Document eZavN84V14xxkD6ejy7ngXpE

.Cancer Mortality in Oil Refi n ery Workers NancyM.Hanis, PhD; KathIeenM. Stavraky, M.D., Ph.D; andJ.L Fowler, M.D. Crude petroleum, its fractions and the polycyclic mortality, compared with nomefinery workers. The sec- aromatic hydrocarbons produced in the refining process ond hypothesis was intended as a rather crudeassess- 'have all been shown to cause cancer in laboratory ani- mals, most commonly in the liver, lung and kidney.' m-efltx' I th-e presence-on-a-refinev-site of carcinogens ' other dian'itiose present in peti%okuiifib-ts.- Studies of humans exposed to petroleum or its fractions have demonstrated increased risks of cancers of the skin, Materialsand Methods -scrotum and lung, and less frequently, the gastrointestinal The Data Imperial Oil Limitedprovidd the follow- tract and bladder.*' ing info:mation for Zl.732 male employees, who were In spite of the evidence suggesting that workers in the either active in t!!e company, or were company annoi- oil refining industry might experiencean increased risk of tarits @oth full annuitants and deferred annuitants, i.c. cancer, few studies of cancer risks among these workers persons no longer with the company after 70.15 years have been made until recently. Ina study of workers in17 service. but members of the pension scheme) during the oil refineries in the UnitedStates, Caffey et ai' found that period 1964to 1973. inclusive:(1)current age; (2) years of the ratio of observed to. expected cancer deaths was service with the company; (3) status (active, retired, ter- greatest inworkers with the lowest exposure to hydrocar- minated, dead1(41date of retirement;(5) dateof death;(6) bons. The risk of death from cancer of the digestive causeof death; (7') job title (e+, chemist, warehouseman1 system decreased with increasing exposure to hydrocar- and, (8) job function and location (kind of activity. e g . bons, whereas death from respiratory and genital cancers producing. chemical plant. and its location by province increased with such exposure. On the other hand. the and reiinery or nowefinery sitel - results of a study conducted by the medical director of Imperial Oil Limited suggested that refinery employees This information regarding job title. location and func- tion was sufficient to-categorize employees by exposure had increased death rates from cancers of the intestine and refinerywork. according to the following definitions. (ICDA 152-1531andother digestiveorgans (ICDA 155-159). 1. Exposure was defined as daily occupational contact The present study was designed to test the following with petroleum or its products; moderate exposure as hypofiiGE7(11Worken exposed to petroleum and _its_ contact less t!!an daily (e-g, it could representas muchas p r o c r e a s d risk of mortality from Cancer, two weeks every monthk and nonexposure as no known es-Iy of the intestines (includinn rectum) and other occupational contact. On the advice of the company's digestivsaans, than do nonexposedworkers. Reference industrial hygienist, the department, site, job titles and to the intestines throughout this article will assume the functions were used to place employees in one of the inclusionof the rectum(ICDA codes 152-154).(21Workers three exposure categories. A small group of employees on a refinery site have a similarly increased risk of cancer (1.63'0 of the total) was exposed for only part of the 10- year period. Since the group was small, those members From the Department ot Epdmlology 6 Preventwe MedKlne. Faculw oi Medicine, Univmty of Western Ontano. London. Canada (suppooned by a grant from the Phvtrunr knraIncwporated.Ontmo. Canada)(Dr HJ~ISL working at least half of the 10-yearperiod inexposed jobs were designated exposed, and those working less than the Ontario Cancer TfeJUVIent and Research foundattonand OrgaNnrnt of half, nonexposd. m.I (. Epidcmddg). and Plevencnchledcme. Facultr of Nedsme. the Umversicy Of WesternOnIuD. London. CJnJdr (0;. SUWakvl dnd Impmalo11Limited, 111 St. a a u Ave. W, Tomato. Onorio. Canada (Of. Foulrfl+L, Hanir is prerenlly with the Revarch and Lnvironmmcal Healch Division.