Document keQByRKX4vOo6ypQZnMjdmdb

Int. J. Cancer: 87, 874 879 (2000) 2000 Wiley-Liss, Inc. Publication of the International Union Against Cancer RECTAL CANCER AND OCCUPATIONAL RISK FACTORS: A HYPOTHESIS-GENERATING, EXPOSURE-BASED CASE-CONTROL STUDY Simone DUMAS1,2, Marie-E lise PARENT1*, Jack SIEMIATYCKI1,3 and Jacques BRISSON2 1INRS-Institut Armand-Frappier, Laval, Quebec, Canada 2Universite Laval, Quebec, Canada 3McGill University, Montreal, Quebec, Canada In 1979, a hypothesis-generating, population-based casecontrol study was undertaken in Montreal, Canada, to explore the association between occupational exposure to 294 substances, 130 occupations and industries, and various cancers. Interviews were carried out with 3,630 histologically confirmed cancer cases, of whom 257 had rectal cancer, and with 533 population controls, to obtain detailed job history and data on potential confounders. The job history of each subject was evaluated by a team of chemists and hygienists and translated into occupational exposures. Logistic regression analyses adjusted for age, education, cigarette smoking, beer consumption, body mass index, and respondent status were performed using population controls and cancer controls, e.g., 1,295 subjects with cancers at sites other than the rectum, lung, colon, rectosigmoid junction, small intestine, and peritoneum. We present here the results based on cancer controls. The following substances showed some association with rectal cancer: rubber dust, rubber pyrolysis products, cotton dust, wool fibers, rayon fibers, a group of solvents (carbon tetrachloride, methylene chloride, trichloroethylene, acetone, aliphatic ketones, aliphatic esters, toluene, styrene), polychloroprene, glass fibers, formaldehyde, extenders, and ionizing radiation. The independent effect of many of these substances could not be disentangled as many were highly correlated with each other. Int. J. Cancer 87: 874 879, 2000. 2000 Wiley-Liss, Inc. included in all years of the study; rectal cancer was included in 5 of the 7 years. Case and control series There were 257 cases successfully interviewed out of the 304 eligible primary rectal cancer cases (84.5% response rate). Two groups of controls were available. The first group (referred to as cancer controls) included some of the patients with a cancer other than cancer of the rectum. Of the total pool of subjects, we excluded 3 groups: patients interviewed in years in which rectal cancer was not ascertained and interviewed, lung cancer cases, and cases with cancer of other intestinal sites (colon, rectosigmoid junction, small intestine, peritoneum). The remaining 1,295 cancer cases, constituting the group of cancer controls retained for analysis, were distributed by site as follows: esophagus (n 65), stomach (n 168), liver (n 25), gallbladder (n 20), pancreas (n 75), pleura (n 7), prostate (n 227), testis (n 19), penis (n 7), bladder (n 277), kidney (n 112), melanoma of skin (n 77), non-Hodgkin's lymphoma (n 149), Hodgkin's lymphoma (n 37), and other (n 30). Cases and controls were drawn from the 19 major hospitals of the Montreal area, thereby assuring a good representation of the base population. A second control group consisted of 533 successfully interviewed (72% response rate), age-stratified, population-based controls, selected using electoral lists and random-digit dialing. Although dietary and alcohol-consumption habits have been associated with rectal cancer risk (World Cancer Research Fund, 1997), these factors are probably not sufficient to explain the incidence gap between developed and developing countries as well as the systematically lower risk among women than among men (IARC, 1982, 1987, 1992, IARC, 1997). There must therefore be other, as yet unidentified risk factors for rectal cancer, and it is plausible that some of these are environmental/occupational risk factors. Nevertheless, the evidence on the role of occupational agents in rectal cancer etiology remains scant or non-existent. In the 1980s, a large hypothesis-generating, population-based case-control study was undertaken in Montreal, Canada, to search for evidence for associations between several types of cancer and hundreds of occupational exposures. Starting from a list of 76 industries, 54 occupations, and 294 substances (Siemiatycki, 1991), the present report describes in-depth analyses of the association between rectal cancer and exposure to the subset of substances, occupations, and industries which have been reported as possible