Document M4dmKLQ3oJ16aoMyZReQNLKGM

I-.* t A Case-Control Study of Lung Cancer and Diesel Exhaust Exposure in Railroad W ~ r k e r ~ ' - ~ . ERIC GARSHICK, MARC B. SCHENKER, ALVARO MUNOZ, MARK SEGAL, THOMAS J. SMITH, r SUSAN R. WOSKIE, S. KATHARINE HAMMOND, and FRANK E. SPEIZER ii * Introduction f A variety of mutagenic and carcinogenic substances have been identified on SUMMARY A cuecontrol rtudy of db.th8 among U.S. ~IlfOadWrlef8WU c0ndUct.dto the hypothesisthat lungcancerI8.uotl.tsdwith exxpoaunto dl..sl exh8u.1. Employ& N k~ r l wdrt h 2 10yrof .micewho m n b o m o n0r.ft.r & n W y 1,1900Md W h o d l d k h s , particles from diesel exhaust emissions Yemh 1,1981 and Fobruacy 28,1982mndlglbk. We co1l.ct.d 8796 of the &8th -b (1). Occupational groups with potential 15,059 death8 reportedto the US. Railroad Retlment h r d (RRB). Cam of lung cn#r0- diesel exhaust exposure include under- w n matched to Control8 by age (f 2.5 yr) and data of death (f 31 dy.). p O t . n t k l t~o ground miners (2), bus maintenance d i m 1 exhaust was adgned b n d on an lndustdal hyglem ev8luatlon of and mrr(r t -i workers (3,4), heavy construction equip- Each subjecf8 work hl8tory wa8 determinedfrom a yearly lob nport flW hy h b mpkyw w 11 ment operators (5). truck drivers (6),and railroad workers (7-11). In a recent study RRB h1959 until && 01 "Oment. OXpo8Un to 1959 WU job hold In 1959, the end of the dram koinotlw era, or hy the I n t lob heldIfdnnnnt (12), diesel exhaust inhaled at high concentrations in rats resulted in lung cancer. There is concern that diesel exhaust before 1959. Smoking hl.torkr wem obtalnod by questlonnaln hnext of kln. u8lng conditional ioglstlc m g m l o n mabl8 to ad/unt tor smoking and arb.ato8 expown. #rcn yr of age 01younger at the U r n of death with work In 8 diesel exhaunt b x p o d job Ibr 20 p a JgnHhntly I n c m d nWvo odds (odd8 ntlo = 1.41,959b CI = 1.06,l.M) of lung~11101:~0 exposure in occupational groups and in effoct of d i m 1 exhaust expourn m a soen Inworimcsn65 yr of .(IO or older bocauw fmnyat- the general population may be a risk fac- nrmntl~d.hoctly~b.nrltlontod~locomot~Thirrtudyw~( tor for lung cancer, but this has not yet that occupational expoaun to dl& exhaust lncmases lung cancer risk. been demonstrated. AN REV RESPtR Ms lssr: l3Mut-m Previous studies of lung cancer and dieselexhaust exposurehave been limited by several deficiencies including inade- quate-statistical power, inaccurate or absent classification of diesel exhaust exposure, and failure to consider when diesel exhaust exposure began (8, 13). Possible confounding factorssuch as as- bestos exposure and cigarette smoking also have not been considered in some studies. We have completed a case- Reports of deaths that occurred between March 1, 1981 and February 28, 1982 were collected from the RRB during the course of the year. Death certificateswerereceived from the RRB for approximately three fourths of the deaths, and the remainder of the death certificates were requested from the appropriate state boards of health. Cause of death was coded according to the International Diesel Exhaust EXposure Esrimma There are more than 150threedigit Iutustue Commerce Commission (ICC) railroadjob codes (14). Job codes for each worker ryrrt availablefrom the RRB starting in 1959and endingwith death or retirement. For workers who retired between 1955and 1959, last road job worked was available Somerat- control study of lung cancer deaths in Classification of Diseases, eighth revision, (Received in originalform August 28, IS6rrlb Elation to dieselexhaust exposure occur- (ICD-8). rrvised form November 20. 1986) ring in U.s. railroad workers that con- The cases were selected from deaths with sidered such factors. primary lung cancer (ICD 162). which was From the Channing Laboratory. Dcp.rtlwr the underlying cause of death in most cases. of Medicine, Brigham and Women's HorpaIk hch-TPopulation The population base for this study was the approximately650,000activeand retired male U. S. railroad workers with 10 yr or more of railroad servicewho were born in 1900or af- ter. The U. S. Railroad Retirement Board (RRB)operatestheretirrment system for these I workers; it is separate from the Social Security System. 'lb qualify for either retirement or survivor benefits, 10yr or more of service must be acquired by the worker. For the next of kin to qualify for benefits, the railroad matched to 2controls who mre within 2. yr of the date of birth and within 31 days of the date of death noted in the lung cancer cases.The controlswere selected ran- domly from workers who did not have cancer noted anywhereon their death certificate, and who did not die of suicideor of accidental or unknown causes. After informingthe certifyingphysician of our intentions, a questionnaire was sent to the deceased's next of kin that includedquestions about the deceased's smoking habits, place of residence, and duration and type of relationship the respondent had with the de- ton; the Environmental Health sdencc Department of Family and CommunityMedLior University of Massachusetts Medical Schod. Worcester, Massachusetts; the O c c u p t h d rad Environmental Health Unit, D e p a r U l l c n t d ~ nal