Mortality and cancer incidence in Ontario glass fiber workers
1 Program in Occupational Health and Environmental Medicine, McMaster University, 1200 Main Street, West Hamilton, Ontario M5G 2E9, Canada
2 Institute for Work & Health, 481 University Avenue, Toronto, Ontario M5G 2E9, Canada
Correspondence to: Harry S. Shannon, Program in Occupational Health and Environmental Medicine, McMaster University, 1200 Main Street, West Hamilton, Ontario M5G 2E9, Canada. Tel: +1 905 525 9140, ext. 22333; fax: +1 905 528 8860; e-mail: shannonh{at}mcmaster.ca
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
Background In a previous cohort study of glass fiber manufacturing, we found a significant increase in lung cancer. This study extends the follow-up period.
Methods We conducted a historical prospective study of 2557 men employed in producing glass wool. We obtained work histories, causes and dates of death, and date and site of cancer diagnoses. We computed standardized mortality and incidence ratios (SMR, SIR).
Results The overall SMR for lung cancer was 163 (95% CI = 118–221). The SMR did not vary consistently by duration of employment and time since first employment. However, plant workers with >20 years' employment and >40 years since first exposure had an SMR for lung cancer of 282 (95% CI = 113–582). The SMR dropped with later date of first exposure, but the trend was non-significant. There was an unexpected overall increase in kidney cancer incidence.
Discussion The increase in lung cancer is greater than in other cohorts of glass fiber workers. Since exposure data are lacking from the early years of the plant, we cannot state if the excess was due to glass fibers, other work exposures or other reasons.
Keywords Cancer epidemiology; cohort; fibers; lung cancer
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
There have been a number of epidemiological (and other) studies of the potential health effects of man-made mineral fibers (MMMFs). The concern is that, because the fibers are structurally similar to asbestos fibers, MMMF may also cause respiratory diseases including especially lung cancer. Two large epidemiological cohort studies have been conducted: one in the USA by Marsh et al. [1] among ‘fibreglass’ workers and the other in Europe by Boffetta et al. [2]. Several smaller studies have also been done.
Importantly, research has distinguished different types of MMMFs: rock and slag wool glass fibers (wool) and glass filament. Synthesizing results from a conference held in 1986 which included results from the US and European cohorts, Doll [3] wrote ‘... we must, I think, conclude that there has been an occupational hazard of lung cancer in the rock or slag wool section of the industry, that there may have been a hazard in the glass wool [fibre] sector and that there is no material evidence of a hazard in the glass filament section’ (p. 811). He also concluded that no hazard for other causes of disease had been established.
A more recent systematic review by Berrigan [4] incorporates updated information. His meta-analysis estimated standardized mortality ratios (SMRs) for lung cancer of 1.32 (95% CI 1.15–1.52) for rock wool, 1.23 (1.10–1.38) for glass fiber and 1.08 (0.93–1.26) for glass filament exposure. He argued that some or all of the increases in mortality could be due to smoking. The review was limited to mortality from respiratory cancer and did not explore other diseases. The International Agency for Research on Cancer also recently conducted a re-evaluation of MMMF [5]. They concluded that glass fibers were ‘not classifiable as to carcinogenicity to humans’.
The evidence does not appear to support a relationship with causes of death (or disease) other than lung cancer. In general, reports study mortality rather than incidence of disease. Since survival from lung cancer is poor, this is likely only to have a modest effect on results. However, for other diseases, the power of a study to detect any increase in risk may be notably greater if incident cases are examined.
In addition to the cohort studies, Berrigan [4] also reviewed 10 case–control studies, nested within their cohort (or at least part of the cohort). Usually, the case–control component aimed to adjust for smoking and other potential confounders. Berrigan concluded that accounting for smoking ‘appeared to have modest effects on the risk estimates’ (p. 360).
One of the studies in Berrigan's review was our previously developed cohort of glass wool production workers. In our last report [6], we found 19 cases of lung cancer compared to 9.5 expected in plant workers. This was a statistically significant increase. However, there were no significantly increased SMRs from other causes. In this report we extend the period of follow-up from the end of 1984 to the end of 1997. In addition, we now have data on cancer morbidity, available from 1969 through 1996.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
We report a study of workers at an insulating glass wool plant in Ontario, Canada. The cohort of 2557 comprises all men who had worked for at least 90 days and were employed at some time between 1955 and 1977.