-E 2. Refinery work was defined as daily work on a refinery site. irrespective of exposure to petroleum. Information provided regarding location of work permitted the -ported Uox 45. Linden.-N by a srdnt from the Ontario Cancer Trearmenc and Research classification of employees as refinery or non-refinety Farnda:ion. workers. Journal of Occupational hledicine/Vol. 21, No. WMarch 1979 167 , .. ' Definitionof the Study Population Following the categorization of employees, the iinac study population was defined by making the following exclusions from the total population: (I) females, because the number was very small; (2) summer students, because Analysis of the Data Since employees were observed for varying lengths of time during the 10-year observation period. penomyears at risk were used as the denominators for mortality rates. Both active employees and annuitants contributed to the the number was .small and the total exposure time for person-years at risk. each student short; (3) short-term employees who termi- Age-adjusted causespecific mortality ratesformed the nated during the study period, with less than five years basis for comparisons of the exposed and nonexposed service with the company, because it was not practicable and the refinery and non-refbeF,iisCThii~di&t to trace these employees; (41 employees active with the m e ~ ~ b f ~ 6 ~ ~ ~with~the w ~ ~ company in the final year of the study, but who had had total study population as the standard. Tests of signifi less than one full year of employment; (5) 19 retired ern- cance have been made by a &I square test of the dis- ployeeswho died early in the study period,but for w& crepancy between the distribution of the observed no t n t o r m a t i o m d be iou n a mn bvnlch to categorize KiZexposure or piace oi W n m ong tne excluded short-term employees and the 19 annuitants, only two number, of deaths in each exposure category and the number expected. assumingthat the number of deaths in each category is proportional to the distribution of cancer deaths were known to haveoccurred. person-years in each category.' This procedurewas used Fig 1details the final study populationof 15,032 male in preference to direct significance tests of differences employees, who had been observed either as active employees or annuitants for all or pa? of the loy.e.ar obser- vation period betweenage-adjustedrates. since the latter tests havenot beenadapted for usewhere the denominators of the rates are person-years at risk. A total of 865 employees were lost to followup. This For examination of mortality rates by duration of e m group was not part of the 6,681 short-termemployees ex- . ployment persorryears in each duration category were 1 cluded from the study population, but was made up of determined by allocatingthe person-years contributedby persons with 5 to 15 years service (vho left the company each employee during the loyear period, to the appro- and with whom no contact was maintained by the corn- priate age and duration category.'O For example, an em pany because they were not members of the pension ployee enteringthe company in 1940at age 20 would be scheme. These untraced employees were distributed as age 44 in 1964, the beginning year of observation, and follows in th6 refinery/exposure comparison groups and would have had 24 years of employment The first fwe their controls: exposed, 4.9%; moderately exposed, 5.9%; years of the loyear observation period would be a b nonexposed, 7.4%; refinery, 2.2% and non-refinery, 7.9%. cated to the appropriate age group in tfw less than 3& f year duration group. The remaining five years, from'age . Verification of