Medicine, University of Califo- Dnb. California;and the Department of EpidcmiobO: Johns Hopkins School of Hygiene and publir Health, Baltimore, Maryland. Supportedin part by GrantNo. R807515Ira the Environmental Protection Agency, IldividsJ Research Award No. ES45157 from the "mJ Institute of Environmental Health Sciclra, ad InstitutionalResearch Award NO.HLmafF the National Heart, Lung and Blood-1 ' Presented in part at the Annual Meamlof* board must be notified of the railroad work- ceased. A second questionnaire was sent if American Thoracic Society, Miami, M V 19% er's death. It is estimated that less than 2% there was no response, and no response to ei- Requests for reprints should be addrarcd1o of eligible retirees and next of kin do not re- ther questionnaireresulted in a phone call to Dr. Eric Garshick,Channing Laborat~fY1,80- i ceive benefits. the next of kin. wood Avc, Boston, MA 02115. 1242 CASECONTROL STUDY OF LUNG CANCER AND DlESEL E X H M EXPOSURE 1243 had additionaljob codesreported between road surveyed did not have a passenger ser- Analysis 1955 and 1959. Most diesel exhaust exposure vice, but the job codes including train atten- To preserve the matching on age and date of in the U.S. railroad industry began when the dantsand passenger car waitersand chefswere death, analysis was done using multivariate industry changed from steam- to diesel- considered to be exposed to diesel exhaust conditionallogisticregression(24). Additional powered locomotives after World War 11. This based on their work on operating trains. regression analysesweredone for various sub- msition occurred primarily during the Years of exposureto diesel exhaust were to- groupswith unconditional(ie, matching not 1950s; and by 1959,9570 of the locomotives taled from 1959to death or retirement for each preserved)logistic regression but adjustingfor in the United Stateswere diesel powered (IS). worker, and summarized as "diesel-years." age. Logistic regression specifies that the nat- n e approximatemidpoint for the transition Years not working for the railroad, or with- ural logarithmof the odds of disease depends todieselpower in the United States was 1952. out reported jobs or unknown jobs, were con- on a set of variables; in this study, diesel ex- We have chosen 1959 as the effectivestart of sidered as unexposed years (2.4% of the total haust exposure,asbestos exposure, and ciga- diesel exhaust exposure for this study, recog- years from 1959 to death or retirement). rette smoking.The relative odds for lung can- nizingthat some workers had additional ear- As a result of the transition from steam to cer attributable to each exposure is derived lier years of exposure to diesel exhaust. If a diesel locomotives in the 1950s. age and the from the exponent of the regression coeffi- -worker retired before 1959, he was considered year a worker retired relative to 1959 in- cient for each exposureparameter. Regression unexposed. fluenced the chance of working in a diesel- coefficientsand their standard deviationswere An industrialhygiene surwyof the U.S. rail- exhaust-exposed job. Workers w&o retired calculated using a Fortran program provided road industry was conducted as part of this prior to death in 1981or 1982were less likely by Thomas (25). All p values presented are invatigation. Selected jobs with and without to have worked in a diesel-exhaust-exposed two-sided. regular diesel exhaust exposure were identi- job (table2). Roughly half of the older work- fied by a review of job title and duties, and amfinnedby industrial hygiene sampling(16, 17). Personalexposurewas assessed in 39job categories representativeof workers with and without dieselexhaust exposure in 4 railroads. ers were unexposed to diesel exhaust, and those exposed had a short duration of exposure For this reason, and because of the suggestion that older workers might have less accuratecertificationof death (18). the cases Results Cause-specificdeath certificates were obtained for approximately 87% of the 15,059 deaths reported by the RRB, and Workers with regular diesel exhaust exposure and accompanyingcontrolswere analyzed in we identified 1,374 cases of lung cancer. werc found in severaljob locations: jobs in- 2 groups, workers 4 64 yr of age (i.e,retire- Twenty cases of lung cancer were ex- side the roundhouse and diesel locomotive ment age) at death and those 2 65 yr of age cluded because of incomplete data, and repair facilities had exposure during the re- at death. 