The mortality of coke workers in Britain: a 20-year follow up

1. Introduction1.1 Background: The present study updates earlier reports on the mortality of two related cohorts of coke workers in Britain. All of the men studied had been employed as coke workers in 1967: 3,883 were employees of National Smokeless Fuels Ltd (NSF; now part of Coal Products Limited), and 2,790 were employed by the then British Steel Corporation (BSC; now British Steel).-Analyses of mortality over 12-year (BSC) and 13-year (NSF) periods had shown some excess lung cancer mortality in both cohorts (SMR). Limited use of occupational histories had also shown some evidence that the excess was related to work on or near the coke oven batteries, but the results were not consistent in this respect. There had been indications of excess mortality from some other causes also, notably bladder cancer, and cancer of the buccal cavity and pharynx.1.2 ObjectivesWe now report on further analyses of the mortality of both cohorts, based on an approximately 20-year follow-up, using improved job classifications, estimates of individuals’ exposures to benzene soluble materials (BSMs), smoking histories of individuals (available for some men only), and more powerful methods of statistical analysis. The objectives of these further analyses were to(a) estimate more precisely the real work-related excess of lung cancer;(b) establish if deaths from any other cause are work-related; and(c) seek to establish whether lung cancer mortality among men exposed mainly since 1970 differs from that in men whose exposure occurred primarily in earlier years.2. Methods2.1 Basic data: Most of the basic data for the study had been gathered earlier, in two phases: 1971-73, and 1975-76. These data were subjected to intensive further validation, to ensure consistency with the cohort definition, and internal consistency between various aspects of the recorded characteristics (eg. date of birth, and age at starting work). In particular, validation of the existing occupational histories of work in the coke industry until about 1976 was a major task. The histories of men in the cohort who were still employed as coke workers in 1976 were also updated, but the new information was not processed, and analyses were based on the men’s work experience in the coke industry up to the start of the follow-up period, in 1967.2.2 Classification of work history information; Another major effort involved devising, with the guidance of senior technical staff of both companies, a new and detailed coding scheme involving all combinations of job-name and place-of-work as recorded on the approximately 6,500 work histories of this study. The various detailed codes were subsequently grouped into a set of about 10 occupational groups, according to broad area of work (eg. coal preparation, screens, by-products etc.) and likely exposure to benzene soluble materials (BSMs) in particular.Thus, work on the coke oven batteries was subdivided into ovens tops, ovens sides, elsewhere ovens and ‘general ovens’ jobs. The allocation of detailed job-names of battery jobs into these four ovens-work groups was greatly assisted by statistical analyses of BSM monitoring data taken by both companies during the 1970s and 1980s. The data consisted of 1,630 measurements from all 13 NSF and from six BSC plants. These analyses identified systematic differences in BSM levels between jobs and between plants, with time throughout the 1970s and 1980s. (This reduction, estimated as a 13% decline on average and a reduction over any five-year period (P<0.01), follows a concerted effort to improve operating practice and maintenance procedures. The length of time worked by individuals, up to 1967, in the various broad categories of coke work was then derived from the occupational histories. This task was complicated by missing dates of job-changes for some men. Some of the difficulties were resolved by assuming that a job-change occurred at the mid-point of the period in question; but, where such assumptions seemed unreasonable, the man concerned was excluded from mortality analyses in relation to work in the industry. Individuals' exposures to BSMs were then derived for men who worked at those plants where BSM data were available.2.3 Smoking histories: Data on smoking habits were available for a limited number of the men studied. The NSF data were obtained during cross-sectional surveys of respiratory symptoms at four of the 13 coke works, in 1968 or in 1973. The BSC data were obtained at medical surveys 1971-73 of men who attended at works medical centres. The analyses involving smoking data reported here are based on 718 NSF and 1,049 BSC men. 2.4 Vital status: Vital status of men in the cohort was established up to 31 July 1987 with the assistance of the Registrar General (Scotland) and the Office of Population Censuses and Surveys (England and Wales). The vital status of 71 (1.8%) of NSF men and of 82 (2.9%) of BSC men was undetermined. Date of death, and primary cause of death coded according to the 8th Revision of the ICD system, were obtained for each man who died.2.5 Statistical methods: Statistical analyses used the person-years-at-risk method to compare the observed cause-specific mortality of the men studied with that expected, using published age-and year-specific death rates for men in the