Document ykeb0M7E9VJe73LQrkgkQZjdV

tea. J. Cancer: 42, 851-856 (1988) 1988 Alan R. Lias. Inc. at tfo knMffnttKsuf tMen *?* ^Miictoofl 6e iwtnwaomM Cswr* PROPORTION OF LUNG CANCERS IN MALES. DUE TO OCCUPATION, IN DIFFERENT AREAS OF THE USA Paolo Vineis1. Terry Thomas2. Richard B. Hayes2-5. William J. Blot2. Thomas J. Mason2. Linda Williams Pickle2. Pelayo Correa5. Elizabeth T.H. Fontham5 and Janet Schoenberg4 1 Unit of Cancer Epidemiology. Dipanimemo di Scienze Biomediche e Oncologia Umana dell 'Universita. Torino. Italy: -Epidemiology and Biostatistics Program. National Cancer Institute. Bethesda. MD: yLouisiana State University. New Orleans. LA: and *New Jersey State Department of Health. Trenton. NJ. USA. Occupational data from 5 casa-conorol studies in the United States involving 1,973 male cases and 3.210 controls were analyzed to estimate the percentage of lung cancer attribut able to well-known and suspected lung carcinogens. The stud ies were conducted in areas heterogeneous In terms of industrial activities. The percentage of lung cancers attribut able to occupations entailing potential exposure to welt-rec ognized carcinogens ranged, by study area, from 3 to 17%. The further inclusion of occupational groups with suspect car cinogenic exposures changed these estimates very little. Ex clusion of data derived from next-of-kin interviews influenced the estimates of attributable risks, but not in a systematic fashion. The estimates also varied according to ethnic group, smoking status and birth cohort, with higher values in nonwhites. non-smokers and among members of more recent birth cohorts. Possible errors in exposure classification, which may make these estimates conservative, are discussed. tively (OMB. 1972). In Louisiana, the Dictionary of Occu pational Titles (1977) was used. For the other study areas, study-specific job coding schemes were developed. In a recent review. Simonato and Saracci (1983) classified industries/occupations with well-established carcinogenic ex posures (List A) and industries/occupations with suspect car cinogenic exposures (List B). With minor modifications, we have adapted these categories for well-established and suspect lung carcinogens. List A industries or occupations (Table III) are generally associated with specific chemical exposures. List B (Table IV) includes mainly jobs or industrial activities in which an excess of lung cancer has been described in epide miologic studies, but for which the degree and specificity of the associations with specific exposures are uncertain. Al though coding schemes differed across the 5 studies, the in dustry/occupation dictionaries were used in a standard fashion Some occupational exposures are among she known causes of lung cancer, and, under certain circumstances or in specific geographic areas, their contribution to the total lung cancer burden may be substantial. In order to quantify the role of occupation in lung-cancer development and to examine its to identify Lists A and B jobs. The codes were reviewed independently by three of us (P.V.. T.T. and. R.B.H.). A consensus was reached on all job-history categories except for 20 cases and 28 controls, with uncertain List A exposures, who were deleted from the analysis. variability across several geographic areas, we analyzed data from 5 case-control studies of lung cancer conducted in the 1970s and 1980s in the United States (Bloc et ai., (980. 1982 and 1983: Correa et at., 1984; Schoenberg et at.. 1987). Each study collected lifetime occupational histories, as well as in formation on smoking. We estimated the age- and smokingadjusted proportion of lung cancers due to occupations asso ciated with recognized and suspected lung carcinogens. Analysts Adjusted odds ratios (OR) were calculated using Garc's maximum-likelihood method (Martin and Pickle. 