Intervening to Reduce the Future Burden of Occupational Cancer in Britain: What Could Work?

We have estimated that 8% of cancers in men and 2.3% in women are caused by work, giving more than 8,000 deaths and 13,600 cancer registrations in Great Britain (1) for all occupational carcinogens and occupational circumstances classified by the International Agency for Research on Cancer (IARC) as Group I (established) or IIA (probable) carcinogens that had either "strong" or "suggestive" evidence of carcinogenicity in humans (2).

The methodology has been extended to estimate the future burden of occupational cancer and to forecast the impact of alternative policy decisions affecting future workplace exposure levels (3). This article presents estimates of the future burden of occupational cancer under a series of scenarios of change for 14 occupational carcinogens and circumstances in Great Britain that each contribute at least 100 occupation attributable registrations to current burden and account for 86.3% of the total burden (Table 1).

A full description of the methodology for estimating the current (4) and future burden (3) of occupational cancer can be found elsewhere. For our current burden estimation, Levin's formula was used to estimate the attributable fraction (AF), that is, the proportion of cases caused by occupational exposure [(5, AF = p(E) × (RR-1)/{1 + p(E) × (RR-1)} in its simplest form]. This requires an estimate of the risk of disease, generally as relative risk (RR) which we obtained from published literature, and the proportion of the population exposed [p(E)], which we derived from national data sources, accounting for employment turnover and life expectancy, and adjusted for employment trends. To account for cancer latency a risk exposure period (REP) was defined for each carcinogen as the exposure period relevant to a cancer appearing in a specific target year (10–50 years for solid tumors, 0–20 years for lymphohaematopoetic tumors). As exposure-response risk estimates and proportions exposed at different levels are not generally available, risk estimates and proportions exposed were obtained wherever possible for “high,” “medium,” and “low” exposure levels with a “background” level, where appropriate, assumed to have 0 excess risk [these categories have been expanded from “high” and “low” only which were used to estimate current burden (1)]. Estimated AFs were applied to total British deaths (for 2005) and registrations (for 2004) to give attributable cancer numbers.

To estimate future burden AFs were estimated for a series of forecast target years (FTY), that is, 2010, 2020, …, 2060 (3). A REP projected forward in time was defined for each FTY with the contribution of past exposure to future cancer risk decreasing for each FTY (see Supplementary Fig. S1). Adjustment factors were applied to newly recruited workers (assumed to be aged 15–24 years) in separate 10-year estimation intervals to adjust for changing numbers used in broad industry sectors, for example, an increase in the service industry sector and a decrease in manufacturing industry [based on Labour Force Survey data (6), Supplementary Fig. S2]. Where data were available adjustment was also made for declining exposure levels [see Supplementary Table S2 for estimated data on exposure levels and future annual declines in levels required for these adjustments to be made, plus Workplace Exposure Limits and current compliance levels to these limits. Full details of the method of adjustment can be found elsewhere (3)]. Where suitable exposure data were not available, RRs could be adjusted to represent reduced risk scenarios, for example, excess risk was reduced successively by 25% per decade for painters, or workers could be shifted arbitrarily from higher to lower risk categories, for example, shift workers at risk of breast cancer were moved from longer to shorter duration of exposure.

For the current burden estimate for mesothelioma uniquely associated with asbestos exposure we used numbers directly from the UK register of mesotheliomas for 2005, as we believe this captures practically all cases in Great Britain and is, therefore, more appropriate for use as a basis for burden estimation for this disease than our standard methods, which greatly underestimate current mesothelioma incidence. Asbestos related lung cancer was also estimated for current burden from mesothelioma register numbers using an assumption of a 1:1 ratio (1). We excluded only a small number (30–70 men and women a year in the UK) of background cases (spontaneous or from naturally occurring asbestos) from the register numbers. For future burden, we have allocated the current attributable mesotheliomas to 3 exposure levels, high and medium for occupational exposure and a low category for domestic exposure and environmental exposure believed to have been experienced by the post 1940s birth cohort and additional to the background rate (see footnotes to Supplementary Table S1 for the details of the methodology).

