Document 8YrMBbLLwb1qYMJbqRNgL64Z

Cancer Causes and Control 1995, 6, 361 ~ 368 High bladder cancer mortalit-y in rural New England (United States): an etiologic study Linda Morris Brown, Shelia Hoar Zahm, Robert N. Hoover, and Joseph F. Fraumeni, Jr. (Received 31 January 199s; accepted in revised fbrm 23 March 1995) An interview study of next-of-kin of 325 persons who died of bladder cancer and 673 individuals who died of other causes in Vermont and New Hampshire (United States) was conducted to assess reasons for the persistent pattern of elevated bladder cancer mortality for both genders in rural New England There was some evidence of elevated risks for both leather and textile workers that rose to over twofold for workers who also lived near these industries and for persons with French-Canadian ancestry. Occupational exposures in the textile and leather industry may explain at least a portion of the excess bladder cancer riskin rural New England Cancer Currses and C W t d 1995, 6, 361-368 Key words: Bladder neoplasm, case-control study, etiology, leather industry, textile industry, occupational exposure, United States. Introduction A s d m g feature of the United States cancer-mortality maps of 1950-69' (Figure 1) is the clustering of hq$ rates of bladder cancer in rural New England for men and women. Women in this area experienced the highest bladder-cancer mortality rates in the US, whereas the rates for men in rural New England and adjoining areas of New York State were second only to New Jersey. In contrast to New Jersey, where concentration of the chemical industry may contribute to the excess risk, the reasons for the high rates in rural New England are unc1earFJ Updated cancer maps by the US National Cancer Institute have revealed persistence of the excess bladder cancer rates in this region for the period 1970-89 (Figure 2). To evaluatepossible risk factors prevalent in this area such as French-Canadian ancestry, consumption of French-Canadian foods and fiddlehead or bracken fern (a known animal carcinogen): occupational exposures, and residence near leather, textile, or paper/puIp factories, we analyzed data from a caseconVol .study of bladder cancer conducted in Vermont and New Hampshire. Materials and methods The study included all White residents of Vermont and New Hampshire who died during 1975-79from bladder cancer (ICD-85 code 188; ICD-96 code 188.0-188.9), identified from the Vital Records and Health Statistics Office of New Hampshire and the Public Health Statistics Office of Vermont. Two controls per case, matched on state, gender, race, age (& 2 years), and year of death, were selected randomly from all other resident deaths (excludmg suicides). One of the two controls per subject was matched also on county of residence. A mortality-based s t u d y design was used because no population-based cancer registries covered the region at Ms Brown, and Drs Zahm, Hoover, and Fraumeni are with the Epidemdogy and Biostatistics Program, National CancerInstitute,Betbesda, hfD, Aa'dress correspondence to Ms Brown, Epidemiology and Biostatistics Program, National Cancer Institute, National Institutes of Health, hecutive Plaza North, Room 415,6130 Executive Blv4 MSC 7368, Bethesdu, MD 20892-7368, uscl Cancer Causes and ConmL Vol b 1995 361 White Men 1.533.96(lowest 10%) - White Women 0.68-1.63 (lowest 10%) .- Figure 1. Bladder cancer mortality rates by state economic area for White men and women (1950-69). that time, making the logistics of prospectively identifying incident cases prohibitive. Details of subject selection procedures are presented elsewhere.3 During the time period 1975-79, there were 440 White residents of Vermont and New Hampshire who died of bladder cancer. Interviews were completed for the nextof-kin of 224 males and 101 females (74 percent). A total of 923 White controls were selected and the next-of-kin of 73 percent were interviewed (459 males, 214 females). 