Document mBkxZ4j4kGQr4mxYGLxN5N19d

AMERICAN JOURNAL OF INDUSTRIAL MEDICINE 35:9-14 (1999) Peritoneal Cancer and Occupational Exposure to Asbestos: Results From the Application of a Job-Exposure Matrix Pierluigi Cocco, md, and Mustafa DosemecI, PhD' Background Because ofthe rarity ofperitoneal mesothelioma, occupational risks associated with it have seldom been studied, particularly among women. In this respect, death certificates databases may provide numbers large enough for analysis, although the International Classification of Diseases, 9th revision (ICD-9) does not single out mesothelioma from the rest ofperitoneal cancers. The aim of this paper is twofold: to explore occupational risks of peritoneal cancer among men and women, and to test the performance of a job-exposure matrix in detecting its association with asbestos exposure using the occupation and industry reported in the death certificate. Methods From a large database containing information on the 1984--1992 death certificates of 24 U.S. states, we identified 657 deathsfrom peritoneal cancer and 6,570 controls who died from non-malignant diseases, 1:10 matched by region, gender, race, and 5-year age group. Results Occupations at risk included insulators among men, and machine operators among women. Among men, we found a significant increase in risk associated with employment in manufacturing industries, such as industrial and miscellaneous chemicals: miscellaneous non-metallic mineral and stone products: construction and material handling machines; and electrical machinery, equipment, and supplies; as well as in services to dwellings and other buildings. Industries at increased risk among women included elementary and secondary schools; miscellaneous retail stores; and publishing and printing. Our job-exposure matrix classified 17 male cases and 3 controls in the high probability category of exposure to asbestos (OR = 61.6). Among men, risk of peritoneal cancer increased significantly by probability and intensity of exposure to asbestos. No such pattern was observed among women. The job-exposure matrix did not classify anyfemale subjects in the high probability or intensity ofasbestos exposure. Discussion This study provides evidence that death certificate data andjob-exposure matrices are useful tools to observe well-established associations, such as the one existing between peritoneal cancer and asbestos exposure among men, in spite of crude information, disease misclassificaiion, and occupational misclassification. Tliese factors are more likely to- preclude meaningful results among women. Am. J, Ind. Med. 35:9-14,1999. Published 1999 Wiley-Liss, Into KEY WORDS: peritoneal cancer: asbestos; insulation workers; epidemiology; jobexposure matrices Occupational Studies Section, National Cancer Institute, Bethesda, MO 'Correspondence to: Mustafa Dosemeci, Ph.D., Occupational Studies Section, National Cancer institute, 6130 Executive Boulevard, EPN room 41S, Bethesda, MD 20632. E-mail: dosemecm@epndce.nci.nih.gov Accepted 6 July 1998 INTRODUCTION Malignant mesothelioma arises from the lining cells of the coelomic cavities [Mack, 1995]. Malignant mesothelio mas of the pleura and the peritoneum share exposure to asbestos as the major etioiogic factor [Spirtas et al,, 1994]. Published 1999 Wiley-Liss. Inc. 'This article is a US Government work and, as such, is in the public domain in the United States of America. 10 Cocco and Dosemeci However, they differ in respect to sex-ratio, route of exposure (inhalation vs. ingestion) [Mack, 1995], and fiber type (chrysotile vs. amosite asbestos) involved [Peto et al,, 1995; Ribak et al., 1989]. In contrast with pleural mesothe lioma, which predominates among men, mesothelioma of the peritoneal cavity occurs with roughly equal frequency in men and women [Mack, 1995]. This could imply that the pertinent exposure is distributed with little reference to gender. The geographical pattern of peritoneal mesothe lioma in the United Kingdom parallels the past distribution of the asbestos industry, but unlike pleural mesothelioma, it did not cluster in seaports where shipbuilding and ship repairing were carried out [Gardner et al., 1985], The contrasting features of peritoneal and pleural mesothelioma suggest that different routes of exposure to asbestos and/or different fiber types may contribute to distinct epidemiologic patterns [Leigh et al., 1991]. The underlying cause of death in death certificates is coded using the Internationa) Classification of Diseases, 9th revision (ICD-9), which does not single out mesothelioma from the rest of peritoneal cancers. Besides, the proportion of mesothelioma among all peritoneal tumors is unclear. In a recent review