risk factors in previous studies or which appear to be related to rectal cancer. Exposure assessment The interview questionnaire was in 2 parts: a structured section requesting information on important cancer risk factors, in particular lifestyle characteristics, and a semi-structured section designed to obtain a detailed description of each job the subject had held in his working lifetime (Gerin et al., 1985; Gerin and Siemiatycki, 1991). For each job in each subject's history, questions were asked about the company, its products, the nature of the work site, the subject's main and subsidiary tasks, and any additional information (e.g., equipment maintenance, use of protective equipment, activities of co-workers) that could furnish clues about possible exposures. A team of chemists and industrial hygienists examined each completed questionnaire and translated each job into a list of potential exposures by means of a checklist that included 294 substances (Siemiatycki et al., 1987; Siemiatycki, 1991). This type of retrospective exposure assessment had not been done previously in the context of a community-based study, and it took several years to develop a satisfactory methodology. The final codes given to a file were based on consensus among the chemists and indus- MATERIAL AND METHODS The design and data collection methods have been described elsewhere (Gerin et al., 1985; Siemiatycki, 1991; Siemiatycki et al., 1987; Fritschi and Siemiatycki, 1996). In brief, subjects were men, aged 35 to 70 years, residing in Montreal, who were diagnosed with a new, histologically confirmed cancer at 1 of 19 anatomical sites. Between 1979 and 1985, there were 4,576 eligible cancer patients, and 3,730 of these (82%) were successfully interviewed. Eighty-one percent of subjects responded for themselves; proxies provided information for the rest. Not all sites were Grant sponsors: Health Canada; National Cancer Institute of Canada; Institut de recherche en sante et securite au travail du Quebec; Fonds de la recherche en sante du Quebec; Medical Research Council of Canada. *Correspondence to: Marie-E lise Parent, INRS-Institut Armand-Frappier, 531 Boul. des Prairies, Laval, Quebec, Canada H7V 1B7. Fax: 450-686-5599. E-mail: marie-elise.parent@inrs-iaf.uquebec.ca Received 13 December 1999; Revised 24 March 2000; Accepted 27 March 2000 OCCUPATIONAL EXPOSURES AND RECTAL CANCER 875 trial hygienists. Chemical coding was carried out completely blind with regard to the subject's disease status. For each product thought to be present in each job, the coders noted 3 dimensions of information, each on a 3-point scale: degree of confidence that the exposure had actually occurred (possible, probable, definite), frequency of exposure in a normal working week (5%, 530%, 30% of the time), and concentration to which the worker was exposed (low, medium, or high). Given the retrospective nature of the coding, we did not consider it possible to attribute more precise quantitative values to the exposure levels. Non-exposure was not interpreted as absolute 0; rather, it was considered as exposure up to the level that can be found in the general environment and to which all subjects may be exposed. Occupational circumstances retained for analysis The present in-depth analysis focused on occupational variables reported to be potential risk factors as well as on a subset of occupational circumstances earmarked from our database. Those retained on the basis of the database were selected by screening analyses looking at rectal cancer in relation to all 54 occupations, 76 industry titles, and 294 occupational substances (Siemiatycki, 1991). We retained those occupational variables which showed, using either population or cancer controls, some evidence of being associated with rectal cancer, e.g., the odds ratio (OR) was at least 1.3 and the level of significance was 0.15 when adjusting for 6 literature-based (a priori) and data-based (significantly associated with rectal cancer) non-occupational potential confounders: age in years (3 levels), number of years of schooling (2 levels), respondent status (self vs. proxy), cigarette smoking as the average number of cigarettes smoked per day times the number of years smoked (3 levels), beer drinking as the average number of beers drank per day times the number of years of drinking (3 levels), body mass index calculated as weight (kg)/height (m)2 (2 levels). We further excluded from analysis any occupational variable that had fewer than 5 subjects in any cell