American Journal of Epidemiology Advance Access originally published online on July 27, 2006
American Journal of Epidemiology 2006 164(5):405-420; doi:10.1093/aje/kwj252
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Meta-Analysis |
Meta-Analysis of Mortality and Cancer Incidence among Workers in the Synthetic Rubber-Producing Industry
1 Centre for Statistics in Medicine, University of Oxford, Oxford, United Kingdom
2 Division of Primary Care, School of Community Health Sciences, University of Nottingham, Nottingham, United Kingdom
3 MRC Institute for Environment and Health, Leicester, United Kingdom
4 Department of Health Sciences, School of Medicine, University of Leicester, Leicester, United Kingdom
5 Department of Epidemiology and Public Health, Faculty of Medicine, Imperial College London, London, United Kingdom
Correspondence to Dr. Alexander Sutton, Centre for Biostatistics and Genetic Epidemiology, Department of Health Sciences, University of Leicester School of Medicine, 22–28 Princess Road West, Leicester LE1 6TP, United Kingdom (e-mail: ajs22{at}le.ac.uk).
Received for publication October 25, 2005. Accepted for publication February 28, 2006.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
Production of synthetic rubber involves exposure to several potentially harmful chemicals. The authors carried out a systematic review and meta-analysis of cohort studies of workers in the rubber-producing industry. Data were obtained from computerized literature searches of several databases from their inception through December 2003. The reference lists of identified articles were inspected for further relevant articles. The authors conducted random-effects meta-analyses of log standardized mortality ratios (SMRs)/standardized incidence ratios. Heterogeneity between study results was explored through subgroup analyses and meta-regression on cohort demographic factors and study quality indicators. The authors identified 36 published articles reporting information on 31 different cohort groups. The meta-SMR was 0.86 (95% confidence interval (CI): 0.82, 0.91) for all-cause mortality (28 cohorts) and 0.94 (95% CI: 0.89, 1.01) for all malignant neoplasms (27 cohorts). Heterogeneity was observed for these endpoints and for the majority of disease-specific outcomes. Statistically significant excesses were observed for diabetes (meta-SMR = 1.36, 95% CI: 1.17, 1.59) (five cohorts) and leukemia (meta-SMR = 1.21, 95% CI: 1.03, 1.43) (16 cohorts), the latter particularly for persons working exclusively in nontire manufacturing (meta-SMR = 1.70, 95% CI: 1.14, 2.54) (four cohorts). Excesses highlighted in previous narrative reviews were not substantiated. Interpretation of these results is complicated by substantial unexplainable heterogeneity; small excesses in specific mortality outcomes may have been disguised by the healthy worker effect.
chemical industry; meta-analysis; mortality; neoplasms; occupational exposure; review [publication type]; rubber
Abbreviations: CI, confidence interval; SIR, standardized incidence ratio; SMR, standardized mortality ratio
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
Two narrative reviews of the many epidemiologic studies investigating the risk of cancer in the rubber industry have been carried out (1
, 2). These reviews highlighted excess risks of lung, bladder, and stomach cancers and leukemia. To our knowledge, there has been no formal systematic review or meta-analyses of mortality rates and cancer incidence specific to workers in the rubber industry. Thus, we set out to conduct such a review and meta-analysis.
The results reported here build on those of a previous meta-analysis by Greenberg et al. (3), who examined mortality and cancer incidence in cohort studies of workers in the entire chemical industry. Focusing on the synthetic rubber industry, we have updated and extended their findings. A supplementary document providing further details is available on the Journal's website (http://aje.oxfordjournals.org/). Figures and tables suffixed with an "S" can be found in that document.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
This systematic review and meta-analysis was performed in accordance with the MOOSE (Meta-analysis of Observational Studies in Epidemiology) guidelines (4
) for meta-analyses of observational studies. Results are reported here according to those guidelines.
Search strategy
In their meta-analysis, Greenberg et al. (3) used an existing American Chemistry Council database of papers published between 1966 and 1997 on cohort studies of health effects in chemical workers in the United States or Western Europe. We used the relevant references from this database to develop our search strategies (see appendix A of the supplementary material). We searched the following databases from their inception through December 2003: MEDLINE (US National Library of Medicine), ToxFile (Dialog, Cary, North Carolina), CancerLit (US National Cancer Institute), EMBASE (Excerpta Medica Database; Elsevier BV, Amsterdam, the Netherlands), CA SEARCH (Chemical Abstracts; Dialog), BIOSIS Previews (Ovid Technologies, Inc., New York, New York), SciSearch (Science Citation Index; Thomson Scientific, London, United Kingdom), PASCAL (French National Research Council, Vandoeuvre-lès-Nancy, France), and the NTIS database (US National Technical Information Service, Springfield, Virginia). Searches were conducted between January and May of 2004. We identified potentially relevant references by examining the titles and abstracts of all references obtained through electronic searching and procured these references for closer examination. We further checked that our searches included all relevant papers in the American Chemistry Council database, and we scrutinized the reference lists of identified papers for additional relevant publications.