Total AFs for cancer sites associated with multiple exposures have been estimated using the product equation [AF_Σi = 1 − Π_i(1−AF_i)] for independent multiplicative estimates (7). To estimate attributable cancer numbers the forecast AFs were applied to an estimate of total cancers for that site based on current age-specific rates applied to Great Britain population projections (Supplementary Table S3). For mesothelioma however associated only with asbestos exposure recent projections based on past mortality rates were used (8).

Table 1 gives the carcinogens and occupational circumstances from the current burden estimation for which future cancer burden was estimated, and the cancer sites that were affected. The carcinogens have been categorized as: (i) those for which no appropriate exposure measurements were available to use in standard setting; (ii) exposures defined by occupational circumstance; and (iii) carcinogenic agents for which standards exist or can be set.

Supplementary Table S4 gives the scenarios tested for each carcinogen and occupational circumstance. Unless otherwise stated 2 baseline scenarios have been evaluated: baseline scenario 1 historic employment and exposure level trends until 2010, no change thereafter, and baseline trend scenario 2 historic and predicted employment and exposure trends included up to 2030, constant thereafter. Intervention scenarios have been compared with baseline 1.

For those agents where standards can be set the scenarios test the introduction of or reductions in current occupational exposure limits (OEL) and improved compliance to these standards. For arsenic and tetrachloroethylene the existing workplace exposure limits (WEL), and for strong inorganic acid mists an earlier WEL, were much greater than estimated current average exposure levels (Supplementary Table S2). For these and also for TCDD, the estimated boundary level between the 2 lowest exposure categories was used as a starting point for a possible exposure standard (Supplementary Table S1). For TCDD the low/background boundary was used, representing a threshold below which excess risk for the agent was 0 (background exposed). No such threshold is generally recognized for genotoxic carcinogens, so the high/low- or medium/low-boundary level was chosen for the other 3 substances.

For asbestos and diesel engine exhaust (DEE), no industries were categorized as background exposed with zero excess risk, so an estimate of the threshold level for background exposure was obtained from independent data. For DEE, 0.001 mg/m³ as elemental carbon was chosen to reflect exposure levels in daily life, based on background urban and suburban exposure measurements in Britain (12). For asbestos an upper boundary of 0.00001 f/mL was assumed for background exposure, based on urban exposure levels from which mesothelioma cases considered to be caused by "background" exposure may arise (13).

For occupational circumstances such as painters and welders, where no specified carcinogen has been identified, only a single RR was available. A decline in exposure level has therefore been assumed to translate linearly to a fall in excess risk. This approach was also adopted for dermal exposure to polycyclic aromatic hydrocarbon (PAH) in coal tars and pitches for which no exposure level RRs were available. For shift work, limits on the total time spent on night shifts over a lifetime were used as the intervention.

Where levels of exposure were not amenable to WEL setting, proportions of the exposed workers were moved to lower exposure categories (e.g., solar radiation) or a reduction in total numbers exposed (e.g., radon) was used in the forecasting. For environmental tobacco smoke (ETS), the effects of different levels of compliance to the current indoor smoking bans were tested.

Table 2 gives for all the carcinogens and occupational circumstances tested the attributable numbers of cancer registrations for 2010 and 2060 together with the numbers of cancer registrations avoided in 2060 for each of the scenarios in Supplementary Table S4. Attributable cancer registrations are shown per year for each target year. Supplementary Table S5 also gives AFs and combined results by cancer site across the 14 carcinogens or occupational circumstances. Full results can be found elsewhere (14).

As historic and forecast exposure levels decline, the AFs generally decline for baseline scenarios 1 and 2 to about 1.5% of all cancer by 2060 (Table 2, and for example Figs. 1A(i) and A(ii) for DEE and lung cancer). However, for breast cancer associated with shift work, rising employment in service sector industries (Supplementary Fig. S2) leads to rising occupational AFs [Fig. 1B(i)].