of the contingency (2 2) tables based on any exposure level or duration. As a result of these selection criteria, we retained for the present analysis 33 substances, 12 industries, and 12 occupations. Statistical analyses Exposure to each substance was analyzed first as a dichotomy of any exposure vs. no exposure and then as a dichotomy of substantial vs. no exposure. For any given substance, the unexposed group comprised subjects never exposed to that substance plus a small number who were considered possibly exposed or whose exposure occurred only in the 5 years preceding the interview. The substantially exposed group comprised those who the chemists were confident had been exposed (probable or definite exposure) and who had more than 5 years of exposure at medium or high frequency and concentration. Occupation and industry titles were analyzed first as a dichotomy of ever having been employed in the occupation/industry of interest vs. not employed and then as a dichotomy of employed for 10 years or more vs. not employed. The 5 years preceding the interview were not counted in the duration of exposure. Occupational circumstances were first included one at a time in the unconditional logistic regression model along with the 6 potential non-occupational confounders. Substances showing evidence of an association with rectal cancer were then examined using regression models that took into account the presence of other occupational substances. Two parallel sets of analyses were carried out using the 533 population controls and the 1,295 selected cancer controls. To simplify the presentation, we mainly show results based on cancer controls. This set of results was chosen because it constitutes a larger set than that of the population controls and, thus, provides more stable estimates; the response rate for rectal cancer cases was closer to that of cancer controls than to that of population controls, and we surmise that there is a greater opportunity for information bias when using population controls than when using cancer con- trols. Nevertheless, for substances showing the strongest evidence of an association with rectal cancer based on cancer controls, we also present risk estimates based on population controls. RESULTS Table I shows the distribution of selected non-occupational characteristics among rectal cancer cases, cancer controls, and population controls. There were no statistically significant differences for these characteristics between cases and either of the control groups. Population controls tended to be slightly older, to have smoked less, to have consumed less beer, and to have had fewer proxies respond for them than cancer controls. Substances Table II presents ORs between rectal cancer incidence and exposure to each of the 33 selected substances at any level of exposure and at the substantial level, using cancer controls. ORs are based on statistical models treating each substance independently and adjusted for the 6 non-occupational potential confounders only. Among the various substances selected on the basis of previous publications, only rubber dust and rubber pyrolysis products showed any association with rectal cancer. Eighteen of the other substances selected for analysis had statistically elevated (or borderline significant) ORs in either the "ever-exposed" or the "substantially exposed" group. Exposures to cotton dust, wool fibers, and rayon fibers were associated with elevated risks. All of the solvents under analysis, i.e., mononuclear aromatic hydrocarbons (MAHs), toluene, xylene, aliphatic ketones, carbon tetrachloride, aliphatic esters, methylene chloride, trichloroethylene, acetone, and styrene, also showed some excess risks. Other substances associated with elevated risks included formaldehyde, extenders, glass fibers, polychloroprene, and ionizing radiation. For all substances which had reasonable numbers exposed at the substantial level, ORs were higher among subjects substantially exposed than among subjects ever exposed. Seven of the substances showing some excess risk in Table II had more than 20 exposed cases, and these were subjected to more in-depth analyses with respect to frequency, concentration, and duration of exposure. The dose-response relationships suggested in Table II for cotton dust and wool fibers were primarily due to increasing risk with increasing years of exposure. For aliphatic ketones, the increasing risk with increasing exposure level observed in Table II reflected dose-response trends in all 3 