Article Text
PDF
Abstract
Objective To investigate the long-term mortality of a cohort of Italian asbestos miners.
Methods The cohort included 1056 men employed in a chrysotile mine between 1930 and 1990, who were followed up during 1946–2014, for a total of 37 471 person-years of observation. Expected deaths and SMRs were computed using national and local (after 1980, when available) reference.
Results A total of 294 (27.8%) subjects were alive and at the end of follow-up, 722 (68.4%) were dead and 40 (3.8%) were lost to follow-up. The SMR for overall mortality was 1.35 (95%CI 1.25 to 1.45). The SMR for pleural cancer, based on seven observed deaths, was 5.54 (95% CI 2.22 to 11.4) and related to time since first exposure, but not to duration of employment, cumulative exposure or time since last exposure. The SMR for lung cancer was 1.16 (95% CI 0.87 to 1.52; 53 observed deaths), with no excess among workers with cumulative exposure below 100 fibre/mL-years (SMR 0.82; 95% CI 0.44 to 1.40).
Conclusions The update of the follow-up of this cohort confirmed an increased mortality from pleural cancer mortality in miners exposed to chrysotile and a lack of significant increase in lung cancer mortality.
- Asbestos
- chrysotile
- mining
- pleural cancer
- lung cancer
Statistics from Altmetric.com
What this paper adds
Data on long-term mortality of chrysotile miners and millers according to indicators of exposure are limited.
We identified an excess mortality from pleural cancer in this mature cohort of asbestos (predominantly chrysotile) miners and millers.
No significantly increased mortality from lung cancer was apparent, nor an association with indicators of asbestos exposure.
The study strengthens the evidence of an association between chrysotile mining and pleural cancer and provides no supporting evidence for a similar association with lung cancer.
Introduction
The Balangero mine was the largest open-air asbestos mine in Europe. It was located in the foothills of the Alps in Northwestern Italy. It started operation in 1916 and the production increased until an yearly average of 130 000–1 60 000 tons of chrysotile asbestos in the 1970s. The mine closed in 1990, shortly before the 1992 ban on the mining, marketing and use of all types of asbestos, including chrysotile, in Italy. The mortality of the cohort of asbestos workers in this mine was reported in 19791 and updated in 19902 and 2009.3
We continued to follow-up the mortality experience of this workforce, which numbered over 1000 men, 68% of whom are now deceased. Compared with the previous report,3 this report comprises 29% more expected deaths, 18% more observed lung cancer deaths and 75% more observed pleural cancer deaths; in addition, we provide information on the role of various time-related factors related to asbestos exposure in contributing to the mortality of this uniquely mature cohort of asbestos miners.
Materials and methods
The initial cohort included 1182 men: after excluding contract workers, those who were employed for less than 1 year, those with inconsistencies in the data and those known to have died before 1946, a total of 1056 men (for a total of 37 471 person-years of observation) employed in the Balangero mine for at least 1 year between 1930 and 1989 were included in the analysis.
Information on date and place of birth and period of hire and of last employment was obtained from the personnel records of the mine. Since no data were available about the employment history after 31 December 1987, we hypothesised that subjects still working on that date remained employed until 1990 when production ceased, unless deceased in that period.
Details of jobs held by cohort members were obtained from factory records and coworkers. Jobs were classified as mining, crushing, waste dumping, screening and fibre separation, bagging and storage and maintenance. For each worker, cumulative exposure was calculated by summing across all jobs the products of estimated exposure and duration of employment in that job.
Fibre counts at the plant were first carried out in 1969. In order to categorise jobs by dust exposure levels before 1969, exposure circumstances occurring between 1946 and 1969 were simulated at the plant. Factory files were examined for information on daily production, equipment used, characteristics of the job and number of hours worked per day, and workers employed since 1935 helped to reconstruct the appropriate conditions.
Weights were introduced to make comparable simulated and measured data: for example, many of the more dusty operations, such as drilling, had a duration of only 1–2 hours in a typical working day; on the other hand, working shifts were longer in the past (up to 48 hours per week until 1963). Simulated and measured fibre levels were reported in a previous publication1: for example, mean concentrations in drilling were 37±10 fibre/mL up to 1950 and 5±1 fibre/mL in 1971–1976.
Cumulative exposure was calculated for each worker by summing the exposure level for each year of employment in the specific job category and expressed as fibre/mL-years. For 38 workers for whom no details on job history were available and for maintenance workers, the mean value of all jobs was used.
Data on the prevalence of tobacco smoking were obtained for 171 members of cohorts (14.5% of the total) from medical records compiled between 1970 and 1979 (median year, 1971).