362 Cancer Causes and Control. Vol 6. 1995 Refusal was the most common reason for nonresponse for proxies of both cases and controls. However, decedents for whom proxy interviews were obtained were similar to decedents without proxy interviewswith respect to case-control status, gender, age, and county of residence. Questionnaires sought information on demographic characteristics, lifetime occupational and residential histories, history of tobacco and beverage use, Bladder cancer in rural Npru England variables of interest. Since both methods yielded similar results, analyses are presented for the unconditional approach because it allowed the inclusion of data from all subjects. Separate analysis for the two control series also yielded similar results, so only the analysis for the combined control series is presented. The variables for residence (Vermont, New Hampshire),age ( <65,65-74, 75-84, > 84), and history of bladder infection (yes, no) were included in models to adjust for potential confounding. Age-adjusted population-attributable risk (PAR) estimates of the proportion of bladder cancer due to employment in textiles and/or leather industries were calculated separately for men and women by using the method of Whittemore."*'2 White Women DeaIhs/100,OM]( U S l . 9 2 ) 2.34-4.29 (highest 10%) 1.81-2.33 (90-89.YI) 1.33-1.80(10-49.9'R~~ O.6U 1.32 (lowest 10%) A Figure 2. Bladder cancer mortality rates by state economic area for White men and women (197CL89);Connecticut, Maine, Massachusetts, New Hampshire, New York, Rhode Island, and Vermont. consumption of selected dietary items including bracken fern, and medical history including badder infection. For each job held for one year or longer since age 12, the name of the employer, job title, duties, and years worked were collected and coded using the SIC' and SOC' systems. Occupational and industrial categories were formed based on these codes and a check list of 31 occupations and industries of interest. Only those categories for each gender reported by five or more cases or controls are included in this report. Both conditional and unconditional logistic regres- ion"'^ were used to calculate odds ratios (OR) for the Results Table 1 presents ORs for bladder cancer for known or suspected risk factors. Since the risk estimates were similar for men and women, the combined results are presented History of bladder infection at least five years prior to death (OR=3.7, 95 percent confidence interval [CI] =2.55.4) and French-Canadian ancestry (OR= 1.3, CI =1.0-1.8) were associated with statistically significant excess risk. However, there were no signhcant elevations in risk for consumption of traditional French-Canadian foods such as pig's feet (rugout de putte de cochon),pork pies or other meat pies (tourtieres), or salmon pie (turte uu suurnon).Risk was not affected by consumption of bracken fern, type of water supply used (ie., community or well), or by consumption of coffee, including heavy consumption of greater than 21 drinks per week (OR = 1.0, CI =0.6-1.7). Risks associated with cigarette smoking were not elevated for ever-smokers (OR = 1.0, CI =0.7-1.4) and only slightly elevated for heavy smokers of 40 or more cigarettes per day (OR= 12, CI=O.7-1.8). When the smoking-related causes of death (eg., cancers of the buccal cavity, pharynx, esophagus, pancreas, and respiratory system; ischemic heart disease; pulmonary heart disease; symptomatic heart disease; cerebrovascular disease; bronchitis, emphysema, and other diseases of the respiratory system) were excluded from the control series, risk due to smoking rose slightly to 1.2 (CI =0.8-1.7). The ORs for employment in selected industries or occupations among men are shown in Table 2. Risks were elevated nonsipficantly