dimensions of exposure. The apparent dose-response relationships seen for MAHs and their constituents, toluene and xylene, appeared to be due primarily to increasing risk with increasing frequency and concentration of exposure. For formaldehyde, the overall dose- TABLE I SELECTED CHARACTERISTICS OF RECTAL CANCER CASES AND CONTROLS Characteristics Rectal cancer Cancer Population cases controls controls (n 257) (n 1,295) (n 533) P Mean age (years) Mean number of years of schooling Proxy respondent (%) Mean cigarette-years smoked (cig/day years) Mean beer-years consumed (beer/ day years) Mean body mass index (kg/m2) 58.7 9.4 15.2 801 47 25.4 58.4 9.8 19.7 893 59.6 0.05 10.1 0.05 12.6 0.01 802 0.05 60 33 0.01 25.4 25.6 0.05 1P value for test of the null hypothesis of equality of the 3 groups on this parameter. 876 DUMAS ET AL. TABLE II OR AND 95% CI FOR THE RELATION BETWEEN RECTAL CANCER AND SELECTED SUBSTANCES, USING CANCER CONTROLS Substances Exposure level Any N1 OR2 95% CI Substantial N OR 95% CI Selected from the literature Metal Metallic dust 69 0.9 0.71.2 30 0.8 0.51.3 Iron compounds 63 0.9 0.71.3 23 0.7 0.41.1 Mild steel dust 46 1.0 0.71.5 17 0.7 0.41.2 Stainless steel dust 10 0.8 0.41.7 2 0.4 0.11.5 Iron dust 9 0.7 0.41.5 5 0.9 0.42.5 Rubber Rubber dust 12 1.7 0.93.4 2 1.8 0.49.0 Rubber pyrolysis products 10 3.0 1.46.5 1 1.7 0.217.3 Combustion products Soot 17 0.8 0.51.3 4 0.8 0.32.4 Coal 8 0.6 0.31.3 6 1.3 0.53.3 Wood 9 0.7 0.31.4 4 1.0 0.33.2 Others Asbestos (chrysotile) 30 0.7 0.51.0 3 0.5 0.21.6 Asbestos (amphiboles) 11 0.7 0.31.2 2 1.5 0.37.6 Ink 12 1.3 0.72.6 6 1.2 0.52.9 Added based on screening analyses Textile Fabric dust 28 1.0 0.71.6 19 1.3 0.82.2 Cotton dust 25 1.0 0.61.6 18 1.5 0.92.6 Wool fibers 26 1.6 1.02.6 19 1.9 1.13.3 Rayon fibers 10 1.6 0.83.2 7 2.6 1.06.7 Solvents MAHs3 101 1.2 0.91.6 34 1.6 1.02.5 Toluene 50 1.4 1.02.0 17 1.7 1.03.0 Xylene 39 1.3 0.91.9 7 2.9 1.17.3 Aliphatic ketones 21 1.4 0.82.3 15 2.4 1.34.6 Carbon tetrachloride 16 2.0 1.13.5 2 0.5 0.12.0 Aliphatic esters 12 1.4 0.72.6 10 3.0 1.46.8 Methylene chloride 7 1.2 0.52.8 5 3.8 1.112.9 Trichloroethylene 12 2.0 1.03.9 3 0.9 0.33.2 Acetone 11 2.3 1.14.7 8 4.8 1.813.0 Styrene 6 1.7 0.74.5 5 3.9 1.212.9 Others Formaldehyde 36 1.2 0.81.9 13 2.4 1.24.7 Extenders 20 1.4 0.82.3 8 3.6 1.49.2 Glass fibers 14 0.9 0.51.6 8 4.3 1.711.3 Hypochlorites 17 1.0 0.61.7 12 1.3 0.72.6 Polychloroprene 14 2.0 1.13.8 2 7.1 1.051.8 Ionizing radiation 6 3.8 1.311.4 0-- -- 1N, number of exposed cases.2OR adjusted for age, education, respondent status, cigarette smoking, beer consumption, and body mass index.3MAHs include all aromatic compounds that have only 1 benzene ring, including substituted products such as xylene, toluene, styrene, phenol, and ethyl benzene. response pattern appeared to reflect increases by concentration and by duration. The results in Table II ignore the possible confounding among occupational risk factors. Since there are no well-established occupational risk factors for rectal cancer, it is not clear that any other occupational variables should be included in any model for a given substance. However, if 1 or more of these substances is in fact a risk factor for rectal cancer, then the correlated exposures could lead to confounding and misleading inferences. In such a case, it might be appropriate to include the occupational risk factor in all models. Since the carcinogenicity of these substances is not established, the results in Table II, unadjusted for occupational variables, represent our best estimates of the relative risks for each substance. Still, it is of some interest to include multiple occupational variables in the same model, to see which variables seem the most susceptible to confounding by other occupational variables. To accomplish this, we selected a subset of variables in Table II, namely, those with a lower confidence limit of the OR 1.0 at the "any exposure" level. The 8 variables which satisfied this criteria were included, together with the standard 6 non-occupational co-variates, in a single model. This was done only at the "any exposure" level, and the results are shown in Table III. Among this set of variables, those that seem least affected by the presence of others in the model are rubber pyrolysis products, wool fibers, polychloroprene, and ionizing radiation. Most substances showing excess risk based on cancer controls also showed elevated risk based on population controls, but the precision in risk estimates was lower using the latter group