The following inclusion criteria were used for this review.
- The study had to have had a cohort design including employees working in synthetic rubber manufacturing (this excluded workers exposed to chemicals who were not involved in rubber manufacturing).
- The article had to have reported original results in the form of standardized mortality ratios and/or standardized incidence ratios based on an external comparison group (or to have presented data allowing such outcomes to be derived).
- The article had to have been published in English between 1966 and 2003.
- The article had to have been published in either the peer-reviewed literature or the non-peer-reviewed public-domain literature, including Web-based publications.
): 1) the years searched were expanded to 2003; 2) the location of studies was expanded from North America or Western Europe to worldwide; and 3) the search included non-peer-reviewed literature in the public domain and "gray literature." We used the "best available" data from each cohort or subcohort related to each outcome. This usually meant using the most recently published reference for each cohort, but where data on a particular outcome for either the whole cohort or a subcohort were reported only in an earlier paper, this information was included.
Data extraction
For each cohort, we extracted data on as many mortality and cancer incidence outcomes as available (see figure 1 for mortality outcomes and figure 2 for incidence outcomes).
|
|
The causes of death and sites of cancer were coded according to the International Classification of Diseases, Ninth Revision; we recoded them if earlier versions of the International Classification of Diseases had been used. We extracted data on observed and expected numbers of cases for calculation of the standardized mortality ratio (SMR) and/or the standardized incidence ratio (SIR), together with associated standard errors for the loge(SMR) and/or the loge(SIR). In some instances, it was necessary to derive these standard errors from a reported 95 percent confidence interval, assuming that the 95 percent confidence interval for loge(SMR) or loge(SIR) is given by loge(SMR) or loge(SIR) ± 1.96/(
O), where O is the observed number of deaths/events (5). Data on the following cohort/study characteristics were also extracted where available: dates on which the study was carried out, inclusion and exclusion criteria, comparison population, percentage of the cohort that was male, average age of the cohort, average duration of employment, country and geographic area of the cohort, study sponsorship, author(s)' affiliation(s), study design, industry sector, chemicals produced and used, industry processes, and study quality.
Assessing study quality
We initially intended to use the Newcastle-Ottawa Scale (G. A. Wells et al., University of Ottawa (www.ohri.ca/programs/clinical_epidemiology/oxford.htm)) for assessing the quality of observational studies, but following a pilot process, we found this scale difficult to interpret for these occupational cohort studies. We therefore developed a modified version including five quality components: representativeness of the exposed cohort, exposure assessment/reporting, comparability of exposed and nonexposed cohorts, assessment of outcome, and adequacy of follow-up. (See appendix C of the supplement for a copy of this modified instrument.)
Methods for quantitative synthesis
Overall pooled estimates of the SMR and SIR, denoted "meta-SMR" and "meta-SIR," respectively, together with associated 95 percent confidence intervals, were obtained using random-effects meta-analysis (6). Tests for heterogeneity between study results were also performed, and associated p values are reported. For studies in which there were zero observed and/or expected events, 1 was added to both the observed and expected number of events so that an estimate of the loge(SMR)/loge(SIR) and its associated standard error could be obtained. Internal comparisons using various modeling techniques and focusing on various exposures and diseases were available for only a few cohorts; thus, no meta-analyses of these cohorts were carried out.
A subgroup of cohorts of workers involved exclusively in the manufacture of tires was identified as clearly definable a priori for separate subgroup analyses.
No formal correction for multiple comparisons was made; thus, results should be interpreted cautiously because of the number of outcomes examined. To limit such problems, we restricted the meta-regression analyses, described below, to all-cause mortality, all malignant neoplasms, and all outcomes with meta-SMRs or meta-SIRs that were significantly inflated at the 5 percent statistical significance level. We used subgroup analyses and meta-regression techniques (7) to attempt to explain any observed between-study heterogeneity. The covariates considered were geographic region, exposure level, percentage of the cohort that was White, percentage of the cohort that was male, percentage of the cohort traced, percentage of death certificates traced, midcohort year, length of follow-up, and total cohort size.
For sensitivity analyses, we analyzed key outcomes separately by categories of the assessed study quality variables to ascertain whether there were any relations with quality and outcome. Additionally, we assessed the influence of individual studies on the overall meta-SMR (or meta-SIR) by reestimating the overall effect after omitting each study in turn. We also assessed publication bias graphically by means of funnel plots and Egger's test (8).
