Document yrJqNdMGmBBK2vJeDbEjL9wZ2

Inr. J. Cancer: 13,5-ij-559 (1989) 1989 Alan R . Liss. Inc. --%.fh ?ubllcaliOn of me laternattonal Unmn Apatnsr CdnCOr Publication de !'Union ;nternattanala Contra 10Cancar A CASE-COKTROL STUDY OF MULTIPLE MYELOMA NESTED L i THE AMERICAN CANCER SOCIETY PROSPECTIVE STUDY P BOFFETTA2:.3, S.D. STELL.MA.Ua:nd L. CARFIXEL' :Depurmtmt of Epuiemolop a d S[arLs:ics.. - \ m e r u n Cmcer Soaen. I I80 Avenue ojthe .-\mericus. ,Vew York. .W 10035. L'SA: urn'2Dipum'menrodi Scienze Srornrdiche e Oncoloqiu Umana. Canedra di Epidernioiugiu dei T m o r i . L'niwrsita di Torino. 7 ~ i a h u e n u 7 . 10126 T z n . [.mi\- Among the subjects enrolled in che American Cancer Sociecy Cancer Prevention Study II. a large nation-wide prospec- tive study, 282 died from multiple myeloma (MM)during the first 4 yean of follow-up. These were divided into incident cases who were initially free from disease and prevalent cases who reported M M or related symptoms at the time of enroll- ment. For each case, 4 controls matched for age. sex, ethnic group and residence were randomly selected. Previous history of diabetes [odds ratio (OR) = 2.01 and employment as farmer (OR = 2.7) were the risk factors consistently sug- gested by the results of the analysis. The risk associated with farming showed a linear trend with duration of exposure. Pesticide or herbicide exposure was not a risk factor per K but, when combined with farming, it increased the OR to 4.3. Low education, occupational exposure to dyes, and employment in a bank, on the railroad or as a maid were also suggested risk facton, but either they were not statinicaily significant or they were based on small numbers of exposed subjeca. No association &tween M M and asbestos exposure was observed. .\.lultiple mlleloma j M M )i s a malignant neoplasm of plasma cells, usually generalized. mainly involving bone. bone marrow and ludney. It results from JII unbalanced proliferation of a clone of B-lymphocyte-derived ccIIs. that normally synthesize and secre:e immunogiobins. An sstimated 5.100 .-\mencans .will have died of .MM in 1988 1.iCS. 1988); the junlval sxpenence is quite p r : median jumivd time ranges between 11 md 15 months. depending on rumor mass ,.Cfc!nrye. 19851. M.CI is more frequent in the elderly. and among men and Blacks iPonern md Blattner. 19851, and is among :he few cancers whose . n u d i t y rates have jteadily increased over recent decades 1 Devesa f r of., i 987). Several epidemiological studies on the etiology of .LfM nave recently been published. They investigated the role of environmental and occupational exposures-mainly radiation. paper and wood industry. farming. petroleum industry-as well as familial and genetic factors. Two reviews of these studies have dso been published (Blattner. 1982 POKern and Blattner. 1985,. Results presented here are based on a case-control analysis nested in a nationwide. o n - g m g , prospective rnonaiity study. --- -_ \I.ATERI;iL AND \lETHGDS In 1382. more than --.(3XI.imenc3n C m x r Secie? I ACS J volunteers tnrolled OVC: I . 2 million msn m d women i n a prospectiLe mortality j:udy. Cancrr Prevention Siudy I1 (CPSII). Tnose enrolled. who were usually fnends. neighbors, or relatives of the volunteers. compiled 2 1-page confidential questionnure on history of cancer and other diseases; occupational history and exposures: current and past use of medicines and vitamins: menstrual and reproductive history; current diet: history of dnnlung, smolung and numerous other habits. Enrollment was by family groups with 3t ! e s t one person aged 45 or more. and all members of :he household over 30 years of age were asked to participate. Enroilment of subjects was carried out in all 50 of the United Sates. the Dismct of Columbia and h e n o Rico. Every second year. s:olunterrs re asked yo check if the enrollees u e dive or dead. ind IO record date m d place of death. Death ,certificates were jubsequenrl? obtained hrouyh S a t e health depanments. The first 3 follow-ups took place in I984 u d 1986. Three questions on occupation were =ked: current job. last occupation if retired. and any other job held for the longest penod of time. Occupations were coded using an ad-hoc jysrem with 65 different categones. .A more specific classification. such as a j-di_pit Standard Indusmal Classification code. was impractical for the large number of ambiguous responses denved from a self-ddministered questionnaire. Eve? subject was classified according to reported employment in any pend of life ("ever emp1oyed"j. as well as according to "usual" occupation. The laner was obtained either from current job. if held for 10 or more years. or from the remaining information. Exposure in the workplace or in the daily life to m y of 12 zroups of substances iasbestos. chemicalsiacids/solvents. coal ;r stone dusts. coal tar;pitch/asphalt, diesel tngine exhaust. dyes. formaldehyde. gasoline exhaust. pestic:des:herbicides. textile tibersidusts. wood dust. X-rays,radioactive materials) wxs investigated in the following section of the questionnaire. The duration of every employment and exposure was also asked. Causes i f death 'were derived from death certificates and c..coded : a s.i.I i-Cby2 t J io r n r in ~t ed nosologist Diseases. 