Document nj5Mwe9g3y615vXoOpEb943a

L -19J- PuL)Iication01 the InternationalUnion Againsl CartCel Publicationde VUnlon InternationaleContre le Cancer LYMPHOMA, MYELOMA AND OCCUPATION: RESULTS OF A CASE-CONTROL STUDY Lin FRITSCHaInd Jack SIEMIATYCKI' Epidemiolog?' urid Biostatistics Unit, Institiit Armand-Frappier. Montreal. Canada. The known risk factors for lymphoma and myeloma cannot account for the current incidence rates of these cancers, and there is increasinginterest in exploringoccupational causes.We present results regarding lymphoma and myelomafrom a large case-control study of hundreds of occupational exposures and 19 cancer sites. We examine in more detail those exposures previously Considered to be related to these cancers, as well as exposures which were strongly related in our initial analyses. Lymphoma was not associated in our data with exposure to solvents or pesticides, or employment in agriculture or woodrelated occupations, although numbers of exposed cases were sometimes small. Hodgkin's lymphoma was associated with exposure to fabric dust, and non-Hodgkin's lymphoma was associated with exposure to copper dust, ammonia and a number of fabric and textile-related occupations and exposures. Employmentas a sheet metal worker was associated with developmentof myeloma. r 1996 Wile\)-Liss.I t i c The actiology of mycloma and lymphoma arc still unknown. although therc arc scvcral suspcctcd risk factors for these canccrs. The incidence of non-Hodgkin's lymphoma (NHL) appears to be increasing in most dcvclopcd countries (Hartgc er al.. 1094). and cstablishcd risk factors such as HIV infection. rare gcnctic syndromes and immunosuppressant therapy cannot account f o r the magnitudc of this incrcasc (Hartgc and Dcvcsa. 1YY7). Similarly, a proportion o f Hodgkin's lymphomas arc associatcd with viruses, particularly the Epstein-Barr virus (Jarrct. 1902). but this virus is lcss commonly associatcd with the subtypcs of Hodgkin's lymphoma which have been incrcasing in incidcncc (Hartge er nf., 1994). The actiology of multiple mycloma also remains elusivc. although a numbcr of occupational factors arc being considcrcd (Ricdcl and Pottern. 1992). Wc carried out a Iargc case-control study o f 19 canccr sites. including lymphoma and mycloma (Siemiatycki, 1991) to gcneratc hypotheses about thc occupational ciluses of canccr. Information was obtained on exposure to each of a large number of occupational substances using a new method of assessing occupational exposures which decreases misclassification of exposure status. We have previously published initial analyses from this study. In this report we present further and more detailed examination of positive findings from our study regarding occupational risk factors for lymphoma and myeloma. subjects) was constituted from among all the other canccr patients, excluding lung cancer patients. During thc same period. 740 population controls wcrc selected, in some years from electoral lists and in somc years by random digit dialing. Of these, 533 (71%) werc successfully interviewed and constituted a second control group. Wc also created a pooled control group made up of population and cancer controls. So as to weight the two control groups equally, we selected a random sample of 533 subjccts from among the set of cancer controls. These included subjects with cancers of thc following sites (number of cases in parenthescs): c o b rcctum ( 1 18). bladdcr (101). prostate (101).stomach (52).kidney (4h), mclanoma (29). pancrcas (25) and oesophagus (24). A structurcd questionnairc was used to obtain information on a numbcr of potcntial confoundcrs. These included age. cthnic origin. fathcr's and mother's origin. birthplace. restdcncc for first 15 ycars o f life. height and wcight. ycars 01' schooling. prcscnt income. smoking and alcohol habits. and currcnt family income. For the 