The computer tape with data from our earlier report was unfortunately corrupted, so we had to re-extract both personal information and work histories. Since there were few historical data on dust levels, we used duration of exposure as a proxy for quantitative information such as cumulative dose. We analyzed data by date of first exposure as well, since all anecdotal reports note that fiber levels were higher in the earlier days of the plant. We divided the cohort into three groups: plant-only workers, office-only workers and the remainder who had worked in both plant and office (‘mixed’ exposures). All jobs done up to the closing of the plant in 1991 were recorded.
The plant began operation in December 1948 and closed in April 1991. Limited exposure measurements for some of the chemical agents were made by the company during 1977–1990. They included ammonia, formaldehyde, phenol, carbon monoxide, solvents, asphalt fumes, total dust, crystalline silica and glass wool fibers. The reported concentration of glass fiber ranged from 0.01 to 0.32 fibers/ml with an average value of 0.03 fibers/cc for the period 1977–1990.
As we conducted our study, an independent effort was in progress to quantify past exposures in the plant (A. Virji, personal communication). This used the direct measurements noted above, and other information on changes in production, automation, presence or absence of ventilation and other factors, to estimate the multiples by which to adjust measurements from 1977 to 1990 to reflect historical exposures. The multipliers were typically less than two, and reached a maximum of six. These still led to estimates of <1 fiber/cc for glass fibers. Other exposures were also still less than (in most cases, much less than) the current threshold limit values (TLVs) [7].
Since the US and European studies had used various records and methods to derive estimates of fiber concentrations in different jobs and information about other possible exposures [8–10], we considered whether to use such information in this analysis. We noted the massive effort for the US cohort [8], yet they concluded that despite 35 person-years of effort (more than half of which was time of senior researchers), quantitative extrapolations for periods before around 1960 were ‘highly uncertain’. They also noted that there was little measurement data for validation of the estimates before the mid-1970s (although what was available supported the estimates). Since similar concerns would also apply to this study, we concluded that use of quantitative estimates would provide unjustifiable weight to those estimates, and that it was preferable not to rely on them.
The absence of quantitative data means that exposure–response relationships using our surrogate measures will likely understate such relationships. Furthermore, overall estimates of any health effects at higher exposures will be diluted, since we cannot separate out those with lower exposures.
Follow-up was conducted by computerized record linkage (CRL) at Statistics Canada [11]. Their National Mortality Data Base includes information on all deaths in Canada since 1950, and the Cancer Registry has records on those with newly diagnosed primary tumors since 1965. CRL uses a probabilistic method to compare individuals in the cohort to those in the mortality database and cancer registry [12]. By calibrating the ‘scores’ using those known to be alive or dead, or to have been diagnosed with cancer, the process identifies those who have died and those who have developed cancer.
Statistics Canada appended to the work history file the date and cause of death for those who had died. Similarly, they added the date of diagnosis and cancer site for those found to have cancer. Individuals' names were removed from the file, which was then returned to us. We have found this linkage procedure to be accurate [13].
We analyzed the data using the OCMAP-PLUS program from the University of Pittsburgh [14]. This computes the duration of time each subject spends in age- and calendar-specific periods. These durations are added across all individuals in the cohort. Those men not identified as dead through the CRL were assumed to be alive on 31 December 1997. Based on the mortality rates for Ontario, SMRs were calculated adjusting for age and calendar period. For cancer incidence, comparable calculations produced standardized incidence rates (SIRs). Further analyses allowed for length of employment and time since first employment (TSFE), while others took into account the date first worked at the plant. The cut-points for use in these analyses were chosen in advance, before we did any calculations. The program computes 95% CI for the SMR or SIR based on Poisson probabilities. P-values are two sided.
