burden of cancer in high-income countries such as France. Information on types of exposures, affected jobs, industries and cancer sites affected is key for prioritising policy and prevention initiatives.
- Occupational Exposure Review
- Comparative Risk Assessment
Statistics from Altmetric.com
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.
What is already known about this subject?
Exposure to various carcinogenic agents in occupational settings is known to impact the risk of cancer.
The impact of occupational exposures on the national burden of cancer is often quantified using population-attributable fractions, the calculation of which requires knowledge of the prevalence of exposure in the working population and relative risk estimates for each carcinogen linking it to specific cancer sites.
What are the new findings?
We provide the rationale for selection of occupational carcinogens and corresponding relative risk estimates for cancer burden studies in high-income countries.
We map the most comprehensive and up-to-date list of carcinogens at occupational settings and their impacted cancer site pairs.
There are still prevalent and diverse carcinogens present in occupational settings: 73 exposure/cancer site associations, out of which 44 with sufficient evidence, were identified as relevant to assess cancer burden related to occupational exposures in high-income countries.
How might this impact on policy or clinical practice in the foreseeable future?
This paper provides key information for the evaluation of cancer burden, thus facilitating the set-up of such studies, in particular in high-income countries.
Our results provide an important database to evaluate the national burden of cancer related to occupational exposures for setting public health priorities and labour regulations as well as for developing clinical practices to guide screening and diagnosis of cancers related to occupational carcinogens.
The International Agency for Research on Cancer (IARC) Monographs have classified almost 200 exposures as carcinogenic or probably carcinogenic to humans, with a large proportion of these exposures found in occupational settings. Therefore, the impact of occupational exposures on the burden of cancer is a pressing public health issue for many countries.1–4 In particular, studies that estimate the number of cancers attributable to historical occupational exposures, such as chemical, physical or circumstantial carcinogens, are key to guiding public health and prevention priorities and for developing and enforcing labour regulations for different occupational exposures.
The number of cancers attributable to an occupational exposure is commonly estimated using the population-attributable fraction (PAF), defined as the proportion of an outcome that would not have occurred in the absence of the chemical, physical or circumstantial carcinogen.5 However, estimation of a PAF is complicated by many factors. In particular, the PAF reflects the strength of the association between the occupational risk factor and cancer as well as the prevalence of exposure to this factor in the population. Accordingly, determining the proportion of cancers attributable to occupational exposures requires knowledge of the relative risk (RR) estimates for each carcinogen linking it to specific cancer sites as well as the prevalence of exposure to these carcinogens in the working population over the time period considered. Therefore, due to the large number of carcinogenic occupational exposures, the estimation of the PAFs for cancers due to every occupational exposure is challenging particularly due to the lack of summarised information on existing exposures and/or the associated cancer risks.6
Accordingly, the aim of this paper is to present the first steps of the PAF estimation process, that is, a rationale for the selection of occupational exposures and a determination of their causally associated cancer sites, taking into account the IARC Monographs and recent epidemiological studies as well as the availability of RR estimates. Second, this paper provides the list of relevant occupational carcinogens identified and corresponding RR estimates, using France as an example of a high-income country.
The rationale and framework for selecting carcinogenic occupational exposures and RR estimates were determined by a working group composed of international experts in occupational health and epidemiology.
Identification of occupational risk factors
Occupational exposures (chemical substances, physical agents, organisational or occupational circumstances) were identified using a previous list of exposures prepared by Siemiatycki et al,7 updated through industry sectors or jobs identified as exposed by IARC Monographs volumes 1–114. Specifically, exposures identified for inclusion were classified by IARC Monographs as either carcinogenic (group 1) or probably carcinogenic (group 2A).
Occupational risk factors were included using the following selection criteria: (i) the factor must have been present in the occupational environment at some time from 1965 to 2015 (ie, the reference exposure period during which an occupational agent affects the risk of cancer) and (ii) corresponding exposure and cancer incidence data must exist for France.