40% ol'subjccts who refused lo divulge family income. this was inferred on the basis of the relationship between self-rcportcd income and mean income in the census tract of residence. available lor ;ill suhjccts. A scmi-structurcd probing interview was used l o obtain detailed descriptions o f all jobs the subject had had in the working lifetime. A tcam of chemists and industrial hygienists examined each completed questionnaire and translaled each job into a list ol' potential substance exposures by means of ;I checklist form which included 204 substances. For each suhstance. the chemists noted their dcgrcc o l confidence that the cxposurc had actually occurred (possible. probable. definite). thc frcqucncy of cxposurc during a normal workweek ( < 5%. 5-30% or > 30%) and the level of Concentration 0 1 the agent in the working cnvironmcnt. The 3 dimensions were combined as previously dcscribcd (Siemiatycki el 01.. 1094) [o produce 3 groups: uncxposcd. non-substantial cxposurc and substantial exposurc. Thc unexposcd group compriscd those ncvcr cxposed to a substance. Thc substantially exposcd group comprised those whom the chcmists wcrc confidcnt had been exposcd (probable or definitc cxposurc) and who also had more than 5 years of exposure a t a high frcqucncy and concentration. The rest were considered non-substantially exposed. Each job was also categorized as to the Canadian classifications of occupations and industries. MATERIAL AND METHODS The design and data collection methods of the case-control study of occupational exposures and 19 cancer sites have been described previously (Siemiatycki, 1991; Siemiatycki et al., 1994)and will be summarized here. Subjects were males aged 35-70 years, resident in Montreal, and with a new, histologically confirmed cancer. There were 4,576 eligible cancer patients between 1979 and 1985, and 3,730 of these (82%) were successfully interviewed. Eighty-one percent of subjects responded for themselves, and proxies provided information for the rest. There were 215 NHL cases interviewed out of 258 eligible cases (83% response rate), 54 Hodgkin's lymphoma cases interviewed out of 59 eligible cases (92% response rate) and 23 myeloma cases interviewed out of 27 eligible cases (85% response rate). A pool of potential controls (2,357 Unconditiond logistic regression was used to model the risk of developing each of the 3 cancers (myeloma, Hodgkin's lymphoma and NHL) using each of the 3 control groups (cancer, population and pooled). Age, proxy status, income (quintiles) and ethnic group (French/Anglo/other) were entered into all models as apriori confounders. Each of the other non-occupational covariates were added to the model in turn, and if there was a significant ( p < 0.10) change in the fit of the model with any of the 3 control groups, that covariate was ! 'To whom correspondence and reprint requests should be addressed, at the Epidemiology and Biostatistics Unit, lnstitut ArmandFrappier, 531, boulevard des Prairies, Case postale 100, Laval, Quebec, Canada H7N 423. Fax: (514)686 5599. 1Received: January 22,1996 and in revised form March 30, 1996. _ __ _ .-.. .. includcd in thc tinal modcl and uscd for all 3 control groups. makcrs. and rubber and plastics workers. and the industries I For Hodgkin's lymphoma. a variable for cvcr rcgularly drink- tcxtilcs and non-fcrrous casting and rolling. I ing alcoholic spirits was added to the modcl. For mycloma and non-Hodgkin's lymphoma. no additional covariatcs attained the level of significancc rcquired for inclusion. RESULTS Therc wcrc 469 occupational variablcs in our study: 294 substances. 98 occupations and 77 industries. For the present analysis. wc concentrated on occupational variables which satisfied either of the following two conditions. First wc included suspected risk factors for lymphoma or myeloma, based on thc IARC Monograph program and published reviews (Pearce and Bethwaite. 