The disease of primary interest a priori was lung cancer. We intended to focus some attention on respiratory diseases as well, including pneumoconiosis.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
Person-years for the whole cohort by age and calendar period are shown in Table 1, with each axis divided into 5-year categories. The extended follow-up added
30 000 person-years of observation, or 68% of the previous number. The cumulative follow-up beyond age 65 had increased 4-fold.
|
We examined mortality for major cause groups separately for the three categories of exposure (Table 2). For the largest group, plant-only workers, there were 299 deaths in total, compared with 320 expected (SMR = 93, 95% CI = 83–105), a deficit perhaps due to a healthy worker effect. The SMRs were lower in office-only workers (69, 95% CI = 45–100) and in those with mixed exposures (67, 95% CI = 45–95). Deaths from cardiovascular diseases were below expected but not significantly so (e.g. in plant-only workers, SMR = 90, 95% CI = 74–109). The 94 cancers in plant-only workers represented a non-significant excess (SMR = 115, 95% CI = 93–140). There were slightly more respiratory disease deaths than expected (SMR = 119, 95% CI = 74–182), while most other comparisons showed SMRs <100, but not significantly so. Further analyses concentrate on plant-only workers.
|
We also examined mortality for selected specific causes by exposure category (Table 3). There was a significant increase in lung cancer deaths (observed deaths = 42, SMR = 163, 95% CI = 118–221). A non-significant increase was observed for chronic obstructive pulmonary disease (COPD) and bronchitis (12 deaths, SMR = 137, 95% CI = 71–239). The SMRs for all other selected causes were (non-significantly) <100.
|
Since lung cancer was the only cause of death with statistically significant evidence of an increase in risk and since it was the main disease of a priori concern, we examined mortality due to it by duration of employment and TSFE (Table 4). While no clear pattern emerges, it is notable that plant workers with long employment, at least 20 years, experienced a significant increase in SMR (observed deaths = 17, SMR = 189, 95% CI = 110–303). In the subgroup that also had >40 years' TSFE, the SMR reached 282, based on seven deaths (95% CI on SMR = 113–582).
|
Overall cancer morbidity for selected sites is shown in Table 5. There was a significant increase in lung cancer incidence in plant-only workers (SIR = 160, 95% CI = 119–211). For other sites, the SIRs were generally close to or below 100, with one exception. The number of kidney cancers was nearly double that expected, both for plant-only and all workers. For the whole cohort, the 14 cases represented a significant excess (SIR = 192, 95% CI = 105–321).
|
The final set of analyses examined the data in plant-only workers by the date of first hire (Table 6), since anecdotal evidence suggests that dust concentrations declined over time. The overall SMR (for all causes) was lowest in the middle period of first hiring (1960–1970), significantly <100 (observed deaths = 41, SMR = 55, 95% CI = 40–75). The SMR for lung cancer declined from a significant 172 for those first hired before 1960 to 101 (based on just two cases) for those first hired at the site after 1970. With small numbers, the trend was not significant.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
We have presented results from an extended follow-up of a cohort of glass fiber workers. Both the observed and expected numbers of deaths were more than doubled from the previous analyses. As well, we now include data on cancer incidence—which is available in Canada from 1969 onwards.
Overall, mortality remains below expected. The overall SMR was 88, a statistically significant shortfall. In general, analyses for the specific causes of death (or groups of causes) showed no significant increases. Indeed, in many instances, the number of deaths (or cancers) was less than expected.
The notable exception is lung cancer. This was our a priori cause of major interest, and our previous report had shown a significant increase, with 19 cases compared with 9.53 expected in plant-only workers. With the additional data, the SMR for lung cancer in this group was 163, remaining statistically significant. If we analyze only the additional data (based on person-years and mortality after 1984), there were 23 lung cancer deaths, compared with 16.2 expected, an increase but not a significant one (SMR = 142, 95% CI = 90–213). Overall, lung cancer incidence was also significant, and the SIR was comparable (observed deaths = 50, SMR = 160, 95% CI = 119–211). For both mortality and incidence, there were no clear patterns by duration of employment and TSFE. Nevertheless, the highest SMR (and highest SIR for incidence) occurred in the longest duration period, and in any event, duration of exposure is only a surrogate for proper quantitative measures of exposure. Also, there was a tendency for lung cancer mortality to decrease with later first hire, although with small numbers for the more recently exposed, the trend was not significant.
The increase in lung cancer is greater than that found in other cohorts. Even for the extended period of follow-up, the point estimate of the SMR is high (albeit not significant). It is possible that fiber levels were particularly high in this plant, although we have no evidence of this. A detailed retrospective exposure assessment was beyond the scope of this study. In any event, it is doubtful if any quantitative estimates of fiber concentrations would be accurate, given the lack of data before 1977. Alternatively, there may have been other exposures present that were lung carcinogens, although, as noted in Methods, best estimates of other known exposures were well below current TLVs.
