CAREX Canada: an enhanced model for assessing occupational carcinogen exposure

Selection of carcinogens

Known and suspected carcinogens were selected for inclusion in CAREX Canada (where the workplace was identified as a potential source of exposure) from the list of IARC-classified Group 1 (known human carcinogens), Group 2A (probable carcinogens) and Group 2B (possible carcinogens). The occupationally-related agents were then evaluated based on three criteria: (1) the agent was likely to be present in Canadian workplaces in the year 2006, and exposure was relatively prevalent (>10 000 likely to be exposed); (2) toxic effects, including carcinogenicity (IARC Group 1 agents were given higher priority); and (3) the feasibility of assessing exposure based on available data. Occupational exposure was defined as inhalation and/or dermal exposure at work to levels above environmental background levels. We combined some separate IARC agents into groups for exposure assessment when they were expected to be encountered together in workplaces. This grouping occurred for the polycyclic aromatic hydrocarbons and the antineoplastic agents.

Data gathering

A major component of the CAREX Canada occupational project was the creation of the Canadian Workplace Exposure Database (CWED), a repository of occupational exposure monitoring data from across Canadian jurisdictions. This database was used in two ways: (1) to identify the potential for exposure in particular industries and occupations and (2) to create exposure level estimates among the data-rich priority agents. At the time these estimates were developed, the CWED contained approximately 100 000 exposure measurements for known and suspected carcinogens collected between 1981 and 2004 for regulatory purposes in two of Canada's largest provinces. The CWED has been continuously expanded as other Canadian provinces agree to contribute data, and at time of writing contained 460 000 measures of exposure (carcinogens and non-carcinogens).12 Other primary sources of data include the EU CAREX project (for proportions of workers exposed and other general methodology), published literature on exposure to carcinogens in the workplace, government reports and grey literature (figure 1). Population data on industry, occupation, province/territory and sex were obtained from Statistics Canada based on their 2006 Census of Population.13

Figure 1

Generalised methodology used to generate estimates of occupational exposure to known and suspected carcinogens.

Occupational estimates approach and simple adaptations

A generalised approach (outlined in figure 1) was developed to ensure comparability among substances, methodological transparency and ease of interpretation. This framework was flexible and was adapted for agents with unique data sources or particular assessment challenges.14 Primary data sources were reviewed for relevance to the Canadian context and the proportion of workers likely to be exposed by industry and occupation in Canada was assigned using expert assessment and agreement by a minimum of two occupational hygienists. All procedures and estimates were externally reviewed by a scientific advisory committee composed of research and practice professionals in occupational hygiene and epidemiology.

These proportions were then linked to the census data to calculate a simple prevalence (number of workers exposed) by agent. Prevalence estimates were generated by industry (North American Industry Classification System or NAICS, 2002 version), occupation (National Occupational Classification—Statistics, 2006 version), province or territory, and sex.

Where data are available in the CWED, exposure levels were also estimated using a semiquantitative method based on the concepts used by the Institut de recherche Robert-Sauvé en santé et en sécurité du travail in Quebec, Canada (IRSST).15 We used these criteria for exposure levels to match the methods and reporting structure of the IRSST because Quebec is the largest current exposure data collector in Canada, but these data were publically unavailable. Using their method allowed estimates to be comparable across jurisdictions. Two concentration thresholds were selected for each agent based on workplace exposure limits and cancer-relevant health outcomes. In general, we set high exposure thresholds to current threshold limit values from the American Conference of Governmental Industrial Hygienists (ACGIH), unless the limit was not set for cancer prevention reasons. In these cases (eg, for wood dust, where the current threshold limit value (TLV) is set to prevent non-malignant respiratory disease), we selected a threshold based on the epidemiological literature. We set the moderate thresholds as half of high exposure thresholds. For a given industry/job combination, high exposure was assigned when there were at least 25 samples available for that agent in a particular job and industry, and 20% or more of those samples had concentrations that were at or above the high threshold. The criteria for moderate exposure were met in one of two ways. Either (1) there were at least 25 samples available for that agent in a particular job and industry, and 20% or more of those samples had concentrations above the moderate threshold or (2) there were at least 10, but less than 25 samples, and 20% or more of those samples had concentrations above the high threshold. Workers were assigned into the low category of exposure where there were either no samples (but expert assessment deemed exposure as plausible) or the samples did not meet the criteria for moderate exposure (ie, were typically lower than half the exposure limit).