1992; Hartge er nf., 1994: Riedel and Pottern. 1992). Among these, we included only Tablc I shows sclcctcd charactcristics of thc 3 casc scrics and 2 control scrics used in the various analyses. Mycloma and Hodgkin's lymphoma cases had. on average. lower incomes than the other groups. and Hodgkin's lymphoma cases werc significantly younger than the others. Population controls werc less likcly to have a proxy rcspond for them. There werc significantly morc population controls and significantly fewer canccr controls who had never regularly drunk beer or spirits. those suspected risk factors for which there were 4 or more For each of the 3 site series and each exposurc variablc. exposed cases in our data. Because of the small numbers of separatc analyses were carried out with the cancer controls cascs of mycloma and Hodgkin's lymphoma, this represcnted a (n = 533). thc population controls (n = 533) and the pooled fairly stringcnt inclusion criterion: only onc variable (the controls (n = 1.066). Also, scparate analyses were carried out occupation of farmers and horticulturists) satisfied thc criteria. with a univariatc modcl and a multivariatc model. In general. For NHL. which had a larger number of subjects. the following the results did not diffcr much by control group: we thereforc cxposurc circumstances satisficd thc inclusion criteria and present in the tables only thc results using the pooled controls could bc cxamined: thc substances pesticides, wood dust, and discuss selccted rcsults from the separatc control groups bcnzenc and solvcnts: thc occupations farmcrs and horticultur- in the text. ists and carpenters: and the industries agriculture and produc- tion of household furniture. Mveloma Thc sccond group of occupational variablcs included wcrc those which had bccn associatcd with lymphoma or myeloma in our initial scrccning analyses (Sicmiatycki. 199I ) . Thcsc analyses had evaluated the risk of each type of canccr in rclation to cach of the 400 cxposurc circumstances using the Mantel-Hacnszcl method. For cach possiblc association. X different odds ratios (ORs)were calculated. being dctincd by: lcvcls o f cxposurc (any cxposurc o r high cxposurc): type of control group uscd (population o r cancer): and type of targct population (all subjects or Frcnch-Canadians only). Thc critc- Thcrc were only 23 cascs of myeloma: consequently. the numbcrs of cascs cxposcd to any individual occupational variablcs wcrc vcry low. Tablc I1 shows the estimatcs of rclativc risk o f mycloma in rclation t o the sclcctcd cxposurcs controlling for age. proxy status. incomc and cthnicity and using the poolcd controls. Thc only cxposurc choscn o n thc basis of litcraturc was work as a farmer o r horticulturist and thcrc wcrc only J cascs with this history. None of the ORs f o r this occupation were signiticantly raised. and thcrc was no suggestion of increasing risk with increasing duration. ria for sclcction o f variables f o r this report were thc I'ollowing: Of the empirically choscn exposures. work as a shcct metal the occup;itional circumst;incc was associated with the cancer with an OR o l a t least 1.3 and a one-sidcdp value < 0.10 in at lcast 4 o f the S ORs. and thcrc wcrc at least 4 cxposcd cascs. Using thcsc critcria, for myeloma. we carmarkcd for cvaluation thc substancc clcaning agents. the occupations of janitors and clcancrs (the broad group of othcr personal scrvicc workcrs was dropped as most of thcsc subjects wcrc janitors workcr for I O o r morc years was the only cxposurc for which the ORs wcrc signiticantly abovc 1.0 in this modcl. Although thcrc wcrc only J cxposcd cases. all o f thcm wcrc long- duration workers. and in this cxposurc category the OR was 6.5. This association rcmaincd significant irrcspcctivc of whcthcr canccr o r population controls were used, and