For the purpose of this analysis, cohort members contributed person-time of observation starting 1 year after first employment at the mine or in 1946 for those first employed earlier and present in 1946, and ending at death, date of last contact for those lost to follow-up, 85th birthday or 31 December 2014. We excluded deaths and person-years above the 85th birthday, because this is an open category and because of lower validity of causes of death in this age group. The cohort had been previously followed up to the end of 2003 for mortality3; we extended the follow-up to 2014 by ascertaining vital status and cause of death, including contributory causes for deaths that occurred after 1988, from population registers and municipal registration offices and collecting death certificates from local health authorities. The causes of death were coded according to the Ninth Revision of the International Classification of Diseases (ICD-9).4
The expected numbers of deaths from selected cancers and other relevant causes were computed using age-specific and calendar-year-specific (5-year categories) male death rates from the whole country before 1981 and for the Piedmont Region, where the mine was located, from 1981 onwards (regional rates were not available before 1981). National mortality rates were available since 1955 from the National Institute of Statistics (ISTAT) and the WHO; we applied the 1955–1959 rates to the period 1946–1954; this choice might have led to an underestimate of expected deaths of causes, which showed an increased rate during this period, such as lung cancer. However, given the small numbers of deaths in that period, the impact on the final results is unlikely to be relevant. In addition, mortality rates were not available for years 2004–2005, and from 2011 onwards. Therefore, we applied rates for 2001–2003 to the period 2000–2004, and rates for 2006–2010 to the period 2005–2014. We applied the version of the ICD in use at the time of death and reported the results according to ICD-9 for consistency with previous reports.3
We computed SMRs as the ratio of observed and expected numbers of deaths. For selected causes of death, we also computed the SMRs according to several indices of exposure, according to categories chosen a priori. The Poisson distribution was used to compute the CIs for up to 10 observed deaths and the normal approximation for >10 observed deaths.5 The Poisson trend statistic was used to detect trends in the SMRs.5 In the case of lung cancer mortality, we performed an internal analysis based on Poisson regression models to compute mortality rate ratios (MRRs), and the corresponding 95% CIs, according to the duration of exposure, age at first exposure and time since first exposure.5 To assess the influence of the assumption that workers employed in 1987 remained employed until 1990, we conducted a sensitivity analysis in which employment histories were truncated on 31 December 1987.
Results
At the end of the follow-up, 294 (27.8%) subjects were alive and 722 (68.4%) had died; for six of them (0.8%), the cause of death was not available. A total of 40 (3.8%) subjects either emigrated or were lost to follow-up. Overall, the 1056 men in the cohort contributed a total of 37 471 person-years of observation.
Table 1 gives the observed and expected numbers of deaths from selected causes and the corresponding SMRs. We observed an excess overall mortality, based on 722 observed and 533.4 expected deaths (SMR 1.35, 95% CI 1.25 to 1.45). Mortality from all cancers was close to expectation (176 observed vs. 165.3 expected deaths, SMR 1.07, 95% CI 0.91 to 1.23). The SMR for pleural cancer was 5.54 (95% CI 2.22 to 11.4), based on seven observed deaths. With reference to lung cancer mortality, the SMR was 1.16 (95% CI 0.87 to 1.52), based on 53 observed versus 45.5 expected deaths. Only one case of peritoneal cancer was detected, resulting in an SMR of 1.09 (95% CI 0.03 to 6.08).
SMRs for selected causes
The SMRs for cancers in organs of the upper aerodigestive tract were increased, without however reaching statistical significance (oral cavity and pharynx, 1.35; 95% CI 0.54 to 2.77; oesophagus, 1.23; 95% CI 0.40 to 2.87; larynx, 1.58; 95% CI 0.68 to 3.11). The results for mortality from other cancers were unremarkable. Considering other major causes of death, we observed 62 deaths from liver cirrhosis (21.2 expected, SMR 2.92; 95% CI 2.24 to 3.74) and 58 from accidents and violence (33.1 expected, SMR 1.75; 95% CI 1.33 to 2.26). Mortality from ischaemic heart diseases (SMR 0.89; 95% CI 0.69 to 1.12), cerebrovascular diseases (SMR 0.82; 95% CI 0.60 to 1.09) and chronic obstructive pulmonary diseases (SMR 1.02; CI 0.65 to 1.51) was close to expectation.
Table 2 provides details on the number of deaths from selected causes in the different follow-ups of this cohort. No separate results on pleural cancer have been reported in the first study. The first survey showed a significant excess for laryngeal cancer, but no additional deaths from laryngeal cancer occurred after the second follow-up.