Some agents were grouped frequently within epidemiological studies, for example, aromatic amines or polycyclic aromatic hydrocarbons (PAHs), and, in such cases, this review only considered the group of carcinogenic agents and not the individual agents that composed these groups. Drugs were excluded from the selection of agents, as the corresponding risks were based on treatment with these drugs, rather than on occupational exposures. Finally, using the IARC Monographs, we identified the relevant exposure-cancer site pairs, that is, cancer sites associated with each occupational exposure, where there was sufficient or limited evidence of carcinogenicity in humans.
Identification of relative risk estimates
After finalising the list of exposure-cancer site pairs, we retrieved and summarised risk estimates which are reported as RRs, odds-ratios (ORs) or hazard-ratios (HRs) (ORs and HRs were used as an approximation of a RR). Priority was given to adjusted RR estimates from systematic reviews and meta-analyses of studies that were mainly performed in high-income countries or settings similar to France. Adjusted RR estimates were preferred to take account of confounding or interaction with the considered exposures. Furthermore, RR estimates were included if they (i) were relevant in terms of the exposure levels in France (from 1965 to 2015) and (ii) matched available exposure data, in terms of circumstances to exposure measurement. For example, RR estimates were not included if they were available only for levels of exposure heavily exceeding the levels observed in France or if they were not based on exposures reported within occupational settings, for example, arsenic in drinking water and the risk of bladder cancer.
We used the following sources to identify the most relevant RR estimates from high-quality studies:
RR estimates used in previous occupational cancer Comparative Risk Assessment studies, in which a systematic literature review was performed for each exposure stratified by cancer site, by Rushton et al 4;
RR estimates from studies listed in the IARC Monographs (published after the above-mentioned systematic literature review by Rushton et al 4);
If no RR estimates were available from the first two sources, we performed a search in Medline for recent meta-analysis or systematic reviews. If none was available, RR estimates were obtained from large cohort studies.
Pairs for which no relevant RR estimates could be retrieved were excluded.
The expert working group reviewed each identified risk relation estimate and corresponding study for study quality and for relevance to France in terms of exposure categorisations. For each occupational exposure-cancer site pair, the best RR estimate and the corresponding 95% CIs were extracted by level of evidence (sufficient or limited) in humans.
Selection of occupational exposures
Figure 1 illustrates the selection process for occupational agents. Of the 118 exposures and 75 exposures which are classified by IARC as carcinogenic (group 1) and probably carcinogenic in humans (group 2A), respectively, 87 agents were judged to be occupational (68 group 1 exposures and 19 group 2A exposures). Among them, 12 group 1 and 5 group 2A exposures had no prevalence data in occupational settings (including biological agents such as hepatitis B virus and human papillomavirus 68), and for 3 group 1 exposures there was no known exposed population over the period 1965–2015 (such as aflatoxins and sulfur mustard). The exposures excluded because there were no corresponding incidence data were those related to non-melanoma skin cancer, namely, shale oils, mineral oils and coal-tar distillation. Non-arsenical insecticides were not selected because since the corresponding Monograph in 1991, some exposures included in this group such as malathion and diazinon, have been evaluated in more recent Monographs, and this more recent information was used. Furthermore, we did not identify any available exposure data for ‘non-arsenical insecticides’ as a group or for glyphosate. All excluded exposures and the corresponding detailed reasons for their exclusion are presented in detail in the online supplementary material list S1. Of the 64 carcinogens relevant to France, 31 (across both groups 1 and 2A) were categorised into four larger carcinogen groups. First, different types of ionising radiation were combined into one group termed ‘all types of ionising radiations’. Second, to match exposure categorisations used in epidemiological studies, the eight aromatic amine agents were combined into one larger group, and the six PAH agents or industries with exposure to PAH were merged into another large group (see online supplementary material list S1). Lastly, for the association between petroleum refining and leukaemia, we considered this exposure as part of benzene exposure, as exposure to benzene was identified as causing leukaemia in petroleum refining workers.8 Finally, of the original 193 carcinogenic exposures, 37 were reported in this study (see figure 1 and table 1).