For some agents, CWED data were not available, but we were able to estimate exposure levels via some simple adaptations to the general framework. Two examples of this are artificial ultraviolet radiation (UVR) and solar UVR exposure to outdoor workers. Since there are no systematic measures of these physical hazards in Canada, we developed modified methods in order to assess both prevalence and level of exposure. The detailed methods and results of the solar UVR exposure assessment have been published elsewhere.16 Briefly, exposure categories were developed based on expected time spent outdoors. Groups were categorised as having high exposure when ≥75% of their workday was expected to be outdoors, and these jobs were identified using skin cancer prevention materials from Australian campaigns.17 For the moderate and low categories of exposure, Canadian career-selection websites were used to identify jobs with outdoor components.

To develop estimates of the levels of artificial UVR exposure, we used a report by the International Commission of Non-Ionizing Radiation Protection (ICNIRP).18 This report classified exposure to artificial UVR into categories of low, medium, high and very high with respect to the potential for overexposure. The only source classified with a low potential was exposure to general lighting, which CAREX Canada defined as environmental exposure (rather than occupational). Therefore, we excluded ICNIRP's low potential for overexposure classification and defined our exposure groups based on their medium, high and very high potential for overexposure categories. A group of workers was considered to have low exposure if the ICNIRP found that there was a medium potential for overexposure, moderate exposure if the ICNIRP classified the job as having a high potential for overexposure and high exposure if the ICNIRP classified the job as having a very high potential for overexposure.

Unique data sources: shift work, antineoplastic agents and ionising radiation

It is relatively simple to adapt the generalised occupational estimates approach to other exposures that either a) are non-traditional from an industrial hygiene perspective or b) have unique challenges or exposure monitoring practices that make the general approach less desirable. The three main exposures where this occurred in CAREX Canada were shift work, ionising radiation and antineoplastic agents.

Briefly for shift work, we used Statistics Canada's Survey of Labour and Income Dynamics public-use micro-data file for the year 2006.19 This survey was selected because it is a large, nationally representative survey that has information on industry and occupation for all respondents, as well as data on the type of shift that people work in Canada. Exposure was defined as regular work between the hours of 12:00 and 05:00, which was captured by the self-reported shift types of ‘A rotating shift’ and ‘Regular night or graveyard schedule.’ Population counts were calculated by using Statistics Canada provided weights for use in their survey data. The granularity of codes available for industry and occupation in the public-use micro-data file was less than that for census-derived exposure estimates. Therefore, in order to compare exposure estimates by industry, some basic assumptions were made, which are noted where applicable in the results tables.

For ionising radiation, we consulted complementary data available in the 2007 and 2008 annual reports from the National Dose Registry (NDR)2021 in order to reconstruct exposure estimates for 2006. The NDR is a centralised Canadian registry of occupational radiation doses which contains records for over 100 000 currently monitored Canadians.

We calculated a range in the number of workers exposed to ionising radiation to account for uncertainties in how regulations are administered across different Canadian jurisdictions. The low end of the range is the number of monitored workers with a whole-body dose above the limit of detection (ranges from 0.1 to 0.2 mSv (millisievert)) over 1 year. The high end of the range is an extrapolation based on a detailed review of each NDR defined job title with respect to the number of workers likely to be missing from the population of monitored workers. We also examined the proportions of workers exposed from European CAREX projects where we were unable to estimate the proportion of exposed workers directly. For levels of exposure, workers were classified as having low exposure if their annual whole-body radiation dose was between the limit of detection and 1 mSv; moderate exposure if their dose was >1 to <20 mSv; and high if their dose was >20 mSv (which is also the ACGIH occupational exposure limit for an annual dose).