All of the above methods were implemented in Stata 8.2 (Stata Corporation, College Station, Texas) using a combination of available macros (9) and a suite of specifically developed macros (available from the corresponding author upon request).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
Literature search
We identified 76 references that contained potentially relevant information for the meta-analysis: 49 from the original American Chemistry Council database and 27 from our electronic searches. Data on mortality were extracted for 28 separate cohorts from 33 separate papers, and data on cancer incidence were extracted for eight separate cohorts (five also having mortality data) from nine papers (six also having mortality data)—that is, 31 separate cohorts from 36 papers (M1–M36; see Appendix). Table 1 indicates which papers relate to each cohort grouping and summarizes the characteristics of the populations studied. (Each paper is described in detail in appendix B of the supplement, and earlier and related papers, including nested case-control studies, are also listed there.)
|
Cohort studies had been carried out in North America, Europe, and China during the time period 1974–2002. While some cohorts contained only males, a few included both males and, generally, a small proportion of females.
The industries investigated included the manufacture of tires, footwear, general rubber goods, and plastics. Some included only one factory or industrial site; others included a large number of factories manufacturing a wide range of goods. Information given on the chemical processes involved and the chemicals used and produced was patchy and inconsistent. Processes mentioned included curing, vulcanization, emulsion, and polymerization. Chemicals highlighted were generic groups such as accelerators, vulcanizing agents, and antioxidants and a wide range of specific chemicals, including benzene, styrene, formaldehyde, ammonia, carbon black, nitrosamines, ß-naphthylamine, and butadiene. Quantitative estimates of exposure were presented for only one cohort, and this exposure was to one substance, benzene (M23, M25).
The results of the study quality assessment are presented in table 2. Quality assessment indicated that: 1) the majority (n = 29) of the 36 papers had reported data on a representative exposed cohort (two unrepresentative, five nonspecified); 2) where exposure was described (22/36), the information had been ascertained from formal records; 3) where outcomes were described (30/36), they had been assessed using formal records (death certificates, etc.); 4) with regard to the comparability of the groups compared, 12 out of 36 studies had used standard matching/adjustment methods, while the majority (20/36) had not used such methods by the criteria defined in our instrument; and 5) follow-up was virtually complete (
5 percent of the cohort remained untraced) in the majority of studies (26/36) but was less adequate in a portion of them (6/36).
|
Quantitative synthesis
Figure 1 presents the meta-SMR estimate for each mortality outcome. (Outcome data from each specific study that contributed to each meta-SMR are available upon request.) The number of papers reporting mortality data varied greatly depending on the outcome considered. Not surprisingly, the most frequently reported outcomes were all causes of death (28 cohorts) and all malignant neoplasms (27 cohorts). Data on cancers of the digestive system, respiratory system, bronchus, trachea and lung, stomach, bladder, and the lymphatic and hematopoietic systems were all well reported (
20 cohorts). Data on other outcomes were reported much less frequently, with the lowest reporting rate being that for thyroid cancer (two cohorts). Detailed matrices describing the reporting patterns for mortality and cancer incidence are provided in tables S1 and S2. For all-cause mortality, there were 14 percent fewer deaths observed than expected (meta-SMR = 0.86, 95 percent confidence interval (CI): 0.82, 0.91). Mortality from all malignant neoplasms had a small but nonsignificant deficit (meta-SMR = 0.94, 95 percent CI: 0.89, 1.01). Heterogeneity was observed for the majority of outcomes. A deficit of mortality was found for a number of other outcomes: cancers of the buccal cavity and pharynx, digestive organs and peritoneum, esophagus, rectum, liver and biliary passages, pancreas, skin, breast, and brain and central nervous system, Hodgkin's disease, cardiovascular disease, coronary heart disease, nonmalignant respiratory disease, bronchitis, emphysema and asthma, cirrhosis, external causes, accidents, suicide, and homicide. An excess was observed for several outcomes; statistically significant excesses for found for leukemia (meta-SMR = 1.21, 95 percent CI: 1.03, 1.43) (16 cohorts) and diabetes mellitus (meta-SMR = 1.36, 95 percent CI: 1.17, 1.59) (five cohorts). Although the risk for diabetes was reasonably consistent across the five studies evaluated, caution should be exercised in interpretation because of the small number of studies reporting data for this outcome.
The precision of the meta-SIRs for the cancer incidence outcomes was lower than that for the corresponding mortality outcomes because of the smaller number of studies reporting data on them (figure 2). The meta-SIR for all malignant neoplasms was 0.94 (95 percent CI: 0.82, 1.07). Cancer of the pancreas was the only outcome with a significant excess cancer incidence (meta-SIR = 1.52, 95 percent CI: 1.17, 1.98) (six studies). The meta-SIR for leukemia was raised but, unlike that for mortality, not statistically significantly (meta-SIR = 1.16, 95 percent CI: 0.67, 2.03).