9th Jccordiny Rsvision. ro [he International The present mdysis is based on 508.63- men and 676.613 women who provided usable questionnaires. Four-year mor:alitl; q x n e n c e was available for 18 out of 58 ACS divisions in which :he study is o r p u z e d idivisions <orrespond to States or to some large metropolitan arexr. while IO divisions had data for 1 years only. Subjects who had died. with 41hi reponed on death certificate as either the underlying or contnbuting cause of death. formed the case senes: 1 randomly selected subjects were matched to each case for sex. ACS division. year of birth and ethnic group. to form the control group. The inc:dent nature of the cases was ascertained throush a manual check of the questionnaires in which any intomation about pnor or present cancer %as reported: Selected cases were thertfcre divided into "incident" vld "prevalent" cases. and % e r r xiaikzsci Xp"ratL.i> Dara from p e ~ a l e n tcases. '*hose recall $IC past exposures ma;: be hlased by knov. ledge of disease status. 2nd *hose current habits m d conditions may reflect the disease itself. were analyzed in order to confirm or disprove results from incident cases. .i?-step procedure of analysis was used: crude unmatched odds ratios ( O R ) . hlantel-Haenszel matched OR (htantel and Haenszel. 1959). and conditional logistic reFession f Breslow 'To whom repnnt requests ,hould s m . 31 :he Lruversity of Tunn. Received: July 4. 1988 m d :n revcsed fonn Yovernber 3. 1988. STUDY OF MULTIPLE MYELOMA 555 and Day. 1980). In order to control for potential confounding factors other than matching variables (education, history of diabetes. X-ray treatment, farming. pesticide and herbicide exposure). the regression was carried out using the PECitV program (Lubin. 1981). Test-based 9 5 9 confidence intervals (CI)were calculated according to Miettinen (1976). Linear trends were tested according to Mantel (1963). The ,MannNkitney test was used to test differences between distnbutions. Adjusted population atmbutable risks were calculated using logistic regression coefficients iBruzzi ef al.. 1985). ESULTS After 4 years of follow-up. 98.5% of SubJeCtS were traced. and death certificates here obtained for about 84% of deceased subjects when this analysis was started. At the '-year followup. 99.1% of subjects were traced and 96.8% of death certificates were obtained. In order to assess the quality of death certificate coding, about 25% of cases' death certificates were manually checked, and no errors were detected. A further check of selected items from about 10%of cases' and controls' questionnaires showed a very low data entry error rate for most variables. with the exception of the information on previous cancer: this variable was manually checked for all subjects. MM was mentioned on the death certificates of 281 subjects. Nine of these did not have 1controls satisfying all marching cntena: therefore 8 were matched within 2 years of age. and I within 5 years. The number of MM cases on which this analvsis is based (282) is slightly hgher than the number of expected deaths according to 1983 age. sex and ethnic group- specific US death rates (261). but only 249 cases had MM as the underlying cause of death. the information on which national mortality statisrics are based. Of the cases. 149 reponed a previous or current history of MM. and 5 more subjects stated symptoms which can be related to .MM. without mentioning the disease. These were all combined as "prevalent" cases. The remaining 128 cases, with no mention of IMM or related symptoms. were considered to be "incident" cases. Unless otherwise specified. all results reported below refer to incident cases. The numbers of incident and prevalent cases and controls by sex. ethnic group. age and year of death are shown in Table I . As may be expected, prevalent cases tended to die earlier than incident cases. Forry-nine variables were used to investigate factors other than occupation. Table II shows those associated with an in- creased risk in this analysis or in previous studies. Low education was significantly associated with MM; subjects without a high school degree had an OR = 1.5 compared with subjects with such a degree. Previous history of 8 non- malignant conditions (diabetes, kidney disease. chronic bronchitis. stomach ulcer, tuberculosis. hay fever. asthma, and arthritis, was analyzed. Only diabetes showed a significant association with MM. a finding that was c o n f i e d among prevalent cases (OR = 1.8. 