For antineoplastic agents, a varied expert-based approach was used depending on the job/industry being considered. Data sources included: (1) peer-reviewed literature, (2) human resources databases, (3) reports on industry practices, (4) unpublished reports on the potential for antineoplastic agent exposure and (5) personal communication with experts in academia and industry.22 Estimates were produced separately for hospital nurses and physicians, hospital pharmacy staff, community pharmacy staff and veterinary workers.

Data quality definitions

After we calculated prevalence and level of exposure estimates, we assigned a data quality score to each agent. We classified agents as having high quality data where we were able to produce quantitative estimates of the levels of exposure using at least 1000 CWED observations per agent across a variety of industries. An important limitation to this classification is that much of our data are over 10 years old. For agents where we had evidence of stable exposures over that time period (ie, silica), this is less of an issue; however, for agents like tetrachloroethylene, where regulations have changed substantially in the last 5–10 years, we are less convinced that our historic data truly reflect current exposure in the workplace.

We classified agents as having moderate data quality when there were few data available, but where other proxy information on exposure allowed us to assign exposure levels. A good example of this situation is the antineoplastic agents, where few objective exposure measures are available since regulatory exposure monitoring does not occur in Canada. Moderate data quality was also assigned for substances with less than 1000 (but greater than 200) exposure data points available.

We classified agents as having low data quality where we had insufficient exposure data points (<200) to attempt to assign exposure levels, and also in the case of asbestos, where exposure situations have changed significantly in past decades so that assigning accurate exposure levels was difficult.

General results: prevalence and level of exposure, and data quality in Canada

Overall, CAREX Canada produced occupational prevalence estimates for 44 known and suspected carcinogens, and semiquantitative estimates of exposure levels for 18 of these agents. Table 1 outlines the results, categorised by IARC classifications. The data quality variable for each agent is also noted.

Overall, the most prevalent exposure in Canada was shift work (1.9 million exposed), followed by solar UVR exposure to outdoor workers (n=1.5 million), diesel engine exhaust (n=781 000) and benzene (n=374 000). In terms of data quality, 24 (55%) of the 44 agents had low quality data (not enough information to assign exposure levels, or data too outdated to use) and 8 (18%) had moderate data quality (200–1000 workplace samples in the CWED or good quality qualitative data on exposure levels). Only 12 of the agents reviewed (27%) were classified as having high quality data (quantitative exposure level estimates possible, with >1000 samples available in the CWED). More men than women were exposed to carcinogens; the mean percentage of men exposed by carcinogen was 79%, ranging from 27% for antineoplastic agents to 97% for polychlorinated biphenyls (see online supplementary table S1 for more details).

Results are shown in table 2 for the 18 agents where exposure level estimates were calculated. For most agents, the greatest proportion of exposure tends to be located in the low exposure category. However, agents such as solar radiation, wood dust and artificial UV radiation display large proportions of workers in the high category of exposure.

Exposure prevalence by two-digit NAICS 2002 industry codes is shown in table 3. The total number of workers employed in each industry is also presented, as well as the top three exposure agents (ordered by prevalence). The number of exposures column indicates occurrences of exposure as opposed to individuals exposed. Although one person may very well be exposed to more than one agent, the exposures-per-worker metric gives an indication of the overall prevalence of exposure to known and suspected carcinogens. The mining and oil and gas extraction industry (NAICS 21) along with the construction industry (NAICS 23) are two notable instances where the exposures-per-worker exceed 1.0.

By way of an illustration, table 4 shows the same information as table 3, focusing in a more detailed level on a single sector (manufacturing). This table does not include exposures from ionising radiation or shift work, since the industry classifications systems used for these agents were not available at a sufficient level of detail to link directly to the other estimates. Therefore, in many cases, one of these two exposures could be in the top three, but we cannot address this for the manufacturing industry. The exposures-per-worker metric here is highest in primary metal manufacturing and wood product manufacturing. A similar table with results presented by occupation (job title) is available in the online supplementary table S2. In addition, results for the 12 most prevalence exposures in Canada are presented by province in the online supplementary table S3.