Observed and expected deaths from selected causes in subsequent mortality follow-ups
Table 3 presents the number of deaths and the corresponding SMRs from selected causes, according to selected measures of asbestos exposure. No trend in SMRs was found for pleural cancer mortality according to duration of employment, cumulative exposure and any other measure of exposure considered. All pleural cancer deaths occurred 30 or more years since first exposure and five of them 40 or more years since first exposure. Four of the deceased workers experienced 20 or more years of employment. There was, however, no significant duration–risk relationship. Mortality from pleural cancer remained elevated 30 or more years after cessation of exposure. Cumulative asbestos exposure and other time-related factors considered did not show a clear pattern of risk with mortality from pleural cancer. In particular, the SMR for pleural cancer among workers with cumulative exposure below 100 fibre/mL-years was 5.84 (95% CI 1.20 to 17.1). The SMR for lung cancer was 0.82 (95% CI 0.44 to 1.40) among workers with cumulative exposure below 100 fibre/mL-years. A separate analysis of workers with cumulative exposure below 25 fibre/mL-years resulted in a SMR equal to 2.40 (95% CI 0.49 to 7.01, three deaths).
SMRs from selected causes according to indicators of asbestos exposure
The results of the Poisson regression analysis on lung cancer mortality are reported in table 4. This internal analysis showed no significant trend for any of the exposure variables considered. The sensitivity analysis in which the employment history of workers employed on 31 December 1987 was truncated on that date produced results very close to those of the main analysis (not shown in detail).
Mortality rate ratios (MRRs) from lung cancer according to selected indicators of asbestos exposure
The analysis of laryngeal cancer was limited by small numbers. Out of eight workers who died from this cause, seven were first employed before 1946, and a trend was suggested according to duration of exposure and cumulative exposure (table 3). SMRs for ischaemic heart disease did not show an association with indicators of exposure, although there was a trend of increasing SMR according to period of first employment (table 3).
Among other causes of death at increased mortality are ‘ill-defined descriptions and complications of heart disease’ (62 deaths; no reference rates available), asbestosis (24 deaths) and tuberculosis (16 deaths).
The prevalence of tobacco smoking in the 1970s was 42.9%, 65.5%, 54.8%, 75.0% and 37.5% in the age groups 25–34, 35–44, 45–54, 55–64 and 65–74, respectively.
Discussion
The present update of mortality in a cohort of asbestos miners from Italy confirmed the increased risk of pleural mesothelioma, which persisted more than 30 years after cessation of exposure, and the lack of a significant excess of risk of lung cancer and other neoplasms, notably laryngeal cancer, which was elevated in a previous report.1–3 The study is not informative with respect to peritoneal mesothelioma. Thus, the update of this cohort of asbestos chrysotile miners provides additional quantitative evidence on the role of chrysotile on mesothelioma risk.6
There is anecdotal evidence that crocidolite was occasionally present at the Balangero mine for material testing and mixture preparation.7 Piolatto et al 2 reported that the examination of several samples of chrysotile from the mine ruled out the presence of contamination with amphiboles at detectable concentrations and that a new fibrous silicate, named balangeroite, was characterised (0.2%–0.5% of the total mass samples of asbestos commercialised from the Balangero mine). Balangeroite fibre dimensions are similar to those of amphiboles, and in some studies this fibre appears to be even more durable than crocidolite.8–10 We previously considered the possible roles of balangeroite and amphibole fibres used in the plant as an explanation for the increased mortality from pleural cancer.1–3
In the first analysis of this cohort, no deaths from pleural cancer were reported,1 and two deaths were observed in the second follow-up. This did not allow carrying out in-depth analyses.2 In the third analysis, four deaths from pleural cancer and one from peritoneal cancer were reported3 (table 2). Pleural cancer deaths occurred among cohort members with high cumulative exposure, which is compatible with an effect of level of exposure,11 although no linear trend in risk was evident, and the small number of deaths did not allow allowance for latency and other relevant risk determinants. In the present update, deaths from pleural cancer were also observed among workers with low cumulative fibre exposure, and while applying due caution in the interpretation of the results because of the small number of deaths from pleural cancer, a role of cumulative dose was no longer apparent. The small number of deaths from pleural cancer also reduced the ability to disentangle the effect of various indicators of asbestos exposure, for example, through a multivariate internal analysis. Our results confirmed the need for long latency in order to detect the full effect of asbestos exposure on mesothelioma risk: all deaths from pleural and peritoneal cancers in this cohort were observed 30 or more years from first exposure, with no suggestion of a decline in risk after cessation of exposure.12 13
Our analysis confirmed the results of previous analyses of this cohort3 of a lack of substantive increased mortality from lung cancer in this cohort; in particular, no increase in risk was observed in the group of workers with a cumulative dose of less than 100 fibre/mL-years. Our study has 90% statistical power to detect a SMR of 2.0 for lung cancer among workers with less than 100 fibre/mL-years cumulative exposure.5 On the other hand, the power of the analysis on lung cancer risk among workers with less than 25 fibre/mL-years was low.