95C CI: I .C-3.2). Kidney disease was strongly associated with M.M among prevalent cases (OR = 3.1. 95% CI: 1.2-8.3). but not among incident cases. No association was observed with current or former nge---, rette. cigar. pipe. or total tobacco smohng, nor with consumption of milk. caffeinated or decaffeinated coffee, tea, diet or non-diet sodas. beer. wine. hard liquors or total alcohol. Significantly more cases than controls reponed a private well as the source of drinlung water. while no difference in the consumption of food nch in protein or animal fat. of fned foods, fruits and vegetables or artificial sweeteners was observed. No significant difference in the history of X-ray examinations or X-ray treatment was observed among incident cases and controls. while both exposures were strongly associated with M M among prevalent cases: OR = 4.0 ( 9 5 9 CI: 2.65.9) and OR = 5.9 ( 9 5 9 CI: 4.C-8.S), respectively. No excess nsk was found according to mantal status, religion. or indicators of social activity (number of mends to bhom the subject was feeling close, church or temple attendance. participation in group or club activities). .Moreover. no difference was found in subjects' or parents' place of birth: specifically. 5 cases and 70 controls either were born or had parents who were born in Nordic countries (OR = 1.0). NO association was found for weight. change of weight in the year prior to enrollment. height. history of operations. or family history of cancer: specifically. the number of reported MM cases among parents and siblings was 0 out of 40.5 relatives for incident cases. 32.435 for :her controls. 2'765 for prevalent cases. and 02.919 for their controls. : TABLE I - SUBJECTS ENROLLED IN THE LV4LYSIS B Y SEX. ETHXIC CROUP. AGE, YEAR OF DEATH. . l V D NC!DE4T UATL'RE OF THE 3lSEASE Incident cycs Conmlr Prcvalenr C3X5 Conmls Sex Male Female Ethnic group White Black Other Age 30-54 55-63 -65-74 75 Yex of death 1982 1983 1984 1985 I986 Tord 'Nm applicable 74 54 116 9 14 29 59 26 4 24 10 28 32 118 296 716 M1 36 I' 56 I I4 238 I04 SA' 512 83 71 -143 4 I' 61 49 32 21 61 37 21 I4 I54 332 234 572 28 16 48 245 195 I28 NX 616 TABLE U - Y L V l P L E MYELOMA RISK A N D LXPOSCRES OTHER OCCUP4TION I Y C I D E X C.\SES Education' 45,137 1.5 1.0-2.3 Histor) of disease': hay fever l4;38 1.6 0.8-2.9 asthma 5/21 1.0 0.3-2.7 a;rhntlS 3911fA 0.9 0.6-1.5 diabetes bdney dis. li38 51 14 1.9 I1 o1..-1c-J3..o4 Tobacco srnolung' 69'289 0.9 0.6-1.3 .Alcohol itnnking' 201133 0.5 0.%0.9 Dnniung water supply8 31j80 1.- 1.1-2.7 D t e q haDlKS" animal fat 598,242 1.0 0.6-1.4 fned f d s 43, I76 1.0 0.6-1.5 fruits-veg. 6 3 1 2 3 1.2 0.8-1.8 X-ray treament'O 14/40 1.1 0.8-2.6 X - n y examinations" 62'259 0 . 9 0.6-1.4 'Excused a s e s md con~ls.-:~anrel-Haenstelmatched odds rauo.-'95% confidcncc :ntewd - ' L p s e d . less than high-school grade Nor cxpoud: hlsh-uhool g n a e or higher -'Exposed. positive hrsrorv Not ex@. no h;story.-'bjipowd: ever smoker Nor exposed. never smoker - ' E x p o d : ever regular dnnker Nol ex- posed never rcplar dnnker.-'Expoud: pnvau well. Nor c x p o d : public walcr suppl? i)r other -"Exposed: consumption idays'wek] of faods above the median of the :onsumption of combined cases and c o n m l s . %elxposed: consumption below the median ,Animal fat. mepi. sausages. cheex. buntr. eggs.-'"Gposd: posiove hislorv of m3mcni ,Vor expos&: no hislory of ucarmeni.-~iEnposed: number of X-ny'examinarions above b e median of the number oi cxarmnanons of combined cases m d mmols Nor exposed: r o d number below the median. 556 B O F F E ~ T AET A L Among women, no significant difference was found according to use of hair dyes. oral contraceptives. or exogenous estrogens. or according to menopausal status and difficulty in TABLE IV - MULTIPLE MYELOMcA.LREIS.sK ASD OCCUPATION: INClDLq (kcupaoon' CUC2 OR' 95% cr conceiving. Table 111 shows the association between >lM and 12 occu- Teacher Manager 10159 0.6 0.3-1.3 6/5 1 0.5 0.2-1.1 pational exposures. Self-reported exposure to pesticides or her- Office worker 14i45 I .3 0.7-2.7 bicides was just significantly associated with MA4 @ = 0.05). Bookkeeper 3/16 0.8 0.2-2.5 and a non-significant increase was also present for exposure to textile fibers and dusts. wood dust, X-rays and ndioactive materials. and dyes. The absence of an ssociation with asbestos was c o n f m e d for prevalent c s e s (OR = 1.3. 9 5 8 CI: 0 . 6 2 . 9 ) . The risk for exposure to pesticides or herbicides was higher among prevalent than among incident cases (OR = 3.4. 9 5 8 CI: 1.5-6.3. 19 cxposed cases). Bank employee Real estate agent Registered nurse Sales worker Pharmacist .Auto mechanic Truck h v e r Rylroad worker 514 5.0 1.1-16.3 3i8 1.s 0.15.4 3 10 0.8 0.2-3.6 12/17 I .o 0.5-2.0 35 I .6 0.3-8.1 31I3 0.9 0.2- 3.5 ?I3 4.0 O . S l 7 . 