Industry subsector analysis
Figure 3 displays the meta-SMR mortality estimates for the exclusively tire-producing subsector. Results for several of the causes of death were only reported in one study. However, the meta-SMRs for causes reported in more than one study showed a pattern similar to that for the synthetic rubber industry as a whole. The only statistically significant excess mortality among tire producers was due to diabetes (meta-SMR = 1.46, 95 percent CI: 1.11, 1.92), which is consistent with the result for the whole industry.
|
Figure 4 presents forest plots showing SMRs from the individual studies for all causes, leukemia, and diabetes (parts a–c, respectively), by industry sector. An excess of mortality from diabetes was found among workers manufacturing both tires and other goods (figure 4, part c). For leukemia (part b), in the six (out of 16) cohorts of workers manufacturing only tires, little excess mortality was observed (meta-SMR = 1.03, 95 percent CI: 0.75, 1.41). The results for subgroups manufacturing both tires and other rubber goods (meta-SMR = 1.12, 95 percent CI: 0.93, 1.34) and exclusively manufacturing other rubber goods (meta-SMR = 1.21, 95 percent CI: 1.03, 1.42) indicated that the majority of the excess mortality was in these two study subgroups, particularly the latter.
|
There were, at most, two studies presenting incidence results for any one outcome in the tire subsector, and all results were inconclusive (see figure S1).
Exploring between-study heterogeneity
Meta-regression analysis was performed where heterogeneity was evident. There were eight or more studies that had data available for the appropriate regression using the covariates listed in "Methods." Multiple testing issues, problems of low power of meta-regression, known inflated type I error rates, and highly influential outlying studies influence the interpretation of these results. Results for total overall mortality and leukemia mortality (the only outcome with statistically significant excess mortality including a sufficient number of studies) are described below.
For all-cause mortality, positive and statistically significant associations were observed for proportion of the cohort that was male (p < 0.001), cohort size (p < 0.01), and midcohort year (p < 0.001). A negative significant association was observed for length of follow-up (p < 0.001). Regression plots of these associations are provided in figure S2. Since many of the studies were of all-male populations, interpretation of the regression on percentage of males was difficult: This is likely to have been heavily influenced by the Solionova et al. (M27) study, which was the only one that had a majority-female population. There was a tendency for the deficit in mortality to be inversely proportional to cohort size. Midcohort year had a more separated distribution of values, with a suggestion that the deficit had turned to excess mortality sometime after 1980.
There was a suggestion of an increased deficit of mortality for longer durations of follow-up; this might be an indication of a healthy worker survivor effect (10).
For leukemia mortality, a significant positive association was observed for proportion of Whites (p = 0.007) and a significant negative association was observed for total cohort size (p = 0.003); midcohort year (p = 0.08) had a marginally positive association. Graphs of these associations are shown in figure S3.
Sensitivity analysis
Influence of individual studies.
Investigation of the influence of individual studies (via systematic "leave one out" exclusion) was undertaken. The conclusions for all-cause mortality and all-cancer mortality were robust to the exclusion of any one study from the meta-analysis (figure S4). The magnitude of the excess mortality observed for leukemia in the meta-analysis varied, with the Rinsky et al. (M25) study having the largest excess and appearing to be the most influential in terms of reduction in the point estimate for the pooled SMR, but inferences remaining unchanged (the 95 percent confidence interval becomes narrower when this study is removed because the between-study heterogeneity is reduced when it is excluded) (figure 5). However, inferences did change at the 5 percent level when other studies were removed, which was expected given that the result was of borderline significance initially.
|
Study quality.
We attempted through subgroup analyses to explore whether study quality influenced outcome (table 2). Figure 6 presents the results, for total mortality, of excluding all studies but those with the highest quality rating for each quality component separately. Results were robust to the exclusion of those studies perceived to be of lower quality. For leukemia mortality, pooled SMRs still indicated elevated mortality rates for all quality components, but inferences were no longer statistically significant at the 5 percent level for the "exposure quality" and "comparability of groups" components (figures S5 and S6).