of SEER incident data, cancer of the perito neum was only considered jointly with cancer of the retroperitonemn [Mack, 1995], Mesothelioma accounted for about 20% of these two cancer sites combined and it did not vary' by gender. However, the proportions of epitheliumderived cancers (including mesothelioma) and connective and other soft tissue-derived cancers might be different between the peritoneum and the retroperitoneum. Neverthe less, a recent study in the United Kingdom showed that mortality from asbestosis was more closely related to that from peritoneal cancer than pleural cancer [Coggon et al., 1995], which supports using mortality data to explore the association between asbestos exposure and peritoneal can cer. The same report showed that the risks for cancers of the pleura and peritoneum distribute quite differently by occupa tion, with construction workers being at much higher risk for peritoneal cancer than pleural cancer, vehicle body builders and plumbers showing excesses at both sites, and various other occupations, such as metal plate workers, upholsterers, electricians, and welders showing only pleural cancer ex cesses [Coggon et al., 1995]. To further investigate the association of peritoneal cancer and occupational exposure to asbestos by gender and to explore the reliability of both a priori designed jobexposure matrices and occupational information in the death certificate, we conducted a case-control study using data from a national surveillance program of occupational dis eases developed since 1984 by the National Cancer Institute, the National Institute for Occupational Safety and Health, and the National Center for Health Statistics [Burnett and Dosemeci, 1994], We first designed a job-exposure matrix based on the 1980 U.S. Census list of occupations and industries, and subsequently applied it to the codes on the death certificates of cases and controls. METHODS Since 1984, the National Cancer Institute, the National Institute for Occupational Safety and Health, and the National Center for Health Statistics have supported the coding of occupation and industry titles on death certificates from a number of U.S. -states according to the 1980 U.S. Census occupation and industry codes [Burnett and Dosem eci, 1994]. Briefly, it currently consists of a total of 4.5 million death certificates from 24 U.S. states, covering the years 1984-1992. Only one occupation and industry combi nation is reported for every subject, and no duration of employment is available. Data reported on the death certifi cates of subjects aged 20 years or more were used to evaluate the risk of peritoneal cancer associated with occupational exposure to asbestos by gender. Cases were 657 subjects (249 men and 408 women) who died from cancer of the peritoneum (ICD-9 codes 158.8 and 158.9). Eighteen men and 20 women were African-American. Because of even smaller numbers, subjects of Asian origin and native Americans were excluded from the analysis. All the remaining were noted as Whites in the death certificate. Ten controls per case were selected from among subjects who died from non-malignant diseases, frequency-matched to cases by geographic region, race, gender, and 5-year age group. Risk of peritoneal cancer was first explored by industry and occupation. To evaluate risk in relation to occupational exposure to asbestos, we designed a job-exposure matrix based on the 1980 U.S. Census list of occupations and industries, and subsequently applied it to the occupation and industry codes in the death certificates of cases and controls. An estimate of intensity (none -- 0, low = 1, medium = 2, high = 3) and probability (none - 0, low = 1, medium 2, high = 3) of exposure to asbestos was developed for each 3-digit occupation and industry code. Intensity of exposure was estimated based upon literature information [Parmeggiani, 1985], computerized databases (OSHA files, NIOSH inspection data base), unpublished industrial hygiene re ports, and personal experience. The probability index associ ated with a given occupation or industry 1980 Census code was estimated based on the proportion of exposed workers within the job title or industry under consideration, and the number of other occupations or industries coded likewise. In addition, occupations were characterized into two groups depending upon the sources of exposure. If exposure was determined by the occupation itself- regardless of industry (e.g., insulators), final intensity and probability scores were obtained by squaring the respective occupational levels. If exposure was determined by both occupation and industry (e.g., maintenance workers in shipyards), intensity and u 4 Peritoneal Cancer and Exposure