35 were excluded because 2 states would pair and maintenanceof diesel locomotives; locomotiveengineers and firemen who oper- atcd the diesel locomotives; and conductors and brakemen who worked on or near runnhg trains. Hostlershad exposurewhile driv- 4locomotivesin and out of shops. We found thattheworkers wereexposed to other respira- Asbestos Exposure Estimates Asbestos exposure in the railroads occurred primarily during the steam engine era. Ex- posure was related mostly to the repair of locomotive steam boilers, which were insulated with asbestos. Otherpotential exposures not provide permission to use informa- tion on the death certificate to contact the next of kin. Of the remaining cases, <335 (%To) of the 349 subjects 64 yr of age at death and 921 (95%) of 970subjects 2 65 yr of age at death were suc- ble partidate, including cigarette particulate, were in the repair of asbestos-insulatedpas- cessfullymatched to at least 1control and sueswhich required adjustment of the measured senger cars, in the railroad bridge and build- had work histories available for analy- of respirable particulateobtained (16. ing construction trades, in steam pipe repair, sis. Ninety percent of the cases in both 17). Mean estimates of current exposure songthe dieselexposedgroups adjusted for Ckamtcsmoke ranged from 83 to 168 &m'. For the shopworkers, mean estimates in the SviduaI railroadsvaried between 87 and 322 Wm'. In contrast, the clerks and station wts(nonexposed)had values between 9and 76 Clg/m'. Detailed discussionof the interpre- and in the repair of stationaryboilers. We developed asbestos exposure categories based on the resultsof a survey of current employees of onerailroadthat askedabout prior asbestos exposure (19) and on a review of railroad, medical, and industrialLiterature(20-23). As- bestos exposure category was based on job held in 1959(or before if available) or on the age groups had 2 controls. Of the 2,385 controls in the study, 98% were matched to cases within f 31 days of the date of death, and the remaining 2% were matched within 100 days. Eighteen con- trols who died within 10 days before the end of February 1981 were included in tation and use of these values and other indi- lastjob held if the subject retired before 1959. the study. Ninety-five percent of the con- -ofdiesel exhaustexposurewillbe presented As a result of the transition to diesel-powered trols were matched within 2.5 yr of the future work. In this study, diesel exhaust engines,subjectsstarting work in or after 1959 dateof birth of the case, and the remain- apoS~ries considered to be a dichotomous W n o ) variable Job codes for which no personal sampling performed were considered exposed or lfnarposedbased on similarities in job activiWsand work locationsto sampled jobs, and d a of contact with operating diesel *=equipment. For example, carmen (car mainworkers)generally worked away from would not have had opportunity for significant asbestos exposure and were considered unexposed regardless of job code Workers categorized with likely regular past asbestos exposure were subjects who had worked in and around boiler and locomotive repair facilities. Exposureto uniformly high levels of asbestos would have occurred in a subset of shop workers responsible for boiler lagging ing 5% were matched within 6 yr. Initial date of railroad service was estimated from the RRB reported years of service for most workers and was only reliable since the late 1930s because of the limi- tations of computerized RRB records. Some workers starting work in 1947 or later had firstyear of railroad services u p O-ting trains and were classified as unex- removal; others in the shop area would have plied by the RRB. Thus, the total num- to diesel exhaust. Maintenance-of-way had less exposure Workers responsible for the ber of years worked is underestimated worken and others with potential or occa- ''Om diesel exhaust exposure were considnot to be significantlyexposed. The ap- hhteness of these classificationswas veriaquestionnairesurvey of 534 workers Survey railroad. Workers indicated ap- *-&ly how much time was spent in sev- y e a d locations(Le, near or away from ms,trucks, or equipment).The rail- maintenance of railroad structures and car repair had more intermittent asbestos exposure but were also categorized as exposed. Among activeworkers older than 50yr of age in the surveyed railroad (19), past asbestosexposure was short, a median of 3 yr.' A detailed list of ICCjobcoda and diesel exhaust and asbestos exposure assignmentsarc available from the authors upon q u a t . and inaccurate for the older group, and could be slightly underestimated for the younger group. Although we lacked specific information on jobs worked outside the railroads, on the basis of the retire- ment yea^ and initial years of railroad service, most workers in this study appeared to be lifelong railroad employees. 1 1244 There were no significant differences between cases and controlsin each of the 2 age groups for age, race, year first worked, retirement year, and total years worked, only available for the younger group (table 1). Of the controls in the younger age group, 80% had disease of the circulatorysystem (ICD 390-458)as the underlying cause of death, 7% died of disease of the respiratorysystem (ICD 460-519).and 7% died of disease of the digestive system (ICD 520-577). In the older age group, 74% of the controlshad a disease of the circulatory system as the underlying cause of death, 15% died of diseaseof the respiratorysystem, and 4% died of digestive diseases. In the 2 age groups, only 5 of the controls(0.2%) were coded to illdefined causes (ICD7%). For the younger age group, the proportion of cases with asbestos exposure (25%) was slightly, but not significantly greater than in the controls(22%). In the older group, slightly more controls had asbestos exposure than did cases. Approximately 60% of all the younger workers had some exposureto