|
Assessing publication bias
Figure 7 presents funnel plots and results of Egger's corresponding asymmetry test for all-cause, all-cancer, leukemia, and diabetes mortality outcomes (parts a–d, respectively). Egger's test for all-cause mortality gave a p value of 0.15, which was low but not formally significant. However, few of the studies fall within the guidelines for a homogeneous funnel supplied on the graph, confirming visually the large degree of heterogeneity in estimates (figure 7, part a). There is perhaps a suggestion of a "tunnel" effect, suggesting that imprecise studies with estimates close to 1 are "missing" towards the center of the plot. The funnel plot for all-cancer mortality (part b), as for all causes, does not conform to a classic funnel shape. The associated Egger's test produced a p value of 0.1, which could be interpreted as some evidence of a selection mechanism. However, the trend is perhaps the opposite of what would be expected a priori, as the "missing" studies are those with SMRs greater than 1 rather than less than 1. The funnel plot for leukemia mortality (part c) appears to conform more closely to the expected shape (Egger's test: p = 0.96) compared with those for all-cause mortality and all-cancer mortality, but there are four studies that are much less precise than the rest, and these are all on the left-hand side of the center of the funnel. The results also appear more homogeneous, with all studies but the one by Rinsky et al. (M25) being contained within the fixed-effect funnel guidelines. For diabetes (part d), the small number of studies makes interpretation of the funnel plot difficult, and no formal test was performed.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
In this meta-analysis, we found a significant deficit of all causes of death and deficits of many other specific causes of death. Significant excesses were found for leukemia and diabetes. The results from the meta-analysis differ in some respects from the conclusions drawn in the two previously published narrative reviews of cancer incidence (1
, 2). Both reviews highlighted excess risks of lung, bladder, and stomach cancer among workers in the rubber industry. In this meta-analysis, we found no evidence to support such increases. The reviews also found strong evidence for excess risks of leukemia, and that finding is supported by the results of this meta-analysis. The review conducted by the International Agency for Research on Cancer (1) included mostly a subset of the studies we included that had been published by 1982; the second review, by Kogevinas et al. (2), also included a subset of the cohort studies that were included here but published beyond that date. Excess risks in various occupational subgroups and departments were highlighted, and the conclusions for diseases such as cancer of the bladder appear to have been derived from these, as well as from the large number of case-control studies and studies based on administrative data which were reviewed but not included in our meta-analysis. The reviewers drew attention to the limitations of these studies, particularly a potential lack of statistical power and imprecise and inaccurate exposure classification.
Many of the limitations inherent in the design, implementation, and interpretation of individual cohort studies in occupational settings are relevant to the studies included in this meta-analysis and thus should be considered when interpreting the results. Many of the individual studies found a healthy worker effect as a result of comparing a relatively healthy workforce (both selected into the industry and surviving in the industry) with the general population. The reduced mortality from all causes of death, cardiovascular disease, and nonmalignant respiratory disease in this meta-analysis is likely to be attributable, at least in part, to the healthy worker effect. It has been suggested that the healthy worker effect is of less consequence in the interpretation of cancer mortality (11).
The papers included in our meta-analysis were limited to those published in the English language and in the peer-reviewed literature between 1966 and 2003. It is therefore possible that some relevant studies were not included, which may have biased the results. For example, papers written in languages other than English may be more likely to be translated into English if they show significant results (12). Although few problems were flagged by our assessments of publication bias, because of the low power of such assessments, the possibility of publication bias, particularly outcome-reporting bias, cannot be ruled out, especially for outcomes for which we had only a small number of studies, such as diabetes. With the availability of analysis software that routinely produces results for a large number of causes of death, it is easy to see how selectivity of reporting could have occurred (13).
Sensitivity analyses exploring the influence of study quality proved inconclusive. This is often the case in meta-analysis because of 1) the complexity of disentangling multidimensional quality issues, 2) problems in assessing the quality of a study as opposed to the quality of reporting, and 3) the low discriminatory power of quite uniform quality components.
We carried out meta-regression to investigate the influence of several variables, some of which could be considered confounding variables or effect modifiers. However, in most of the individual studies, investigators did not collect information on the smoking habits of their study populations. Only two studies (M29, M36) presented results adjusted for smoking. Although smoking is a potential risk factor for leukemia (found in excess here), it is not thought to be a strong confounder for this disease; results have varied, with only a few studies finding up to a 50 percent increase in risk (14).
The exposure of the rubber workers is likely to have varied greatly both between individuals within each cohort (depending on their jobs and tasks) and between the workforces of the different cohorts. The lack of information on exposure makes it difficult to attribute any observed associations with mortality to specific exposures (as noted in previous reviews (1, 2)). In this meta-analysis, duration of employment was the only measure of exposure available in enough papers to produce meaningful results, and this is only a proxy measure of exposure. Internal analyses exploring dose-response relations were available for only a few cohorts, often in additional, related publications (see appendix B of the supplementary material). One cohort, the Pliofilm cohort, had quantitative estimates of exposure for benzene and leukemia only (M23, M25). This cohort has most often been used to calculate risk assessment for leukemia and exposure to benzene; several publications have given a variety of mortality results and risk assessments for this cohort (15–17). Leukemia was the only cause of death that showed a clear excess risk in the meta-analysis, but this result was clearly influenced by the inclusion of the Rinsky et al. study (M25), which had the highest SMR. Benzene exposure is associated most strongly in both mechanistic and human studies with increased risk of acute nonlymphocytic leukemia, but there were insufficient data in the studies included in the meta-analysis to carry out a separate analysis for this diagnostic subgroup.