to Asbestos IT '.e probability scores resulted from multiplying the respective TABLEI. Occupations and Industries With at Least 3 Cases Associated levels of occupation and industty. The final scores of With an Increase in Risk of Peritoneal Cancer (OR s 1.5): probability and intensity of exposure were further grouped Study of Death Certificates in 24 US. States, 1984-1992 in four categories (none = 0, low = 1-2, medium = 4, high { 6). Cut points were selected a priori, with the highest Census Cases/ il category defined by a score of 6 to increase the statistical cade Occupation or Industry title controls OR <95% CJ.) power. As excluding any probability of exposure to asbestos was possible only for 1.5% of the study population (70/ Men 7227), due to the widespread use of asbestos in the past, the Occupations unexposed reference group included also subjects with low 549 Nat specified mechanics and 3/14, 1.8 (Q.5-6.5) probability and low intensity exposure in order to provide repairers ) more stable risk estimates. 593 Insulation workers 17/1, 180(23,5-1375) Odds ratios (ORs) were estimated by logistic regression 779 Machine operators, not speci 6/26, 2.0 (0.8-5.0) and 95% confidence intervals (95% C.I.) by the Wald fied . method using the GMBO program in the Epicure software 869 Construction laborers 8/57, 1.7 (0.7-3.8) package [Preston et al., 1990], Govariates in the logistic industries regression model included age (continuous), marital status 60 Construction 43/286, 1.7 (1.2-2.4) (never married versus ever married), socio-economic status 192 Industrial and miscellaneous 4/15, 2.3 {0.7--7.0) (five categories, based on the Green's score for specific chemicals occupations [Green, 1970]), metropolitan vs. non-metropoli 262 Miscellaneous non-metaiiic 5/6, 7.6(2.3-25.5) tan residence, and ethnic origin (North America and Europe mineral and stone products vs. South America and Africa). The statistical significance of 312 Construction and material han 3/10, 3.3 (0.9-12.4) the linear trend by increasing intensity and probability of dling machines exposure to asbestos was tested by dividing the regression 342 Electrical machinery, equip 6/12, 5.1 (1.8-13.9) coefficients of the variables assumed as non-categorical by ment, and supplies their standard error to generate a 2 statistic. Under the null 722 Services to dwellings and other 3/9, 3.9(1.0-14.8) hypothesis, this test behaves as a normal standard deviate buildings [Breslow and Day, 1980], Two-tailed p values were consid Women ered throughout this article. Occupations 156 Teachers, elementary school 22/106, 1.5(0.6-26) RESULTS ' 407 Private household cleaners and servants 6/73 2.0(0.6-6.51 Average age at death of peritoneal cancer cases was 64.3 13.5 among men (64.3 13.5 among Whites, and 64.8 8.S among African-Americans), and 70.1 12.0 among women (70.2 z. 11.8 among Whites, and 67.2 14.9 among African-Americans). Only 15.9% of male cases (36/226) and 7.7% of female cases (28/362) for whom information was available had autopsy, compared to 14.2% of male controls (269/1900) and 9.1% of female controls (266/2910). Seventeen male cases (6.8%) and one male control were reported as insulation workers in the death certificate (Table I). Machine operators also showed a risk 458 779 industries 172 682 772 842 Hairdressers and cosmetolo gists Machine operators, not speci fied 6/37, 6/25, Printing and publishing, except newspapers Miscellaneous retail stores Beauty shops Elementary and secondary ' schools 3/14. 4/9, 6/37, 33/179, 1,7 (0.7--4.1) 2.6(1,0-6.31 3.8 (1.3-10.8) 5.0(1.5-16.6) 1.7(07-4.2) 1.5(1.0-24) increase in both genders, which was significant among men. Non-significantly elevated risks of peritoneal cancer were observed for mechanics and repairers and construction ent and they included: publishing and printing, miscella laborers among men, and elementary school teachers; pri neous retail stores, and elementary and secondary schools. vate household cleaners and servants; and hairdressers and Table II describes the results obtained with the job- cosmetologists among women. Industries with a significant exposure matrix. Risk of peritoneal cancer increased signifi excess risk among men included construction; miscella cantly by probability (test for trend: all men: p = 0,0004; neous nonmetallic mineral and stone products; electrical white men p = 0.016) and intensity (test for trend: all men: machinery, equipment and supplies; and services to dwell p -- <0.0001; white menp = 0.0006) of exposure among all ings and other buildings. Industries associated with a men and white men. A significant upward trend was also significantly elevated risk among women were quite differ- observed among African-American men by intensity of 12 Cocco and Dosemeci TABLE II. ORs tor Peritoneal Cancer by Probability and Intensity of Exposure to Asbestos: Study of Death Certificates in 24 U.S. States. 