same 'regions' of Britain. Lung cancer mortality was compared also with death rates for semi-skilled workers in England and Wales. Results are presented in terms of Standardised Mortality Ratios (SMRs).In a second series of analyses, trends in death rates (per 1,000 person-years-at-risk) were examined in relation to age, years in various coke industry occupations, and estimated exposure to BSMs (up to 1967), using Poisson regression methods for grouped occurrence data. These analyses were supplemented by investigation of some possible associations using proportional hazards ('Cox regression') modelling.3. Results3.1 Lung cancer: The SMR for all-cause mortality over 20 years was 92 in both the NSF and the BSC study groups compared with general population ('regional') death rates. The lung cancer SMR was 125 and 127 in the two groups respectively, a statistically significant excess. Compared with, semi-skilled workers rather than with the general male population, the lung cancer SMRs were lower, at 104 (NSF) and 110 (BSC), but nevertheless, higher than the SMRs for all causes (89 in both groups, using Social Class IV death rates).Analyses of lung cancer mortality in relation to time worked in the various coke oven (battery) jobs strongly suggested an association between lung cancer death rates and time worked on 'general ovens' jobs in the NSF study group. There was however a statistically significant association between lung cancer mortality and estimated exposure to BSMs in the NSF workers, having taken age into account also. The relative risk was estimated as 1.18 for an increase of 10 mg.m~3 years of BSM exposure. Though the fact of the association is not in doubt, the magnitude of the relative risk as estimated is not similarly reliable, because of limitations to the reliability of the BSM exposures as estimated (see Discussion below).Similar analyses in the BSC study group showed a statistically significant association between lung cancer mortality and length of time as an 'ovens tops' worker. The relative risk was estimated as 1.81 per 10 years' as an ovens top worker. Analyses in relation to estimated exposure to BSMs also showed statistically significant differences in lung cancer death rates between exposure groups. The differences arose primarily from the contrast between the highest exposure category, and the two intermediate exposure groups. Limited data from both cohorts gave results consistent with a positive interaction between cigarette smoking and exposure to BSMs in the occurrence of lung cancer, but the data were not so strong that the interaction can be considered proven.3.2 Cancers at other sites: Earlier indications of excess bladder cancer mortality among NSF workers, based on the 13-year follow-up results, were not supported by extending the follow-up to 20 years. The 20-year SMRs for bladder cancer were 111 (NSF) and 88 (BSC). Further analyses suggested some associations between bladder cancer mortality and coke ovens (battery) work for BSC (but not for NSF) workers. These BSC results are based on eight deaths only. There was no evidence of excess kidney cancers.A statistically significant excess mortality from laryngeal cancer was found in the NSF cohort (SMR 260, based on nine deaths); the slight excess in BSC workers (SMR 132, three deaths) was not statistically significant. Laryngeal cancer was not found to be related to battery work or to exposure to BSMs.Excess mortality from stomach cancer in BSC workers was not statistically significant (SMR 130, 95% CI 89-183), and the NSF SMR for this cause was 88. There was however a statistically significant positive association between occurrence of stomach cancer and time worked in the by-products jobs.Mortality from cancers of several other sites was also considered and results are reported.3.3 Non-malignant causes: Summed over age-groups and cohorts, the 20-year mortality from ischaemic heart disease (IHD) showed an SMR of 95 (based on 790 deaths). There was however a clear decline in SMRs with age in the NSF cohort, from an SMR of 190 at attained age less than 45, to an SMR of 79 at age 75 or more. There was some, less striking, evidence of a corresponding age-related decline in SMRs in the BSC cohort. Analyses of IHD mortality in relation to time worked in the various occupation groups studied, for men at attained age of less than 65 years, showed a strong association in NSF workers of IHD mortality with time spent on ovens tops and sides, and with estimates of BSM exposures. There was some, weaker, evidence of an association with ovens work/BSM exposure in the BSC cohort, and an indication of higher risks in relation to time in coal preparation also.There was no evidence of excess mortality from non-malignant respiratory disease (bronchitis, emphysema and asthma), with an SMR of 90 (126 deaths observed) for the two cohorts combined, over a 20-year follow-up. Further analyses did not suggest an association between mortality from these causes and length of time in the dusty occupations of coal preparation or screens.Results of mortality from the other non-malignant causes studied did not indicate a problem among coke workers in particular.4. Discussion4.1 Strengths and limitations of the study: The present study is based on a large study group of about 6,500 men who were coke workers in 1967, and with a 20-year