1980: Gan. 1970) as measures of the relative risk of lung cancer associated with ever-employment in the selected occupations. The odds ratios were adjusted for age. in 4 groups, birth cohort, in 4 groups, and usual cigarette use. in 5 groups (Table II). In 3 areas (Louisiana. New Jersey and Pennsylvania) it was possi SUBJECTS AND METHODS ble to use 9 categories of smoking, but this did not materially affect the estimates. Ninety-five percent confidence limits were Study designs The designs of the studies, carried out in the states of Louisiana. Florida, Pennsylvania. Virginia and New Jersey, are summarized in Table I. The study areas were originally selected because of high lung-cancer rates and/or because of the presence of certain industries. Subjects with a diagnosis of lung cancer resident in selected geographic areas of the respec computed using Gan's maximum likelihood method (Gant. 1970). Population-attributable risks (PAR), adjusted for age. birth cohort and cigarette use. were estimated using the pro cedure of Whittemore (1983): AR = I - (EtykXjk/nyjk). where n = total number of cases. yk = number of controls in stratum k. xik and yIk = number of unexposed cases and controls in stratum k. ' tive states were identified from existing cancer registries, from co-operating hospitals or from death certificates. Ail the stud RESULTS ies included at least some interviews with next-of-kin of the enrolled subjects. Next-of-kin were the only source of infor mation in Pennsylvania and Vireinta. Response rates for the studies ranged from 63 to 93%. Overall. 2.973 male cases and 3,210 male controls were included in the analyses. Distribu tions of the case and control samples according to demo graphic characteristics and smoking habits are shown in Table Twenty-five per cent of the cases and 20% of the controls had ever held a job included in List A. Among those with List A jobs (Table HI), the highest odds ratios (OR) were for boilermakers (1.8) and for asbestos production workers (2.7). Approximately 30% of both cases and controls had held a job included in List B. Among individual List B occupations, significantly elevated ORs were reported for talc miners, welders and leather workers (Table IV). Choice and coding ofoccupations Information was collected by interview for each job held for 6 months or more, including the place and type of work and the duration of employment. In New Jersey, occupations and industries were coded according to the 1970 Census classifi cation and to the Standard Industrial Classification, respec -`To whom reprint requests should be sent at the National Cancer Insti tute. Executive Plaza North. 4I8-D. Bethesda, MD 20205. USA. Received: May 2^1988. [I......... EXHIBIT............. .. .............................. . HWBUI0010204 852 VSNE1S ET AL. SutcafeJ vstf Study base Controls and exclusions Matching Response rates Number of cases and controls References SOK * TABLE J - STUOY DESIGNS OF THE 5 STUOtES iNCLUOEO m THE ANALYSIS Ltdistm ftnntk W?6-l*79 Pcnmivanci Vifcrew* ?Y76 iWMttl y'-* Hospital-based in 29 parishes Hospital-based and death Death certificates Death certificates certificates in four in three counties n one county counties Cases ascertained through ' hospital records, cancer^ registry and death certifi cates in 6 areas Randomly selected from hospital pa tients--lung disease and tobacco related cancers excluded Randomly selected from: Randomly selected a) hospital recotds and decedents--lung hi death certificates-- disease and lung and psychiatric suicide excluded diseases excluded Randomly selected from death certificates NJ licensed drivers, death certificates Sex. age. hospital, ethnic group Age. county, source of ascertainment Sex. age. county, year of death Ethnic group, age. Age. area of residence, year of death and ethnic group, vital status county 76% leases) 89% (controls) Subject: 760/937 NOK: 294/134 84% (cases) 84% (controls) Subject: 163/290 NOK: 178/166 93% (cases! 