1984-1992 Exposure metric Whites Cases/controls Ob (95% C.L) , African-Americans Cases/ controls OR (95% Ci.) Ml Cases/ controls OR (95% CX) Men Probability of exposure Baseline Low Medium High Test for trend (p value) Intensity of exposure Baseline Low Medium High Test tor trend (p value) Women Probability of exposure Baseline Low Medium Likely Test tor trend (p value) Intensity of exposure Baseline Low Medium High Test tor trend (p value) 159/1,661 17/172 38/474 17/3 159/1,661 20/288 31/320 21/41 1.0-- 1.1 (0.6-1.9) 0.9 (0.6-1.4) 60.0 (17.1-211) 0.0016 ' 1.0 -- 0.8 (0.5-1.4) 1.1 (0.7-1.6) 5.1 (2.9-9.1) 0.0006 8/114 3/12 7/54 0/0 8/114 2/30 7/29 1/7 369/3,655 4/20 15/205 0/0 1.0 -- 2.5 (0.8-7.5) 0.7 (04-1.3) -- -0.75 369/3,655 12/185 7/40 0/0 . 1.0 -- 0.7(0.4-1.31 2.0 (0.8--4-6) -------0.33 17/175 0/2 3/23 0/0 17/175 3/20 n/5 o/o . 1.0 -- 3.9 (0-8--19.0) 2.3 (0.7--7.5) -- 0,15 1.0 -- 1.3 (Q.2-7.5) 3.6(1.1-12.0) 2.7 (0.3-27.4) 0.04 167/1,775 20/184 45/528 17/3 167/1,775 22/318 38/349 22/48 1.0 -- 1,2(0.8-2.0) 1.0 (0.7-1.5) 61.6(17.5-216) 0.0004 1.0 -- 0.9 (0.5-1.4) t.3(0.9-16) 46 (2.6-85) 0.0000 1.0 -- -- 2.2(05-10.3) -- 0.91 1.0 -- 2.8(06-14.2) -- -- 0.27 386/3.830 4/22 18/228 0/0 386/3,830 15/205 7/45 0/0 1.0 -- 2.2 (0.7-66) 06(0.5-1.4) -- -0.48 1.0 -- 05(0.4-1,4) 15(06-4.1) ---- 0.40 asbestos exposure (P = 0.04), although the trend was not monotonic. No African-American men had high probability of exposure. Risk of peritoneal cancer did not increase with probability ofasbestos exposure among women, although no female cases had high probability of exposure. Risk was non-significantly elevated in the medium intensity of expo sure for all women and white women. Among women, trends were not statistically significant, and the highest risk was observed for moderate intensity of exposure to asbestos. However, it was modest and not significant. Also, among women, neither cases nor controls were assigned to high probability or intensity' of exposure to asbestos. Seventeen male cases and three controls had high probability of exposure to asbestos, which was therefore associated with a 62-fold increase in risk of peritoneal cancer. All were white men. Although our matrix classified other occupations in the high probability of exposure, the excess was entirely driven by insulators. Most insulators worked in the construction industry (13/17), two in electric utility companies, and one each in manufacturing of industrial chemicals and in an unspecified manufacturing industry. Among other occupa tions in the construction industry, risk was not elevated for supervisors and carpenters, and small non-significantly elevated risks were observed for construction laborers and painters (OR = 1.3, not shown in Tabic I). A fivefold increase in risk was associated with high intensity of exposure to asbestos. Cross-classification of study subjects by exposure probability and intensity did not add much further information, because of small numbers and empty cells. DISCUSSION Our results confirm that occupational exposure to asbestos is associated with a very high risk of peritoneal cancer. Insulators showed the highest risk, and they contrib uted the most to the significant increase in risk in the construction industry. However, elevated risks were also observed for machine operators in both genders, in a few manufacturing industries, and among workers employed in Peritoneal Cancer and Exposure to Asbestos 13 services to dwellings and other buildings. The test for trend was significant either by probability or by intensity of - exposure to asbestos. Among'women, risk was elevated in the publishing and printing industry, in retail stores, and in elementary and secondary schools. Trends by probability and intensity of exposure to asbestos were not significant As Selikoff pointed out (Selikoff, 1992a), the finding of new high-risk groups for asbestos-related diseases in studies based on death certificates should be explained with the tendency to list the most recent occupation in the death certificate, while prior or short-term asbestos work would not be expected to be recorded. This might account for most of the positive findings among women in the present study, such as elementary and secondary schools. Our matrix classified teachers in the reference category, although asbes tos was widely used for insulation purposes in schools. In fact, as explained in the Methods section, our unexposed reference group also included subjects with low probability and low intensity exposure. However, cases