diesel exhaust. In contrast, only 47% of all the older workers had some diesel exhaust exposurein their jobs, p <0.001(table 2). Smokinginformation (table 3) was obtained from next of kin. Ninety-two percent of the questionnaire or phone respondents for the younger group were female, and in most (83%)they were the spouse In the older group, 88% of the respondents were female and the spouse in 76%. Information on smoking (yes/ no) was missing for 17% of the cases and 16% of controls in the younger age group, and 13Vo of the cases and 18% of the controls in the older group. Most smokers were long-term smokers who had continued to smoke within 5 yr of their deaths. More cases than controls were cigarettesmokersand had a greater pack-year history of smoking. When age started smokingwas not obtained (<6% of both cases and controls), we assumed a Starting age of 16. Diesel exhaust exposure (table 2) was summarized as the number of years in dieselexposed jobs. Years of diesel exposurewas used asa continuousexposure variable in regression models. In the younger agegroup, the unadjusted coefficient for diesel-years was 0.01648,with an odds ratio of diseaseover20yr of 1.39 (95% C1 = 1.05,1.83). When adjusted for any asbestos exposure (yeslno) and lifetime smoking (pack-years), the estimate of the effect of diesel exhaust ex- posure remained close to the crude esti- TABLE 1 AGE, RACE, W O R K HISTORY SUMMARY. AND ASBESTOS EXPOSURE a)CATEGORIES FOR BOTH YOUNGER (< 64) AND OLDER (a CASES AND THEIR CONTROLS Younger Workers CaSe<64 Older Workers CaSe.85 CaSe Control CaSe Control Number Age. Median Range Race, white 335 80 35-84 310 (93)' 637 60 37-69 (92) 921 73 65-82 837 (91) 1748 73 62-82 1576 (91) Started work 3 1949 < 1949 Years worked mean t SO Retirement year median Asbestos exposure Regular intermittent None 109 (33) 226 (W 29.2 3 7.4 1980 47 (14) 37 (11) 251 (75) 164 (29) 453 (71) 29.4 i 7.0 1979 78 (12) 63 (10) 496 (78) 68 (7) 853(93) NAt 1972 129 (14) 108 (12) WV4) 126 0 1622(93) NAt 1971 289 (17) 193 (11) 1286(72) * Numbam in pllnnthesg am perwnmgce. imt ~ oavtail- f o w~orl~arastarting beton lato 1- mate, with a relative odds of 1.41 (95% ratio (OR = 1.64,95% CI = 1.18,2.29)). CI = 1.06,1.88) (table 4). In the older Subjects with 5 to 19diesel-yearshad an age group, the unadjusted coefficient for odds ratio of 1.02(95% CI = 0.72,1.45). diesel-yearswas 0.00084 (p = 0.89),with Only 3% of the older workers had the adjusted value (table 4) also not sig- diesel exhaust exposure for 20yr or moR nificant and near the null value For comparison, the odds ratio for 5 to Toverify that risk in the younger group increased with increasing years of diesel 19 yr 0.95 of exposure (95% CI = in the older 0.79, 1.13) agnrodufpowr a>s exposure, diesel-yearswas categorized as 20 yr of exposure was 0.94 (95% CI = 320 diesel-ytars, and 5 to 19die~el-year~, 0.56, 1.59), adjusting for smoking and with 0 to 4 yr as the referent group. Ad- asbestos exposure justing for asbestos exposure and pack- Because workers in the diesel engine year category of smoking, subjectswith repair shop were the dieselexposed work- 3 20 yr of diesel exhaust exposure had ers with the most potential for past as- the highest, significantly elevated odds bestos exposure, a regression model in the TABLE 2 DISTRIBUTION OF YEARS OF DIESEL EXHAUST EXPOSURE IN THE THREE W W S E D JOB LOCATIONS AND SUMMARY CAlEGORY (DIESEL-YEAAS) Younger workers older worken P Years of WC64 CaSe.65 Exposure case control ca$e control No Exposure Shopworken Engineer and f i r e m Brakemen M d conductors DieseCyears 0 1-10 11-19 .M 1-10 11-19 .2u 1-10 11-19 320 1-10 11-19 320 lM(36)' 39 (12) 15 (5) 21 (6) 7 (2) 13 (4) 35 (10) 11 (3) 27 (6) 58 (17) 45 (13) 53(16) 1 1 7. M. 272(43) 53 (8) 38(6) 24 (4) 14 (2) 31 (5) 40 (6) 36 (6) 58 (9) 96 (15) 77(12) 12?(19) ISS(26) seO(52) 89 (10) 620 6 (1) 27 (3) 83 (9) 5 (0) 56(6) 114 (12) 14 (2) 155 (17) 260(28) B(3) W(S) 148 (8) 166 (10) 6 (0) 48 (3) 132 (8) 22 (1) 119 (7) 183 (11) 25 (1) 275 (18) as(=) 54(3) 1246 and death 5 yr later. For workers with 5 to 14 yr of cumulative exposure, the odds ratio was 1.07 (95% CI = 0.69, 1.66). with a referent category 0 to 4 diesel-years. Discussion We have conducted a population-based case-controlstudyof U. S.railroad work- ers to test the hypothesis that diesel exhaust exposure results in an increased risk of lung cancer. Because diesel-powered locomotives were not fully introduced into the railroadsuntil the late 195Os, only in a group who worked sufficiently long after that time could an effect of diesel exhaust exposurebe expected to be seen. For cases greater than or equal to 65 yr of age at death (in 1981 and 1982) and their matched controls (Le., retired workers), no effect of diesel exhaust exposure was seen. In the younger age group (d 64 yr of age at death), there was both an increased prevalence and duration of diesel exhaust exposure. Controllingfor cigarettesmoking and for potential past asbestosexposure, a significantiyelevated odds ratio for lung cancer was observed. Workers with 20yr of more of exposure had the highest, significant odds ratio when exposure was analyzed with categorical rather than withcontinuous variables. The