Objectives Increased cancer risks have been reported among workers in the rubber manufacturing industry employed before the 1960s, but it is unclear for workers hired subsequently. The present study focused on cancer incidence among rubber workers first employed after 1975 in Sweden and the UK.
Methods Two cohorts of rubber workers employed for at least 1 year were analysed. Standardised incidence ratios (SIRs), based on country-specific and period-specific incidence rates, were analysed for all cancers combined (except non-melanoma skin), bladder, lung, stomach cancer, leukaemia, non-Hodgkin's lymphoma and multiple myeloma. Exploratory analyses were conducted for other cancers with a minimum of 10 cases in both genders combined.
Results 16 026 individuals (12 441 men; 3585 women) contributed to 397 975 person-years of observation, with 846 cancers observed overall (437 in the UK, 409 in Sweden). No statistically significant increased risk was observed for any site of cancer. A reduced risk was evident for all cancers combined (SIR=0.83, 95% CI (0.74 to 0.92)), lung cancer (SIR=0.74, 95% CI (0.59 to 0.93)), non-Hodgkin's lymphoma (SIR=0.67, 95% CI (0.45 to 1.00)) and prostate cancer (SIR=0.77, 95% CI (0.64 to 0.92)). For stomach cancer and multiple myeloma, SIRs were 0.93 (95% CI (0.61 to 1.43)) and 0.92 (95% CI 0.44 to 1.91), respectively. No increased risk of bladder cancer was observed (SIR=0.88, 95% CI (0.61 to 1.28)).
Conclusions No significantly increased risk of cancer incidence was observed in the combined cohort of rubber workers first employed since 1975. Continued surveillance of the present cohorts is required to confirm absence of long-term risk and confirmatory findings from other cohorts would be important.
- occupational exposure
- cohort study
Statistics from Altmetric.com
What this paper adds
Occupational exposure in the rubber manufacturing industry is an International Agency for Research for Cancer group 1 human carcinogen. However, recent studies suggested that, following improvement in occupational hygiene, cancer risks were no longer observable.
In a prospective cohort of 16 026 workers employed since 1975 in the rubber manufacturing industry in the UK and Sweden, and followed for 23 years on average, no consistent increased cancer incidence was observed.
Although the findings from this study are reassuring, continued surveillance of the present cohorts is required to confirm absence of long-term risk.
Occupational exposure in the rubber manufacturing industry, that is, production of tyres and general rubber goods and process of retreading, is an International Agency for Research for Cancer (IARC) group 1 human carcinogen.1 The IARC evaluation, based mainly on observational studies on workers mostly employed before the 1960s, concluded that there was sufficient evidence of an increased risk of bladder cancer, leukaemia, stomach, lung cancer and lymphoma. The carcinogens involved in these excesses are as yet unknown but exposure to aromatic amines and solvents have been suspected to play a role.2
The rubber manufacturing industry has undergone radical technological changes since the 1950s, entailing major reductions in rubber dust3 and fume exposure and the decrease of known carcinogenic agents like benzene and β naphtylamine, although others, such as nitrosamines, are still present. A recent study from five European countries on 38 457 workers employed since 1975, with nearly a million person-years, showed no increase of cancer mortality for bladder cancer, leukaemia, lung cancer and lymphoma.4 However, this study suggested an increased risk of stomach cancer and multiple myeloma in the general rubber goods (GRG) sector, but not in the tyre sector. These findings were driven by an increased risk in one of the five contributing cohorts.
Incidence data were also collected in the UK and Sweden, part of the European cohorts and provide a helpful complementary evaluation of the association between occupational exposure in the rubber manufacturing industry and cancer. While easier to collect and widely available in several countries, mortality data are a heterogeneous mix of cancers diagnosed in the preceding years with various latencies between exposure and outcome. In addition, incidence data are less affected by misclassification problems as compared to mortality data and are not affected by trends in curability of some forms of cancers.5
This study reports data on incidence of cancer in the UK and Sweden with two objectives: first, to confirm whether no increased risk of cancer was also observed for cancer incidence; second, to evaluate whether the suspected increased risk of stomach cancer and of multiple myeloma were also present in cancer incidence.