of referents. Risk estimates based on population controls for those 8 substances with a lower confidence limit of the OR 1.0 at the "any exposure" level based on cancer controls are shown in Table IV. Occupations and industries We investigated the effect of having been employed in 24 selected occupations or industries. Occupations retained for analysis included metal processors, metal machinists, sheet-metal workers, welders and flame cutters, general machinists, metal products fabricators, service station attendants, printers, masons and tile setters, leather workers, performing and graphic artists, and police and firefighters, whereas industries included iron, steel mills and foundries, metal fabricating and machining, rubber prod- OCCUPATIONAL EXPOSURES AND RECTAL CANCER 877 TABLE III OR AND 95% CI FOR THE RELATION BETWEEN EVER EXPOSURE TO SELECTED SUBSTANCES AND RECTAL CANCER, USING 2 STATISTICAL MODELS, BASED ON CANCER CONTROLS Substances Model 1 Adjusted for nonoccupational factors N1 OR12 Model 2 Adjusted for non-occupational and occupational factors OR23 95% CI Rubber pyrolysis products Wool fibers Carbon tetrachloride Trichloroethylene Acetone Toluene Polychloroprene Ionizing radiation 10 26 16 12 11 50 14 6 3.0 1.6 2.0 2.0 2.3 1.4 2.0 3.8 2.6 1.25.9 1.6 1.02.6 1.4 0.72.7 1.6 0.83.5 1.6 0.73.6 1.1 0.81.7 1.8 0.93.7 3.4 1.110.5 1N, number of exposed cases.2Adjusted for age, education, respondent status, cigarette smoking, beer consumption, and body mass index.3Adjusted for the non-occupational variables listed above as well as for all of the other substances in this table. ucts, petroleum and coal products, mineral fuels, printing and publishing, pulp and paper mills, sawmills, clothing, leather goods, defense services, and household furniture. Results based on occupations and industries were less directly informative about the carcinogenicity of the occupational substances per se, but they were more comparable to results from other studies based on job titles. The number of exposed cases was often very low, resulting in imprecise OR estimates. None of the occupations exhibited very strong and persuasive associations with rectal cancer. There were suggested associations for prolonged employment among leather workers [OR10 years 1.6, 95% confidence interval (CI) 0.9 2.8, 16 exposed cases] and performing and graphic artists (OR10 years 2.3, 95% CI 0.9 6.1, 6 exposed cases). Four industries exhibited some elevated risks: household furniture (ORever 2.7, 95% CI 1.35.8, 11 exposed cases; OR10 years 4.8, 95% CI 1.6 14.0, 7 exposed cases), rubber products (ORever 3.6, 95% CI 1.111.7, 5 exposed cases), iron, steel mills and foundries (ORever 1.8, 95% CI 0.9 3.7, 12 exposed cases), and sawmills (ORever 2.7, 95% CI 0.9 8.1, 5 exposed cases). DISCUSSION Our study was based on a reliable, case-by-case, expert-performed, detailed exposure assessment (Siemiatycki et al., 1997) regarded as the best approach for population-based retrospective studies (Bouyer and Hemon, 1993). In addition, we studied histologically confirmed incident cases and obtained information on several potential confounders. Nevertheless, these had a relatively modest impact on risk estimates. The choice of control group is often problematic as different methods have advantages and disadvantages (Hennekens and Buring, 1987). We had access to population and cancer controls, and both led to the same conclusions. Although this study benefited from an expert's translation of lifetime occupational history into specific exposures, it is still subject to exposure misclassification. The misclassification was non-differential, leading to attenuation of OR estimates (Hennekens and Buring, 1987). Another limitation of our study was the limited statistical power due to the low exposure prevalence for some substances and a modest number of study subjects. Still, notwithstanding these limitations, this is the largest case-control study of rectal cancer with relatively high-quality exposure assessment. We included several non-occupational variables as co-variates, but we had very limited data on dietary variables, some of which are likely risk factors for rectal cancer (World Cancer Research Fund, 1997). In the last 3 years of field work, our questionnaire included a brief checklist of foods rich in -carotene. We used this information to derive a -carotene index and analyzed it in relation to rectal cancer. There was no association in our data, and the -carotene index was not retained as a confounder. In future