alternative regression model using the categorical variables of 5 to 19 and 2 20 yr of diesel exposure suggests an apparent nonlinear relation between number of years of exposure and lung cancer risk. However, the confidencelimits are wid%and it is not possible to definethe preciseshape of this relationship. Asbestos exposure, a possible confoundingvariable, was in the final model because of its biologic importance. We were not ableto individuallyidentifythe small group of workers with past heavy asbestos exposure. In addition, some of the ICC job codes used to assign potential asbestos exposure were broadly defined and contained some workers without asbestos exposure. To estimate stability in the asbestos exposurecategories over time, we examined job mobility during the years between 1959 and 1979 (although exposure occurred before 1959). Of the workers less than or equal to 44 yr of age in 1959and stillworking in 1979 (approximately 50% had retired), 72% who would have been categorized in the regular asbestos exposure group 20 yr later were in the regular group in 1959, and 17% were in the intermittent asbestos exposure category in 1959. Of workers categorized as unexposed to asbestos, 96% would have been in the unexposed category 20 yr earlier. Of workers in the intermittent category, 20 yr earlier 67% would have also been in the intermittent categoryand 21% in the regular category. Thosewith any exposure(regularand in- termittent), 20 yr earlier 88% would have been in an exposed category. The misclassification from assigning past as- bestos exposure based on the 1959 job would be expected to have an effect apparent more readily in the older age group, and would result in a reduction of the apparent asbestos effect. In the younger group, there would be less misclassification but also less real exposure. Despite the asbestosexposuremisclassification,the effect of asbestosexposure on lung cancer in the younger age group is within an expected range. These younger railroad workers, who were in their thirties during the 195Os, if exposed to asbestos would have had relatively short exposure Studies of lung cancer among former shipyard workers with mostly wartime exposure have demonstrated oddsratios for asbestos exposureof similar magnitude as found in this study (26-28). In a study in Swedish railroad shop workers with more than 30 yr service, the standardized mortality ratio was 192 (29). In our study, workers located in the diesel repair shop were most likely to have had past asbestos exposure. Workers in the other diesel exhaust jobs exposure did not iiiclude workers with significant potential for prior asbestos exposure. Only the hostlers had the potential for asbestos exposure, and they were a small subset (10 workers in the younger group in 1959) of the remaining dieselexhaustexposed workers. With the shop workers excluded from the other diesel-exposed workers, a significantly elevated odds ratio for the effect of diesel exhaust was still obtained. Next of kin were used to obtain inform a t i o h n cigarettesmoking, and the estimates for the smoking effect on lung cancer arewithin the expected range (30). Rogot and Reid (31) found that surrogates identified 91% of ex-smokers or nonsmokers compared to previously obtained personal questionnaire response. There was a tendency for the surrogate to overestimate the amount smoked. Koloneland coworkers (32)compared the response of subjects (mostly husbands) with wives interviewed separately.Ninetysix percent of those identified as ever smoking were in agreement. Age started smoking was in agreement 84% of the time to within f 4 yr. These surrogates also tended to overestimate daily ciga- rette consumption. Thus, surrogate re- sponse data appear to overestimatecigarette consumption. This could exaggerate the contributionof cigarette smoking to lung cancer risk if the next of kin of subjects dying of lung cancer were more likely to report smoking histories than were those of controls. Seveml models relating cigarettesmok- ing to lung cancer have been described. Doll and Pet0 (33)in their study of Brit- ish doctors expressed the dose-response relationship as the product of a Power of duration of smoking and a power of dose for lifelong smokerssmoking las than 2 packs per day. Others have used estimated pack-years (34) and cigarettes per day (26). We used pack-years ex- pressed as a categoricalvariable to con- trol for the effect of cigarette smoking, The category of pack-years "missing" Served as an additional level of cigarette smoking. Further regression analyses in the subgroup of workers with pack-year data available showed no advantageover any other modeling of cigarette smoking in this population. Another potential source of error in our study is the misclassification of diesel exhaust exposureusing the ICCjob classification as a surrogate for exposure Jobs with potential diesel exhaust exposure included railroad express drivers, railroad police, bus drivers, and operators of diesel powered equipment. Such exposure was likely to be low intensity and intermittent,and as a result thesepersons were considered unexposed. Misclassificationby assigning exposureto an ICCcodejob group that contained