6 312 6.0 1.3-28.9 Among 16 occupations with at !east 1 exposed cases and 2 exposed conaols (Table IV), farming had an OR of 3.4 (95% CI: 1.5-7.5). Other occupations showed an increase in risk. based on a few cases and conwls: maid. bank worker and railroad worker cp < 0.05). truck driver and janitor ( p > 0.05). Among prevalent cases and their controls, there were 13/27 farmers iOR = 2 . 5 , p < 0.05). 014 bank employees. Fmer Janitor .Maid Food worker !628 U6 4/4 111 I 3.4 1S 7 . 5 2.7 0.3-9.0 5.0 I .2-2I . 1 I .5 0.5-5.0 ' M ( Yo~ccupanon dunng life i y e u x t for dera~lst..YOC exposed: for m y occupao m . dl other subjects. Only occupanons .with at least 2 exposed c u u and conmls are listed jcpmtcly-'Expowd cases 4contmls.-'Manul-Haenuel mtcMd odds 1ano.-95'% conridencc mterval. 118' truck drivers, 011 railroad workers, and 213 janitors !OR = 1.7. p > 0.05). One incident case and 3 controls were assemblers. as were 3 prevalent cases and 1 controls (OR for prevalent cases = 6.0, p < 0.05~.When this analysis was repeated. classifying subjects in every occupation that was re- ported in the occupational history I "ever employed"), only minor changes in the results were detecred. The relationship between farming and .MM was specifically addressed by seveni further analyses. If incident and prevalent cases are combined. the OR for farmers is 2.9 (95% CI: 1.75. I); 18 incident cases and 38 conrrols were "ever employed" as farmers (OR = 2.7, 95% CI: 1.3-5.7). Xn association with pesticide and herbicide exposure is suggested among farmers. but not among non-farmers. In Table VII. results from a conditional logistic regression analysis on selected risk factors are reponed. After controlling for possible confounders. the ORs for level of education, drinking water supply. and occupational exposure to pesticides or herbicides. textile fibers. wood dust and X-rays were lowered: therefore these associations appear to be confounded by the factors controlled for. namely employment as farmen. while the ORs for previous history of diabetes, exposure to Table V shows the association between farming and blhl dyes, farming and other high-risk occupations were not sub- among incident cases, by duntion of fanning md blnh cohort. stantially modified. It was not possible to determine the duration of exposure for 5/16 exposed cases. and lli38 controls: a trend in risk with increasing duration is nevertheless suggested. When fanners DISCUSSION are analyzed by birth cohort. m increase in risk is present only among subjects born before 1910. Given the small number of subjects, it was impossible to separate the tffects of age and duration of exposure: they may merely retlect a higher risk in older persons. The relationship between f m i n g . exposure to Follow-ups in CPSII are carried out very carefully: the proportions of subjects who were traced and for whom death certificates were obtained u e among the highest ever reponed in large prospective studies. pesticides or herbicides, and Mhl risk is shown in Table VI. Questionnaires were filled out by subjects themselves. data were entered by specialized personnel. and causes of death were coded from death certificates by a trained nosologist; several manual checks showed that the overall quality of data -, TABLE Ill MULTIPLE .MYELOMA RISK LVD OCCPaTIONAL E.YPOSLRES is high. INCIDE.LT C.4SES Characterization of cases as incident or prevalent was an LpoSUfC CJiCJ OR' 9% CIJ important step in this analysis, since subjects diseased at en- -Pesticides and =. . -.hrrbicidrs Asbestos Diesel exhaust Tmile fibers And dusts Wood dust 12'25 31 17 1413 8; 19 923 rollment might recall exposure history in a different way from 2. I 1 . 0 4 . 2 subjects free from disease. When results from incident and 0.7 I .4 1.3 0.2-2.4 0.7-2.7 O.M.4 prevalent cases are compared. however. it seems that major findings of this analysis are consistently present in both groups. and that discrepancies in the results reflect actual differences betwesn the 2 groups. Prevalent cases were more likely to 1 .s 0 7-8 3 report a history of kidney disease, which is a major clinical X-nys. radioactive 7 ; l j I .? 0 . 8 4 . 8 manifestation of MM. They were also more likely to have materid Chemicals, acids. solvents Coal or scone dusts Coal tar. pitch. asphalt Dyes undergone multiple X-ray examinations and X-ray treatment. 8158 0.5 0.2-1.1 as is expected with .MM patients. 612d 1/10 I .o 1.7 - 0.5-5.6 This analysis did not confirm the association between chronic immunologic disorders. such as rheumatoid arthritis and asthma, and .Mhl, that was suggested by some case reports 5,8 2.7 0.9-8.6 (Imahori and Moore, 1972) and epidemiological studies (Iso- Formaldehyde 4 9 I S 0.6-5.7 malu et al., 1978; Hakulinen et a!.. 1985). but not confirmed Gasoline exhaust 1461 0.9 0.5-1.6 by another study (Linet er al.. 1987). Only a generic definition ' E x p o d CJKS and sonaols.