Material and methods
A protocol specifying inclusion criteria and a detailed statistical analysis plan was prepared between local principal investigators of this study prior to data analysis. Details on the methods and mortality data have been published elsewhere.4 The initial cohort consisted of rubber workers employed since 1975 for at least 1 year in rubber manufacturing industries in Germany, Italy, Poland, Sweden and the UK, but incidence data were available only in Sweden and the UK. In Sweden, data from a cohort of workers first employed in 1975 or later were extracted from an initial cohort using personnel records from Swedish rubber manufacturing plants, situated in 11 different places all over the country. Vital statistics were obtained from Statistics Sweden and with linkage to the Swedish cancer registry, information on up to two tumours per worker were extracted. In the UK, an initial cohort of workers employed for the first time in the rubber manufacturing industry within the period 1982–1991 was established from 41 rubber factories in England, Wales and Scotland. Incidence of cancer was obtained from the UK Health and Social Care Information Centre. For both Sweden and the UK, the follow-up with vital statistics and cancer incidence was conducted up to 31 December 2011.
Primary, secondary and exploratory outcomes were defined in the study protocol, prior to study conduct, based on the IARC's evaluation and on the strength of association as reported in the systematic review of Kogevinas et al.6 The primary outcomes of interest were therefore incidence of bladder cancer and lung cancer. Secondary outcomes were incidence of all cancers combined (excluding non-melanoma skin cancer), stomach cancer, leukaemia, multiple myeloma and non-Hodgkin's lymphoma (NHL). Data from other cancer sites, not previously defined, were also included in exploratory analysis but only if more than 10 cases were observed in men or women combined. Non-melanoma skin cancers and benign neoplasms were excluded from the analysis. The list of international classification of diseases (ICD) codes used in the present article is reported in online supplementary table S1. ICD-7 and ICD-8 codes were used in Sweden, whereas ICD-9 and ICD-10 codes were used in the UK.
The observed numbers of cases for each cancer site were compared with the expected numbers calculated on the basis of national gender-specific, age-specific and period-specific incidence rates. Five-year age groups were used for age and time period. Reference rates were obtained from the CI5+ Database (Revision of February 2014). In Sweden, national reference rates were obtained. In the UK, reference rates were obtained separately for Scotland and England/Wales. Rates of Scotland and England/Wales were pooled as all UK factories were located in these three areas of the UK. Patients were followed from 1 year after date of hire until the earliest of the following: date of death, date of loss to follow-up/emigration or right censored at 31 December 2011. For each country, standardised incidence ratios (SIRs, ie, the ratio of observed to expected cases) were calculated together with their CIs based on the Poisson distribution of observed cases.7 ,8 Country-specific SIRs were combined using random-effects models9 which take into account potential heterogeneity among cohorts.
Measures of heterogeneity were reported using I2 statistics10 as well as tests for heterogeneity based on Cochran's Q statistic, although this test is known to have poor statistical power.11 In a sensitivity analysis, an induction-latency period of 10 years was applied.
The role of the duration of employment on the risk of all cancers combined was investigated in a Poisson model with a smooth function of duration of employment as an explanatory variable, and the logarithm of the expected number of cancers as ‘offset’. Duration of employment was modelled with cubic natural splines and three degrees of freedom. Cutpoints (knots) of duration of employment were built from duration of employment of participants diagnosed with cancer, such that at least eight cases occurred between two points. This enables a stable estimation of SIR while keeping enough points for modelling the splines for the parameter of duration. Since duration could not be estimated for several workers still employed at the last job history update, this analysis was restricted to the subset of 8100 workers (51% of the cohort) with complete job history.
All data were anonymised prior to statistical analysis. This study did not require a specific IRB approval as performed on fully anonymised secondary data. p Values below 5% were considered as statistically significant.