for men ever employed in the leather, textile, or lumber industry; elevated nonsigniscantly for men employed for more than five years in the leather or lumber industry; and elevated significantly for men employed for more than five years in the textile industry (OR=2.2, CI = 12-3.7).The ageadjusted PAR for employment in the textile and/or Cancer C a w s and Control. Vo16.1995 363 L M.Brown et al Table 1. Risk of bladder cancer for known or suspected risk factors among White men and women in New Hampshire and Vermont Factor History of bladder infection French-Canadian ancestry Ever ate bracken fern Regularly ate bracken fern Ever used community water Ever used well water Ever drank coffee Ever smoked cigarettes Cases (n = 325) 79 (24.3) 78 (24.0) 24 (7.4) 15 (4.6) 294 (90.5) 215 (63.1) 293 (90.2) 196 (60.3) Controls (n = 673) 57 (8.5) 137 (20.4) 71 (10.6) 38 (5.6) 604 (89.8) 441 (65.5) 606 (90.0) 409 (60.8) Adjusted ORa (CUb 3.7 (2.5-5.4) 1.3 (1.0-1.8) 0.6 (0.4-1.O) 0.8 (0.4-1.4) 1.1 (0.7-1.8) 1.o (0.7-1.3) 0.9 (0.6-1.5) 1 .o (0.7-1.4) a All ORs are relative to risk for subjects who were never exposed to that factor. Exceptfor history of bladder infection which was adjusted for age and state, all ORs are adjusted for age, state, and history of bladder infection in a logistic analysis. Percentages are in parentheses. Cl = 95% confidence interval. leather industry was 6.5 percent (CI = -1.6/o-14.60/o). Although based on small numbers, ORs were elevated significantly for men employed in the auto industry, including those ever employed (OR=3.5, CI=1.1-11.0) and those employed for more than five years (four exposed cases, zero controls). Risks also were elevated significantlyfor men employed for more than five years as truck drivers (OR =2.4, CI = 1.2-4.8). Among women, bladder cancer was not elevated significantly in any occupation or industry (Table 3). Table 2. Risk of bladder cancer associated with employment in selected industries or occupations among White men in New Hampshire and Vermont (224 cases, 459 controls) Industry Leather Textile Paper/pulp Lumber Printing Dairy farming Other farming Construction Asphalt Auto industry Shipbuilding Occupation Painter Electrician Carpenter Truck driver Cases 27 (12.1) 40 (17.9) 14 (6.2) 39 (17.4) 13 (5.8) 35 (15.6) 38 (17.0) 52 (23.2) 12 (5.4) 9 (4.0) 16 (7.1) 5 (2.2) 8 (3.6) 19 (8.5) 32 (14.3) Ever worked Controls Adjusted ORa (Cub Cases Worked more than 5 years Controls Adjusted ORa (Wb 38 (8.3) 61(13.3) 36 (7.8) 65 (14.2) 30 (6.5) 81 (17.6) 105 (22.9) 119 (25.9) 39 (8.5) 5 (1:l) 37 (8.1) 1.5 1.4 0.8 1.4 0.9 0.8 0.7 0.9 0.6 3.5 1.1 (0.9-2.6) (0.9-2.2) (0.4-1.6) (0.9-2.2) (0.4-1.7) (0.5-1.3) (0.4-1.0) (0.6-1.3) (0.3-1.3) (1.1-11 .O) (0.6-2.0) 11 (4.9) 29 (13.0) 7 (3.1) 22 (9.8) 6 (2.7) 22 (9.8) 21 (9.4) 23 (10.3) 4 (1.8) 4 (1.8) 7 (3.1) 18 (3.9) 32 (7.0) 20 (4.4) 34 (7.4) 17 (3.7) 58 (12.6) 67 (14.6) 67 (14.6) 12 (2.6) 0 (0.0) 17 (3.7) 1.4 (0.6-3.2) 2.2 (1.2-3.7) 0.7 (0.3-1.6) 1.5 (0.8-2.6) 0.8 (0.3-2.0) 0.8 (0.5-1.4) 0.6 (0.3-1.0) 0.6 , (0.4-1.1) -0.7 (0.2-2.3) ry) 1 .o (0.4-2.5) 12 (2.6) 10 (2.2) 39 (8.5) 53 (11.6) 0.9 1.5 1 .o 1.4 (0.3-2.7) (0.5-4.0) (0.6-1.8) (0.8-2.2) 1 (0.4) 5 (2.2) 10 (4.5) 19 (8.5) 3 (0.6) 6 (1.3) 27 (5.9) 18 (3.9) 0.6 1.7 0.7 2.4 (0.06-6.6) (0.5-5.9) (0.3-1.6) (1.2-4.8) aAll ORs are relative to risk for subjects who were never employed in that industry or occupation. The ORs are adjusted for age, state, and history of bladder infection in a logistic analysis. Percentages are in parentheses. CI = 95% confidence interval. 