The Pliofilm cohort has been used for assessing benzene exposure partly because this cohort was exposed to few other substances that may have been confounding factors. Several published analyses have attempted to evaluate simultaneous exposure to chemicals in the rubber industry, including benzene, styrene, and 1,3-butadiene, for which excesses of lymphohematopoietic cancers have been observed (18–20).
Both the Pliofilm and the butadiene-styrene cohorts have been followed up over a long period of time, and each has generated a large number of analyses and publications. This can present persons carrying out a meta-analysis with a potential problem regarding the choice of articles and results. Care is needed to determine whether the study populations in different series of papers are the same, and elucidation of the reasons for any differences is not always straightforward. In addition, the most recent, specific papers may report results for only a restricted number of causes or in different formats, so that earlier papers reporting results from an earlier follow-up may have to be used. In the meta-analysis presented here, the papers by Matanoski et al. (M16), Meinhardt et al. (M18), and Sathiakumar et al. (M26) provided the data for various causes and subcohorts later analyzed by Delzell et al. (18, 20, 21).
In summary, this meta-analysis shows that overall the mortality experience of workers in the rubber industry is similar to or better than that of the general population. The observed excess rate of death from diabetes may be cause for concern. We found an excess of leukemia in cohorts exclusively manufacturing rubber goods other than tires that was potentially associated with exposure to benzene and/or 1,3-butadiene. Other excesses highlighted in previous narrative reviews were not substantiated quantitatively here. A further systematic review of nested and population-based case-control studies and administrative cohort studies, together with possible meta-analyses including occupational and exposure subgroups, would be valuable.
|
APPENDIX |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
Papers Included in the Meta-Analysis
- M1. Andjelkovic DA, Taulbee J, Symons M. Mortality experience of a cohort of rubber workers, 1964–1973. J Occup Med 1976;18:387–94.
- M2. Bernardinelli L, De Marco R, Tinelli C. Cancer mortality in an Italian rubber factory. Br J Ind Med 1987;44:187–91.
- M3. Carlo GL, Jablinske MR, Lee NL, et al. Reduced mortality among workers at a rubber plant. J Occup Med 1993;35:611–16.
- M4. Chen J, Wei X, You X. A retrospective cohort study on digestive cancer in the rubber tire industry in Shanghai. J Occup Health 1997;39:302–12.
- M5. Delzell E, Louik C, Lewis RJ, et al. Mortality and cancer morbidity among workers in the rubber tire industry. Am J Ind Med 1981;2:209–16.
- M6. Delzell E, Monson RR. Mortality among rubber workers. III. Cause-specific mortality, 1940–1978. J Occup Med 1981;23:677–84.
- M7. Gustavsson A, Hogstedt C, Holmberg B. Mortality and incidence of cancer among Swedish rubber workers, 1952–1981. Scand J Work Environ Health 1986;12:538–44.
- M8. Holmberg B, Westerholm P, Maasing R, et al. Retrospective cohort study of two plants in the Swedish rubber industry. Scand J Work Environ Health 1983;9:59–68.
- M9. Holmes TM, Buffler PA, Holguin AH, et al. A mortality study of employees at a synthetic rubber manufacturing plant. Am J Ind Med 1986;9:355–62.
- M10. Ietri E, Belli S, Comba P, et al. Cohort mortality study of rubber and plastics product makers in Italy. Occup Med 1997;47:417–22.
- M11. Johnson CA. Mortality of workers in the styrene-butadiene rubber polymer manufacturing industry. In: International Institute of Synthetic Rubber Producers—proceedings of the 23rd annual meeting, New Orleans, LA, April 19–23, 1982. Vol 1. Houston, TX: International Institute of Synthetic Rubber Producers, 1982:1–27.
- M12. Kilpikari I. Mortality among male rubber workers in Finland. Arch Environ Health 1982;37:295–9.
- M13. Kilpikari I, Pukkala E, Lehtonen M, et al. Cancer incidence among Finnish rubber workers. Int Arch Occup Environ Health 1982;51:65–71.
- M14. Li K,Yu S. Mortality in a Chinese rubber factory: a prospective cohort study. J Occup Health 2002;44:76–82.
- M15. Matanoski GM, Santos-Burgoa C, Schwartz L. Mortality of a cohort of workers in the styrene-butadiene polymer manufacturing industry (1943–1982). Environ Health Perspect 1990;86:107–17.
- M16. Matanoski GM, Schwartz L. Mortality of workers in styrene-butadiene polymer production. J Occup Med 1987;29:675–80.
- M17. McMichael AJ, Spirtas R, Kupper LL. An epidemiologic study of mortality within a cohort of rubber workers, 1964–72. J Occup Med 1974;16:458–64.
