Document N2GV3aK1VQqyjJgDXr7y6q2V

. * * I' D . c, +=p--P. Solvent Exposure and Leukemia , Among Rubber WorkP ers: An Epidemiologic Study r A. 1. McMichael, M.D., Ph.0; R Spirtas, M.S.; L L Kupper, Ph.D.; 1. F. Gamble, M.S.P.H. T h e findings of this composite epidemiologic study indicate an association between leukemia and jobs entailing exposure to solvents. A literature review on the epidemiologic characteristics of these cancers yields little firm evidence of chemical exposure as an aiologic factor. A four-fold increase occurred in the US. leukemia mortality rate between 1920-1960, followed by a gradual decline. This decline has occurred for whites of all ages, but not for non-whites.1 2 leukemia incidence rates within the U.S. have generally been higher for males than females. This male-female acess has declined in childhood cases, while the ratio has been increasing for adult u s e s . 3 Internationally, males show a similar excess with the exception of only a few nations.2 The age-specific moctllity rate rises to a minor peak during the first five years of life, falls to a minimum during young adulthood, then ascends to a maximum peak in later adulthood and d d age.4 Geographically, the mortality rate within the US. is highest for the white population in the rmthwatern and northcentral states @rrt not including Ohio, the state in which the cohorr study k c r i b e d MOWwas conducted). A deficit exists in the northeast and southern states. lhe mottali rate kr the non-white population exhibits a null cluster of elevated incidence rates in the mid-southern states.4 Various emmigrant group studies in the U.S. have not revealed any consistent trend, and there seems to be little consistent evidence of an association between leukemia and either social class or industrialiution.2 5 Of the four major cytological types of leukemia, defined by the 8th Revision of the International Classification of Diseases, myeloid and lymphatic con- stitute the greatest percentage of cases. The proportionof myeloid cases peaks at 60% of all leukemia cases between the ages of 30 to 40, then declines. The lymphatic type has the highest childhood dative prevalence (50%). declines until the ages of 30 to 40. and then rises to its peak of 60%between the ages of 80 to 90. The lymphatic type surpasses the myeloid type in percentage near the age of 70. Myeloid leukemia exhibits its highest incidence in whites living in certain northcentral and northeastern states. The lympatic type demonstrates an elevated incidence among whites in northcentral and west coast states Non-whites do not show geographic clusters of elevated rates for either of these types of leukemias. lhe ratio of male cases to h a l e cam is highest for the c h i c lymphatic type and lowest for the chronic mycloid.2 4 These reported differences in s e ~ratios could suggest either the p o s s i b i l i of occupational environmental factom in the ctiobgy of the lymphatic type, with maks being K k a i y aposed to such factors; or the possibility of differential susceptibility of rmks and females to environmental factors. Considerable evidence suppons the hypothesis that chromosomal aberrations are associated with the occurrence of Ieukemia.2 6 Ionizing radiation may achieve its myeloid leukemogenic effect via chromosomal damage.' Chemicals such as benzene, phenylbutazone, and chloramphenicol are suspected leukemogens, although ---there is little sound _epide?ohic evidence.2 7 There is also evidence that benzene can produce chromosomal abemtions.6 However, the evidence regarding the leukemogenic effects of these chemicals can only be viewed as suggestive. . Them are some indications that the other major cancers of the lymphatic and hematopoietic systems (Hodgkin's disease. lymphosarcoma, and multiple myeloma) may have some environmental etiology. Epidemiological evidence on Hodgkin's disease indicates that hgh social class white males, 50 'yews or dder. have the greatest in- cidence rate.' Akauki supporu the environmental hypothesis with evidence indicating a greater incidence of tiod@n's disease in kpancw migrants in Hawaii than in japan The evidence supponing cmrironmcntal factan in lym~ ~ O U K O ~ ~ aWnSd muhipk myelomas is meagcr although Spanese studies have shown radution and migration to effect the incidcncr of these dircasa.m 