non- dieselexposedworkers (asin someof the dieselshop codes) would reduce the likelihood of seeing a significant effect of diesel exhaust exposure. ?krotypes of misclassificationmay n- sult from the use of death certificatesto identify cases and controls in this study. Controls not suspected of having lung cancer may actually have had undiag- nosed pulmonary malignancies, and identified lung cancer cases might be other lesions misdiagnosed as primary lungcancer. It is unlikely that many case or controlswere misclassified for sevem: reasons. First, we excluded as eligible controls subjects with death caused b! accidents, by other cancers, or by uncer tain causes. Second, lung cancer, partic ularly in younger men, is usually a clini cally obvious disease that is rapidly fa tal. It was in this group of yoUW workers that a significanteffect of dies' 1245 TABLE 3 than or equal to 50 pack-years of smok- SMOKING HISTORY FOR CASES AND CONTROLS IN THE ing in the younger group. Analysis using YOUNGER AND OLDER AGE GROUPS unconditional logistic regression on the Younger Workers C-<64 Older Workers Case366 subset of younger workers (n = 758) with known pack-years was done to explore case Control Case Control the impact of other smoking models in the magnitude of dieselexhaust lungcan- Smoker YeS No Missing Pack-years *so cer risk. Adjusting for age and asbestos exposure, there was no significant difference in lungcancer risk among the different models that included pack-years expressed as a single continuous variable, >50 2 independent variables, cigarettes per Missing Nonsmoker day and years of smoking, or pack-yearcategory categorized by the number of Yeam smoked. 3 20 years the subject had stopped smoking C i e w per day prior to death. Adjusting for age and as- 1-15 16-25 26-35 >36 Years stopped smoking before death 0-4 5-14 bestos exposure, with diesel years used as a continuousvariablein these models, the odds ratio for 20 yr of diesel exposure ranged from 1.50 to 1.53 (p < 0.02). Ex- pressed as a continuous variable, the coefficient for pack-years was 0.01278 (p < 0.001), and used together the coeffi- 3 15 cient for cigarettes smoked per day was 0.01604 (p = 0.006) and for years of smoking was 0.02484 (p = 0.001). Smok- muagergroup was constructed,entering of exposure based on estimated years ing 4 50 pack-years 4 yr or less-before the diesel exposure of the shop workers v e l y from the other dieselexposed worked prior to 1959. None of these alternatives gave a stronger association death had a risk of lung cancer of 4.08 (p < 0.001), smoking > 50 pack-years 4 lob categories. When adjusted for past abestos exposureand smokingas in ear- with risk than the dichotomous variable used. yr or less before death had a lung cancer risk of 6.08 (p < 0.001). smoking 4 50 t kr models, the odds ratio for 3 20 yr Because of the incompleteness of the pack-years5 yr or more beforedeath had of arposure(exclusiveof the shop work-)was 1.55 (95% CI = 1.09,2.21, with next of kin responses, we were not able to calculate pack-years smoked for all a risk of 2.47 (p = 0.027), and smoking >SOpack-years 5 yr or morebefore death 0 to 4 yr used as the referent category). those in the study. To preserve the had an odds ratio of 3.93 (p = 0.004). Alternative models were explored to matched nature of the study and mini- In the entire younger data set (n = 972), krtet explain past asbestos exposure mize exclusions, we treated "pack-years mining the categoryof pack-yearsmiss- They included separate dichotomous missing" as a category of smoking. This ing but categorizing the remaining sub- Variables for the regular and intermittent gave an odds ratio between the value for jects with known smoking information npoSUngroups and an estimateof years subjects with greater than 50 and less as ex-smokers (quitting 5 yr or more be- fore death) and as current smokers in TABLE 4 REGRESSION RESULTS USING DIESEL EXHAUST EXPOSURE AS A SINGLE CONTINUOUS VARIABLE (DIESEL-YEARS) ADJUSTED FOR CIGARETTE SMOKING AND ASBESTOS EXPOSURE -re c.1.gocv odds COeffiCiWlI Ratio c-aga*w --Y-m -, y m c 50 pack-year8t >50p.ck*t P.ck-yOU8 mimingt -qp>8!5 kkator. YIN < 50 p.ck-ylm4t > 50 pack-ymrat 0.01719 0.18111 1.19lgB 1.73606 1.37975 -0.00161 -0.01807 1.47W 2.21321 p.cky.vr miaingt 1.35378 ;- an th.b.r* ol20 Vol urpaun. 1.41' 1.20 3.29 5.88 3.97 0.91* 0.98 4.38 9.14 3.87 95% CI 1.06,1.88 0.873.65 1.57.6.93 2.73.11.80 1.86.8.51 0.71.1.17 0.81.1.20 2.90,6.60 8.11,13.70 2.58.5.84 p values 0.02 0.27 < 0.01 < 0.01 < 0.01 0.47 0.86 < 0.01 < 0.01 < 0.01 daily cigarette consumption categories, adjusting for asbestosexposure, gave an odds ratio for 20 yr of diesel exposure of 1.46 (p = 0.01). The interaction between diesel exposure and the 3 pack-year categorieswas assessed by constructing3 cross-product terms between the categories< 50, >50, and missingpack-year indicatorvariables and the indicator variable for 320 diesel yr. The cross-product terms were not sig- nificant (Chi-square = 2.05, df = 3, p = 0.56). To exclude the effect of recent diesel exhaustexposure, a model was tested that excluded exposure occurring with the 5 yr before death. The regression coefficient, adjusted for cigarettesmoking and asbestos