Causeof Death 49 onwards, would be allocated by age to the30and over The Provincial Registrars, in all but one province, and duration group. Statistics Canada assist4 the company medical director by Drovidinp, from the death certificates. verification of Results the-underlying cause of.death. Death cekification infor- In the study population during the l o y e a r observation mation on underlying cause was not located for l d period there were 1311 deaths of which M1 were dugto he1,511 deaths which occurred dunng the period of ob- cancer. Table 1 shows the percentage distribution of servation. Analyses in thii study are based upon those deaths by major cause. Disease of the circulatory system deaths citing cancer as the underiying cause on-the death and malignant neoplasms together account for nearly certificate. 80% of all deaths. An analysis of deaths from causes other than cancer will be the subject of a separate paper. An eXafI'IiMtiOrI ofthedistribution of deaths by causeand age revealed that only f i e cancer deaths occurred in employees under the age of 40.Consequently, the under L 40 age group was excluded from subsequent rnodity comparisons. The distributions of perswears at risk by age contributed by employees age 40 and over in the -Table 1. Perceatage Distributiond Deaths inthe Study Populationby Major Cause. 1964-1373. I ICrCxD0.d MalignunMoplaanr (140.209) O&iswc~01 circub(0Cyrvstem(390-458) Di%asesofreSpintavsystem(460-519) 301 19.9 905 59.9 90 6.0 Oisease~01dii-wtm (520-57n 51 3.4 Accidnm.poisoninpsand v m (800.999) 59 3.9 i AI! other 10s 6.9 411 313. Total 1.511 100.0 -Fig 1. Study population. lmpmal Oil Limited. 1964-1973. " X A Codes. 8th Revision (1965) shown inparentheses 168 Cancer Morta!ily in Oil RefineryWorkers/Hanis. Stavraky. and Fowler -Table 2. Pmon-Years at Risk by Age for the Study Population, 1964-1973.' Age Tolal Nonexposcd 40-44 45-49 50-54 55-59 6064 65-69 70-74 75-79 . 80.84 85+ Total 16.750.5 15.1 14.0 12.661.5 10.793.0 9.065-0 7.1 10.5 5.074.0 3.1 78.5 1.667.5 642.5 82.057.0 4.661.0 3.932.5 3.321.5 2i944.5 2349.0 1.718.5 1.1 43.5 1.111.5 342.5 133.5 21.21 2.5 'Ase groupslessthan 40 exdudedfmmlhitaMe Moderately Goosed 2.980.0 2.396.0 1.606.5 1.209.5 830.0 456.0 352.0 292.5 152.5 '34.5 10,309.5 EXDOSed ~ ~~~ 9.1 09.5 8.785.5 7.733.5 6.639.0 5.a86.0 4.936.0 3.578.5 2.220.0 1.17 2 5 474.5 50.535.0 Non-Rclinm 10.961.0 9.258.5 7.354.5 6.005.0 4,741.0 3445.5 2.326.0 1.404.0 686.0 271.0 46.452.5 Rcfinew 5.789.5 5.855.5 5.30t.O 4.7aao 4.324.0 3.665.0 2.748.0 1.714.5 981.5 371.5 35.604.5 refinery and exposed groups and their correspondingcon- erylnon-refinery groups were not as consistently compa- trot groups are shown inTable2 rable geographically. Ten percent of the refinery group compared with 7% of the nomefinerygroup were in the Comparison of Cancer Mortality in the Exposed and Re- Atlantic provinces; the refineqcnorwefinery comparisons finery Croups and Their Controls * Before cancer mortality was examined, the compara- bility of the exposed and refinery groups and their respec- in Ontario and Quebec and in the western provinceswere 60%:47% and 27%:40%. respectively. The remainders . were in scattered sites. These obvious differences dictat- tive controls was assessed for the following variables: year of first observation in the study, age at last observation, and geographic location. A high percentage of both study groups and their controls enteredthe study in7964, the first year of observation. Over 70% of the exposed, moderately exposed and nonexposed groups and about 80% of the refinery and nowefinery groups entered in 1964. In all groups. small percentages entered observation in each subsequent year. Thus, the vast majority of each study group was employed with the company in 1%. the beginningof the observationperiod, The geographic comparison revealed that the distribution of the exposed/moderately