- M18. Meinhardt TJ, Lemen RA, Crandall MS, et al. Environmental epidemiologic investigation of the styrene-butadiene rubber industry. Mortality patterns with discussion of the hematopoietic and lymphatic malignancies. Scand J Work Environ Health 1982;8:250–9.
- M19. Monson RR, Fine LJ. Cancer mortality and morbidity among rubber workers. J Natl Cancer Inst 1978;61:1047–53.
- M20. Monson RR, Nakano KK. Mortality among rubber workers. I. White male union employees in Akron, Ohio. Am J Epidemiol 1976;103:284–96.
- M21. Negri E, Piolatto G, Pira E, et al. Cancer mortality in a northern Italian cohort of rubber workers. Br J Ind Med 1989;46:624–8.
- M22. Norseth T, Andersen A, Giltvedt J. Cancer incidence in the rubber industry in Norway. Scand J Work Environ Health 1983;9:69–71.
- M23. Paxton MB. Leukemia risk associated with benzene exposure in the Pliofilm cohort. Environ Health Perspect 1996;104(suppl 6):1431–6.
- M24. Pell S. Mortality of workers exposed to chloroprene. J Occup Med 1978;20:21–9.
- M25. Rinsky RA, Hornung RW, Silver SR, et al. Benzene exposure and hematopoietic mortality: a long-term epidemiologic risk assessment. Am J Ind Med 2003;42:474–80.
- M26. Sathiakumar N, Delzell E, Hovinga M, et al. Mortality from cancer and other causes of death among synthetic rubber workers. Occup Environ Med 1998;55:230–5.
- M27. Solionova LG, Smulevich VB. Mortality and cancer incidence in a cohort of rubber workers in Moscow. Scand J Work Environ Health 1993;19:96–101.
- M28. Sorahan T, Parkes HG, Veys CA, et al. Mortality in the British rubber industry 1946–85. Br J Ind Med 1989;46:1–10.
- M29. Straughan JK, Sorahan T. Cohort mortality and cancer incidence survey of recent entrants (1982–91) to the United Kingdom rubber industry: preliminary findings. Occup Environ Med 2000;57:574–6.
- M30. Szeszenia-Dabrowska N, Wilczynska U, Kaczmarek T, et al. Cancer mortality among male workers in the Polish rubber industry. Pol J Occup Med 1991;4:149–57.
- M31. Veys CA. Bladder cancer in rubber workers. A phenyl beta-naphthylamine (PBNA) exposed workforce. Prog Rubber Plast Technol 1996;12:258–73.
- M32. Wang HW, You XJ, Qu YH, et al. Investigation of cancer epidemiology and study of carcinogenic agents in the Shanghai rubber industry. Cancer Res 1984;44:3101–4.
- M33. Waterhouse JAH. Current status of cancer risk in the rubber industry. In: Birch JM, ed. Epidemiology. (Advances in medical oncology, research, and education, vol 3). Oxford, United Kingdom: Pergamon Press, 1979:97–105.
- M34. Weiland SK, Mundt KA, Keil U, et al. Cancer mortality among workers in the German rubber industry: 1981–91. Occup Environ Med 1996;53:289–98.
- M35. Wilczynska U, Szadkowska-Stanczyk I, Szeszenia-Dabrowska N, et al. Cancer mortality in rubber tire workers in Poland. Int J Occup Med Environ Health 2001;14:115–25.
- M36. Zhang ZF, Yu SZ, Li WX, et al. Smoking, occupational exposure to rubber, and lung cancer. Br J Ind Med 1989;46:12–15.
- M2. Bernardinelli L, De Marco R, Tinelli C. Cancer mortality in an Italian rubber factory. Br J Ind Med 1987;44:187–91.
|
ACKNOWLEDGMENTS |
|---|
This work was funded by the American Chemistry Council (Arlington, Virginia).
The authors are grateful to the members of the advisory panel, who gave advice throughout the project, and to participants in the workshop held by the American Chemistry Council in Washington, DC, in January 2005 for their suggestions and input. The authors also thank the members of the MRC Institute for Environment and Health (Leicester, United Kingdom) and the University of Leicester Department of Health Sciences (Leicester, United Kingdom), including Dr. Len Levy, Alex Capleton, and Karen Bradley.
Conflict of interest: none declared.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONAPPENDIXReferences |
|---|
- International Agency for Research on Cancer. The rubber industry. (IARC monographs on the evaluation of the carcinogenic risk of chemicals to humans, vol 28). Lyon, France: International Agency for Research on Cancer, 1982.
- Kogevinas M, Sala M, Boffetta P, et al. Cancer risk in the rubber industry: a review of the recent epidemiological evidence. Occup Environ Med 1998;55:1–12.