irrcspcctive of whcthcr univariatc or multivariatc models wcrc uscd. and clcancrs) and shcct mctal workers and the industry of Thcsc shcct metal workers had becn cxposcd to complcx defencc scrviccs. For Hodgkin's lymphoma, wc includcd the cnvironmcnts. including iron compounds, mctallic dust, mctal substanccs cooking fumes, fabric dust, ccllulose, and waxes oxide fumes, chlorine dioxides, arc welding fumes, abrasives and polishes. the occupations of pcrforming and graphic dust, alumina, aluminium compounds, mild steel dust and artists, chefs and cooks, and leather workers, and the indus- polycyclic aromatic hydrocarbons. tries government as well as flour, feed and bakery products. The only other exposure variables which manifested evi- For NHL, we examined the substances bronze dust, copper dence of an association with myeloma were exposure to dust, alkali and caustic solutions, ammonia, hydrogen chloride, cleaning agents and work as a janitor or cleaner. However, plastics pyrolysis products, fur dust, cotton dust, plastic dust, these associations were not quite statistically significant. Three formic acid, and fluorocarbons. We also included the occupa- of the 4janitors and cleaners with myeloma had been heavily tions salesmen, janitors and cleaners, textile processors, textile exposed to cleaning agents such as hypochlorites and ammo- TABLE I - DISTRIBUTION OF NON-OCCUPATIONAL VARIABLES BETWEEN CASES AND CONTROLS I i I. Age (mean) Years schooling (mean) Income (mean, Canadian $) Proxy respondents (%) Ethnic grou.p (.%) I French Anglo Never regular smoker f%) Never d&k beer regujariy (%) Never drank sDirits reeularlv (%) 61.1 9.1 23,518 39 52 17 13 57 74 48.0 11.2 22,736 19 69 11 20 63 80 55.2 10.1 26,120 22 62 12 17 70 81 59.5 9.6 25,264 19 58 18 20 71 85 59.6 10.1 26,430 13 64 14 17 70 81 ' p value for test of the null hypothesis of equality of the 5 groups on this parameter. < 0.01 > 0.05 < 0.01 < 0.01 > 0.05 > 0.05 < 0.05 < 0.05 500 FRITSCHI /\NDSlEMlATYCKl TABLE I1 - ODDS RATIOS ( O R ) AND 9.53 CONFIDENCE INTERVALS (95'i CI) FOK DEVELOPING MYELOMA. F R O M A CASE-CONTROL S T U D Y IN M O N T R E A L Exmw,surc Level (vcnrsl No. O R ' 95;'A CI Substances Cleaning agents Non-substantial 2 1.5 0.3-7 .O Substantial 5 2.3 0.8-6.5 Occupations Janitors. cleaners 1-9 2 2.9 0.6-13.4 IO+ 2 3.8 0.8-18.3 Sheet metal workers 1-9 0 10+ 4 6.5 2.0-21.9 Farmers. horticul- 1-9 2 1.4 0.3-6.6 turists IO+ 2 1.1 0.3-5.4 Industries Defence services 1-9 7 1.9 0.74.9 10+ 1 3.1 0.4-26.4 'Adjusted for age. proxy status, income (quintiles). ethnicity (French/Angloiother). TABLE Ill - O D D S RATIOS ( O R ) AND 9 5 7 CONFIDENCE INTERVALS (9.5%CI) F O R DEVELOPING HODGKIN'S LYMPHOMA. F R O M A CASE-CONTROL S T U D Y IN M O N T R E A L Exposure Levcl (yearr) No O R ' 95% CI -~~ Exnosure ~ Level (vearsl No. O R ' 95% CI Substances Vinyl chloride Non-substantial 3 3.7 0.8-17.4 Substantial 0 Pesticides Non-substantial 9 1.2 0.6-2.0 Substantial 6 0.0 0.4-2.3 Wood dust Non-substan tial 17 0.5 0.3-0.8 Substantial 22 0.8 0.5-1.3 Benzene Non-substantial 20 0.7 0.4-1.1 Substantial 6 0.8 0.3-2.1 Solvents Non-substantial 26 0.6 0.4-1.0 Substantial 48 0.9 0.6-1.3 Occupations Farmers and horti- 1-9 8 0.7 0.3-1.5 culturists IO+ 10 0.8 0.4-1.7 Carpenters 1-9 0 10+ 4 0.8 0.3-2.2 Industries Agriculture 1-9 in+ 10 0.8 0.3-1.6 I 0.9 0.4-1.7 Household furniture 1-9 1 0.4 0.1-3.4 I O + 3 1.6 0.4-6.1 Substances 'Adjusted for age. proxy status, income (quintiles). ethnicity Cooking fumes Non-substantial 4 2.0 0.6-7.3 (FrenchiAngloiother). Substantial 3 2.1 0.5-9.4 Fabric dust Non-substan tial 2 0.6 0.1-3.1 Cellulose Substantial 5 6.0 1.9-19.1 Non-substantial 5 1.2 0.4-3.7 Substantial 3 4.0 0.9-17.7 but thcrc was n o incrcase in risk with incrcasing lime in the occupation of lcathcr workcr. Four of the 7 cascs cxposcd to Waxes or polishes Non-substantial 3 2.0 