11 Cohort Mortality Study In 1970, the United Rubber Workers Union negotiated with each of the six ^. maiw US. rubber companies for I prqram of comprehensive research into work-related health problems The khod of Public Health, at the Univer. sity of North Carolina, i s conducting the research program within four of these companies. The data presented here derive from seven plants within one of these four companies. and refer adusivcly to deaths among the hourly (i.e. factory) employees, either active or retired. The initial mortality study was carried out at the company's largest plant, in Akron, Ohio, a n d has been fully rew e a m A cohort of 6,678 male, hourly rubber worken was identified (and crosschecked) retrospectively, and mortality data for the period 1964-72 were obtained from the'compaw's life insurance records. This cohort. of-active and retired workers, aged %to & at Cnuary I , 1964. and with an average employment duration of 25 years. was wed for nine years, with 1%of per- %an-d with com- plete death ascertainment on the remain- der of the population. An estimate of racial composition (basedon payroll records and death certificates) indicated that 13% of the cohort was black. compared to 10% within the U.S. male population of the same age-range. The sirnilariw of these two figures allowed mortality comparison of the cohort with the U.S. male population. By comparing the observed deaths to the number of deaths expected (calculated from 1968 US. male age- specific death rates), standardized mortality ratios were calculated. For all derths, the S M R was 93 in the age-range 40 to 64, and 101 in the age-range 65 to 84. These age ranges were chosen because 65 is the age of normal retirement k should a l G be noted that these SMR's are conservative bemuse of the use of a standard population that in- ttw lnternationri Classification of To examine the mmality from lym- Chsea5e4 (8th Revision) ategoria 200 to phatic and hematopoietic cancers in 209, includes the leukemias (see Figuw more detail. SMRs were calculated for &1). with a threefold excess in the aw the individual ICD categonn The mults -range 40 to 64 11 deaths OMare presented in Figure 2. The greatest j when 3.5 were expected. It also in- mortality excesses are for lym- -clude the commonest types of Ivm- phosarcoma. Hodgkin's Disease. I= phoid tissue cancer, that is, the lym- phrtic leukemia. and mbeloid leukemia. + i phosarcomas. reticulum cell sarcomas. There was a >wenfold eaccss of deaths and Hodgkin's Disease (Figure 1). for from lymphatic leukemia in thc younger -which there was, overall. a two fold ex- age-range. Six of these eight deaths were cess across the full age-range. 40-84 from chronic lymphatic leukemia. 16 deaths observed where 6.2 were ex- Thus, this cohort mortality analysis at pected. the Akron plant focussed attemtion upon I. ,r C' 'I II $ck, disabled and institutionalized. who camFigaut rhec1lpwnpKte!rsennd-&mtohect-criai-inIvcerirs k K u x t i i i SMR's, for these same two age-nnges, 40 to 64,65 to 84. While the SMR for all uncers is approxirrutcly 100. sevcrrl -type d cancer showed a statistically significant &ation in patlicukr, cancersof the stamch, the prostate, and the lymphatic and hematopoietic systems. Thir latter group. comprising cancers o f the lymphatic and hematopoietic systems. For reasons of c s' cost and time, information about the sourcepoDulationr-at-nskat the smaller. p p p - A k r o n e t s was obtained. Instead, a proporti03 m a t a w a l v i s was carried out. using an age-. standardization procedure exactly analogous to that used in calculating S M R s . Within the age-range 20 to 90. prOpOrtiOMl mortality excesses were observed, at these six non-Akron plants combined, of 15% for lymphosarcoma and Hodgkin's Disease, and 45% for leukemia These figures suggest a rather smaller mortality excess from these cancers at the non-Akron plants, compared to the Akron plant. (The considerable variation observed in age-standardized proportional mortality ratios between the individual non-Akron plants could reflect either a real difference in cancer risk because of differing in-plant work environment. or it c w l d simply reflect the "instabiltty" of the values of the ratios because of the small numbers of deaths involved.) The question then arises: Are these ob- ' served