-'Mantcl-Hacnsrcl matched odds ntio.-95% confidence intcwal. of "arthritis" was available from the questionnaire. No information was available from this study in order to test the hy- STUDY OF .MULTIPLE .MYELOMA -TABLE V MULTIPLE MYELOMA RISK .AND FARWL~G.E m C T OF DCRAnON OF MPOSL'RE AND Y M R OF BIRTH: LXCIDENT C.4SES Durations Yeu of birth <I910 I9Io-IY19 >I919 Not exposed' 1-20 years 2 1 1 0 years 41 * yeus .Missing duration Not exposed' Exposed Not exposed' Exposed Yot exposed' Exposed CUCO' OR: 9 5 9 CI' 1 12'483 016 -1.o - 25 1.7 0.3-8.9 6 6 4 . 3 1.5-12.5 '8;l I 3 4 154 3.1 I .o 1.3-- 7.7 IO 13,158 3.2 1.o 1.2--5.6 5'!3 35.142 '-.I f, I .o 1.1-6.7 - I >5 0.8 0.1-7.2 'Cases and connois.-'MYliel-Haensrel marchedd d s ntio.-%lt confidence inumd.-'Reierrnce caiegory.-'Tesi for lincv mnd: p-value <0.05. 557 --.--- pothesis, derived from case reports. of a role of vaccinations epidemiological studies (Milham, 1971: Gallagher er af.,!983; (Woodroffe. 1973, or sarcoidosis (Petterson et af.. 1987) in Burmeister et af.. 1983; Cantor and Blair, 1984; Pearce er af., the etiology of MM. 1985; Delzell and Grufferman. 1985; Steineck and Wiklund, The association between MM and previous history of dia- betes has not been previously reponed; in fact. Gallagher et d . (1983) reported a negative association. The present result may be due to chance only. The increase is consistent among both incident and prevalent cases. and the prevalence of history of diabetes is not lower among controls than in the entire CPSII cohort. However, results from other studies are needed before one can speculate about a possible biological interpretation of 1986; Nandakumar er al., 1986; Rodin et af.,1987; Mctaugh- lin er af.. 1988; Cuzick and De Stavola, 1988). Three other studies. however. did not c o n f i i this finding (Tollerud er ul.. 1985: Linet er ai.. 1987; Brownson and Reif. 1988). This association is strongly confirmed by the present analysis for subjects born before 1930: it is present for both incident and prevalent cases. and there is a trend toward increasing risk with duration of exposure. this result. which may be related to ;igeneral alteration of the Two different hypotheses have been proposed in order to immune system. identify a specsfii ddogx agent: expoupc to chermcals, such No association was found between M M and origin from as pesticides. insecticides and herbicides (Bumeister er af.. Y Nordic countries. as suggested by an ecological study (Blattner 1983; Cantor and Blair. 1984; Moms et uf.. 1986; Flodin et et uf..1981). but not confirmed by an analytical study (Nan- af..1987). and exposwe to some viruses (Pnester and Mason, dakumarer d.. 1986). nor was the suggestion of familial clus- 1974). Both explanations have some bioiogd plrtrslbdity, tering of MM,derived from case reports (Kyle and Greipp, given the known carcinogenic effect of agSaiE.dmmjds. 1983) and one case-control study (Bourguet er af..1985). con- such as knzcnc. 011 hemampkw d l s . the B CT- firmed by the present analysis. Subjects who had not aaained high-school education were at hi- rislc of MM in cotnpm~onwith subjects having received cino enicity of wvcrai PQStlcib DD?. and the suggestion of a carci herbicides on softtissues (IARC. 19%. hsrbig nic effect of phenoxy education. but the difference was not statistically sig- The evidence of vval myeloma-related diseases in laimpls t. No association with education was found in 2 studies (Shanna er ai.. 1973). and of horizontal transmission of their (Johnston er af., 1985; Nandakumar er al.. 1986). while an increase of M M risk among subjects from lower social or cco- nomic class has been reported (Pearce er af., 1985: Nandaku- TABLE VU - ML'LllPLE MYELOMA RISK .AND SELECTED RlSK FACTORS. mar er uf.. 1986). It is likely that education level is a proxy REXLTS OF LOGISTIC REGRESSION .U4ALYSIS1I.NCIDt37 CASES variable for environmental or occupational factors other than those analyzed. The lack of M association between MM and life-style facton such Y smkmg, drinking and diet confirms the findines from mvious studies (GallagGer et ai.. 1983; Linet et af.: 1987; Fiodin er af.. 1987). bst f r i & Education Previous history of diabetes Tobacco stnolung ..Ucohol dnnlung Dnnlung water suppiy urnno 1.4 2.0 0.9 0.6 1.3 954~1' 0.9-2.2 1.1-3.8 0.6-1.4 0.3-1 .o 0.8-2.' Fanning was suggested as a risk factor for MM from ecological I Agu er af.. 1980: Blame: et af.,1981) and analytical Fruits and vesetables in diet X-ray matmenr Occupational exposurc IO: 1.2 o . a i . 