364 Cancer Causes and ConvoL Vol6.1995 Bladdw cancer in rural Nev England Table 3. Risk of bladder cancer associated with employment in selected industries among White women in New Hampshire and Vermont (101 cases; 214 controls) Industry Leather Textile Paper/pulp Lumber Printing Dairy farming Other farming Cases 8 (7.9) 33 (32.7) 3 (3.0) 1 (1.0) 1 (1.0) 4 (4.0) 9 (8.9) Ever worked Controls Adjusted ORa 18 (8.4) 1.1 (0.4-2.8) 63 (29.4) 1.2 (0.7-2.0) 6 (2.8) 1.2 (0.3-5.2) 8 (3.7) 0.2 (0.03-2.1) 5 (2.3) 0.3 (0.04-3.3) 9 (4.2) 0.5 (0.1-2.0) 11 (5.1) 1.8 (0.7-4.9) Worked more than 5 years Cases Controls Adjusted 3 (3.0) 14 (13.9) 1 (1.0) 0 (0.0) 0 (0.0) 2 (2.0) 7 (6.9) 10 (4.7) 31 (14.5) 2 (0.9) 1 (0.5) 3 (1.4) 6 (2.8) 7 (3.3) ORa 0.7 1.2 0.7 0 0 0.5 2.5 (Cub (0.2-2.9) (0.6-2.5) (0.1-8.4) - (0.9-2.8) (0.6-8.1) aAll ORs are relative to risk for subjects who were never employed in that industry. The ORs are adjusted for age, state, and history of bladder infection in a logistic analysis. Percentages are in parentheses. CI = 95% confidence interval. Slight, nonsignificant elevations were observed for employment in the leather, textile, and paper and pulp industries and for farming other than dairy faiming. The age-adjusted PAR for employment in the textile industry was 4.4 percent (CI= -11.3/c-20.10/o). The OR increased further among women employed for more than five years in farming other than dairy, but the ORs decreased or stayed the same for longer duration of employment in the other industries. Table 4 presents ORs associated with working at or living near a leather, textile, or paper/pulp factory. Signhcantly elevated risks of greater than 2.0 were seen for men who both worked at and lived one mile or less from a leather or textile factory. Similar associations were seen for women, but the ORs were not significant. Little or no risk was seen for subjects who worked at a leather or textile factory but did not live near the plant, and vice versa Small nonsigfllficant excesses were seen Table 4. Risk of bladder cancer associated with employment at andlor residence within one mile of selected types of industries among White men and women in New Hampshire and Vermont Industry Leather Textile Paperlpulp Worked at No Yes No Yes No Yes No Yes No Yes No Yes Lived near No No Yes Yes No No Yes Yes No No Yes Yes Cases 182 15 15 12 161 23 22 17 202 6 7 8 Men Controls Adjusted ORa (Cl)b 388 1.o - 26 1.1 (0.5-2.2) 30 1.1 (0.6-2.2) 12 2.6 (1.1-6.1) 354 1.o - 43 1.2 (0.7-2.1) 43 1.1 (0.6-2.0) 18 2.4 (1.2-4.8) 406 1.o - 20 0.6 (0.2-15 ) 15 0.9 (0.4-2.4) 15 1.2 (0.5-3.0) Cases a4 3 7 5 62 19 6 14 91 2 7 1 Women Controls Adjusted ORa (Cub 182 1.o - 11 0.8 (0.2-3.0) 14 0.9 (0.3-2.5) 7 1.7 (0.5-6.0) 138 1.o - 42 1.0 (0.5-1.9) 13 0.9 (0.3-2.7) 18 2.0 (0.9-4.6) 197 1.o - 5 1.1 (0.2-6.0) 11 1.4 (0.5-3.9) 1 1.7 (0.1-33.7) aAll ORs are relative to risk for subjects who were never employed in each occupation and who never lived one mile or less from each type of factory. The ORs are adjusted for age, state, and history of bladder infection in a logistic analysis. CI = 95% confidence interval. Cancer Caws and ConmL Vol 6.1995 365 L M.Brown et a1 for women working at and/or living near a paper/pulp facility; but among men, an excess risk was observed only for the combination of occupational and residential paper/pulp exposure. Discussion Mapping cancer mortality at the county level has been useful in generating etiologic clues and targeting epidemiologic re~earch.".'~For example, high rates of lung cancer mortality observed among men along the southeastern US Atlantic coast were found in case- control studies to be associated with asbestos exposures in shipbuilding, particularly during World War II.'5-'7 High rates of oral cancer among women in rural counties of the South were found to be related to the long-standing use of smokeless tobacco." Bladder cancer shows considerable geographic variation, especially among men, with elevated rates in urban areas of the northeastern and upper midwestern US. Occupational factors appear to contribute to the high rates in areas such as New Jersey, where the chemical industry is concentrated and workers have been