exposure, yielded an odds ratio of 1.43 (95% CI = 1.06, 1.94) for work- ~cuoaY~~p.sk~~~). ers with 15 yr of cumulative exposure, 1247 exhaust was demonstrated. In a study ing active workers, subjects with the References b& on the Third National Cancer Sur- highest likelihood of past asbestos ex- 1. National Research Council. Health effects of vey wheregreater than 90% of thecancers were histologicallyproved, the death certificate for lung cancer was coded appropriatelyin 95% of cases (35). Because there would be no reason for physicians to link job category with death-certifi- posure were studied, and w xkers likely to have the longest duration of diesel exhaust exposurewere excluded. Although these investigators acknowledged some of these limitations,they found a significant trend for lung cancer for workers exposure to dieselexhaust. Washington, D.C.: National Academy Press, 1981. 2. Reger R, Hancock J, Hankinson J. Hear1 F, Merchant J. Coal miners exposed to dieselexhaust emissions. Ann Occup Hyg 1982; 26799-815. 3. Waller RE. Trends in lung cancer in London in relation to exposure to diesel fumes. Environ Int atediagnosed lung cancer, resulting who were probably exposed to diesel ex- 1981; 5~479-83. misclassification would be random and haust. Another recent study in a cohort 4. Rushton L. Alderson MR, Nagarajah CR. would reduce the magnitude of the diesel effect observed. Other sources of confounding that could influence these results are the effect of other, unknown occupational or environmental exposures and passive of heavy equipment operators ( 5 ) utilizing SMR analysis found increased lung cancer by duration of union membership, but not by the limited exposure data provided. Cigarettesmokingwas not considered in the analysis. . f Epidemiologic survey of maintenance workers in London transport executive bus garages and Chiswick Works. Br J Ind Med 1983; 40.340-5. 5 . Wong 0. Morgan RW, Kheifets L. Larson SR. Wharton MD. Mortality among membersof a heavy construction equipment operators union with potential exposure to diesel exhaust emissions. Br smoking.It is unlikelythat either of these The relative risk obtained in this cur- J Ind Med 1985; 42:435-48. are systematically related to diesel ex- haust exposure. Recently, completed epidemiologic studies have found contradictoryresults rent study is low. Some believe that at relative risks of this magnitude, epidemiologic methods cannot be used to establish risk. Critics would indicate that one 6. Luepker RV, Smith ML. Mortality in unionized truck drivers. J &cup Med 1978; 20677-82. 7. Kaplan 1. Relationshipof noxiousgases to carcinoma of the lung in railroad workers. J A M A 195% 17k2039-43. for the association of diesel exhaust ex- could never eliminate the effects of con- 8. Schenker MB,Speizer FE. A retrospective co- posure and lung cancer. A mortality founding factors (36).The pilot study (9) hort study of diesel exhaust exposure in railroad study of maintenance men employed for at least 1 yr between 1%7 and 1975 in the London Transport bus garages and repair facilities was done, with follow-up conducted by our group also indicated that the relative risk of lung cancer would likely be of this magnitude. Risk assessment using past studies (i.e., London workers: study design and methodologic issues. In: Pepelko WE, Danner RB, Clark NA, eds. Health effects ofdiesel exhaust emissions. Cincinnati: En- vironmental Protection Agency, 1979. EPA-6001 9-80-057b; 114-26. to 1975 (4). No excess of lung cancer Transport workers study)have suggested 9. Schenker MB, Smith TJ, Mufioz A, .Wuskie S, deaths was seen. However, exposure to an increased risk of similar magnitude Speizer FE. Diesel exposure and mortality among diesel exhaust was poorly characterized, with both a short duration of exposure and of follow-up. Waller (3)published a study of London Transport workers, men45 to 64 yr of age, from 1950through (below 2) (37).Known confoundershave been considered, and we have identified no sourceof systematic bias in the study that would increase the chance of seeing an effect of diesel exhaust exposure. railroad workers: results of a pilot study. Br J Ind Med 1984, 41:320-7. 10. Howe GR. Fraser D, Lindsay J, Pmnal 9, Yu SZ. Cancer mortality (1965-1977) in relation to diesel fume and coal exposure in a cohort of retired railway workers. J Natl Cancer lnst 1983; 1974, examining lung cancer deaths In the present analysis, we have as- 70:1015-9. among different categories of workers and comparing the rates to expected rates basedon London lungcancerdeath rates. The highest mortality ratio was among those with the highest potential diesel ex- sumed that each year of diesel exhaust exposure amongallexposed job locations was equivalent. We plan to expand these results to consider the different levels of exposureamong diesel-exposed railroad 11. Hall NE, Wynder EL. Dieselexhaust exposure and lung cancer: a case-control study. Environ Res 1984; 3477-86. 