82% (controls) Subject: 0/0 NOK: 335/332 89% (cases) 82% (controls) Subject: 0/0 NOK: 347/308 70% (cases) 63% {controls! Subject: 542/666 NOK: 378/377 Correa el at.. 1984 Blot et at.. 1982 Slot era!.. 1983 Bioter at.. 1980 Schoenbere et of.. 1987 Table V reports the population-attributable risks (PAR) by study area for occupations in List A and in List A or B. For jobs included in List A. estimates, by study area, were be tween 3 and 17%. Inclusion of List 8 jobs as exposed had no substantial effect on the PARs. The 95% upper confidence limits of the PARs ranged from 8 to 35%. The lower confi dence limit for each PAR reached 0%. The high variability of the attributable risk among study areas suggests that it may be an oversimplification to produce a single estimate for all 5 study sites combined. seemed to be present (Table V(LI), since higher attributable risks were found in people born in 1930 or later. The occupations showing thehighest ORs among non-smok ers were all asbestos-related, including asbestos production workers, pipefitters, boilermakers and shipyard workers. For these occupations combined, the OR among smokers was 1.4 (l.1-1.9). and among non-smokers 2.4 (0.9-6.8).considering data from personally interviewed subjects only. The corre sponding PARs were 5.6% and 15.9%. Table VI shows the ORs and PARs. for List A and List A or DISCUSSION B according to respondent type. For New Jersey and Florida, both the ORs and PARs were higher for data collected in personal subject interviews, compared to next-of-kin inter views. For Louisiana, the opposite occurred. Race and smoking status influenced the estimates (Table VU). Non-whites tended to have higher ORs and populationattributable risks than did whites. Non-smokers consistently showed higher ORs than did smokers, with the PAR as high as 42% for non-smokers in Pennsylvania. However, the num bers of non-smokers were small. An effect of birth cohort The population-attributable risk is a function of the risk of disease associated with exposure, as here measured by ;--i. odds ratio, and the proportion of exposed subjects in j J population, as here estimated by the proportion of expos---' controls. Differences in PARs among study areas are due to variability in either of these measures. The proportion of controls who ever held any of the occupations in List A was . 14% in Louisiana. 27% in Virginia. 20% is New Jersey. 13% in Florida, and 34% in Pennsylvania. These results primarily reflect differences in industry and occupation patterns between the study areas. In fact, the study areas were initially selected TABLE It - SELECTED CHARACTERISTICS OF THE STUD* POPULATION in part because of their employment patterns. Some further &sx NbibUct 2 Cnnil\ *? differences in occupational patterns may be associated with the study group sources and inclusion criteria. Only the control series in New Jersey was population based. Controls selected Cigarette use1 Never < 1/2 pack/dav 1/2-2 >2 Unknown Birth cohort <1910 1910-1919 1920-1929 1930-r in 198 2.177 435 52 888 1.129 726 230 4- 783 7 376 73 1.748 14 256 2 47 30 950 38 1.193 786 8 28) from hospital patients and among the deceased may be over 24 represented with subjects who have held hazardous occupa 12 54 8 9 tions. This possibility is. however, not supported by available data (McLaughlin et at.. 1985). The insignificant contribution of suspect exposures in List B is probably due. at least in part, to the uncertainties of exposure classification for occupations 30 37 in which the actual carcinogens have not been well char acterized. 