[Abstract/Free Full Text] - Greenberg RS, Mandel J, Pastides H, et al. A meta-analysis of cohort studies describing mortality and cancer incidence among chemical workers in the United States and Western Europe. Epidemiology 2001;12:727–40.[CrossRef][Web of Science][Medline]
- Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis of Observational Studies in Epidemiology (MOOSE) Group. JAMA 2000;283:2008–12.
[Abstract/Free Full Text] - Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol 2. The design and analysis of cohort studies. (IARC scientific publication no. 82). Lyon, France: International Agency for Research on Cancer, 1987.
- Sutton AJ, Abrams KR, Jones DR, et al. Methods for meta-analysis in medical research. Chichester, United Kingdom: John Wiley and Sons Ltd, 2000.
- Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: a comparison of methods. Stat Med 1999;18:2693–708.[CrossRef][Web of Science][Medline]
- Egger M, Davey-Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. Br Med J 1997;315:629–34.
[Abstract/Free Full Text] - Sterne JAC, Egger M, Sutton AJ. Meta-analysis software. In: Egger M, Davey-Smith G, Altman DG, eds. Systematic reviews in health care: meta-analysis in context. London, United Kingdom: BMJ Books, 2001.
- Arrighi HM, Hertz-Picciotto I. The evolving concept of the healthy worker survivor effect. Epidemiology 1994;5:189–96.[Web of Science][Medline]
- Li CY, Sung FC. A review of the healthy worker effect in occupational epidemiology. Occup Med 1999;49:225–9.
[Abstract/Free Full Text] - Gregoire G, Derderian F, Lelorier J, et al. Selecting the language of the publications included in a meta-analysis—is there a Tower-of-Babel bias? J Clin Epidemiol 1995;48:159–63.[CrossRef][Web of Science][Medline]
- Collins JJ, Acquavella JF, Esmen NA. An updated meta-analysis of formaldehyde exposure and upper respiratory tract cancers. J Occup Environ Med 1997;39:639–51.[CrossRef][Web of Science][Medline]
- Linet MS, Cartwright RA. The leukemias. In: Schottenfeld D, Fraumeni JF, eds. Cancer epidemiology and prevention. 2nd ed. New York, NY: Oxford University Press, 1996:841–92.
- Crump KS. Risk of benzene-induced leukemia: a sensitivity analysis of the Pliofilm cohort with additional follow-up and new exposure estimates. J Toxicol Environ Health 1994;42:219–42.[Web of Science][Medline]
- Paustenbach DJ, Price PS, Ollison W, et al. Reevaluation of benzene exposure for the Pliofilm (rubberworker) cohort (1936–76). J Toxicol Environ Health 1992;36:177–231.[Web of Science][Medline]
- Schnatter AR, Nicholich MJ, Bird MG. Determination of leukemogenic benzene exposure concentrations: refined analysis of the Pliofilm cohort. Risk Anal 1996;16:833–40.[CrossRef][Web of Science][Medline]
- Delzell E, Sathiakumar N, Hovinga M, et al. A follow-up study of synthetic rubber workers. Toxicology 1996;113:182–9.[CrossRef][Web of Science][Medline]
- Macaluso M, Larson R, Delzell E, et al. Leukemia and cumulative exposure to butadiene, styrene and benzene among workers in the synthetic rubber industry. Toxicology 1996;113:190–202.[CrossRef][Web of Science][Medline]
- Delzell E. The chemical industry's impact on the health of its workers. (Editorial). Epidemiology 200;12:602–3.
- Delzell E, Macaluso M, Sathiakumar N, et al. Leukemia and exposure to 1,3-butadiene, styrene and dimethyldithiocarbanate among workers in the synthetic rubber industry. Chem Biol Interact 2001;135–136:515–34.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D. McLean, A. t Mannetje, E. Dryson, C. Walls, F. McKenzie, M. Maule, S. Cheng, C. Cunningham, H. Kromhout, P. Boffetta, et al. Leukaemia and occupation: a New Zealand Cancer Registry-based case-control Study Int. J. Epidemiol., April 1, 2009; 38(2): 594 - 606. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D R Jones, A J Sutton, K R Abrams, J Fenty, F Warren, and L Rushton Systematic review and meta-analysis of mortality in crop protection product manufacturing workers Occup. Environ. Med., January 1, 2009; 66(1): 7 - 15. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Monson RE: "META-ANALYSIS OF MORTALITY AND CANCER INCIDENCE AMONG WORKERS IN THE SYNTHETIC RUBBER-PRODUCING INDUSTRY" Am. J. Epidemiol., July 15, 2007; 166(2): 236 - 236. [Full Text] [PDF]
|
|
|
|
|
|
N Alder, L Rushton, A. Sutton, and D. Jones FOUR AUTHORS REPLY Am. J. Epidemiol., July 15, 2007; 166(2): 236 - 236. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||