Asbestos exposure has been associated with a few other cancers, but the available evidence is not consistent.14–17 This study confirmed previous follow-ups that did not support an association with stomach and colorectal cancers. No new cases of laryngeal cancer were observed in the current update of the follow-up; the SMR for this cause of death remains elevated, although the excess is lower than in earlier analyses (table 2). The increased mortality from this cause detected in the first analysis1 may have been due to random fluctuation.
The results of the mortality analysis of this cohort are persistently consistent with an higher mortality from cancers related to alcohol drinking (oral cavity, oesophagus) and other alcohol-related causes (liver cirrhosis, accidents and violence), although the mortality excess from these causes appears to have become less pronounced in previous follow-ups (table 2), likely reflecting improved social and health conditions in these workers over time. Drinking habits of the area could also have played an important role in determining the increased overall mortality of this cohort, which cannot be explained only by cancer mortality.
Among the strengths of this study research are the very long follow-up, the low proportion of subjects lost to follow-up and the relatively large number of lung cancer deaths, which allow detailed dose–response analysis according to chrysotile exposure.
A limitation of this study, common to other occupational cohort studies, is related to the potential confounding role of tobacco smoking, occupational exposures outside the mine and alcohol drinking. We retrieved data on tobacco smoking for 14.5% of cohort members: the prevalence of smoking was comparable with that of the general male population.18 Our study lacks information on other confounders, but these are unlikely negative confounders for lung cancer; however, occupational exposure to asbestos in other jobs might have contributed to the excess mortality from pleural cancer; a possible role of alcohol as a confounder is discussed above. In order to address the problem that environmental measurements were available only since 1969, simulated data were generated for the earlier period. These values have considerable limitations: they represent mean values for large job categories and could not consider specific circumstance (eg, weather conditions), which might have affected exposure levels. Reliance on death certificates is an additional limitation of this study.
The Balangero mine was an open-air quarry: it is unlikely, however, that this characteristic resulted in an important environmental contamination (and therefore non-occupational exposure to the workers in this cohort, many of whom lived in the same area), because the proportion of asbestos in the residual material that was dismissed from the mill was in the order of 0.3%.
As discussed in a previous analysis of this cohort,3 reliance on death certification for outcome assessment could result in low sensitivity and specificity in the diagnosis of mesothelioma, potentially leading to bias in the SMR in both directions. In addition, detailed analysis of mortality from pleural and peritoneal cancers suffered from limited statistical power from small number of deaths.
In conclusion, the extended follow-up of this cohort of asbestos (primarily chrysotile) miners confirmed the presence of an increased mortality from pleural cancer and the lack of a significantly increased mortality from lung cancer. This may be explained by the relatively low exposure level experienced by these workers, although other explanations, such as exclusion of highly exposed (contract) workers, competitive mortality from other causes of death such as non-neoplastic pulmonary diseases, asbestosis, healthy worker effect (though absent in this cohort for total mortality and small and non-significant for cardiovascular mortality) and differences in smoking habits relative to the general population, cannot be excluded. The cohort also experienced an increased mortality from causes linked to excessive alcohol drinking.
Acknowledgments
The authors thank Dr Nunzia Passiatore for help in the enumeration and follow-up of the cohort.
References
Footnotes
Contributors EP, CR, CLV and PB designed the study. EP and CR established the cohorts. EP, CR and CLV were involved in previous analyses of this cohort. FD acquired the additional mortality data. CP contributed to the statistical analysis. PB and EP drafted the manuscript. All authors reviewed and approved the manuscript.
Competing interests EP and CR reported personal fees for the defence from several law firms, outside the submitted work. CLV reported personal fees from expert opinion for the court and the defence in asbestos litigation, excluding mining and manufacturing, outside the submitted work. PB reported personal fees from expert witness for the defence in asbestos-related litigation, outside the submitted work.
Provenance and peer review Not commissioned; externally peer reviewed.