24 Variation in the odds ratio for industry/occupation groupings 9 among study areas may be due to a number of factors. The Ethnic group distribution of jobs within occupational groups may vary from White Non-white 2.339 644 78 2.526 91 684 79 area to area. For example, the increased risk associated with 21 List A jobs for Virginia reflects in large part exposures in Interview tvpe shipyards, while for Pennsylvania the excess risk mainly re Subject Ncxt-ol'-kin 1.444 1.529 49 1.893 51 1.317 59 flects exposures in steel production. The other study areas also 41 have unique occupational distributions. Total 2.973 100 3.210 100 Carcinogenic exposures associated with the same industries 'In .1 Mwvlk--. J'tir vwAinp available. and Pciwuylvamal mure Ocwilal ilUhrmUMKi and occupations may be more prevalent or intense in some study areas. Such differences could be due to variation HWBUI0010205 MALE LUNG CANCER AND OCCUPATION IN USA 833 TAMJE in - OCCUPATIONS IN UST A: CASESrOOWTHOCS. ODDS RATIOS AND VS* CONFIDENCE LIMITS Vineyard workers Roofers/asphalt workers Coke plant workers Gas workers Asbestos production workers Insulators Shipyard/dockyard workers Steam fitters Boilermakers Pipefitters Arsenic, iron ore. asbestos. uranium miners BCMEICMME or chromate pigment production workers Mustard gas production workers Copper smelter workers Chromium plating workers Ferrochromiutn production workers Steel production workers Nickel refining workers Ferrous foundry workers Demolition workers Locomotive building and repair workers Automobile brake workers Cactaonuot 45/37 1/2 -- 9/4 15/20 307/254 44/41 26/17 53/51 13/12 -- -- -- I/I -- 181/166 -- 16/10 3/3 21/21 98/90 Ail aafepBga on' _ 1.4 0.7 -- 2.7 0.9 1.4 1.2 1.8 1.2 1.3 -- -- 0.8 1.2 -- 1.2 0.8 1.0 1.2 m% cxj _. (0.9-2.3) (0.02-21.2) -- (0.7-12.0) (0.4-2.0) (1.1-1.7) (0.8-2.0) (0.9-3.7) (0.8-1.8) (0.5-3.3) -- -- (0.02-20.1) -- (I.0-1.6) -- (0.J-3.0) (0.1-5.2) (0.5-2.0) (0.9-1.7) Son-exposed* 1.383/1.662 1.0 - 'Adjusted iwit Midi ration. and smoking.-:No ases or comnji.-'Not included Ls either Lass A ae List B. industrial processes and in the condition of plant facilities. We. for the PAR (Whittemore. 1983). however, had no direct data on this aspect of exposure. In a study of this type, misciassification of exposure may Sampling error could well account for the variability in odds have an important impact on the estimated ORs. Random ratios across the study areas, since confidence intervals for the misciassiftcaeion of exposure would tend (o bias ORs toward GRs for the S study areas overlap considerably. The precision I. i.f.. no effect (Copeland tt al.. 1977). Next-of-kin probably of statistical estimates of population-attributable risks involve do not report as complete or detailed occupational histories as further uncertainty, as statistical variability in the odds ratio are obtained from interviews of the subjects themselves. In and in the estimate of the proportion of the population exposed deed. in these data the number of jobs reported from next-of- must be considered in the calculation of confidence intervals kin interviews was somewhat less, on average, than that re ported in the first-person interviews. Prior comparisons (Pic TABLE IV - OCCUPATIONS IN UST B: CASES/CONTROLS. ODDS RATIOS AND kle er at., 1983) showed that the number ofjobs reported also ___________________ 95% CONFIDENCE LIMITS varied by type of respondent. The mean number of jobs held Ail mfcjects as reported by subjects, subjects" wives, their children, sib Cajg/coootri OR* m% CLt lings and other respondents was 8.6. 6.3. S.4. 4.4 and 4.2. Pesticide applicators Aluminum production workers Beryllium refining workers Smelter (other than copper) workers " Butchers Painters Welders Oilfield workers Laundry and dry cleaning workers Zinc-lead miners Talc miners Truck drivers (other than delivery and routemen) Priming pressmen Machine room workers in printing Rubber workers Leather workers Pesticide production workers Man-made mineral fiber producers Chimney sweeps Cadmium workers Non-e.xposed3 8/6 4/8 8/15 2.1 (0.5-8.5) --0.5 (0.1-1.9) ___ 0.6 (0.2-1.5) 35/50 201/193 148/127 0/4 68/78 I/O 17/8 433/438 0.8 (0.5-1.3) LI (0.9-1.4) _1.3 (1.0-I.8) ___ 1.0 (0.7-1.4) ___ ___ 2.8 (I.0-7.7) 1.1 (0.9-1.3) 15/26 ___ 0.7 (0.3-1.3) ___ 26121 42/29 4/4 ___ -- -- 1.2 (0.7-2.2) t.S (1.0-3.1) 1--.4 (0.3-8.2) _ ,, 1.383/1.662 1.0 -- respectively. Although usual occupation reported by next-ofkin is generally comparable to usual occupation reported by subjects (Blot and McLaughlin. 