work attempting to replicate our findings, it would be desirable to collect data on dietary variables. Nevertheless, the associations between various dietary habits and rectal cancer, on the one hand, and occupational exposures, on the other, are unlikely to be strong enough to greatly distort the association between occupational exposure and rectal cancer. One dietary co-variate retained in our analyses was beer consumption, which has shown some weak association with rectal cancer risk (IARC, 1988; Stemmermann et al., 1990; Dean et al., 1979; Kabat et al., 1986; Kato et al., 1990; Pickle et al., 1984). We have examined the relation between rectal cancer and 54 occupations, 76 industries, and 294 substances. For the present in-depth analysis, we examined a subset of pertinent occupational circumstances because of suggestive evidence from other research or from our preliminary screening analyses. It can therefore be inferred that for all those substances, occupations, and industries not listed here, our evidence was either negative or insufficiently precise to warrant presentation. A full list of the substances, occupations, and industries examined in our screening analyses has been published (Siemiatycki, 1991). A full assessment of causality for each of the substances examined would also require consideration of experimental evidence of carcinogenesis, and physiologicalmetabolic evidence concerning the fate of these substances in the body. Examination of these lines of evidence is beyond the scope of this report. We found higher risks of rectal cancer among workers exposed to rubber dust and rubber pyrolysis products and in the rubber industry. Excess risks of rectal cancer have been reported among rubber workers in Massachusetts (Dubrow and Wegman, 1984) but not in the German and British rubber industries (Sorahan et al., 1986; Weiland et al., 1996). We also noted elevated ORs among leather workers, who are exposed to rubber dust through rubber sole-buffing activities. There were indications of excess risk for exposure to cotton dust, wool fibers, and rayon fibers. These substances were strongly correlated with one another, therefore preventing mutual adjustments. A study in the Nordic countries found higher incidence rates of rectal cancer among textile workers (Andersen et al., 1999). We also noted significantly raised risks in the household furniture industry, where workers are exposed to fabric dusts through upholstery activities. All 10 solvents subjected to the present in-depth analyses exhibited excess risk of rectal cancer to various degrees. However, in several instances, their effects could not be disentangled due to strong intercorrelations. As a result, mutual confounding is likely to occur among various subsets. One previous study indicated an elevation in risk for solvents as a class (Berlin et al., 1995), but others found no such association (Gerhardsson de Verdier et al., 1992; Peters et al., 1989). Our data are compatible with a slight increase in risk associated with toluene exposure, though adjustment for other occupational factors suggests that the observed risk estimate might be subject to confounding by 1 or more of the occupational exposures. The excess risk observed among styreneexposed workers corroborates results from a Finnish cohort study (Anttila et al., 1998). Many workers potentially exposed to MAHs have been found to be at excess risk of rectal cancer, including Canadian metal mill workers (Gallagher and Threlfall, 1983), US stationary engineers and firefighters (Decoufle et al., 1977), and service station attendants (Gerhardsson de Verdier et al., 1992). Excess risks of rectal cancer have been reported among workers exposed to inks (Gerhardsson de Verdier et al., 1992), as well as among printers (Andersen et al., 1999; Minder and Beer-Porizek, 1992). In our study, there was some weak evidence of higher risks among workers exposed to inks (ORsubst 2.3, 95% CI 0.52.9, 878 DUMAS ET AL. TABLE IV OR AND 95% CI FOR THE RELATION BETWEEN EXPOSURE TO SELECTED SUBSTANCES AND RECTAL CANCER, BASED ON POPULATION CONTROLS Substances Exposure level Any N1 OR2 95% CI N Substantial OR 95% CI Rubber pyrolysis products 10 2.1 0.85.5 1 2.2 0.136.4 Wool fibers 26 2.2 1.23.9 19 3.9 1.88.4 Carbon tetrachloride 16 1.5 0.82.9 2 0.7 0.13.3 Trichloroethylene 12 2.4 1.05.6 3 1.1 0.34.6 Acetone 11 1.6 0.73.4 8 2.0 0.85.3 Toluene 50 1.5 1.02.3 17 1.6 0.83.0 Polychloroprene 14 1.6 0.83.3 2 343.4 0.0 Ionizing radiation 6 2.3 0.77.4 0 -- -- 1N, number of exposed cases.2Adjusted for age, education, respondent status, cigarette smoking, beer consumption, and body mass index. 