8 1.6 0.8-3.0 pesucidevherbicides 1.6 0.7-3.7 -TABLE V I MULTIPLE MYELOMA RISK, F4RMING. CUD PESTICIDE AND HERBICIDE EYWSURE. INCIDEYT CASES texule fibervdusts wood aust X-rays, radioacuve mat I .2 0.4-3.3 1 .2 0.5-3.2 1.4 0.5-3.9 linexpored IO pcsricides and herbicides hpud io pcsucides and herbicides dyes Occupation: farmer truck ilnver 2.9 0.8-10.3 2.7 1.3-5.7 2.8 0.5-16.1 Non-farmers Fmen Cases/conuols OR' 95% c13 Caxs/convols 1081467 -1.O' 8/20 41/.1o7 0.3-3.1 81'8 janitor bank employee maid railroad worker 2.3 0.6-8.5 5.5 1.5-21.1 5.6 1.2-26.0 7.1 1.243.6 OR' 95% CI' 1.7 0.84.0 4.3 I .7-10.9 'ORs controlled for age. sex. ethnic group. ACS division. educmon. history of dnbeies. X-ray manneni. psocide and herbicide c x p o s m and farming -:For deri n i i i m of e x p o d and not exposed ca~cpones.scc Tables 11. 111 and IV - 3 5 9 ihmaichcrl odds niio.-'Referencc C3IegOn . - I 9 5 9 confidence iniervd. confidence intend 5 5 8 BOFFErn.4 ET 4L causal agents (Piper er af..1975~t.he correlation between MiM some ( b i b e n . 1966: Gerber. 1970: Linet et af.. 1987) but not mortality and the poultry industry found at the ecological level all iSelikoff er ai.. 1973: Gallagher e t a / . . 1983: Morris er ai., (Priester and Mason. 1974). and the suggested role of viruses 1986) studies that investigated this association. Positive studies in human leukemias and lymphomas (Evans. 1982) make plau- reported a 3-fold increase in risk associated with exposure to sible the hypothesis of a viral etiology of MM.No information asbestos. This analysis had a power of about 52% to detect on animal breeding within the farm industry was available in OR = 3 among incident cases. and a power of about 90% to this study. and therefore the v i n l hypothesis cannot be further detect an OR = 3. investigated. Occupational exposure to dyes was juggested as a nsk factor Possible interpretations of results concerning exposure to for Mbl: the association was present among men only (5cases pesucides and herbicides (Table V I ) , in which the separate and 6 controls exposed). and confirms 3 previously published effect of each group of chemicals could not be distinguished. result (Moms er of.. 1986). It is impossible from the present are that pesticides or herbicides have an effect among i m e r s analysis. however. to :dentify a speciric causal agent. only. possibly in relation with higher doses (the median dun- Ionizing radiauon is perhaps the best-documented 2tiologic tion of exposure to pesucides or herbicides was 29 years among factor for blM: the strongest evidence comes from studies on farmers and 10 years m o n g non-farmers. the difference being mortality of atomic-bomb SUrViVOrS (Ichimaru et af.,1982) and not statistically significant). or that other ,pecific chermcds or of radiologists and other occupationally exposed categones nsk factors related to farming are involved. ilewis. 1963: Cuzick. 1981 ) . X-ray therapy was found to be a However. self-reported exposures to specific chemicals may suffer more easily from recall bias than information denved from occupational history Exposure to pesticides or herbicides was associated with a higher nsk m o n g prevalent cases than among incident cases: this fact suggests caution in the interpretation of findings based on self-reported occupational exposures The moc:ation between MM and employment as a janitor has been previously reported (Fnedman. 1986). as has the association with engine exhausts, possibly related to truck dnving (Rodin t t af.. 1987). Given the small number of exposed cases and controls in these occupations. any interpretation of the present results may be cautious. Persons employed as cosmetologists iGuidoni er al.. 1982: Spinelli et af.. 1984: Pearce er si.. 1985). in the petroleum industry (Thomas er af.. 1980: Decoutle er af.. 1983) and in the pulp and wood industry (.Milham, 1976: Tollerud et uf.. 1985; Yandakumar et ai.. 1986: Fiodin er af.. 1987) were not at higher risk in the present malysis. since no cases and 4 controls were classified as kautician. cosmetolo_eistor barber, no cases and no controls a oil tield or refinery worker, and one risk factor for MM in one study (Fnedman, 1986). as were "heavy" X-ray cxaminations (Rodin et af., 1987). One other study iCuzick and De Stavola. 1988) was substantially negative. Our study did not provide evidence of an effect of X-ray exposure in the clinical setting, and *e negative result on occupational exposure to X-rays or radioactive materials (OR = 1.4. p > 0.05j must be viewed in the light of the limited power to detect an association [about 5 6 8 to detect an OR = 2 ) . Our study suffered from some limitations: diagnoses of cases were based on death certificates only. the self-administered questionnaire gave only limited information on chronic diseases. diet and occupation. namely farming; the small size gave only a limited power to analyze in detail many associations. However, employment as a farmer and previous history or diabetes x e the factors that consistently appear to play a role in the etiology of .MM in the CPSII cohort. Since few subjects reported these exposures. the population attnbutable risks to them are low (about 4% to each factor). and