Supplementary file 1
Selected occupational carcinogens
The 37 selected exposures corresponded to 94 exposure-cancer site pairs: 51 with sufficient evidence and 43 with limited evidence of carcinogenicity in humans according to the IARC Monographs (see online table S1 of the supplementary material). The sites associated with the highest numbers of carcinogenic occupational exposures were lung cancer (18 exposures), bladder cancer (9 exposures), non-Hodgkin’s lymphoma (NHL) (9 exposures) and nasal cavity cancer (6 exposures). However, for NHL, only two occupational exposures were of sufficient evidence in humans, whereas for lung, bladder and nasal cavity cancers, 14, 6 and 4, respectively were of sufficient evidence in humans. For occupational exposures with sufficient evidence of increased cancer risk by site, ionising radiation was associated with 20 sites, the highest number of associated cancer sites. Asbestos and working in the rubber manufacturing industry were ranked second and third in having sufficient evidence of raising the risk of cancer at four and three sites, respectively.
Figure 2 presents the selection process for exposure-cancer sites pairs. Among the 94 exposure-cancer site pairs, 20 (6 with sufficient evidence and 14 with limited evidence) were excluded, and 2 pairs were merged into another existing pair: mesothelioma in painters was considered to be subsumed into the relationship between asbestos exposure and mesothelioma, as the agent causing mesothelioma in painters was established as being asbestos9; similarly, leukaemia in workers in the rubber manufacturing industry was subsumed into the relationship between leukaemia and benzene, as the agent causing leukaemia in rubber manufacturing workers was established as being benzene.9 The two main reasons for exclusion of pairs were the availability of RR estimates based only on environmental exposures (6 pairs excluded) and the unavailability of valid RR estimates (10 pairs excluded), that is, the existing RR estimates were heterogeneous and no meta-analysis, pooled or large study was available. More details on pairs excluded because no relevant risk estimate was available are presented in online table S2 of the supplementary material. Also, cancer of the basal cells of the skin associated with ionising radiation was excluded because no incidence data were available for this cancer site. Accordingly, 73 exposure-cancer pairs (44 with sufficient evidence and 29 with limited evidence) with a RR estimate matching selection criteria were identified, including exposures to 25 group 1 exposures and 9 group 2A exposures (see figure 2).
Relative risk estimates for occupational exposures
Among the 44 carcinogenic exposure-cancer site pairs with sufficient evidence, the highest risks (RRs above 5) were observed for exposure to bis(chloromethyl)ether associated with the risk of lung cancer (RR=7.6 for all levels of exposure combined), and nickel (RR=8.7 for all levels of exposure combined) and wood dust (RR=5.8 for the highest levels of exposure) associated with the development of nasal cavity cancer (see table 2).
Among the 29 pairs with limited evidence, the highest risk estimate was for chromium VI and nasal cavity cancer (RR=3.4 and RR=5.2 for low and high exposures, respectively) (see table 3). The following exposure-cancer site pairs had the next highest RR estimates: cadmium and kidney cancer in females (RR=2.5 for substantial exposure levels), and cobalt metal with tungsten carbide and lung cancer (RR=1.9 for all exposure levels combined).