12. Mauderly JL,JonesRK,McClellandRO,Henderson RF. Carcinogenicity of diesel exhaust inhaled chronically by rats (abstract). Am Rev Respir haust exposure, but the ratio was not workers. Final confirmation of our re- Dis 1986; 133:A86. afferent from a group with little diesel sults await the completion of other 13. SchcnkerMB Dicsel exhaust. An occupational arhaust exposure. However, causes of studies in diesel-exposedpopulations or carcinogen? J Occup Med 1980; 2241-6. death occurring after a worker left transit authority employment were not morded. There was also no consideration of cigarette habits or of the ethnic studies in similar groups using different methods, Our findingssuggestthat other groups with occupationalexposures(underground workers exposed to diesel 14. United States Interstate Commerce Commis- sion. Bureau of Tiansport economics and statistics list of typical occupations or positions in railroad service, file 21-8-2. Washington, D.C.: US. Government Printing Office, 1951. composition of the workers, and, as a re- exhaust) may be at excess risk of lung 15. United States Department of Labor, Bureau sult, the national rates used to calculate thestandadzedMortality Ratios (SMR) mY not have been appropriate. Howe and coworkers(10) studieda co- hort of retired Canadian railroad work- who were alivein 1965 as well asthose who retired between 1965 and 1977. Deaths among active workers were not nudied, and only last job worked was available Analysis was done comparing -among population subgroups(non- -, possibly exposed,and probably QPOscd to diesel exhaust). In this study, tbc composition of the diesel exposure cancer in contrast to those with "background'' levels. Acknowledgment The writers thank Harry Taplin, Cecelia Bennett, Betty Faucette, and Maureen Ireland for coordinating the collection of death certificates and questionnaires, the secretarial staff of the Channing Laboratory for help with manuscript preparation, and Charles Ford and Michael Cagan for programming assistance. They also thank staff members o f the Railroad Retirement Board for their in- of Labor Statistics. Railroad manpower in the 1970's. Bulletin 1717, 1972; 71. 16. Woskie SR, Smith TJ, Hammond SK, Schenker MB, Ganhick E, Speizer FE. Estimation of the diexl exhaust exposure of railroad workers: I. Current exposures. (Submitted). 17. Woskie SR, Smith TJ, Hammond SK, Schenker MB, Garshick E, Speizer FE. Estimation of the diesel exhaust exposure of railroad workers. 11. National and historical exposures. (Submitted). 18. Doll R, Pet0 R. The causes of cancer. Quantitative estimate of avoidable risks in the United Statestoday. Oxford: Oxford UniversityPress, 1981. 19. Garshick E, Schenker MB, Woskie S, Speizer FE. Exposure to asbestos among active railroad workers. (Submitted). w i n g s were not discussed. By exclud- valuable help with this study. 20. Johns-Manville: Service to railroad. A con- f 1240 tribution to America's progmr New York: Johns- Manvillc, 1923. 21. M a n c u ~ oTF.Mesotheliomaamong machinists in railroad and other industries. Am J Ind Med 1983; 4501-13. 22. Howard ND, ed. Railway enginering and maintrnancecydopcdia7thed. Chicago:s i o m Boerdman corp., 1948. 23. Howson ET, ed. Railway engineering and maintenancecyclopdia5thai.Chicago:s i o n s Boardman corp.. 1942. 24. Brcslow NE, Day NE.Statistical methods in cancer nsearch. W. I: Theanalysisof case-control studies. Lyon: IARC, 1980. 25. Thomas DC.Risk usm manual. M o n t d : Departmentof Epidemiology and Health. McGiU University, 1980. 26. Blot WJ, Harrington JM. Toledo A. Hoover R. Heath CW, Fraumeni JF. Lung cancer after anplayment in shipyardsduring Wrld War 11. N Engl J Med 1978; 2W620-4. 27. Fraumeni JF, Blot WJ. Cancer among ship yard workers. In: Pet0 R. Schneiderman M. eds. Banbury report 9. Quantificationof occupational cancer. Cold Spring Harbor: Cold Spring Harbor Laboratory, 1981. 28. Kolonel LH, Hirohata T.Chappell BV, Viola FV, Harris DE. Cancer mortality in a cohort of naval shipyard workers in Hawaii. J Natl Cancer lnst 1980; a739-43. 29. Ohlson C G Klanson B, Hogstedt C. Mortality among asbestos exposed workers in a railroad workshop. Scand J Work Environ Health 1984, 103283-91. 30. Office on Smolringand Hcalth. Cancer: The health consequences of smoking. A report of the surgeon general, Washington. D.C.: U.S. Government Printing Office. 1982. DHHS PHS (81250179). 31. Rogot E, Rcid DD. The validity of data from nact-of-kinin studies of mortality among migrants. Int J Epidemiol 1975; 451-4. 32. Kolonel LN. Hirohata T.Nomura AMY.Ade quacy of survey data collected from substitute responders. Am J Epidemiol 1977; 106:47&&. 33. Doll R, Pet0 R.Cigarettesmokingand chid carcinoma: doseand timerelatiomhippmollo regular smokers and lifelong n o m m o k m J & demiol Community Health 1978; 32303-11. 34. StevensRG, Moolgavkar SH.A cohort pnrlysis of lung cancer and smoking in British m k . en. Am J Epidemiol 1984, 119624-41. 35. Percy C. Stanek E, Gloeckler L.Accuracy of cancer death certificates and its effect on cpoar mortality statistics. Am J Public Health 1981; 71942-50. 36. Doll R. Occupational a hazPrd for @emiologist.% Int J Epidemiol 1985; 1432-31. 37. Cuddihy. Griffith WC. Clark CR McCIcfLn RO.Potential hcalth and environmentale f f w of light duty vehicles 11. P n p d for the Off- of Health and Environmental Research, USDOE, Contract %DE-ALW76EV01013. Albuquapue, NM:Lovclace Biometrical and Environmental k- search Institutz 1981. i