mortality excesses spurious (due to chance or artefact). or are they real? Further, if real. are they associated with specific categories of jobs within the nibher industry? A n analytic study was thus called for. Case-Control Study Because of the absence of source population information at the non-Akron plants, it was decided to carry out a use-control study. capitalizing on the total of 88 deaths (from cancers of this ape) at the Akron and the non-Akron plants. ' The cases were defined as any active or retired hourly worker, dying in the period 1-73. whose death mtificate cn~ludedreference to a cancer of this type. irrespeaive of whether the cancer =as the u-n-d.erlying of death,= rtmply a comributg- cause. For 77 of r h ~w oiK-tiZcancer was the assigned undedyiq ouse of death. The plant. mce, s a . and age-at-death infkmnation for h s e 86 a s e s is pesented 8 n Table 1. Table 2 shows the dstribution of specific types of cancer, at ohc Akron and non-Akron plants. type of cancer. But. futtk. because we only had information on deceased rub- ber workers a1 the six non-Akron plants, --1- 1. kWl* d (I* Ic a s u control SeieCtion was MCeSuriIy con- Y r r N lryIM, fined to rubber worken dead of other causes. To be eligible for selection. the control had to have died within the same period. 1964-73. and had to have died from a cause other than a cancer or a Mood disorder (such as aplastic WI wM8 Id F U .mh~~ Tu -46 s SI 0 -21 3 24 4 ZI anemia). lo ensure control of potential confounding e f f q . the controls were b m rcI m- us 51.3 individually matched with the cases. in a ratio of three controls per caw. &e. To take account of the diversity of race. sex and, probably. plant were jobs a d exposures, a classification regarded as risk indicators of leukemia sptem was &doped whereby the and related cancers, and were .also several hundred departments and the judged lo be associated with the type of .thousand or so jobs that have existed at - 1job an individual does within the N& industry. Controls were therefore in- dividually matched with tases on sex. race. plant, and age-atdeath (control age-atdeath was required to be 0 to 5 'the various plants, now or in the p s t . were reducedto a set of 70 occupational titles. Each occupational title includes jobs with a commonality of process and materials handled. Each job entry on a years greater than that of the case). work history could thus be assigned an 352 study subjects. were thus iden- occupational title code, and this was tified. and copies of each of these in- done blind, with no knowledge of the dividual's complete work history were individual worker's status as case or obtained from the company. Two control. methodological problems within this To take account of job mobility, data study relate to the treatment of in- analysis techniques were used which in- formation from these work histories: (1) corporated the complete set of job en- there is, within the rubber industry, a tries on each individual's work history. large diversity of job types and a variety That is, the exposure variable was of a of chemical exposures; (2) there is con- multivariate kind; no attempt was made siderable job mobility, whereby the to assign a "most representative job" average worker moves through 10 to 30 classification to each individual, because different jobs during an employment of, of the large information losses this say. 30 to 40 years. would have entailed. 6 4n 6 1B f Ilcukcm' hand, the analysis revealed an data la you^ with a formula b a d on association of educational k e l with the formula discussed by Mantel a d cvpc of job d e . Therefom. by treating Hienuel.1J m e , age. plant, and eduutiorul level as Table 5 lists the valucs d the dative I covariables along with the case-vcnus- risk for two different times of first ex- control variable, it was found that, when posure (first, exposure to salvents at any adjusting for the effects of these time; and first exposure at some time be- covrriables, the lymphatic leukemia tween 1940-1960). The 1940-1960 cases effectively spent a k w l four years period was chosen as a likely "etiologic mom in sdvent-crposun groups than perid' for deaths occumng fromthe did the controls. This adjusted difference mid-1960's onward, on the assumption (although less than the seven years in- of a 5 to 20 yeav latency pcmod, as L I J dicated in the initial tabulation of work- histcry data) was highly statistically significant (pI.005). Although the number of matched Innphatic kwkemia sets is not law.a series of simplified matched set analyses were suggested by the a p x i m r t d y 10 to 12 year latency periodfor radiation-induced leukemia. For both exposure-time definitions, there is a dose-response relationship, with the high solventexposure OTC having the highest .