Predictions of total Great Britain cancers taking account of only demographic changes leads to increasing attributable occupational cancer numbers because of the aging and increasing population [Figs. 1A(ii) and B(ii)]. Between 9,900 (scenario 2) and 10,500 (scenario 1) occupational cancers can be expected per year by 2060 from the baseline scenarios, not much lower than the numbers attributed to occupation in 2010. Breast cancer and nonmelanoma skin cancer (NMSC) from sun exposure account for 60%–70% of these (Table 2).

Without new intervention strategies, the numbers of cancers from low-level exposure will continue to increase even though exposure levels are forecast to decline (scenario 2). Figs. 2A and B illustrate this for lung cancer and DEE, where, although exposure levels are falling by an estimated 7.4% a year, substantial proportions of the population remain exposed at low levels of exposure which still carry a small excess risk (RR = 1.1). Introducing an exposure standard of 0.1 mg/m³ and assuming even 99% compliance does not improve on this (Figs. 2C and D, scenario 6a in Supplementary Table S4). In contrast if an exposure standard could be introduced for DEE at the estimated level below which excess risk was 0, that is, 0.001 mg/m³ as elemental carbon [scenario 6, Figs. 2E and F], lung cancers induced by DEE would nearly disappear by 2060 [Fig. 1A(ii)]. For this level of reduction, however, technology driven intervention may be the only realistic solution.

Low-level exposures will also continue to give high forecast numbers of asbestos related mesotheliomas for both baseline scenarios (63 in men and 208 in women for baseline trend 2) because of the increasing proportion exposed at low levels, even though a 13% annual reduction in average exposure levels has been assumed, and the mesothelioma projections used to estimate our forecast numbers have also taken falling exposure levels into account (8). However, this is not the case for lung and the other asbestos-related cancers as, in the absence of a suitable risk estimate, zero excess risk has been assumed other than for mesothelioma at this low (nonoccupational) level. [Supplementary Figs. S3A–F and G–L show results for lung cancer and mesothelioma]. Even if exposures could be reduced to the levels indicated by the strictest scenario (6) tested, that is, to below 1/100th of the existing standard, some mesotheliomas remain in 2060 (61 men and 182 women), because of continued exposure at the level of additional background risk estimated for the 1940s birth cohort, above the low threshold at which it is believed excess risk will be zero (13). However, these forecasts do not take account of any change to the background risk among later birth cohorts. Our forecast mesothelioma estimates for women are higher than for men because of the higher RRs for medium and low exposure used in their calculation (Supplementary Table S1); they are given separately as we have less confidence in the results for women because of the small number of cases on which the risk estimates were based.

For solid tumor cancers for which long latencies are assumed, no difference is seen between any of the interventions and the baseline scenarios before 2030. The intervention scenarios were developed to be progressively more effective leading to progressive reductions in AFs and attributable cancers. Together the minimum interventions proposed in scenario 3 (the first and least restrictive scenario) would avoid over 2,600 cancers a year by 2060, although only about 38 of these are for the chemicals (arsenic, acid mists, TCDD, and tetrachloroethylene). The current (0.1 f/mL) standard for asbestos or a proposed (0.1 mg/m³) standard for DEE do not result in any “avoided” cancers by 2060 (current compliance to these standards, shown in Table 1, exceeds 90%). For respirable crystalline silica (RCS) an improvement in compliance to the 0.1 mg/m³ 8 hour time weighted average OEL from 33% to 90% could lead to nearly 700 fewer lung cancers annually by 2060 (Table 2). The number of cancers avoided increases as standards are tightened or exposure is progressively reduced (scenarios 4 and 5, Table 2).