1983). the shorter lists ofjobs from next-of-kin interviews would reduce the opportunity for categorization as ever-employed in List A or List B jobs. As this under-reporting probably affects cases and controls equally, the net effect would be to lower the OR. Comparing the results for direct first-person interviews with those for next-of-kin interviews in the present study, the ORs were lower for next-of-kin interviews from New Jersey and Florida, while for Louisiana they were higher. The need to attribute occupational exposures from historical job titles is a common limitation of occupational case-control studies, in which little or no information is generally available about on-the-job exposure (Gerin el at.. 1985). Even though a subject worked in a shipyard or steel mill, we cannot be certain whether and to what extent he was exposed to asbestos, poly cyclic aromatic hydrocarbons, and other lung carcinogens. Misciassification could also have occurred due to the use of different coding schemes in the study areas. Although the sensitivity and specificity of exposure classification will vary depending upon the level of detail provided in the occupational 'Adjusted &>f 3|e. birth cofam sad nofc*f.--No e&ses or conuvb.-'Noi in coding scheme, our panel review of the different coding sys cluded in either Leu A or Uu 8. tems used in the studies helped to standardize categorization HWBUI0010206 854 VINES ET AX. TABLE V . CXXVPATTONS HELD IN LIST A AND UST A MB: EXPOSED CASEyCONTSOtS. ODDS RATIOS (0*1 AND attributable risk IN THE POPULATION (PARi toMPWCT Vitpm New kray Randb list A Exposal cases/comrois OR' 95% Cl PAR1 95% upper confidence lima of PAR 173/154 1.2 (0.9-1.51 3.0 8 110/83 1.3 10.9-2.0) to.o 23 list A or B Exposed cases/controls OR' 95% Cl PAR' 95% upper confidence limit of PAR 440/390 t.2 (1.0-1.4) 60 13 192/152 1.3 (0.9-1.8) 9.0 25 `Adjusted bv wnoL'uif status. age. and bulb cohort. 238/211 1.4 (1.1-1.8) 11.0 19 541/566 1.2 <1.0-1.41 9.8 20 86/87 1.4 <1.0-2.1) 10.0 20 176/215 1.1 <0.8-1.5) 6.1 20 144/114 1.4 (0.9-2.1) 17.0 34 222/197 1.4 (0.9-2.0) 17.3 35 TABLE VI - OCCUPATION IN UST A AHO UST A OR 8 BT TYPE OF INTERVIEW. ODDS RATIO (OKI AND ArnUWTABLE RISK IN THE POPULATION (PARS Type uttemewed Mi LtsM Living subjects Cases/comrois OR <95% Cl) PAR Next-of-kin Cases/eonools OR <95% CT) PAR Lin A or B Living subjects Cases/controb OR (95% CD PAR Next-of-kin Cases/controls OR (95% CD PAR 126/138 t.2(0.9-l.6) 3.8 47/16 1.6 (0.8-3.8) 7.7 328/353 U (t.0-1.5) 7.2 112/37 1.4 (0.8-2.4) 10.8 149/137 1.7 (1.2-2.4) 18.7 89/74 1.1(8.7-1.7) 3.1 338/380 1.3 <1.0-1.6) 15.1 203/186 U (0.8-1.5) 4.9 45/61 1.8(1.1-3.3) 18.3 41/26 1.5 (0.7-3.0) 9.8 98/139 16(1.0-2.5) 22.8 78/76 0.8 (0.5-1.3) I TABLE VO - OCCUPATION IN UST A BY ETHNIC CROUP AND SMOKING STATUS. OODS RATIOS <OR> AND ATTRIBUTABLE RISK IN THE POPULATION IRAK) VjfpH Nr/mev Bonds Pentaytoots Ethnic group Noo-wfsjies Cases/comrois OR (95% CD PAR 42/45 1.1 (0.6-1.9) 4.4 27/23 1.4 (0.6-3.6) 15.0 37/32 1.9 (0.9-4.6) 21.9 Whites Cases/comrob OR (95% CD PAR 131/109 1.2 (0.9-1,7) 3.9 83/60 1.3 (0.7-2.1) 8.9 201/179 1.3 (1.0-1.8) 9.1 Smoking slams2 Non-smokers Cases/comrois OR <95% CD PAR 3(29 14 (0.3--5.8) 4.3 8/16 1.8 (0.5-6.5) 19.7 10/52 1.5 (0.6--4.0) 12.3 Smokers Cases/controis OR (95% CD PAR 170/125 1.2 (0.9-1.5) 3.7 102/67 1.3 (0.8-1.9) 8.3 228/159 1.4 (1.0-1.8) 11.6 `Esturafia adjusted by age. birth cohort and smoking.