exposed to aromatic amines that are carcinogenic to the bladder?I9 In an attempt to explain the persistent geographic clustering of bladder cancer in rural New England, we analyzed data from a case-control study of bladder cancer conducted in Vermont and New Hampshire? Analyses focused on several industries prevalent in the region that may involve exposure to bladder carcino- gens. Of particular interest is the textile industry which used dyes that have been linked to bladder cancer." It once employed substantial numbers of New England men and women both in factories and in their homes as 'cottage industries'?' We found a nonsignificant excess risk for men ever employed in the textile industry and a significant excess risk for men employed more than five years. However, the association was weaker among women, whose exposure to industrial chemicals may have been generally lower due to specific job practices. When the textile industry left New England during the 1920s and 1930s, it moved to the southern US. The National Bladder Cancer Study, which included subjects from Atlanta (Georgia), did not reveal any positive associations with the textile industry?l but the differences in risk may reflect regional differences in exposure. In New England, the textile industry produced primarily wool with some cotton, while in the South, it produced cotton, polyester, nylon, and acrylics, but no wool. The chemicals used to degrease, scour, bleach, and dye vary for each fiber type and may explain the regional variation in bladder cancer risk. 366 Cancer Causes and Control Vo16.1995 The leather industry was also once concentrated in New England,20 and thus evaluated as a possible cause for the high rates of bladder cancer in both genders. In previous studies, elevated risks have been reported for leather workers exposed to leather dust, dyes, and ~olvents.'~In our study, overall risks were elevated for men, but did not increase with duration of employment. Since risks were elevated only slightly for women in leather occupations, the evidence is less persuasive that this exposure explains the high rates of bladder cancer among women in rural New England. However, these lower than anticipated risks may be due to misclassifica- tion of occupational exposures by their next-of-kin. The paper and pulp industry has not been linked previously with bladder cancer, but it is heavily concentrated in New England and a source of air and water pollutants that may be carcinogenic.22 We observed no significant association between bladder cancer risk and employment in this industry for either gender. Risks were elevated significantly for long-term employment as a truck driver, although detailed analyses reported elsewhere3 could not determine whether the excess risk was due to diesel emissions or some other factor associated with this occupation. Truck driving also has been associated with excess risk of bladder cancer in the National Bladder Cancer Study,z3 but it seems unlikely to contribute to the regional excess seen among men, nor can it explain the high rates among women. We also evaluated other potential risk factors prevalent to this area such as French-Canadian ancestry and dietary habits including consumption of French- Canadian foods and bracken fern (fiddlehead). We observed slight, nonsignificant elevations in both men and women associated with French-Canadian ancestry, but risks were not elevated for consumption of various traditional French-Canadian foods. Bracken fern has been shown to be carcinogenic in experimental and observational animal studies, producing bladder tumors in rats, guinea pigs, and ~ a t t l eO. ~ur study revealed no consistent association between consumption of bracken fern used mainly as a condiment in salads and risk of bladder cancer. Cigarette smoking is an established cause of bladder cancer, with