This study provides the most up-to-date comprehensive summary of the occupational exposures, which are classified as carcinogenic or probably carcinogenic by the IARC Monographs, their corresponding cancer sites for which there is either sufficient or limited evidence of a causal association and corresponding RR estimates. Thus, it provides key information to enable the selection of carcinogenic occupational exposures and their risk estimates, which is the first step to conduct national studies to estimate the burden of cancer attributable to these exposures, particularly in high-income countries where occupational settings are similar to France. The online supplementary material also provides details on every agent or pair excluded and reason for exclusion, which allows any country to adapt the presented rationale to his own context. Overall, we found 37 carcinogenic occupational exposures relating to 94 exposure-cancer site pairs that will affect the current burden of cancer. Of these 94 exposure-cancer site pairs, we identified 73 pairs and 25 cancer sites for which the PAFs could be estimated, based on available exposure and incidence data and on corresponding RR estimates. In a similar study performed in the UK in 2004, 73 pairs (for 22 types of cancer) were also reported, and results showed that 3.7% of cancers were attributable to occupational exposures. The UK study’s findings differ only slightly from the present study’s findings, which demonstrate that the presented framework and review can be used by other studies for similar high-income countries.4 However, the selection of exposures and RR estimates and the consequent determination of the number of exposure-cancer site pairs remain dependent on the national context related to the prevalence of occupational exposures to these risk factors, and data availability. For example, occupational exposure to vinyl chloride has been reported to be strongly associated with angiosarcoma of the liver and also with primary hepatocellular cancer, with a long latency time between exposure and disease occurrence. Unfortunately, there are no valid RR estimates matching the substantially decreased levels of exposure in France over the last 40 years, and, hence, this exposure-cancer site pair was excluded from our review. Additionally, oxidised bitumens used by roofers have been classified as probably carcinogenic to humans (group 2A); however, estimates of the proportion of roofers in France using oxidised bitumens are not available. Given the uncertainty of how many workers might be exposed to oxidised bitumens, this exposure-cancer site pair did not meet the inclusion criteria for this review. However, we can assume that only a small number of workers were exposed to such bitumens; consequently, the impact of excluding this exposure on future burden assessment should be minor. Although unlikely, in other industrialised countries the proportion of roofers exposed to oxidised bitumens might be important. The online supplementary material of this paper provides comprehensive lists of agents to allow extension of our study to other contexts.
We present the RR estimates separately for pairs with sufficient evidence in humans and those with only limited evidence in humans. This will allow future studies to modulate agent and cancer pairs according to the currently established level of evidence for each pair. Depending on the option, both the estimations and their interpretation will differ substantially: in the first case, the number of cancer sites considered will be much higher (and so will be the overall attributable fraction), but in the second case, the quality of the estimates will be higher, as they will be based on more reliable epidemiological studies (those which enable the level of evidence in humans to be classified as ‘sufficient’).
A given occupational carcinogen may be linked to several cancer sites, and one cancer site can be associated with several occupational carcinogens. For example, the risk of lung cancer was affected by the greatest number of occupational exposures, with predominantly respiratory exposure. Therefore, preventative labour and health policies, such as improved monitoring coverage and the reduction of both the set limits of air contaminants and actual air contaminant concentrations in various working environments, for example, PM2.5 and PM10 such as diesel exhaust, would lead to a decrease in the large number of new lung cancer cases and subsequent deaths as well as to a decrease in the number of other cancers caused by agents which enter the body via respiratory exposure. Furthermore, the benefit of such policies would extend to other causes of death and disability, such as respiratory diseases which are leading causes of occupational disease in most countries.10
In addition, policy measures should also consider the risk linking occupational carcinogens to cancer sites, as even small exposures to some of these agents cause a large increase in cancer risk. In particular, this review found that the RRs for bis(chloromethyl)ether (in association with lung cancer), and wood dust (in association with nasal cavity cancer) were above 5. In addition, the increased risk of mesothelioma linked to exposure to asbestos is very high and well established. We have not presented the RR for the association between asbestos and mesothelioma, as given the low background rate of mesothelioma in the absence of exposure to asbestos, a certain percentage (sometimes 100%) of all registered cases is usually attributed to asbestos in cancer burden studies, without any risk estimates considerations.11 12 While past efforts have successfully resulted in a complete banning of the use of asbestos,13 further preventative efforts should focus on other high-risk exposures.
