The International Agency for Research on Cancer (IARC, 2010a) has classified 107 agents, mixtures or exposure circumstances as Group 1 (carcinogenic to humans), many of which are encountered in occupational settings, for example, asbestos and cadmium. An additional 58 agents, mixtures or exposure circumstances have been classified as Group 2A (probably carcinogenic to humans). Those with occupational significance include diesel fumes and benzidine-based dyes (IARC, 2010a). Table 1 (adapted from Siemiatycki et al, 2004) shows the most important occupational exposures in these two categories.

Table 1 Occupational exposures linked to cancer risk and industries in which exposures can occur

A comprehensive analysis of occupational exposures, with quantitative estimates of the cancers attributable to them, has been carried out by Imperial College London and the Health and Safety Laboratory on behalf of the Health & Safety Executive by Rushton et al (2007). This analysis has been updated and extended, based on mortality in Britain in 2005 and incidence in 2004 (Rushton et al, 2010). Here, we have applied the population-attributable fractions (PAFs) for Great Britain, as estimated in this paper, to the estimated cancer incidence in UK in 2010.

Methods

The methodology used to estimate PAFs of each cancer is described in the papers by Rushton et al (2007, 2010). The carcinogenic agents or exposure circumstances identified for each cancer were those classified by the IARC as Group 1 or 2A carcinogens (IARC, 2010a). Estimation of PAFs requires data on the relative risk of each exposure, and the prevalence in the general population.

Risk estimates were obtained from key studies, meta-analyses or pooled studies, taking into account quality, such as relevance to Britain, sample size, extent of control for confounders, adequacy of exposure assessment, and clarity of case definition. Where possible, risk estimates that had been adjusted for important non-occupational confounding factors, for example, smoking status, were selected. In general, dose–response risk estimates were not available from the epidemiological literature, nor were proportions of those exposed at different levels of exposure over time available for the working population of Britain. However, where possible risk estimates were obtained for an overall ‘lower’ level and an overall ‘higher’ level of exposure to the agents of concern.

With respect to the latency between exposure and the elevated risk of cancer, a ‘relevant exposure period’ was defined. For solid tumours a latency of 10–50 years was assumed; for haematopoietic neoplasms it was 0–20 years.

The proportion of the population exposed to each carcinogenic agent or occupation was obtained from the total number of people employed and the numbers potentially exposed to the carcinogens of interest in each relevant industry/occupation within Britain. If the study from which the risk estimates were obtained was population based, an estimate of the proportion of the population exposed was derived directly from the study data. If the risk estimate was obtained from an industry-based study, national data sources, the CARcinogen EXposure (CAREX) database, the UK Labour Force Survey (LFS) or Census of Employment (CoE) was used to obtain the proportions exposed to the carcinogens concerned in Britain. Adjustment factors were applied to the data to take account of the change in numbers employed and the employment turnover during the ‘relevant exposure period’.

The studies from which risk estimates were taken were often mortality studies only and PAFs derived from these were applied to numbers of registrations. The PAFs published by Rushton et al (2010) have been applied to the estimated numbers of cancers in the UK in 2010, with the following exceptions:

  • We exclude occupationally induced non-melanoma skin cancers, for the reasons stated in the introduction – primarily that enumeration of such tumours is very far from complete (so that including them among the total cancers attributable to different exposures is misleading).

  • PAFs due to occupational exposure to environmental tobacco smoke are included with other tobacco-related cancers in Chapter 2 (Parkin, 2011).

Results

Table 2 shows the estimated number of new cancer cases in 2010 for 22 types of cancer, the PAFs (from Rushton et al, 2010) and the estimated numbers of cases due to occupational exposures. The total is an estimated 11 494 cases (7832 in men and 3662 in women), representing 3.7% of all cancers (excluding non-melanoma skin cancers). The most substantial numbers are lung cancers (exposures due to asbestos, silica, diesel engine exhausts, mineral oils), mesotheliomas (asbestos) and breast cancer, related to shift work that involves circadian disruption (IARC, 2010b).

Table 2 Estimated fractions of cancer cases in UK, 2010, attributable to occupational exposures

Discussion

Included in the evaluation of Rushton et al (2010) were non-melanoma skin cancers, the calculated PAF of which (4. 6%) was applied directly to the number of registrations in 2004. The latter may have been due to a lack of appreciation by the authors of the incomplete nature of cancer registration for non-melanoma skin cancer. Undercounting of such cancers is a consequence of the relatively trivial nature of the great majority, and many such cancers are treated without hospitalization or, probably, a biopsy. Registration is biased by cell type (basal cell cancers will certainly be undercounted), while some cancers – probably occupationally related ones – may be more completely identified.

Prevalence of exposure in the analysis by Rushton et al (2010) was estimated for a period of 10–50 years prior to 2005 for solid tumours, and 0–20 years for haematopoietic neoplasms. If exposure prevalence has been declining, and these latency periods are correct, it is possible that the numbers of attributable cancers for 2010 will be slightly overestimated (because the time since exposure is 5 years longer than for the estimates for 2005).

Several other studies have been carried out to estimate the proportion of cancers in a given population that are possibly the result of exposure to carcinogens in an occupational setting. Doll and Peto (1981) included occupational factors in their evaluation of the quantitative contribution of different factors to cancer mortality in the United States. In addition to the USA, estimates of the effect of occupational exposures on the burden of cancer have been made for the year 2000 for the populations of the Nordic countries (Dreyer et al, 1997), Australia (Fritschi and Driscoll, 2006) and France (IARC, 2007). Quantitative estimates of the carcinogenic effect of 11 occupational exposures worldwide in the year 2000 were made by Driscoll et al (2005).

These recent studies used a methodology similar to that of Rushton et al (2010) – the CAREX (CARcinogen Exposure) database – to provide national estimates of the proportions of workers exposed to different carcinogens and their levels of exposure (Finnish Institute of Occupational Health, 2009), and estimates of the relative risk of different exposures, drawn from literature reviews. There are differences not only in the choice of such studies (and hence of risks associated with occupational exposures, and different levels of exposure), but also in the precise choice of exposures, which can reasonably be considered to be carcinogenic in an occupational setting.

The Nordic study (Dreyer et al, 1997) estimated that verified industrial carcinogens will account for approximately 3% of all cancers in men and less than 0.1% of all cancers in women in the Nordic countries around the year 2000, while the French study estimated that 2.5% of cancers in men and 0.3% in women were caused by occupation.

See acknowledgements on page Si.

Conflict of interest

The author declares no conflict of interest.