The scenarios with the most extreme intervention (scenario 5 for most of the chemical agents; scenario 6 for asbestos and DEE, shift work and solar radiation, and for radon, painters, welders, and coal tars and pitches) show how close to zero the attributable cancer numbers might realistically be expected to fall (Table 2). Numbers fall to 0 only if an intervention results in all workers moving to exposure categories with zero excess risk, for example, achieving full compliance to no smoking in workplaces. Comparing the results for scenario 5, where excess risk has been reduced by 25% in successive decades for painters, welders, and coal tars and pitches, and (6) where a halving of risk is achieved in the first decade, indicates the importance of early versus delayed intervention. Halving the proportions exposed in workplaces to radon in 2010 (scenario 6), for example by introducing appropriate technology, is far more effective than the gradual reduction shown in the other interventions. Similarly, achieving a reduction in risk from solar radiation to that associated with mixed indoor and outdoor exposure (RR = 1.01, Supplementary Table S1), for example, using appropriate skin protection measures, removes most of the large numbers of predicted NMSCs. Restricting women to a maximum of 5 years on night shift work, for which the epidemiological evidence suggests excess risk is 0, would eliminate breast cancers attributable to this exposure [Fig. 1B(ii)].

Our testing scenarios assume that compliance to exposure standards is less than 100%. Our results indicate that a large reduction in number of cancers can be achieved with 90% compliance to a current or proposed standard (scenario 3), and that 99% compliance (arsenic, RCS, strong acids, TCDD, tetrachloroethylene, scenario 6, and DEE scenario 6a) only avoids an additional 109 cancers by 2060 (including 92 lung cancers from RCS exposure).

Forecasts for industry sectors with a current estimate (2004; ref. 15) of more than 80 attributable cancers are given in Table 3 and Supplementary Table S6. The ranking of the predicted cancers by industry sector in 2060 (scenario 2) remains similar to that in 2010 with construction remaining the most important industry sector for potential risk reduction targeting (21% in 2060), followed by 3 service industry sectors, and with breast cancer associated with night-shift work across all industry sectors also a leading contributor. By 2060, large numbers of workers are still projected to be exposed at low levels to the relevant carcinogens (DEE and asbestos, plus tetrachloroethylene in dry cleaners) in personal and household services and land transport. In land transport more than two-thirds, and in defense (armed forces) nearly all attributable cancers are forecast to be NMSCs because of high level (outdoor) sun exposure (Supplementary Table S6).

Our results have shown the potential for considerable eventual reduction in future occupationally related cancers through a range of interventions, although the long legacy of past exposures will continue for up to 50 years. Even with the most stringent scenario tested, cancers are forecast to continue because of exposure to asbestos, PAHs as coal tars and pitches, work as a painter, and exposure to radon and solar radiation, with construction remaining the prime industry of concern. Expected increases in cancer in general as the population ages contribute to the continuing high levels of some occupational cancers, and predicted increases in numbers working particularly in service sector industries also makes a contribution, for example to forecasts for shift work breast cancers, exposure to solar radiation, DEE and asbestos. In estimating the future burden of occupational cancer, we have included the top 14 carcinogenic agents and occupational circumstances, which account for 86% of the estimated current burden of occupational cancer in Britain. Forecasts for agents currently contributing a further 1,800 cancer registrations, including mineral oils, chromium VI, cobalt, aromatic amines and inorganic lead, nonarsenical insecticides, work as a hairdresser or barber, soots and wood dust exposure, and other agents currently responsible for fewer cancers in Great Britain but classified by IARC as Group I carcinogens [including benzene, benzo(a)pyrene (PAH), beryllium, cadmium, formaldehyde, occupational exposure during iron and steel founding, leather dust, nickel compounds, and rubber manufacturing] have not been included in the projection, but are equally important for cancer prevention. If these had been included, 14% more occupational attributable cancers (an additional 1,600 a year) might be forecast (proportionately) by 2060 without intervention, with about 500 of these avoided with some minimum intervention as described for the estimated agents.

The contribution to the future total burden of large numbers of workers exposed at low levels within several service industries is highlighted, rather than the current more highly exposed manufacturing industry sectors, where interventions appear to be more effective in transferring workers from high to low exposed groups. For asbestos and DEE in particular, although exposure levels have been declining cancers still occur because of the low thresholds below which it is thought that excess risk disappears.