the ovenwhelming majority of btM cases are likely to have been caused by other factors. case and 8 controls as x p e n t e r . lumber or wood worker. The AC?L'4OWEDCEME>TS prevalence of subjects with these uccupations in the entire We gratefully acknowledge the assistance of .MS. M. CPSII cohort is very low. .Mushinslu and .MS..M. Davis in organizing m d training re- In this analysis. asbestos exposure does not appear to be a searchers and managing the quantities of data: Ms. .M.E. Per- nsk factor for MM. l h s result d e s not seem to be related to nas helped in the manual chechng of the questionnaires: Dr. control selection. since the proponion of exposed controls D. Lilienfeld and Dr. B . Terracini provided useful comments closely parailels the overall proportion of exposed subpxts in and advice. We especially thank the 77.ooOAmencan Cancer both sexes in CPSU. Asbestos was a nsk factor for MM in Society volunteers for their role in data collection. REFERENCES A G U . V C.. CHRISTEMENB. L. and BL`FFLER. P .\.. Geog-aphic panems o f multiple myeloma: racial and indusrnal correlates. state oiTexas. 196971. J . nul. Cmcer INI.. 65. 735-738 I 19801. .~?.~EERM-C\Y ~ ~ CSEOCRIETY. Cdncer/bc!s ilnd;iqures-I988 .\CS. New York (1988). 9i.4TTYER. ik' .A,. 4lultiple Tyeloma m d masroglobuiinemia. In. D Sihonenield Md J.F Fraumeni :eds.:. C m r P r ?prdrrnioiog;. d m ' ?reven;!on. pp -95-813. Saunders. Philadelphia L 19821. B L A T T ~ E RW. .A. BLAIR. A. and SIasos. T I.. .LZultiple myeloma :n ihe United States. 1950-1975. Cmcrr. 48. 2547-354 I 1981 I . SOURCCETC. . C . . GRCFFEKWAXS... DELZELL. E.. DELOUC.E.R. and COHEVH. .J . Slultipie myeloma and family history or` cancer. .A case- control Study. C m c e r . 56, ?13?-2l39 I I98S). BRESLOYWE.. and D AY. Y E . . Srurisrical rnerhods in cancer reseurch. vol. 1, The a n a i y s of case-connol studies. [ARC SclentiJic Publication 32. IARC. Lyon f 1980). BROWSON. R.C. and REIF. J.S.,A cancer registry-based rrudy of occu- pational nsk factor for lymphoma. multiple myeloma and leukemia. Inr. J . Epidrrnioi.. 17. 27-32 (1988). Bsczzi. P.. GREEN. S.B.. BYAR.D P . B R I ~ T O NL. ..\. and SCHAIRER. C..Estimating the Fpulation itrnbuuble ~ S fKor multiple nsk f x t o r s using case-concrol data. 4mer J Eprdrmioi.. 122. 9W-914 (19851. BLXWEISTERL. .F. EVERETT. G . D . V \ h L i E R . S F and ISACSO~P. . Selec:ed cancer mortaiity and f m practices in lowa. .4mrr J Epidemioi.. 11%. 72- 77 I 1981). C ~ L T O RK..P and B L . ~ I4R... F v m i n g i n d munality from multiple m>- ?!oms. d case-control jwdv u i t h [he use si Jeath ;enificates. I . u t . CJnctr Insr.. 72. 151-255 I 19821. CCZICK.I . , Radixion induced myelomatosis. .V. Engl. J . .Wed.. u)J. 3 W 3 1 0 11981). C ~ Z I C KI. and DE S T ~ V O L BA.. Sluitiple myelom+a case-control sold!. Brit. J. Cancer. 57, 516-520 ! 1988). D E C O ~ ~ FPL..~B.LATTUER.W A . and BLAIR.A.. Mortality among chemical ,workers exposed to benzene and other igents. Envrron. Res.. 30. 16-25 198:). DELZELL.E. and GRUFFERWASN.. Mortalip among white and non-white farmers in North Cxolina, 1976-1978. . h e r 1. Epdemiol.. 121. 391102 (1985). DEVESA, S S.. SILVERHAX, D.T.. YOUNG.J.L.. POLLACK, E.S.,BROW. STUDY OF MULTIPLE MYELOMA 559 C.C.. HORW. J.W., PERCY, C.L.. MYERS. M.H., MCKAY. F.W. and FRAUMENJI.,F.. Cancer incidence and mortality trends among whites in the United States. 1947-84. 1.MI. Cancer Inst.. 79, 701-770 (1987). EVANS, A S . , Viruses. In: D. Schonenfeld and J.F. Fraumeni (eds.). Cancer epidemiolog? and prevention, pp. 3 6 4 3 9 0 . Saunders. Philadelphia (1982). FLODIN, U.. FREDRIKSONM.. and PERSSONB. .. Multiple myeloma and engine exhausts, fresh wood. and creosote: a case-referent study. Amer. J . mdusr. (Wed.. 12, 519-529 (1987). FRIEDMAN. G.D., .Multiple myeloma: relation to propoxyphene and other drugs. ndiation and occupation. Inr. J. Epidemiol.. 15. 4 2 3 4 2 5 (19861. G.ALLAGHER.P.. SPISELLI.J.J.. ELWOOD. J..V. and SKIPPES. D.H.. .Allergies 3nd agncultural exposure LIS risk factors for mulfiple myeloma Brit. J. C a r e r . 48. 3 5 2 4 5 - 11983). GERBER.M.A...4sbestosisand neoplastic disorders of the hematopoietic system. .-\mer. 1.din. Path.. 53, 2C4-108 I1970). G U ~ I T I S. .. WRIGHT. W . E . and PETERS. J..M.. Multiple myeloma in cosmetologists. Amer. J. indusr..Wed.. 3. 169-171 (1982). HAKCLINEN. T.. ISOMAKI. H. and KNEKT. P., Rheumatoid &tis and cancer studies based on linking nauonwide regismes III Finland. . h e r 1. `Wed.. 78 (Suppl. IAI. 2%32 1985). ICHIMARC. .%ISHIMIA.RC.. T., MiUMI. .M. and .MATSUNAGA,M.. MUltiple myeloma among atomic bomb survivors in Hiroshima and Nagasaki. 195C-76: relationship to radiation dose absorbed by marrow. J. M I. Cancer Inrr.. 69. 323-328 (1982). I.MAHORI. S. and .MOOREG..E.. Multiple myeloma and prolonged stimulation of reticuloendothelial system. .V.Y. Stare 1..Wed.. 72. 