There was no significant increase overall in COPD, a condition of secondary a priori interest. The number of cases, though, was small. The one case of pneumoconiosis was close to the expected number (0.77 in plant workers, 0.99 in the whole cohort).
The finding of a significant increase in the overall cohort for kidney cancer morbidity was unexpected. While an excess had not been observed in the mortality data, the number of incident cases is much larger than the number of deaths (14 versus 3). As with lung cancer, though, we found no consistent pattern when the data were broken down in more detail. The large cohort study in the USA reported in 1990 on follow-up from 1948 to 1985 [15]. While deaths from kidney cancer were very close to the expected, mortality from nephritis and nephrosis was significantly increased. Goldsmith and Goldsmith [16] explored this issue by reviewing other literature and concluded that silica was likely to be the cause of the excess renal disease. In their more recent follow-up up to 1992 of the US cohort [1], the authors found no significant excess. Deaths from kidney cancer were very close to expected, and had not been increased at the earlier follow-up. The European study [17] examined cancer incidence for plants in countries covered by cancer registration. Kidney cancer incidence was below expected in the glass wool cohort, although with just six cases, the SIR of 50 was not significantly low (95% CI = 18–109). It thus appears that our finding of increased kidney cancer was unlikely to be due directly to glass fibers.
There were several limitations to our study. Firstly, we had no direct data on exposure. The use of duration of employment provides a surrogate, albeit one that is imperfect. Our analyses by date of first exposure were designed to partially address this, on the grounds that exposures in the earlier periods tended to be higher. As noted earlier, we have doubts that reconstructions of exposure data can produce accurate estimates.
We also did not have data on smoking habits. The fact that mortality and morbidity for other sequelae of smoking (respiratory disease, ischemic heart disease) were not significantly increased implies that levels of smoking in the cohort were not particularly high relative to the general population. If correct, this in turn implies that smoking did not account for the increase in lung cancer. In any event, the levels of smoking in this cohort would have to have been unrealistically high to produce the observed SMR for lung cancer. Cigarette smoking prevalences in Canada in working-age men even in 1970 were >50% [18], so that even if all lung cancer deaths were caused solely by smoking, the smoking prevalence would have been >80% to produce an SMR of 163. Further research on this cohort should attempt to incorporate smoking data—e.g. by conducting a nested case–control study of lung cancer.
In the US cohort, adjustment for smoking in a nested case–control study of lung cancer reduced the relative risk from glass fiber [1]. However, this may have been an artifact of the data collection. Smoking information was obtained from proxy respondents for the cases. However, controls who were still alive reported their smoking habits in a telephone interview. In these instances, social desirability bias may have led to understatement of smoking. Occurring only in the control group, this bias would have led to overadjustment for smoking, thereby underestimating any effect of work exposures in the plants.
While the extended follow-up has substantially increased observed and expected cases and deaths, the numbers in subcategories remain small. This limits our ability to identify patterns in the data, and we should be careful not to conclude that the failure to find significant trends necessarily reflects an absence of trends.
We did not use local rates for standardization, since they are less stable than provincial rates. Using a CD supplied by Cancer Care Ontario [19], we estimated that age-adjusted lung cancer mortality rates for men were 8% higher in the county in which the plant was located than in the province from 1979 to 1998. Comparable incidence rates were 11% higher. Allowance for this thus slightly reduces the SMR and SIR, but they are still statistically significant.
Finally, cancer morbidity data were only available from 1969. This means that we were unaware of any incident cancers that occurred before that date (unless they were fatal before 1969). If the individual was still alive in 1969, we counted person-years at risk for the particular cancer, even though they should strictly be excluded. The net effect will have been a slight underestimate of the SIR, but too small to have affected our conclusions.
In conclusion, the SMR for lung cancer mortality, which in our previous report was higher than those in other published studies, remains high with the extended follow-up, and it was also increased when we analyzed cancer incidence data.