If numbers exposed from CAREX had been used to estimate cancer caused by asbestos exposure, lung cancer would have been underestimated possibly 12-fold and mesothelioma 2-fold for men and women compared with observed UK mesotheliomas (14). This suggests that numbers exposed to asbestos are underestimated by CAREX; our forecasts based on observed mesothelioma cases take this into account. CAREX-based estimates of numbers occupationally exposed, 2.6% of men and 1.5% of women, contrast with estimates of 65% of men and 23% of women based on the population controls in the UK study from which we have drawn risk estimates for mesothelioma (16).

It is also possible that the asbestos-related lung cancer to mesothelioma ratio is higher than the 1:1 we have assumed. By estimating AFs on proportions exposed from the UK study and lung cancer RRs, Supplementary Table S1, there would be 5,194 attributable lung cancers rather than 1,768 in 2010 falling to 13 (not 5) in 2060 for baseline scenario 1. This would represent a current lung:mesothelioma ratio of about 3:1 in men and 6:1 in women. Similar ratios have been observed in asbestos exposed cohorts elsewhere (17). However, using an alternative, lower estimate of the proportions exposed to asbestos (still higher however than the CAREX estimates), based on the mesothelioma attributable fractions derived from CAREX data and the RRs in Supplementary Table S1 but that have then been uprated to mesothelioma register numbers, gives more modest estimates, of 1984 attributable lung cancers in 2010 falling to 5 in 2060 for baseline scenario 1. This would represent a current lung:mesothelioma ratio of 1.2:1 in men and 0.3:1 in women.

Pragmatic approaches have been developed to take account of limitations in the data available for Britain. Alternative exposure levels or employment trends and different risk estimates could be used for other countries and situations and variations in existing standards can readily be explored.

Only limited intervention options were tested in this study, for example, reducing workplace limits and improving compliance with these limits. Many other potentially effective interventions have not been assessed, such as improving technology, increasing awareness, and changing attitudes and behaviors that are important in exposure control and risk reduction. Translating these interventions into testable scenarios is problematic; our solution has been to show the impact of the results that might be achieved, such as reduction in excess risk or shifts to lower exposure categories. We have shown that intervening to reduce exposure to workplace carcinogens could lead to the avoidance of more than 8,200 cancers per year by 2060. In comparison, with nearly 20% of all cancers (excluding NMSC) currently attributed to smoking (18), about an 8% immediate reduction in smoking levels would be required to avoid the same number of tobacco-related cancers. The removal of a workplace carcinogen entirely is of course the most effective possible intervention, by replacement with a less toxic or nonchemical means of fulfilling the same function, for example, for tetrachloroethylene. In this case, only the legacy of past exposures will remain.

The level of compliance to future OELs, that is the proportion of worker-exposures remaining above these limits, has also been tested. Full (100%) compliance cannot in practice be used with the lognormal distribution assumption and testing compliances approaching 100% gives unrealistically low results; as compliance approaches 100%, even though the standard may be well above the zero risk threshold, the distribution mean and proportions exposed above any zero risk threshold approach zero. In general, 90% compliance has been assumed to represent a realistically achievable target. Testing the timings of the introduction of standards and also the effect of different compliance levels in different industry sectors or sizes of industry has been explored. For RCS in Britain, it has been shown that improvement in compliance in small companies and among the self-employed is more effective at reducing lung cancer than reducing the current standard (3).

Where the current standard was found to exceed the mean of current exposures by up to 2 orders of magnitude, testing values at a half or even a quarter of the standard does not really inform risk reduction strategies, as the estimated proportions exposed at high levels under the test scenario will unrealistically exceed the proportions exposed in the mid 1970s at those levels, leading to increased AFs and negative estimates of cancers "avoided." Although this can be addressed by assuming compliance levels stricter than currently achieved estimates, we have tested standards that are less than the current estimated mean levels of exposure and therefore of more interest, although these may be difficult to achieve in practice.