smokers general1 having two to three times the risk of nonsmokers.''Our study revealed no significant elevations in risk, however, even when controls with smoking-related causes of death were excluded. This lack of a significant association with smokin is not surprising since it has been demonstrated2% that the use of dead controls, even when smoking-related causes of death have been excluded, likely will lead to a biased underestimate of risk. Like Bladder cancer in rural New England most other studies of bladder cancer," we found a posiJive association with bladder infection, and no association with coffee drinking. There are several limitations to this study that should be considered. First, it was based on death certificates; thus, all information was obtained from proxy respondents. Most studies indicate that proxy respondents tend to underestimate exposures such as the number of jobs and residences held, as well as medical conditions, although overreporting also can occur.25 Lack of knowledge or difficulty in recall by the proxy also may have influenced the responses, including those for the. dietary variables where the effects of misclassification could have masked small increases in r i ~ k . 2I~n addition, there may have been differential misclassification of exposure by next-of-kin of controls who died of conditions other than cancer?' even though questions were asked in a similar manner to all subjects by trained interviewers. Second, the control group was representative of people from the general population who had died; thus, controls probably had an overrepresentation of risk factors (e.g., smoking and drinking) that contribute to mortality compared with the general population. Third, the number of subjects, especially women, reporting exposure to any specific occupation or industry was small. Nevertheless, this case-control study did find elevated bladder cancer risks associated with textile and leather work, two industries for which there was u priori suspicion that they might be involved If these risks are causal, then based on the estimates of relative risk and percent of the population exposed from this investigation, approximately seven percent of bladder cancer mortality among males and four percent among females residing in these two states could be attributable to work in these industries. Nationally, little of bladder cancer mortality can be attributed to work in these industries since only three percent of male controls and four percent of female controls from a large, national, population-based studyZoJ6reported such an occupational history (compared with 19 percent of men and 29 percent of women in the current study). Thus, the excess risks attributable to these two industries in Vermont and New Hampshire may account for about one-third of the excess mortality from bladder cancer in these two states compared with that in the total US for the period 1970-89. Again, if these exposures are indeed causal, the actual proportion of the excess risks explained is likely to be higher than one-third since the biases involved in next-of-kin interviews would tend to underestimate both the risks and the proportions exposed Even so, the substantial excess risks unexplained by these two industries, along with the continually high bladder-cancer mortality rate in these two states despite the departure of these industries from the region decades ago, argue in favor of other, as yet unidentified factors contributing to the excess rates of bladder cancer in rural New England. Acknowledgements-The authors wish to thank Joe Barker of Information Management Systems, Inc. for computer support and Dan Grauman, NCI, for map preparation. References 1. Mason TJ, McKay FW. U S Cancer Mortality by County: 1950-1969. Washington, D C US Government Printing Office, 1974; DHEW Pub. No. (NIH) 74-615. 