exposednonexposed groups by province was as follows: about 10% (exposed) .vs. 5% (moderately exposedlw 6% (nonexposed) in the ed the need to consider the geographic variable in examining any differences in disease rates among subgroups. Since there were few deaths in the scattered geographic sites this group was omitted from these analyses. The age distributions of workers over the age of 40 in the exposed and refinery groups and their respective controls were compared. Both comparisons revealed age dis- . crepancies, with the exposed and.refinery groups being older than their respective controls. Thirty-three percent of the exposed vs. 16% o i the moderately exposed a. 25% of the nonexposedand 37% of the refineryvs. 23% of the nomfinery were& years old or more. The exposure and refinery study groups and their controls were not examined for comparabilitywith respect to other variables, such as year of birth and year beginning Atlantic provinces; 50% (exposed) vs. 46% (moderately exposed) w. 60% (nonexposed) in Ontario and Quebec and, in the western provinces. 33% (exposed) w. 48% (moderately exposed) vs. 34% (nonexposed). The refin- employment, because these variables were highly come lated with age at last observation. No information was available. such as income or education, to examine care fully the comparability of the study groups and their com .. I -Table 3. Peneatage Distribution of Cancer Oeatht in the Exposureand Refinery Comparison Croups, 1964.1973.. I Site of O n c u t Buccal cavity and phyynx (140-149) bophapusand stonach(150,151) laxe intestineand rectum(153.154) otherdpative organs(155-159) Tnthea. %- (162) ~5urespirataryor~ans(160.161.163) Bonk wnnecriveti-. fkhr ( I 70-174) Prostate(185) Urinary ofgans(188.1 89) a t h e f a n d u n s ~ l e dsites(190-199) lymphaticand hematopoieticsystem (200-209) Total Manexpored (?f=65) Moderately Exposed (N=l6) -lSY* 4.6 18.5 12.3 20.0 1.5 7.7 10.8 7.7 7.7 7.7 100.0 . 125% 6.2 312 -12.5 --6.2 6.2 25.0 99.8 Q n m Deaths EXpotcd (N=215) 28Y. 13.0 135 93 312 -1.4 2.3 8.4 6.5 5.6 6.0 100.0 NaaRebeq 41=m 231. 11.4 10.6 9.1 29.5 15 5.3 7.6 6.8 4s 11.4 100.0 Rcfinuy (H=164) 2.4% 11.0 17.1 12.8 26.2 12 2.4 9.1 6.1 7.3 4.3 99.9 'Age groupslessthan 40 cxciuded from Mis table. tC0A Codes. 6th Revision(l965) shown in brackets. Journal of Occupational Medicine/Vol. 21. No. 3/March 1979 169 -Table 4. Age-Adjusted' Cancer Mortality Rates for Specific Sites in Exposed and Nomxposed Vhrkers, 1964-1973.t Sile of Cancer# Nonexposd Rata Age-Adjusted filort;lliw Rater per 10,000Perron-Years ModeratelyE x p o d EXpaSd Ra:e RelativeRisk Rate ~ t i v c A * k AU sites (140-209) 0.57 1.235 Esophagusandstomach(150,lSl) lntestinesand rectum(153.154) c Other diqestiveorgans(l55-159) 1.06 3-m 0.26 0.88 0.67 0.90 Trachea. lmrchus.lung (1 62) Prostate(185) 0.39 1.e90 - 0.84 Urinarybladder.kidney(1 86.189) 1.08 Lymphatic hematopoieticsystem(200.209) 1.88 0.88 ~Directadi~entusing&esardypopulation&~andard -. tAge groupslest than 40 excludedfmm this table SICDA codes. 8thRevision(1965) showninbrackets indicatesthattheexposedratesdiffered significantlyfromthe naMxposedat pc0.05.chisquare. d.t=l(s . .1Nwnbersof deaths upon whichratesam haKdshowninparccltheses .. trots for sociai class. Nor was any information available on personal habits, such as tobacco and alcohol use. The distributionof cancer deaths at ages 40and over in the exposure and refinery groups and their controls is shown in Table 3. In all groups the sites accounting for the largest percentages of deathswere trachek, bronchus and lung. and the digestive tract, particularly the intestines in the exposed and nonexposed and other digestive system organs in the moderatelyexposed workers. There were no cancers of the small intestines. Hence, all further mention of intestines will assume the large intestines (ICDA 153)and rectum (ICDA 154). Table 4 shows the ageadjustedcancer