If there are several risk factors contributing to the burden of a disease, a change in attribution for one factor will result in a change in the attribution of the others. For example, if future smoking–related lung cancer falls giving a reduced nonoccupational AF, the relative importance of occupation as a risk factor could increase leading to a rise in the occupational AF, although this AF would now be applied to lower projected lung cancer numbers. Attributable numbers rather than AFs, therefore, represent a more useful estimate of the future cancer burden caused by occupation. In addition, it is for this reason that estimated future occupational AFs have been applied to estimates of future cancer numbers based on current cancer rates applied to projected population estimates, ignoring future changes in other lifestyle or environmental risk factors. Cancer numbers attributable to occupation are then comparable by rank order between agents and industries. However the actual estimates may be considered to be inflated either (i) as the population is aging and numbers are increasing when employment levels and, therefore, exposed numbers are declining or (ii) as other causal factors decline so that occupational agents operating synergistically with an environmental or lifestyle factor (e.g., asbestos and smoking for lung cancer) produce fewer cancers. The effect is illustrated in Fig. 3 for forecast lung cancers attributable to the exposures contributing at least 100 cancers (Table 1), estimated using cancer projections based on no change from 2005, demographic change only and increase to 2030 based on an age-period-cohort modeling approach (19).

All results presented here are subject to the biases to which our estimates of the current burden of occupational cancer are subject, described elsewhere (4). The most important of these are data-based, particularly relating to the matching of RRs to our allocation of industries to exposure level categories, and the reliability of the data contributing to estimates of numbers ever exposed. Some reallocation of industries between exposure categories has occurred between the estimation of current and future burden where additional categories have been introduced. In particular moving large numbers in construction and land transport from “high” to a new “medium” exposed category for DEE has resulted in much lower numbers of lung cancers attributable to DEE than estimated for current burden, as a reduced RR has been used for the medium exposed. For bladder cancer, the current burden “high” exposed RR was retained for this large group and a more specifically targeted higher RR has been used for the new and smaller high exposed group of miners and services allied to transport. Also, introducing a low “nonoccupational” category for asbestos exposure does reduce the forecast estimates for the asbestos-related cancers other than mesothelioma; as no risk estimates were available for this group, zero excess risk was assumed for the workers moving out of the higher occupational risk categories with our estimated annual fall in workplace exposure levels.

In summary, comparison of a range of interventions for the most important current occupational carcinogens has shown the potential for future reduction of occupationally related cancer. Interventions to reduce exposure to carcinogens may often also lead to reductions in other health related conditions in the working and living environment, e.g. reduction of silica exposure will not only reduce lung cancer but will affect respiratory function and other non-malignant respiratory diseases. Our methods can be adapted for different data circumstances, to investigate other types of interventions and could be extended to environmental carcinogens and other chronic diseases. Although the forecasts presented here are for Britain, the methods have been used to test the impact of introducing alternative exposure standards in the countries of the European Union for 25 chemical occupational carcinogens (20), and are readily transferable to other national and occupational settings.

No potential conflicts of interest were disclosed.

Conception and design: S.J. Hutchings, J.W. Cherrie, M.V. Tongeren, L. Rushton

Development of methodology: S.J. Hutchings, J.W. Cherrie, M.V. Tongeren,

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.W. Cherrie, M.V. Tongeren, L. Rushton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.J. Hutchings, M.V. Tongeren, L. Rushton

Writing, review, and/or revision of the manuscript: S.J. Hutchings, M.V. Tongeren, L. Rushton

Study supervision: L. Rushton

The authors thank the participants of the future burden methodology workshop, particularly Drs. John Hodgson, David Kriebel, Hans Kromhout, Damien McElvenny, Kyle Steenland, Kurt Straif, and organizer Gareth Evans. The contributions and advice from the Health and Safety Executive and the rest of the project team is gratefully acknowledged, in particular Andy Darnton for his advice on the issues surrounding asbestos.

The work was supported by the UK Health and Safety Executive (grant number JN 3117).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.