of pleural mesothelioma among teachers have been discussed in rela tion to their asbestos exposure in school buildings [Lilienfeld, 1991; Anderson et al., 1991], and a history of prior employment in other occupations and industries with prob able asbestos exposure was positive only for 3/12 teachers and 5/29 maintenance workers who died from mesothelioma in Wisconsin [Anderson et al., 1991]. It has been suggested that environmental, possibly residential, exposure to asbes tos may account for the majority of female peritoneal mesotheliomas [Dawson et al,, 1993]. Spouse exposure was reported as a risk factor for mesothelioma among women in England and Wales [Greenberg and Lloyd Davies, 1974], Unfortunately, we did not have information on spouse occupation to explore whether a greater proportion of women in the occupations and industries with an increase in peritoneal cancer risk were married to blue collar workers, which might also have explained some of our findings among women. Therefore, difficulties in identifying pre cisely the sources of exposure to asbestos and the possibility that other causative agents are involved [Peterson et al., 1984], warrant a conservative evaluation for most positive findings on peritoneal cancer and occupation among women. On the other hand, the hypothesis of an association for female teachers and other workers in elementary and secondary schools cannot be discarded in light of previous reports of pleural mesothelioma cases among these workers. Spirtas et al. calculated a relative risk (RR) of 3 (95%CI 0. 7.16.5) for peritoneal mesothelioma associated with hav ing ever held any of nine activities where asbestos exposure was suspected, based on 20 male cases and 203 controls [Spirtas et al,. 1994], By applying the same criteria to define exposure to asbestos, we obtained an OR of 1.8 (95%C1 1. 3.2.4) among men in our study. Therefore, results from our job-exposure matrix are consistent with Spirtas et ai. The lower risk in the present study may have resulted from the exposure misclassification related to the availability of information on the sole occupation and industry reported on the death certificate, compared to lifetime work histories obtained with telephone interview in the cited paper, and to the disease misclassification due to the use of death certifi cate diagnoses instead of histopathology data. Workplace environments typically involve exposure to a complex mixture of physical and chemical agents, each of which might be responsible for an observed association. Compared to occupation and industry titles only, using job-exposure matrices in epidemiologic analyses offers the advantages of a clearer definition of the risk factor and of a greater statistical power by assembling subjects with the same exposure in various occupations and industries. How ever, level of detail in the occupation and industry coding system and completeness in working histories of study subjects are critical factors for the performance of jobexposure matrices in identifying the exposure of interest [Dosemeci et al., 1994], In the 24 states death certificates database, only one occupation and industry are registered. These are coded with the 3-digit 1980 U.S. Census code, which may not be specific enough in discriminating among exposed and unexposed subjects. The resulting non differential misclassification would bias the risk estimate toward the null. These disadvantages are more important in studies involving women, as the reliability of the occupa tional information in death certificates is poorer for women than men [Schade and Swanson, 1988]. Therefore, negative findings should be interpreted with caution. ' A further obstacle in detecting associations based on death certificate information is related to disease misclassifi cation. Not all deaths ascribed to cancers of the pleura or the peritoneum in the death certificates are mesotheliomas, and some mesotheliomas are misclassified to other sites [Ribak et al., 1989]. In his detailed review of the use of death certificates to detect occupational associations with asbestos exposure, Selikoff pointed out that due to a very poor sensitivity (20-23% in the United States, 32-51% among asbestos insulators), the use of standard death registrations vastly underestimate the incidence of mesothelioma [Seli koff , 1992b; Selikoff and Seidman, 1992]. However, the specificity of the death certificate diagnosis of peritoneal mesothelioma among asbestos ilnsuiation workers was very high as all 92 cases so indicated were confirmed after best evidence procedures [Selikoff and Seidman, 1992]. Also, no major differences in diagnostic accuracy were found by comparing diagnosis of mesothelioma