Limitations of the presented relative risk estimates
The presented RR estimates have various limitations. First, the method of exposure measurement and categorisation in monitoring and surveillance studies often differs from the exposure measurement and categorisation used in pertinent epidemiological studies. Second, epidemiological studies provided a non-negligible number of RR estimates that were statistically non-significant, especially for low levels of exposures or for exposures among women. This can be explained by the low number of women included in such studies, the thresholds used to define the exposure categories or stochastic variability between studies, but does not necessarily reflect an absence of association. Third, the presented RR estimates do not account for various overlaps in occupational exposures. For example, both the occupation of being a painter and the consequent exposure to multiple carcinogenic agents are risks for a number of common cancers. Furthermore, exposure to multiple individual carcinogenic occupational agents also increases the risk for a given cancer. Specifically, for painters, there are risk overlaps with exposure to PAHs, asbestos and silica for lung cancer, and asbestos for mesothelioma. These overlaps of risk are usually accounted for using two different methods. First, the occupational-based and agent-based exposures are estimated separately, with the total number of occupation-specific attributable cancers being determined through the occupational exposure and the agent-based exposures independently contributing to this total. Second, for cancer sites with multiple agent-based exposures, the number of cancers are estimated assuming independence in terms of risk. Therefore, the presented RRs are limited as they assume independence and do not take into account possible interactions between occupational risk factors which affect the same cancer site, interactions with other carcinogens such as tobacco smoking and interactions with genetic factors. However, when available, we chose to use adjusted RR estimates, especially for cancer sites highly associated with confounding factors such as smoking or alcohol consumption. While we are aware that this approach does not overcome the issue of potential interactions, utilisation of adjusted RR estimates result in a more robust estimating of cancer burden, taking into account the effect of potential confounders on cancer risk.
Limitations of the review
Our selection of RR estimates was based on previously identified information, either from a UK study report (for which a systematic literature review was performed) or from studies detailed in the IARC Monographs. Therefore, this study did not involve a systematic review of the literature, and instead relied on existing systematic reviews, the IARC Monographs and experts to identify relevant studies. The occupational exposures which did not have a corresponding RR estimate or had RR estimates with limited strength of evidence are indicative of the need for more research on these risk relationships. Future studies should consider harmonisation and improvement of occupational exposure measurement to facilitate meta-analysis or pooling of results from several studies.
There are a large and diverse number of occupational exposures that affect the risk of cancer. The results of this study have now been used in a study to estimate PAFs for cancers due to occupational exposures in France (manuscript submitted for publication), as such the outlined carcinogenic occupational exposures and corresponding RR estimates for their association with the risk of cancer can be used as a basis for future research on the quantitative appraisal of the cancer risks associated with past and/or current exposures in high-income countries. Our results are thus useful to facilitate assessment of cancer burden studies, which can be used to inform and support prevention of occupational cancers.
The authors would like to thank Drs D Max Parkin, Jürgen Rehm, Paolo Vineis, Catherine Hill, Agnes Rogel, Jean-Nicolas Ormsby, Laure Dossus and Gwenn Menvielle for their advice on this manuscript and for their participation in the project entitled ’Définition des priorités pour la prévention du cancer en France métropolitaine: la fraction de cancers attribuables aux modes de vie et aux facteurs environnementaux'.
Contributors Conceptualisation: CMM, KDS and IS; methodology: CMM, KDS, IB, BC, BF, AGSG, PG, AO, LR, SH, KS and IS; formal analysis: CMM and KS; resources: IS; writing – original draft: CMM, KS, IB, BC, BF, AGSG, PG, AO, LR, SH, KS and IS; writing – review and editing: CMM, KS, IB, BC, BF, AGSG, PG, AO, LR, SH, KS IS; supervision: IS; project administration: CMM, IS.
Funding This project is funded by the French National Cancer Institute (Institut National du Cancer) for the project entitled ’Définition des priorités pour la prévention du cancer en France métropolitaine: la fraction de cancers attribuables aux modes de vie et aux facteurs environnementaux' (grant number: 2015-002).
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.