mortality follow-up. In principle, this should give a substantial body of information with which to identify any patterns of excess mortality among coke workers in Britain. In practice, the success of the study procedures for cohort identification, for subsequent validation of the data collected, and for establishing vital status have ensured a dependable base of information for analysis.The scale of the study is attested to by the fact that 27 plants were included; that 2,208 men are known to have died over 20 years; that these include 319 deaths from lung cancer; and that more than 20 deaths were expected from a relatively rare but important cause of death such as bladder cancer.The work histories were generally of good quality, though missing dates of job-changes meant that simplifying assumptions had to be made in the course of data processing, to avoid substantial exclusions from the study; and it was not possible to assess reliably the likely exposures to BSM experienced by men in some of the jobs as recorded (eg. general or yard labourer). Measurements of BSMs from routine surveillance programmes through the 1970s and 1980s were made available to the study. These data were of interest in their own right. They also opened up the possibility of analysing mortality relative to estimated exposure to BSMs, and this was done.As well as limitations arising from the occupational histories, two other limitations to the reliability of the estimated BSM exposures should be noted. First, the available concentration data refer to the 1970s and 1980s, whereas the men's exposures occurred primarily in the 1950s and 1960s. However, the more recent measurements are likely to be a good guide to earlier levels, because the production process has not been changed fundamentally over the years, though improved methods to limit emissions were introduced during the early 1970s. Secondly, exposures were estimated up to 1967 only, and were not updated in the light of the men's continued experience through the follow-up period. Nor is this limitation as severe as it might first appear: as well as normal turnover of industrial workers, employment in coke production in both companies underwent severe contraction during the 1970s and 1980s, thereby limiting the further exposure of the men studied.4.2 The results and their implications: Extension of the mortality follow-up from 12 or 13 to 20 years confirmed excess mortality from lung cancer in the two study groups of coke workers. More importantly, the results associating that excess with a history of work on or around the coke batteries were much less ambiguous than previously. In particular, the present study has shown a statistically significant association between mortality from lung cancer, and estimates of exposure to BSMs.Scientifically, the result is not surprising. It has long been known that coke oven emissions contain carcinogens; and a relationship between estimates of exposure, and lung cancer mortality, has previously been shown in North American studies.However, what was not previously known was, that the data from the present study combined with the strength of the effect were together powerful enough to show a clearcut relationship in the present context.The main implication is that we now know that it would be useful to carry out further work to describe more reliably the relationship between exposure to BSMs and lung cancer. The data themselves are both extensive and detailed, relative to the US data on which the scientific evidence for standards is currently based. In addition, we now know that further analyses will be productive, because a relationship has been demonstrated even though the exposures as used could be improved. The case for making those improvements (e.g. improved estimates of BSM concentrations; exposure estimation beyond start of follow-up; analyses in relation to age at first exposure, time since cessation of exposure etc.), and so providing important new information regarding lung cancer risks at various concentrations of BSMs, is therefore very strong.The second main finding is that the study has produced no convincing evidence that cancers of other sites are related to coke oven emissions. It is important however to keep an open mind about the possibility of excess work-related mortality from bladder cancer, where the results, though not convincing, remain suggestive. The study did however identify an increase in stomach cancer risks with increasing time worked in by-products jobs. Earlier studies from the North American series of papers also showed a stomach cancer excess in by-products workers. Though we do not postulate a mechanism, it is most likely that these findings are not simply coincidental, and that instead they point to a real effect of by-products work on stomach cancer mortality.Finally, men in both cohorts, but especially NSF, experienced age-related differences in SMRs from IHD, with excess mortality in younger men. Mortality from IHD was associated with work on and around the ovens, especially for NSF workers. Again, we do not have a plausible mechanism to explain the finding, but we think that the results are nevertheless sufficiently strong that the possibility that excess IHD mortality in younger man is in some way connected with their experience as ovens workers should not be dismissed. ""