--Estimares adjusted by binh cohort and age. 15/16 1.8 (0.7-5.4) 21-5 71/71 14(0.8-2.3) 10.2 5/19 1.9 (0.5-8.9) 19.7 81/68 1.4 (0.9-2-1) 9.8 -- -- -- -- -- 4/23 2.7 (0.4-24.8) 42.0 140/91 !.4 (0.9-2. D 16.0 V J HWBUI0010207 MALE LUNG CANCER AND OCCUPATION IN USA 835 TABLE VIH - OCCUPATION IN LIST A AND LIST A 0* B BY BIRTH COHORT. ODDS RATIOS (OR! AND ATTRIBUTABLE RISK IN THE POPULATION fPARI Birtf? rafter US A List A or 8 <1930 Exposed cases/comrols OR 95% Cl PAR 1930+ Exposed cases/comrols OR 95* Cl PAR 706/609 1.3 1.1-1.6 10.9 45/40 1.9 1.0-3.7 14.3 1.446/1.398 1.3 1.1-1.4 8.8 125/122 1.8 1.2-2.9 24.2 ofjobs in Lists A and B. Contrary to the effects of random misclassificaiion as de scribed above, response bias, in the form of more accurate or detailed reporting from cases or controls, could either increase or decrease ORs. A major impact of such differential misclassification seems unlikely, since the total number of jobs re ported was similar for cases and controls. Such response bias is more likely when specific chemical exposures, rather chan occupations, are ascertained by interview. Furthermore, the odds ratios associated with List A occupations were generally lower than those reported from other published studies. Low risks for well-known carcinogenic exposures are more consis tent with random misclassification, i.e., an error in exposure assessment evenly distributed among the cases and the con trols. than with differential response bias. PARs for List A occupations were higher for non-smokers than for smokers. This was due to higher ORs for occupation among non-smokers, and not to a higher proportion of occu pationally exposed subjects. Because of the small number of non-smokers among lung-cancer cases, these differences must, however, be interpreted cautiously. A statistical test for inter action between the risk associated with cigarette use and with ever-employment in List A occupations was not significant (p = 0.14). Non-whites had higher attributable risks for List A occupa tions. mainly due to higher ORs. Although the ethnic differ ences were not large, the findings raise the possibility that our exposure categories imply higher occupational exposures for non-whites than for whites. As Table VIII shows, subjects bom in (930 of after had higher ORs and PARs both for List A and List A or B jobs. This could be due to better recall and mote accurate information for this group, or due 10 different exposures introduced in more recent years. Doll and Peto (1981) have suggested that 15% of lung cancers in the United States may be due to occupational expo sure. Our data are consistent with this overall estimate, al though the PARs in this study varied considerably according to geographic area. One advantage ofour estimates is that they are based on empirical data from studies containing informa tion about confounding variables and effect modifiers. in highly industrialized areas of Norway and northern Italy, the population-attributable risk was as high as 30-40% (ICjuus ei at., 1986; Pastorino et al.. 1984). Clearly, such small and homogeneous areas with heavy industrialization may yield PARs which are higher than those found for the entire US or for the diverse locations included in the present analysis. PAR estimates may also be higher from studies which rely on direct exposure assessment, as in the Norwegian and Italian studies, rather than on indirect categorization based on job titles (Vineis er al.. 1986). Evert if our estimate that 3 to 17% of lung cancers in the study areas are related to exposure to recognized.occupational carcinogens is conservative, the number of lung cancers in duced by occupational exposures would be large, since lung cancer is the most common cancer in the United States, with nearly 100,000 new cases expected among males in 1987 (Silverberg and Lubera. 1987). The estimates of 3 to 17% refer only to well-known carcinogenic exposures. Estimating the contribution of other, still suspected or unknown expo sures. is more uncertain. Additional associations could only add to the calculated population-attributable risk. ACKNOWLEDGEMENTS Our analyses were made possible by a grant to P. Vineis as an Exchange Scientist under the US-Italy Cancer Program, and by a fellowship awarded to him by the Associazione Italians per la Ricerca sul Cancro. We thank Ms. P.A. Stewart for her participation in job title coding and category assess ment. Dr. R. Hoover and Dr. A. Blair for thoughtful advice and Ms. G. Schoenfeld for assistance in manuscript pre paration. REFERENCES Burr. WJ.. Davies. J.E.. Morris Brown. L.. Norowau, C.W.. Buiatti. E.. No. A., and Fraumkm. J.F.. Jr.. Occupation and the high risk of lung cancer in Northeast Honda. Cancer. 