-ea*h%LIr*mr-gu(ara (,*as)a 4 estimated. using the relative odds a p proximation*' (and ignoring individual matching), the overall value for sdvent- exposed versus solvent-non-exposed carried out to compare the estimated relative r i s k Overall, the relative risk for worken is 2.1. Howevec, it must be relative risk for the three individual solventsxpored worken compared to stressed that these estimates are based solventsxposure groups. Relative risk non-exposed workers appears to be of on modest numbers, and statistical tests can be estimated from the matched set the order of 3. When relative risk is indicate that the distributions of cell frequencies used in estimating the relative risk faures in Table 5 do not achieve statistical signifiunce, atcept for the two high solvent exposure OTG figures (5.50, 9.00). Finally, of the 19 individual oc- cupational titles included within these three solvent-exposure groups. one showed r marked difference between cases and controls. For tire repair jobs, there was a sixfold discrepancy between uses and controls. The repairing of tires involves considerable swabbing of tires with solvent, and, at some time in the past, the solvent predominantly used was benzene. Of further intemt, each of these four uses had staurted working there only at some time since 1945, whereas the two controls had started working there before 1945. This may in- diu& that the leukemogenic agent is a chemical that has bem p r p & r e p 1 --des. and may still be present; rather than one that was eliminated prior to World War II or earlier. Conclusions - Within this case-control study, prompted by an observed u r n s mor- tality from cancers of the lymphatic and henutopoicti systems, the urociation of thee tvpn of arum with specific work-cnvironment exposure has been investbted. Of the various possible dationrhips examined, the one dht stands d h the usociation of death from lymphatic kukemir with a history c- of having world in solvent- \- jobs. Both the mrtchcd quadrrrpkt analysis and the multivariate analysis of -- - - --L . * covariance cImonstrated a Ggnificant tendr to be ofe i t k thc ~ t o b l r s t c c rtom t h t h m crrmc ty*. for the positive association of lymphatic (or stem cell) kind or the mytloblasttc total cohorl d rubber workm. could leukemia with haviv worked in these kind The association of lymphatic well indicate the presence of carjobs. (The fact that both types of analysis leukemia with exposure to organic cinogcnic agents within the work m-f are similarly informative about this chemicals appears lo have no reported vironment responsible for various can- association reflects the fact that of all the precedent. c m of the lymphatic and hetnatopoiaic potential confounding factors included Thirdly, the reconciliation of the lym- system. Further uudy i s necessary. in the series of multivariate analyses, the phatic leukemia SMR (which i s a dative Ones bund to be most associated with risk of mortality from that UUK of derth the exposure variable were three of the for rubber workers versus the general -four that had been initially used in the community) with the relative nsk for matching process that is. sex, race "cxposecr'versrnpaposed"rubber and plant.) The nsk of death from lym- workers poses anz interestiq question. phatic leukemia appears to have been The SMR indiut a seven-fdd mortality 3approximately seven for worken in high risk for all active- e (ages 40to 64) rub- solvent-exposure jobs (compared to b e r workers; yet the relative risk other workers), and approximately twofold for workers in medium or light solvent-exposure lobs .calculated in Table 5 indicates a mere three-fdd ~XCCSSfor "exposed" w o r k ~ References 1. haummi IFand Mille, Rw. Epdemidqy when a11 17 cases are examined (and a of human k u h i r : Recent obse~vations / The detailed scrutiny of lymphatic leukemia work-histories yielded some evidence that the greatest contrast between cases and controls was for iobs in the Tire Repair Occupational Title. Fur- two-fdd excess for ",posed" workers when the 10 active-age quadruplets are examined). If one assumes that "non- exposed" rubber workers are at the same mortality risk as the general community, Nat Gncer /nsr 3B:593-605.April 1%7. 