A total of 16 026 workers (12 441 men and 3585 women) were included in this study (table 1). The median follow-up was 23 years, contributing to a total of 397 975 person-years of observation. The majority of workers (77.6%) were men; more women were recruited in Sweden. Overall, 846 malignant cancer cases (excluding non-melanoma skin cancers) were observed during the follow-up, 437 in the UK and 409 in Sweden.
Table 2 shows observed cases and SIRs for different cancer sites for Sweden, the UK and both countries combined. Based on 45 cases observed during the follow-up, the risk of bladder cancer was SIR=0.88 (95% CI (0.61 to 1.28)) with no evidence of heterogeneity between countries. This absence of increased risk for bladder cancer remained when stratified by gender (see online supplementary table S2 and S3), although the analysis restricted to women is based on <10 cases from most outcomes. The risk of lung cancer was based on 82 cases and was significantly decreased in rubber workers as compared to the general population with an SIR of 0.74 (95% CI (0.59 to 0.93)) with no heterogeneity between countries.
Concerning secondary outcomes, a statistically significant decrease was observed for all cancers combined (SIR=0.83, 95% CI (0.74 to 0.92)) and for non-Hodgkin's lymphoma (SIR=0.67, 95% CI (0.45 to 1.00)). Risks of other secondary outcomes were neither significantly increased nor significantly decreased. At country level, the risk of all cancers combined was significantly decreased in Sweden and the UK. In addition, the risk of leukaemia was significantly decreased in Sweden.
Analysis of exploratory outcomes revealed no significantly increased risk for any cancer site when both countries were combined. A significantly decreased risk of prostate cancer was found: SIR=0.77 (95% CI (0.64 to 0.92)) based on 123 cases.
Results for cancers appearing after an induction-latency period of 10 years are shown in online supplementary table S4. This analysis covered 15 466 participants (7131 in Sweden and 8335 in the UK) representing overall 240 561 person-years of observation. Results remained close to those of the main analysis.
Results of analyses stratified by gender are given in online supplementary table S2 (men) and online supplementary table S3 (women). In men, there were significantly reduced SIRs for lung cancer and prostate cancer. In women, there was a significantly reduced SIR for all cancers combined, and SIRs below unity for most cancers. The only significantly increased SIR was for melanoma, SIR 1.65 (95% CI (1.04 to 2.60). It was based on 21 cases, of which 20 occurred in Sweden (see online supplementary table S3). This association was not found in men, neither in the UK nor in Sweden; the SIR was 0.74 (95% CI (0.44 to 1.24)) and was based on 30 cases (19 in Sweden and 11 in the UK).
The risk of all cancers combined (except non-melanoma skin cancer) was further investigated in a Poisson regression with a spline function applied to the duration of employment (figure 1). This analysis was restricted to 8100 workers (4005 in the UK and 4095 in Sweden) with complete information on job history, that is, representing 51% of the cohort. When compared to the rest of the cohort, these workers only slightly differed with more men (81.2% vs 73.5% for complete vs incomplete job history) and older on average (mean age at recruitment 29.7 vs 26.0 for complete vs incomplete job history). For employment durations under 20 years, the risk of cancer remained close to 0.8, and then the trend went up until about 1.
In this cohort of workers employed for at least 1 year in the rubber manufacturing industry first employed since 1975 in two European countries, there was no consistent indication of an increased risk of cancer incidence among the cancer sites preidentified as primary or secondary outcomes of interest, that is, all cancers combined and site-specific incidence for bladder, lung, stomach, leukaemia, myeloma and non-Hodgkin's lymphoma. In addition, none of the exploratory outcomes had their risk significantly increased when both genders were combined. From the Poisson regression, the risk of all cancers combined remained relatively stable with duration of employment with a slight increase after 20 years; confident interpretation of these regression findings is not possible.
A statistically significantly increased risk of melanoma was identified, but it was limited to Swedish women. With such an increase only observed in one country and one gender, and since rubber manufacture is not an outdoor employment, a difference in job exposure is unlikely to offer a good explanation for the increase. Non-melanoma skin cancer incidence is a tentative marker of solar exposure, but is markedly affected by access to screening and by registration bias. Hence, there is no solid data to help interpret the increase in melanoma incidence in Swedish women by either a role of non-occupational or occupational solar exposure, an overdiagnosis in women within the healthcare system in the rubber factory areas or a chance finding.