mortality rates for the major sites in the exposed and nonexposedgroups and the ratios of the exposedmonexposed and the moderately expdsed:nonexposed. Mortality from all malig- nant neoplasms in the exposed group was significantly above that in the nonexposed. The two sites for which mortality was significantly higher and which accounted for the largest part of the excess cancer mortality in the exposed group were esophagus and stomach, and trachea, bronchusand lung. The risk of death from cancer -of the esophagus and stomach in the exposed group was 3% times that in the nonexposed fortung cancer, the risk in the exposed group was almost twice that h the nonexposed. The risks in the moderately exposed group. P T1 were lower than the nonexposed. The numbers of deaths / 4 -ed ;Jfurther group were small: hence. all ezatel LW rou r-' T m z m s o f increased risk of death from cancers of pi@' the esophagus and stomach, and trachea, bronchus and lung in the exposed group persistedwhen geographic fac- tors were considered. Table 5 shows these comparisons. Although none of the. exposedhonexposed differences were statisticallysignificant, it is noteworthythat the rela- tive risk of esophageal and stomach cancer was elevated among exposedworkers in Ontario and Quebec and to a lesser degree in the western provinces The risk of lung cancer seemed to increasegeometricaliyfrom eastern to western Canada Table 6 shows the age-adjusted stomach and lungcaw -Table 5. Age-Aqusted' Stomach and Lung Cancer MortsIity htet in Exposed and Nonsxposed Workers by Geographk W o n , 19661973.t I Atlantic provinces WestrmpDovintts 11.9 (119 4.6 (14) 7.3 (11) * - 0.8 (1) 4.1 m - 1x8 19.7 0.7 (e) (2) 18 11.6 7.3 1.6 04) (SI 1-8 14.0 3.5 4.0 (21) CL) - .. L, !VJ -Tab2 6. A;z-Adjustcd' Stomach and Lung Cancsr i I,lortality fi;zs ia Expased Workers by Ouration of Employment. 1 1964-1973.t I Ape-Adjusted hlortllity AalcU10.000 U n a r hbtality RatedlO.OJ0 PcrsoaYtan (Exposed Workers) SifeolCancerS . Person-Years Refinery Hon-Relinq Relative Risk Esophaguslad Stomach Trachea. 8mnchur and Lung AOtiii(lK&;204) 40.6 (16411 31.4 (132) 1295 Bophas& L i !zx!aZl 4.5 3.7 1.22 (150.15; 1 Intestines 6 :ea (18) (15) 6.9 3.5 1.975 I nc (153.152) Werdg&-;c r+zs m(155-1:5t Trachea. h-ets (28) 5.1 (21) 10.9 (1 41 2.8 . (12) 9.2 1.82) 1-18 -oirec! a$- -,&!I the study popubtm. age240. > 10 years wvic~~~~ - tApe FF;LCS'af;ill Lo excludedfromthis tab& $"umt+rr 3 h z i upcn which ates am based are shawn in cer mortdiiv r a t a in the exposed workers by duration (162) Pmtatr (155) (43) 3.3 (15) Ulinalybk&i?s. + - - i c y 25 cl;,(iae. IE Z ~ Lymphaticd~ . pcieric qs?m !33J-ZG9) (10 ) (39) 2 6 1.2? (10) 21 1.19 3(.79 0 . 4 6 2 (15) 'of empltnrr..% where duration of employment was the only a v a i E k means of estimating the dose of petroleum or its prcdccs accumulatedover time. Itshould be noted that there w.rz ra cancer deaths for these causes among workers *.+E:70 oc less years ofemployment Clearfy bob! esophag!! z-d s:zmach, and trachea. bronchus and lung cancer ncc3:ky rates increasedwith increasing duration of e m p h-sa The f c e D 3 g znalysis of cancer mortality by major site was r - F z z d for t!!erefinery and non-refinerygroups. Table 7 zk-s the site-specific age-adjusted rates in the hvo gzc-, znd the relative risks. As with the ex- p o s e d j n c c a - d comparison, there was a signiiicandy increased .