2. Blot WJ, Fraumeni JF Jr. Geographic pattern of bladder cancer in the United States.JNCZ 1978; 61: 1017-23. 3. Hoar SK, Hoover R Truck driving and bladder cancer mortality in rurd New England JNCZ 1985; 74: 771-4. 4. Hirono I. Recent advances in research on bracken carcinogen and carcinogenicity of betel nut.J Environ Sci Health 1985; C3(2): 14587. 5. World Health Organization.International Classification of Diseases,Eighth Revision. Geneva, Switzerland:WHO, 1967. 6. World Health Organization.International Classifiation of Diseases, Ninth Revision. Geneva, Switzerland:WHO, 1977. 7. Office of Management and Budget Standard Industrial Classi$cation Manual. Washington, D C US Government Printing Office, 1972. 8. US Department of Commerce. Standard Occuputional Classification Manual. Washington, D C USDC office of Federal Statistical Policy and Standards., 1977. 9. Breslow NE, Day NE. Statistical Methods in Cancer Research. Vol I. Analysis of Case-Control Studies. Lyon, France: International Agency for Research in Cancer, 1980; IARC Sci. Pub. No. 3 2 192-246. 10. Dixon WJ, Brown MB, Engelrnen L, Jennrich RI, eds. BMDP Statistical Sofiw.re Manual, Volume2 . Berkeley, C A (USA):University of California Press, 1990 1013-45. 11. Whittemore AS: Estimating attributable risk from case- control studies.Am J Epidemiol 1983; 117: 76-85. 12. Boice JD, Lubin JH, Preston DL EpidemiologicAnalysis with a Personal Computer {PmMEB).ethesda, MD (USA):National Institutes of Health, 1992 DHHS Pub. No. (NIH)91-3180. 13. Fraumeni JF Jr. Etiologic insights from cancer mapping. In: Miller RW, Watanabe, S, Fraumeni JF Jr., et al, eds. Unusual Occurrences As Clues To Cancer Etiology. London, UK: Taylor & Francis Ltd, 1988 1325. 14. Pickle LW, Mason TJ, Howard N, Hoover R, Fraumeni JF Jr.Atlas of LIS Cancer Mortality Among Whites:19501980. Washington, D.C.: US Government Printing Office, 1987; DHHS Pub. No. (NIH) 87-2900. 15. Blot WJ, Harrington M, Toledo A, Hoover R, Heath CW Jr, FraumeniJF Jr. Lung cancer after employmentin shipyards during World War 11.N Engl J Med 1978; 299: 620-4. 16. Blot WJ, Morris LE, Stroube R,Tagnon I, Fraumeni JF Jr. Lung and laryngeal cancers in relation to shipyard employment in coastal Virginia.JNCZ 1980; 65: 571-5. Cancer Causes and Control Vo16.1995 367 L M Brown et al 17. Blot WJ, DaviesJE, Brown LM,et a1 Occupation and the high risk of lung cancer in northeast Florida. Cancer 1982; 50: 364-71. 18. Wmn DM, Blot WJ, Shy CM, Pickle LW, Toledo A, Fraumeni JF Jr. Snuff dipping and oral cancer among women in the southern United States.NEngl JMed 1981; 304:745-9. 19. Silverman DT, Hartge P, Devesa SS. Epidemiology of bladder cancer. In: Devita VT Jr, Hellman S, Rosenberg SA, eds. Cancer hevention. Philadelphia,PA (USA): J.B. Lippincott Co., 1993. 20. Dublin T. Transforming Women'sWork New England Lives in the Industrial Revolution. Ithaca, NY (USA): Cornell University Press, 1994 77, 119-25. 21. Silverman DT, Levin LI, Hoover RN. Occupational risks of bladder cancer among white women in the United States.Am J Epidemiol 1990, 132: 453-61. 22. Robinson CF, Waxweder RJ, Fowler DP. Mortality among production workers in pulp and paper mills. Scand J Work Environ Health 1986; 12: 552-60. 23. Silverman DT, Hoover RN, Mason TJ, Swanson GM. Motor exhaust-related occupations and bladder cancer. Cancer Res 1986; 46: 2113-6. 24. McLaughlin JK,Blot WJ, Mehl E, Mandel JS. Problems in the use of dead controls in case-control studies. 11. Effect of excluding certain causes of death. Am J Epzdemiol 1985; 122: 485-94. 25. Nelson LM,Longstreth WT, Koepsell TD, van Belle G. Proxy respondents in epidemiologic research. Epidemwl Rev 1990, 12: 71-86. 26. Silverman DT, Levin LI, Hoover RN, Hartge P. Occupational risks of bladder cancer in the United States: I. White men. JNCI 1989; 81: 1472-80. 368 Cancer Causes and Control. Vol6.1995