Analysis of cancer mortality data in five European countries also did not report an increased risk of cancer.4 Results on incidence for Sweden and the UK were in the same order of magnitude as results based on mortality for all sites except multiple myeloma in the UK. The SMR for multiple myeloma deaths in the UK cohort was 2.26 (95% CI (0.97 to 4.44)) for both genders, whereas in the present report on incidence the corresponding SIR was 1.04 (95% CI (0.45 to 2.05)). In addition, the observed increased risk in the mortality study was observed only in men in the general rubber goods sector, which represents only 36% of workers in the UK cohort. This suggests that the initially observed high multiple myeloma mortality observed in the UK could be a chance finding, but specific exposure-related explanations cannot be fully excluded. Risk of stomach cancer mortality was high in Poland in the mortality study. Similar to previously reported mortality findings, in the UK and Sweden the incidence of stomach cancer remained not increased.
The absence of increased risk of cancer in this study, in contrast to findings in the ‘old’ rubber industry, could be the result of changes in the rubber manufacturing industry. Data collected within the European project ExAsRub (improved EXposure ASsessment for prospective cohort studies and exposure control in the RUBber manufacturing industry), which combined comparable exposure information in the rubber industry across Europe in a database, are in line with this hypothesis by showing continuous decreasing time trends of inhalable dust from 1975 to 2005.3
The study has a number of limitations. First, there was a lack of accurate data on duration of employment in the rubber industry, due to no recent update of job histories which limited the analysis of long-term exposures. Second, the cohort of workers employed since 1975 in these two countries is young, with an average age at the end of follow-up of 50 years; hence, the magnitude of impact of cancer in these populations could only be partially evaluated and longer follow-up would be needed to evaluate the potential impact of these occupational exposures on entire life. Third, as for most studies which evaluated the impact of occupational exposure in the rubber manufacturing industry and risk of cancer,1 no behaviour data were available and it was therefore not possible to adjust for tobacco smoking and other potential confounders.
The findings of no observed increased risk of cancer in these cohorts from the rubber manufacturing industry are reassuring. However, it is recommended that these cohorts continue to be monitored regularly to investigate if absence of increased cancer risk is present after longer follow-up, and long-term positive effects of industrial hygiene improvements are maintained. Confirmatory findings from other cohorts would also be important.
The authors would like to thank Beata Świątkowska (Nofer Institute of occupational Medicine, Poland), Juergen Wellmann (University of Muenster, Germany), Dirk Taeger (Institute of the Ruhr-Universitat Bochum, Germany), Enrico Pira (University of Turin, Italy), Paolo Boffetta (Tisch cancer institute, USA), Carlo La Vecchia (Università degli studi di Milanoma, Italy) and Cecile Pizot (International Prevention Research Institute, France), for helpful discussion and comments on the manuscript. The authors would like to thank Kim Coppens (iPRI) for editorial support. Zoli Mikoczy and Ulf Bergendorf were responsible for the assembly of the Swedish database, which was funded by a grant from AFA and from Lund University.
Contributors The study Working Group comprised iPRI staff (MB, AK and PB) and national Principal Investigators, TS (UK) and KJ (Sweden). MB, PB and TS were involved in the planning of the study. TS supervised data gathering for the UK cohort. KJ supervised data gathering for Swedish cohort. MB and AK conducted data analysis. MB prepared the first draft of the manuscript. All authors contributed during the revision phase of the manuscript. All authors approved the final version submitted.
Funding The study was partially funded by an unrestricted research grant from the European Tyre and Rubber Manufacturers' Association (ETRMA) and was conducted and reported in full independence from the sponsor.
Competing interests None declared.
Ethics approval Not required for secondary use of data. IRB approval from both countries (United Kingdom and Sweden) covered already the use of these anonymised, non-identifiable cohort data.
Provenance and peer review Not commissioned; externally peer reviewed.
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.