1625-1628 (1972). INTERNATIONAL .\GENCY FOR RESEARCH O h CANCER. [ARC iW0nogrUPh.S on the evaluation of the carcinogenic risk of chemicais IO humanr. Suppl. 7 , Overall evaluation of carcinogenicip: an update of I M C Monographs Vols. 142. IARC. Lyon (1988). ~ M A K I .H.A.. HAKCLINEN. T. and JOCTSENLAHTIU...Excess risk of lymphomas. leukemia and myeloma in patienrs with rheumatoid &tis. 1.chron. Dis.. 31. 6 9 1 4 % 11978). JOHNSTON. J.M.. GRUFFERWANS. .. BOURGUET. C.C.. DELZELL. E.. DELONG. E.R. and COHEN. H.J.. Socioeconormc status and risk of multiple myeloma. 1.Epidemiol. Comm. Hlth. 39, i75-178 (1985). KYLE, R.A. and GREIPP. P.R.. Multiple myeloma. Houses and spouses. Cancer. 51. 735-739 ( 1983). LEWIS. E.B.. Leukemia. multiple myeloma. and aplastic anemia in .-\merican radiologists. Science. 142, 1492-1494 (1963). LtEBEN. J . . Malignancies in asbestos workers. .Arch. e n w o n . Hlth. 13. 61%621 (19661. LINET. M.S.. HARLOW. S . D . and ,MCLAUGHLtN. J.K.. ..\ case-conml study of multiple myeloma in whites: chronic antigenic stimulation. occupation. and drug use. Cancer Res.. 47, 29713981 11987). LLBlN, J.H.. A computer progain for the analysis of matched case-conuol studies. COT. biomed. Res.. 14. 138-143 I 1981). MANTEL, N.. Chi-square with one degree of freedom: extension of the Mantel-Haenszel procedure. J. .-her Statist. .-\ss..58. 69&700 (1963). MANTEL. N. and HAENSZEL. W.. Statistical aspects of the analysis of data from rrmspecuve studies of disease. J. nut. Cancer Inst.. 22. 7!%748 (1959). MCINTYRE.O.R.. Mveloma. In: P. Calabresi. P.S. Schein and S . A Rosenkrg (eds.). Medical oncology. pp. 4 9 6 5 2 2 . Macmillan. New Yo& (1985). JM. FCL. AUCHLINJ.K.. LINET.M.S.. STONE, B MALKER. H.S.R., WEINER.J.A. and .J.. BLOT. W.J.. FRAUMEXI. ERICSON. J.L.E.. Multiple myeloma and occupation in Sweden. Arch. environ. Hlth. 43, 7-10 i1988). MIETTISES. O., Estimabiliry and estimation in case-referent studies Amcr. 1.Epidemiol.. 103. 226-235 I 19761. MILHAM.S.. Leukemia and multiple myeloma in farmers. Amer. 1.Epi. demiol.. 94. 307-310 11971). MILH.AMS. .. Neoplasia in the wood and pulp i n d u s p . Ann. .V.Y. Acud. SCI , 172. 294-300 19761 .L~ORRIS.P.D.. KOEPSELL. T.D. DALtYG. J.R.. TAYLOR. J.W., LYOS. --J.L.. SWAZSOS.G.M.. CHILD. M. and WEISS.N.S.. Toxic substance exposure and multiple myeloma: 3 case-conuol study. J. MI. Cancer Inst.. 76. 987-994 ( 1986). NANDAKUYAR4... ARMSTRONG.B.K. and DE KLERK.N.H.. MultiG; myeloma in Western Ausualia: a case-conool study in relation to occupation. father's occupation. socioeconomic status. and country of bmh. Inf. 1.Cancer. 37, 223-226 (1986). PEARCEN..E.. SKITH. A.H. and FISHER. D.O.. Malignant lymphoma and multiple myeloma linked with agricultural occupations in a New Zealand cancer regisa).-based study. Amer. J , Epidemiol.. 121, 23-237 (1985). PETTERSOHT. ., KotvtiNEN. E.. ILVONEN. M.. JOUPPILA, J., AALTO.E. and WASASTJERSAC... Sarcoidosis and multlple myeloma: an association. Brit. .Wed. 1..295. 958 (1987). PIPER. C.E.. ABT. D.A.. FERRER,J.F. and MARSHAKR. .R.. Serocpidemiological evidence for honzontal rransmission of bovine C-type vuus. Cancer Res.. 35, 1 7 1 1 2 7 1 6 11975). POTTER%. L.M. and BLATTNERW. .A.. Etiology and epidemiology of multiple myeloma and related disorders. In: P.H. W i e m k G.P. Caneilos. R.A. Kyle and C..4. Schiffer ieds.1. Yeoplatnc disease of the blood. pp. 41T429. Churchill Livingstone. New York I 1985). PRIESTER. W.A. and .MASONT..J.. Human cancer mortality in relation to poulay population. by county. in IO Southeastern states. 1.rulf. Cancer Inst.. 53. 4 _ C 9 119741. SELIKOFF.I.J., HAMMOND.E.C. and SEID.WANH.. Cancer risk of insulation workers in the Uruted States. In: P. Bogoski. J.C. Gilson. V.Timbrell and J.C. Wagner feds.1. Biological efecfs of usbesros. IARC Scicntific Pubiicanons 8. pp. 2W-216. LARC. Lyoo (1973). SHARMA.J.M.. WITTER. R.L.. BURMESTER. B.R. and LANWN. J.C.. Public health imolications of Marek's d i s e s e virus and hemsvirus of rurkeys. Studies ffl human and subhuman pnmates. J. MI. C&er Inst.. 51. 1113-1 128 (1973). SPINELLIJ. .J.. GALLAGHER. R.P.. BAND. P.R. and THRELFALWL,.J.. Multiple myeloma. leukemia. and cancer of he ovary in cosmetologists and haxdressers. . h e r . 1. indusr..Wed.. 6. 97-102 (1984). STEIMCK. G .and WIKLUNDK. .. Multiple myeloma in Swedish agncul- t u 4 workers. Inr. 1.Epidemiol.. 15, 3 2 1 - 3 3 11986). THOMAS. T.L.. D E C O ~ ~ FPL. ~an. d MOURE-ERASO,R.. .Modity among workers employed in pemleum refining and petrochemical plants. 1.occup .Wed.. 22. 97-103 (1980). TOLLERUD. D.J.. BRINTOS. L A . . STONE. B.J.. TOBACYAJN.K. . and BLAITNER.W . A . . Monalitv from multiple myeloma among Nonh C a m lina furniture workers. J. &at. Cancer Inst.. 74. 799-801 (1985). WOODROFFZ.A.J.. .Multiple myeloma associated with long hstory of hyposensitisation with allergen vaccines. h e r . I. 99 ( 197'1).