50.364-37) (1982). Doll. R.. and Peto. ft.. The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J. not. Cancer Inst.. 66. 1192-1308(1981). Burr, WJ.. and McLaughun. J.K.. Practical issues in the design and GaRT. JJ.. Point and interval estimation of the common odds ratio in the conduct of case-control studies: use of next-of-kin interviews. In; WJ. combination of 2x2 tables with fixed marginals. Biometrika. SI, 471-475 Blot. T. Hiiayama and D.G. Hod fcds.l. Statistical methods in cancer (1970). epidemiology. Radiation Effects Research Foundation. Hiroshima f 1985). Gerin. M.. Siemiattcki. J.. Kemper. H.. and Begin. D,, Obtaining Burr. W.J.. Morris Brown. [__Potters. L..VL. Stone. BJ_. and occupational exposure histories in epidemiologic case-control studies. J. FuuuEHt. J.F.. Jr.. Lung cancer among long-term steel workers. Amer. I. Epidemiol.. 117. 706-716 f 1983). Occup. Med.. 27.420426 <19851. Kjuus. H., Skjaevern. R-. Langard. S.. Lien. J.T.. and Aamoot. T,. a Blot. W.J.. Morris. L.E.. Stkoube. R.. Tacnon. [.. and Fraumeni. case-rererent study of (tmg cancer, occupational exposures and smoking. J. F-. Jr.. Lung and laryngeal cancers in relation to shipyard employment ScwuL /. Work Environ. Health, 12, 193-202 (1986). in coastal Virginia. J. nut. Cancer Inst.. 65. 571 -575 11980). Martin. E.J.. and Pickle. L.W.. Odds ratio estimation for potychoto- Copeland. K.T.. Checkowav. H.. Mc.Michael. Al. and Holbrook. tnous data. Proceedings of the Sth Animal Conference of the SIS Users' R.H.. Bias due to misctassiltcaiion in the estimation of relative risk. Amer. Group International, pp. 145-147. 5AS Institute. Cary. NC (1980). J. Epidemiol.. 105. 488-495 (1977). McLaughun. J.H.. Burr. WJ.. Mehl. E.S.. and Mandel J.S.. Prob Correa. P.. Williams Pickle. I__Fontham. E.. Delager. N.. Un. Y.. lems in the use of dead controls in case-control studies. 1. General results. HaEwsxbl. N.. and Johnson, W.D.. The causes of lung cancer in Louisi Amer. 1. Epidemiol.. 121. 131-139(1985). ana. In; M. Mizcli and P. Cornea leds.1. Lunft cancer causes and preven tion. pp. 81-89. Vertag Cbemic. Weinbeim f 19841. DicnoNARV of Occupational Titles. 4th Ed. US Department of Labor. Washington. DC (1977). OMB: Standard Industrial Classification Manual. Office of Management and Budget. Washington. DC < 1972). Pastorino. U-. Bexrino. F.. Gekvasio. A.. Pesexti. V.. Rtaou. E.. nod HWBUI0010208 856 VTKEXS ET Crosignanl R. Proportion of lung cancer due to occupational e*po*ure. !nt. X Cancer. 33. Ui-237 <IW). Pickle. L.W.. Brown. L.M., 2nd Biot. W.J.. Information available from surrogate respondents ia cas6cootn>( interview studies. Amer. J. Epidt' miaC 118. 99-108 (1983). SCHCE-VSESG. J.B.. Stcmkacen. A.. Mason. TJ.. Patterson. J.. 8hx. T.. and Altman. R.. Occupation and lung cancer risk among New Jersey white males. X nor. Gwcer Inst.. 79, 13-2211987). SiLVERBERC. .,, and Lu&Ra. Cancer statistics- C4.37. 2-19 (1987). Simonato. L., and Saracci. R.. Cancer, occupational. Encyclopedia of occupational health and safety, 3rd rev., pp. 369-375. Geneva. |)L0 (1983). Vinos. P.. Fletcher. A., and Simonato. L.. Estimates of (he proportion of Ittrtg cancer attributable to occupation. In: Sth Intenumoncl Symposium: epidemiology in occupational health. UCLA. Los Angelo (1986). WnneM0, A.S., Estimating attributable risk from case-control studies. Amer. J. Epidemiol. 117, 76-85 (1983). !ii ARTICLE IS FOR INOVIOUAL USE ONLY AND MAY NOT 8E FURTHBt 8EPRGDUCE0OR STOREDELECTRONICALLY WITHOUT Y.-Pi"TENPERMISSION FROM THE CCPVR.GHT HOLDER UNAUTHORIZED REPRODUCTION MAY RESUIT INIIWANClAi AND OTHER PENAL TIES. HWBUI0010209