6 exposed cases) but only when population controls were used. We found moderately elevated ORs for performing and graphic artists. This supports results from other studies on professional artists (Miller and Blair, 1985), painters, potters, musicians, actors (Minder and Beer-Porizek, 1992), and artistic workers (Andersen et al., 1999). Our observation of raised risks among workers in the pulp and paper mills and sawmills concurs with that from the Swedish paper industry (Arbman et al., 1993). Working with leather was associated with cancer of the rectum in our data. Excess risks have been observed previously for shoe/ leather workers (Gerhardsson de Verdier et al., 1992). Polychloroprene, when combined with toluene, produces contact cement, which is used by shoemakers. Our results are consistent with a significant 2-fold excess in risk among workers exposed to polychloroprene. An elevation of rectal cancer risk among subjects exposed to combustion gases from coke, coal, and wood has been reported in Sweden (Gerhardsson de Verdier et al., 1992), but our data did not replicate this finding. A review of metal-working fluids concluded that straight oil might be associated with an increased risk of rectal cancer (Calvert et al., 1998). There was no excess risk associated with exposure to cutting fluids in our database nor was there an excess risk of rectal cancer in relation to any of the metallic dusts under analysis, in contrast with a previous report indicating excess risk in relation to non-ferrous metals and metal dust or fumes (Peters et al., 1989). There have been reports of excess rectal cancer in the metal industry (Arbman et al., 1993; Gallagher and Threlfall, 1983) and related occupations (Dubrow and Wegman, 1984; Jakobsson et al., 1997; Minder and Beer-Porizek, 1992), but our findings, based on different types of metal worker, do not support those results. One exception was the iron, steel, and foundry industry, for which there was some evidence of excess risk in our data. Higher risks of rectal cancer were associated with substantial exposure to formaldehyde. This gas is mainly used for plastic and resin manufacture. Evidence of an increased risk of cancer of the rectum was observed in a group of workers in a plastic-producing plant (Marsh, 1983). We found increased risks among workers exposed to extenders, a group of multipurpose additives often found in the same work environment as styrene. No other evidence concerning that association was found in the literature. Lastly, glass fibers are used mainly in insulation, plastic reinforcing material, and special textiles. An excess risk of rectal cancer was associated with substantial exposure to glass fibers in our data set. Exposures to glass fibers and styrene were moderately correlated. In summary, this hypothesis-generating study earmarked a number of possible occupational risk factors for rectal cancer. However, the available evidence remains too limited to justify firm conclusions, especially since, for most of these occupational circumstances, there was no or very little corresponding literature with which to compare our results. ACKNOWLEDGEMENTS The field work was supervised by Ms. L. Richardson, and the chemical coding was carried out by Drs. M. Gerin and L. Nadon, Mr. D. Begin, and Mr. R. Lakhani. REFERENCES ANDERSEN, A., BARLOW, L., ENGELAND, A., KJAERHEIM, K., LYNGE, E. and PUKKALA, E., Work-related cancer in the Nordic countries. Scand. J. Work Environ. Hlth., 25 (Suppl 2), 1114 (1999). ANTTILA, A., PUKKALA, E., RIALA, R., SALLMEN, M. and HEMMINKI, K., Cancer incidence among Finnish workers exposed to aromatic hydrocarbons. Int. Arch. occup. environ. Hlth., 71, 187193 (1998). ARBMAN, G., AXELSON, O., FREDRIKSSON, M., NILSSON, E. and SJODAHL, R., Do occupational factors influence the risk of colon and rectal cancer in different ways? Cancer, 72, 25432549 (1993). BERLIN, K., EDLING, C., PERSSON, B., AHLBORG, G., HILLERT, L., HOGSTED, B., LUNDBERG, I., SVENSSON, B., THIRINGER, G. and ORBAEK, P., Cancer incidence and mortality of patients with suspected solvent-related disorders. Scand. J. Work Environ. Hlth., 21, 362367 (1995). BOUYER, J. and HEMON, D., Retrospective evaluation of occupational exposures in population-based case-control studies: general overview with special attention to job exposure matrices. Int. J. Epidemiol., 22, S57S64 (1993). CALVERT, G., WARD, E., SCHNORR, T. and