The unanticipated increase in kidney cancer incidence may be a chance finding, or may reflect exposures to contaminants other than glass fiber. This warrants further investigation.
|
Conflicts of interest |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
None declared.
|
Acknowledgements |
|---|
Funding was provided by the Workplace Safety & Insurance Board of Ontario, the organization that adjudicates claims for occupational diseases in Ontario. We thank Statistics Canada for conducting the CRL. Daniela Ghiculete helped in the early stages of the project.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflicts of interestReferences |
|---|
- Marsh GM, Youk AO, Stone RA et al. Historical cohort study of US man made vitreous fiber production workers: I. 1992 fibreglass cohort follow-up: initial findings. J Occup Environ Med 2001;43:741–756.[Web of Science][Medline]
- Boffetta P, Saracci R, Andersen A et al. Cancer mortality among man-made vitreous fibre production workers. Epidemiology 1997;8:259–268.[CrossRef][Web of Science][Medline]
- Doll R. Symposium on MMMF, Copenhagen, October 1986: overview and conclusions. Ann Occup Hyg 1987;31:805–819.
[Abstract/Free Full Text] - Berrigan D. Respiratory cancer and exposure to man-made vitreous fibers: a systematic review. Am J Ind Med 2002;42:354–362.[CrossRef][Web of Science][Medline]
- International Agency for Research on Cancer. Man made vitreous fibres. In: IARC Monographs on the Evaluation of Carcinogenic Risk to Humans. Vol. 81. Lyon, France: IARC Press, 2002.
- Shannon HS, Jamieson E, Julian JA, Muir DCF, Walsh C. Mortality experience of Ontario glass fibre workers—extended follow-up. Ann Occup Hyg 1987;31:657–662.
[Abstract/Free Full Text] - American Conference of Governmental Industrial Hygienists. 2004 TLVs and BEIs. Cincinnati, OH: ACGIH, 2002.
- Smith TJ, Quinn MM, Marsh GM et al. Historical cohort study of US man-made vitreous fiber production workers: overview of the exposure assessment. J Occup Environ Med 2001;43:809–823.[Web of Science][Medline]
- Cherrie J, Dodgson J. Past exposures to airborne fibers and other potential risk factors in the European man-made mineral fiber production industry. Scand J Work Environ Health 1986;12(Suppl. 1):26–33.
- Cherrie JW, Schneider T. Validation of a new method for structured subjective assessment of past concentrations. Ann Occup Hyg 1999;43:235–245.
[Abstract/Free Full Text] - Smith ME, Newcombe H. Use of the Canadian Mortality Data Base for epidemiological follow-up. Can J Public Health 1981;73:39–46.
- Newcombe HB. Handbook of Record Linkage: Methods for Health and Statistical Studies, Administration, and Business. New York: Oxford University Press, 1988.
- Shannon HS, Jamieson E, Walsh C, Julian JA, Fair ME, Buffet A. Comparison of individual follow-up and computerized record linkage using the Canadian Mortality Data Base. Can J Public Health 1989;80:54–57.[Web of Science][Medline]
- Marsh GM, Youk AO, Stone RA, Sefcik S, Alcorn C. OCMAP-PLUS: a program for the comprehensive analysis of occupational cohort data. J Occup Environ Med 1998;40:351–362.[CrossRef][Web of Science][Medline]
- Marsh GM, Enterline PE, Stone RA, Henderson VL. Mortality among a cohort of US man-made mineral fiber workers: 1985 follow-up. J Occup Med 1990;32:594–604.[CrossRef][Web of Science][Medline]
- Goldsmith JR, Goldsmith DF. Fiberglass or silica exposure and increased nephritis or ESRD (end-stage renal disease). Am J Ind Med 1993;28:873–881.
- Boffetta P, Andersen A, Hansen J et al. Cancer incidence among European man-made vitreous fiber production workers. Scand J Work Environ Health 1999;25:222–226.[Web of Science][Medline]
- Pechmann C, Dixon P, Layne N. An assessment of US and Canadian smoking reduction objectives for the year 2000. Am J Public Health 1998;88:1362–1367.
[Abstract/Free Full Text] - Cancer Care Ontario. Cancer Incidence, Mortality and Survival in Ontario. Release 1, April 2001, 1964–1998. CD-ROM. Toronto, Canada: Cancer Care Ontario 2001.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||