%k of &I malignant neoplasms in the refinery group. Tkz s k s ivhich here accountedfor the largestpart of the e x . s - e r e intestines and other digestive orgznj. the risks k c ~ 7 n e r yworkers being almost twice 25 great as those frc m e f i n e r y workers. The mortality in the refinery 5;-);p from neoplasm of the lymphatic and hematopt'e5.c systems was about half that in the non- refinery The exes ci intestinal and other digestive system 7. age-specifii nortaiity rates fa; both sites were com- cancers ii 5 reiiiery group was examined specificzllv pared. No cwdistent age-speciiicdiiierences betweenthe for g e o g z i k location and for duration of employmen; refinery grw?and its control were seen for either site. Table 8 L-CYASh\e gwgraphic analysis. The increased Stomach azo -;1. cencer mortality in the refinery and risks of m d i t y from intestinal cancer, and from inter- . non-refiner,. g o u p were not explored further. -Table 8. Age-Adjusted' Intestinal and 0 t h OIgwtivs Orpn Cancer MortaSty In Refinely and Nan-Refinery Workers by Geographic L0eat;r~.196b1973.t - - - I I103 5.0 22 8.8 5.8 (2) 9.7 1.5 I Journd s: C - z J S i o n a l Medicine/Vol. 21. No. WMarch 1979 171 .- -Table 9. Age-Adjusted' Intestinal. Other Digestive Organ; Stomach and Lung Cancer Mortality Rates in Relinery/Exposura Subgroups. 1964-1973.t - Mortality Rates in Specific Subgroups of the Employee Population Ageadjusted mortality rates forintestinaland other di- Age-Adjusted Cancer ModaliQ Ralss/10.000 gestive organ cancers and esophageal and stomach, and trachea. bronchus and lung cancer, the sites which were Trachea, DipesUre Esophagus Emnebus, elevated in the refinery and exposed groups. respectively, are shown in Table 9 in subgroups of the refinery exposure categories. The slightly elevated stomach and lung Subgroup 153.154)# (155159) (154 151) (162) cancer mortality in thk exposed and moderatelyexposed groups relative to the nonexposed groups supported the earlier evidence that increased risks of stomach and lung cancer occurred inworkers exposed to crude petroteurn and its products. and suggested that the increased risks were independent of place of exposure. The pattern of intestinal and other digestive organ cancer risk by stp surdrefinery subgroup is less clear. If workers ata refinery site were exposed to a carcinogen other than petroleum, +one would expect the rates for the refinery subgroups to be greater than rates for nomefinery groups, but there was no such consijtent pattern. Howevef, the rates 1I 'Direct adw.mmt usingthe studypopulationasstandard tAge groups less h n 40 excluded from this table I were baseduponvery small numbers of des* Table 10shows the ageadjustedcancer mortaliwrates . W i l A codes. 8th Revision(1965) shownin brackets by job titles for the following sites: intestines. other Numbersof deathsuponwhich ratesare basedshown inparemhcres digestive system organs, esophagus and stomach, and trachea, bronchus and lung. The ratios of the job-specific mortality rates to the study population rate are included. Job me# -Table 10. Age-Adjusted' Cancer Mort& Rates bySita and by Specific Job. 1964-1973.t ~ ~~ Age-Adjusted Canwr Martaw RaW10.000 Perton-Yarn ~~~~~ Intestines (153.154) Rate RiskAano$ 11.91 2.3 0. mu DigestiveOrgans (155-159) Rate Risk Ratto -- Esaphagssand stoarrch (150.151) m RtJlPuo 11.9 3.0 (3) Trachea, Btonchusand Lung (162) Rata Risk Ratio -- 9-8 1.9 5.4 1.i 4.0 18 12.7 1.3 (13) ' (8) 0. . (17) 6.7 1.3 4.5 1.1 m (5) 4.a . 12 9.1 0.9 (5) (10) % 'Direcl adpsmmt using the study popu(ationasstandard mitAge proups less than 40 excludedhan table . tReCneryand eqoune groupsare shown inparentheses:krefinery. NR=Nonrefinery. =exposed. ME=moden:ety exposed. NE=nanexpmed Ratio of rate insubgrou;,to studypopubtionrate ~ N W T I ~o~f dNeathsuponwhich ratesare Used are shown in parentheses 172 Cancer Mortality in Oil Refinery Workers/Hanis. Stavraky, and Fowler .. -._ employees with less than f i e &ice. since this group a ' c s young and very few c m e r deaths occurred in employes under the age oi a.Hmsever. inability to follow empioyees with f i e to 75 service is a more serious matter. In this study. rkxe pccportions of lost employes with fiveto IS yeas --nice were Iawr inthe hvo JN& grmps than in their ?