in death certificates of asbestos insulation workers and former non-unionized asbestos factory workers, whose death certificate had virtu ally no indication of occupational exposure to asbestos [Selikoff, 1992a]. In our study, we had available only the iCD-9 code and not the original certificate, which means that we were unable to discriminate mesothelioma from the rest 14 Cocco and Dosemed of peritoneal cancers. Therefore, specificity in detecting mesothelioma was also lowered in our study. Nonetheless, the observed strong association between asbestos and peritoneal cancer in men provides evidence that information on cause of death and usual occupation and industry on the death certificate is reliable. In addition, the results of this study suggest that using job-exposure matrices is an appropriate approach in a first-step evaluation of occupational carcinogens with the case-control design. The use of histopathologic data to define cases and a careful evaluation of complete work and residential history are of paramount importance in evaluating whether other risk factors besides occupational exposure to asbestos are in volved in tiie etiology of peritoneal mesothelioma among women. REFERENCES Anderson HA, Haorahan LP, Schirmer J, Higgins D, Sarow P. 1991. Mesothelioma among employees with likely contact with in-place asbestoscontaining building materials. Ann NY Acad Set 643:550-572. Breslow NE, Day NE. 1980. "Statistical methods in Cancer Research. Volume I. The analysis of case-conrrol studies. Lyon, France: International Agency for Research on Cancer (1ARC Sci Pub No. 31). Bureau of the Census. 1982. Alphabetical index of industries and occupa tions. Washington; DC; U.S. Department of Commerce, PHC80-R3. Burnett CA, Dosemeci M. 1994. Using occupational mortality data for surveillance of work-related diseases of women. J Occup Med 36:1199 1203. Coggon D, Inskip H, Winter P, Panned B. 1995. Differences in occupational mortality from pleural cancer, peritoneal cancer, and asbestosis. Occup Environ Med 52:775--777. Dawson A, Gibbs AR, Pooley FD, Griffiths DM, Hoy J. 1993. Malignant mesothelioma in women. Thorax 4B:269-274, Dosemeci M, Cocco PL, Gomel M, Stewart PA, Heineman EF. 1994. Effects of three features of a job-exposure matrix on risk estimates. Epidemiology 5:124--127. Gardner MJ, Jones RD, Pippard EC, Saitoh N. 1985. Mesothelioma of the peritoneum during 1967-82 in England and Wales. BrJ Cancer 51:121--J 26. Green LW. 1970. Manual for scoring socio-economic status for research on health behavior. Pubirc Health Rep 85:815-827, Greenberg M, Lloyd DaviesTA. 1974. Mesothelioma Register. 1967-1968. Br find Med 31:91-104, Leigh J, Rogers AJ, Ferguson DA, Mulder HB, el til. 1991. Long asbestos fiber content and mesothelioma cell type, site, and survival Cancer 68:135-141. Lilicnfcld DE. 1991. Asbestos-associated pleural mesothelioma in school teachers: a discussion of four cases. Ann NY Acad Sci 643:454-486. MackTM. 1995-Sarcomas and other malignancies of soft tissue, rctroperi- toneum, peritoneum, pleura, heart, mediastinum, and spleen. Cancer 75:211-244. ' Paimeggiani L., editor. 1985. "Encyclopedia of occupational health and safety," 3rd ed. Geneva, Switzerland: International Labour Office. Peterson IT Jr, Greenberg SD, Buffier PA. 1984, Non-asbestos-related malignant mesothelioma. A review. Cancer 54:951-960. Peto J, Hodgson JT, Matthews FE, Jones JR. 1995. Continuing increase in mesothelioma mortality in Britain. Lancet 345:535-539. Preston DL, Lubin JH, Pierce DA. 1990. Epicure. Seattle, WA: Hirosoft International Corp. Ribak J. Seidman H. SelikoffD. 1989. Amosite mesothelioma in a cohort of asbestos workers. Scand J Work Environ Health 15:106-110. Schade WJ, Swanson GM. 1998. Comparison of death certificate occupa tional and industry data with lifetime occupational histories obtained hy interview: variations in the accuracy of death certificate entries. Am J Ind Med 14:121-136. Selikoff D. 1992a. Death certificates in epidemiological studies, including occupational hazards: inaccuracies of occupational categories. Am J Ind Med 22:493-504. Selikoff U. 1992b. Use of death certificates in epidemiological studies, including occupational hazards: discordance with clinical and autopsy findings. Am J Ind Med 22:469-480. Selikoff U, Seidman H. 1992. Use of death certificates in epidemiological studies, including occupational hazards: variations in discordance of different asbestos-associated diseases on best evidence ascertainment. Am J Ind Med 22:481-492. Spinas R. Heineman EF, Bernstein L, Beebe GW, Keehn RJ, Stark A, Harlow BL, Benichou J. 1994. Malignant mesothelioma: attributablerisk of asbestos exposure. Occup Environ Med 51:804--811.