FINE, L., Cancer risks among workers exposed to metalworking fluids: a systematic review. Am. J. indust. Med., 33, 282292 (1998). DEAN, G., MACLENNAN, R., MCLOUGHLIN, H. and SHELLEY, E., Cause of death of blue-collar workers at a Dublin brewery, 1954 1973. Brit. J. Cancer, 40, 581589 (1979). DECOUFLE, P., LLOYD, W. and SALVIN, L., Mortality by cause among stationary engineers and stationary firemen. J. occup. Med., 19, 679 682 (1977). DUBROW, R. and WEGMAN, D., Cancer and occupation in Massachusetts. Amer. J. indust. Med., 6, 207230 (1984). FRITSCHI, L. and SIEMIATYCKI, J., Lymphoma, myeloma and occupation: results of a case-control study. Int. J. Cancer, 67, 498 503 (1996). GALLAGHER, R. and THRELFALL, W., Cancer mortality in metal workers. Can. med. Assoc. J., 129, 11911194 (1983). GERHARDSSON DE VERDIER, M., PLATO, N., STEINECK, G. and PETERS, J., Occupational exposures and cancer of the colon. Amer. J. indust. Med., 22, 291303 (1992). GERIN, M. and SIEMIATYCKI, J., The occupational questionnaire in retrospective epidemiologic studies: recent approaches in community-based studies. Appl. occup. environ. Hyg., 6, 495501 (1991). GERIN, M., SIEMIATYCKI, J., KEMPER, H. and BEGIN, D., Obtaining occupational exposure histories in epidemiologic case-control studies. J. occup. Med., 27, 420 426 (1985). OCCUPATIONAL EXPOSURES AND RECTAL CANCER 879 HENNEKENS, C. and BURING, J., Epidemiology in medicine, 1st ed., p. 383, Little, Brown and Company, Boston (1987). IARC, Cancer incidence in five continents, IARC Sci. Publ. 42, IARC, Lyon (1982). IARC, Cancer incidence in five continents, IARC Sci. Publ. 88, IARC, Lyon (1987). IARC, Cancer incidence in five continents, IARC Sci. Publ. 120, IARC, Lyon (1992). IARC, Cancer incidence in five continents, IARC Sci. Publ. 143, IARC, Lyon (1997). IARC, Monographs on the evaluation of carcinogenic risks to humans. Alcohol drinking, p. 416, IARC, Lyon (1988). JAKOBSSON, K., MIKOCZY, Z. and SKERFVING, S., Deaths and tumours among workers grinding stainless steel: a follow up. Occup. environ. Med., 54, 825 829 (1997). KABAT, G., HOWSON, C. and WYNDER, E., Beer consumption and rectal cancer. Int. J. Epidemiol., 15, 494 501 (1986). KATO, I., TOMINAGA, S. and IKARI, A., A case-control study of male colorectal cancer in Aichi Prefecture, Japan: with special reference to occupational activity level, drinking habits and family history. Jpn. J. Cancer Res., 81, 115121 (1990). MARSH, G., Mortality among workers from a plastic producing plant: a matched case-control study nested in a retrospective cohort study. J. occup. Med., 25, 219 230 (1983). MILLER, B. and BLAIR, A., Mortality patterns among professional artists: a preliminary report. J. environ. Pathol. Toxicol. Oncol., 6, 303313 (1985). MINDER, C. and BEER-PORIZEK, V., Cancer mortality of Swiss men by occupation, 1979 1982. Scand. J. Work Environ. Hlth., 18, 127 (1992). PETERS, R., GARABRANT, D., YU, M. and MACK, T., A case-control study of occupational and dietary factors in colorectal cancer in young men by subsite. Cancer Res., 49, 5459 5468 (1989). PICKLE, L., GREENE, M. and ZIEGLER, R., Colorectal cancer in rural Nebraska. Cancer Res., 44, 363369 (1984). SIEMIATYCKI, J. (ED.), Risk factors for cancer in the workplace, CRC Press, Boca Raton, FL (1991). SIEMIATYCKI, J., FRITSCHI, L., NADON, L. and GERIN, M., Reliability of an expert rating procedure for retrospective assessment of occupational exposures in community-based case-control studies. Amer. J. indust. Med., 31, 280 286 (1997). SIEMIATYCKI, J., WACHOLDER, S., RICHARDSON, L., DEWAR, R. and GERIN, M., Discovering carcinogens in the occupational environment; methods of data collection and analysis of a large case-referent monitoring system. Scand. J. Work Environ. Hlth., 13, 486 492 (1987). SORAHAN, T., PARKES, H., VEYS, C. and WATERHOUSE, J., Cancer mortality in the British rubber industry: 1946 80. Brit. J. indust. Med., 43, 363373 (1986). STEMMERMANN, G., NOMURA, A. and CHYOU, P., Prospective study of alcohol intake and large bowel cancer. Digest. Dis. Sci., 35, 1414 1420 (1990). WEILAND, S., MUNDT, K., KEIL, U., KRAEMER, B., BIRK, T., PERSON, M., BUCHER, A., STRAIF, K., SCHUMANN, J. and CHAMBLESS, L., Cancer mortality among workers in the German rubber industry: 198191. Occup. environ. Med., 53, 289 298 (1996). WORLD CANCER RESEARCH FUND, Food, nutrition and the prevention of cancer: a global perspective, p. 670, World Cancer Research Fund and American Institute for Cancer Research, Washington, DC (1997).