*-me controk. it is un1ike:y &at losses con*3ut& ;3 the obsewed differ- ences. It is possible. however. +z: S k e n t i a l loss could have contributed to iailure to &:XZ real &fierencer if occupational exposure were rek+zi bob to :caving em- .d. wment and siibseauendv to Cez5 %Another difficult dro5temwi?s%: acctrrare categoriza- <&&t Cand its pDrroodduuccttss.- E r r m in CIE--JI;;iiCZ:;On cannot h a v e 173 ,. . . * wasenhanced by the lack of availabilityof ernploj&ejob historie prior to the 10-year observation period. Thus. it was possible that employees. older ones particularly. could have moved from the exposed or refinery classes prior to the start of observation, and were classified as nonexposed or nomefinery regardless of where the majority of their working years had been spent Move- ment in the reverse direction may have occurred also. Duringthe Wyear period, only 1.6% of the study population changed exposure status, and this problem may have been no morefrequent prior to1964. The cause of death information on the death certifii cate is always subject to inaccuracy. The consequence for this study of using death certificate information should not have been bias. since the determination of cause af death was done independently of knowledge of exposurdrefinery status. Howevk, any inaccuracy which reduced the number of cancer deaths in the study increased the difficulty in detecting real differences. Morbidity data were not available to us, but their use would be preferable, because access to MI morbidity data can provide some indication of accuracy of diagnosis. Fur- ther, the study of cases rather than deaths is desirable when studying a cancer (e.g, of intestines and rectum) with an appreciablesurvival rate. The problemswhich have beset this study are common in studies of occupational hazards in man. They cannot be overcome until industry accepts responsibility for the maintenance of complete records. including changes in job for each employee throughout hi periodof employ- ment and developing and using objective methods of measuring exposure, both interms of quantity and quatii ty. It is important for industty, with government and union cooperation. to develop record linkage methods for following worken from one employer to another and for unions and industry to cooperate in obtaining information on demographic factors, hobbies and personal . habits, necessary for the rational interpretationof occupational studies. Imperial Oil is presently updating this mortality study and is in the process of designinga cancer morbidity study. of more complete informationon the similarity of the aand nonexposed employees, these results cannot ascribed with certainty to a carcinogenic effect of petroleum However, there was no other clear explanation forthe results. and further study is required. When refinery workers were compared with non- refinery workers without consideration of exposure to p e t & u m in either group. the refinery workers were found to have twice the risk of cancer of the intestines (including rectum) and other digestive organs. No r e lationship with duration of employment was evident Although the increased intestinal cancer mortality in the refinery workers \vas not consistent, bias was not an obvious explanation for the observed relationship. Therefore, the presence on a refinery site of a carcinogen other than petroleum has not been ruled out, and further study is wed. Summary. The.cancer mortality experience from 1964 to 1973 of employees of the ~mperiaOi il Limitedwas examined in a cohortstudy. EmpIoyeesinjobs which exposed them on a ' daily basis to crude petroleum or its products, compared with nonexposed employees, were found to have more , than three times the risk of esophageal and stomach cancer and about twice the risk of lung cancer. In .the ex- posed group, increasing risks of both cancers occurred with increasing duration of employment In the absence 174 Cancer Mortality in Oil Re!inery WorkerslHanis, Stavraky, and Fowler . t i !