2 Kesskc KI a d Lilimfeld AM. Perspectives in the epidemiology of kukema. Advances in Canm Research 12:225-302. 1%9. 3 Clarkson BO and Burchnal IH: Progress in leukemias. Proems in Clmcal G n c o ther. it appears that the relevant atposure then, logiully, one expects the "ea- 1W5-663. 1965. agent (probably an organic solvent) may posed" subgroup of workers to be at a be one still present, if not first in- risk somewhat greater than seven-fdd. 4 Bubnk F: Patterns in cancer mortality in the US.:1950-1%7 National Cancer Insrirue Momgraph 33. US. Covrmment Pnnting Of- troduced, in the period since 1945. That the calculated figure is less. rather fice. May. 1971. Modern solvents used in the rubber than greater, than seven could either 5..MacMahon 6 W n p h i c variation in industry include xylene, toluene, trichlorethylene, a d various aliphatic hydrocarbons. While these are generally reflea the limitations imposed by the size of this data set. or, it could indicate the presence of a more widely leulamia mortality m the U.S. Public Health Repon,. 72:39-46.1957. 6. Fomi A el al: Chromoromc studies in workers exwed io benzene or tducm or i believed. to be non-carcinogenic, this distributed leukemogen within the rub- both kch Envim Hlfh 22:373-378,March belief may partly have arisen because in- ber industry. involving many of the lobs 1971. sufficient epidemiologic study of the carcinogenic potential of these solvents has in the "no solventaposure" groups. That is, there might be a heightened 7 V i l i n i I C and k i t a C: & n r w a d leukemia. Ncw Enel / Mcd 271:872-876.1964. 8.MacMahon 6. Epidemiology of ye^ been done "background" risk of leukemia that af- Hodgkin`s disslse. Cancer Research 26.1189- Four other observations should be fects most, if not all. rubber workers. 1m.1966. made regarding these results Firstly. these results call for several follow-up tasks It is clearly important to while. for certain sub-groups of workers, there is an additional super-imposed risk that arises from t k r specific job chirrac- 9.Nmdl C and Rawlings W: EviMce foc enwronmental faaon in the e t d o g y d Hodgkin`s disease. / Chon D i s 25.261-7.May 1972. determine whether a similar association teristics. This question is currenth under IO.Hishtyarna H et al: The incidence of exists within the data availabte from the other rubber companies within the research program. Meanwhile. information must be obtained about the follow-up. and necessarily includes further environmental assessment. Finally. although this study has focussed upon lymphatic leukemia and malignant mphoma and multiple mvelomr in HirOShimJand Nagasaki a t m K bomb sur- vivm. 1945-1965 C i n c ~ r32:lSl-9.1973. 11. Akauki K and Wakru H: frequmcy d Iytnphomicular tumors and leukcmias in . .. details of chemicals (apecialty solvents) solvent exposure, it is not possible to lapan. / Nat Cancer lnsr 9:339-343. 1974 currently and previa& wtd in each of discount an association of lym- 12 McMic)uel AI. % M 5 R a d K u W 11: the jobs within the solwnt-etposure phosarcoma. Hodgkin's Disease and An epidemiologic study d mortality wittun a cohorr of ~ b b t tworkm. 1964-72.IOM - groups. myeloid leukemia w i t h work- 16:458364.kly 1974. Secondly. it is of interest-that it is lym- ehvironment aposun within the rubber 13.Miettinen 0s. Estimrtion d dalivC risk .I phatic rather than myeloid kukemu that industry. Within the data set described from individually matched Icricl. Uiometrics appears to be associatedwith solvent a- above. there were u sof these posure jobs. lhe spame literature in h i s three c a w types to allow any con- a m indicates that. at least for case of clusion of no association to be drawn. 2b:7566. 1970. 14.Comfdd I: A m*hod d estimating comparative rates from clinical data. & pbai' rn to cancer d lung. b m r t and cervix. gross benzene poisoning, the leukemia Indeed. the ObKlVed acess mortality / Nat%f~er /rut 11:12h9-1275.1951. - e- -- d.