0.5-7.7 fabric dust had bccn leather workcrs and 3 of thcsc same 7 . Occupations Substantial 2 7.2 1243.1) cascs had hccn cxposcd 10 ccllulosc (including 2 lcathcr workcrs). Performing or graphic artists Chefs or cooks 1-9 IO+ 1-9 IO+ 1 1.7 0.1-1 9.h 3 5.6 I .0-32.j 4 1.6 (1.44.2 1 2.1 0.2-17.2 Thcrc wcrc 4 pcrforming and graphic artists with Hodgkin's lymphoma. The risk increased with increasing duration of time in the occupation. The ORs wcrc significantly abovc 1.0 in thc Leather workers 1-9 4 7.8 2.2-28.0 unadjustcd modcls. but aftcr adjustmcnt for possible confound- I O + 2 3.2 0.5-19.7 ers thc confidcncc intcrvals includcd 1.0. Thcrc wcrc 4 cascs Farmers or horticul- 1-9 5 2.2 0.7-6.6 who had workcd in the industry producing flour. feed and turists IO+ 1 0.5 0.14. I bakcry products. Thcrc was a consistcnt duration-response Industries Flour, feed or bakery 1-9 products 10+ Government 1-9 10+ 1 1.4 0.2- I 1.6 3 4.0 0.9-17.3 4 1.3 0.44.5 5 2.4 0.8-7.O pattcrn. but only 7 of thc h ORs wcrc significantly above 1.0. For the substancc cooking fumcs, the occupation chcf o r cook and thc industry govcrnmcnt thcrc wcrc no significantly raised ORs in any of the 6 models. 'Adjusted for age. proxy status, income (quintiles), ethnicity (French/Anglo/other) and spirit drinking (yesino). nia. The 8 cases who had worked in the defence industry included 3 subjects who had also worked in occupations of janitors and cleaners. Hodgkin 's lymphoma There were 54cases of Hodgkin's lymphoma. The multivariate model contained age, proxy status, income, ethnicity and spirits drinking. Table I11 shows the relative risk estimates for selected exposure variables, based on the multivariate model, and using the pooled cancer and population controls. There were only 6 cases who had worked as farmers or horticulturists. None of the ORs were significantly raised, and there was no suggestion of increasing risk with increasing time in the occupation. Of the empirically chosen exposures, there were ORs significantly greater than 1.0 for substantial exposure to the substances fabric dust as well as waxes or polishes, and the occupation of leather workers. These statistically significant ORs remained when using different control groups and in the univariate as well as the multivariate models. Dose-response patterns were seen in all 6 models for each of the 2 substances, Non-Hodgkin's lymphoma There were 215 cases of NHL. Results using thc poolcd controls and adjusting for age, proxy status, incomc and ethnicity are presented in Table IV for exposures selected from a basis in literature and in Tables V and VI for empirically chosen exposures. For the exposures chosen on the basis of the literature (Table IV), there were no significantly raised OR< .. It has been suggested that solvents may have a late-stage effect on non-Hodgkin's lymphoma (Vineis et al., 1992); we consequently examined this hypothesis in our data set by estimating the risk of NHL for those with recent exposure to solvents. Exposure to solvents in the 5 years before diagnosis of cancer (or interview time) was not associated with NHL (OR = 0.8; 95% CI, 0.6-1.3). Restricting the exposure period to 2 years prior to diagnosis gave almost identical results. Of the empirically chosen exposures (Tables V and VI), significantly raised ORs were seen in the adjusted model and using the pooled controls for substantial exposure to the substances copper dust, ammonia, plastics pyrolysis products, fur dust, and cotton dust, in the occupations of textile processor and textile maker and in those with long exposure in the textiles industry. A dose-response pattern was present for all the substances except plastic dust and formic acid. For the LYMPHOM/\ /\ND OCCUPATION so I TABLEV - ODDS RATIOS (OKs) A N D Y5'"i CONFIDENCE INTERVAL5 tional variables at a timc. but the rcsults werc difficult to (YSrr CII FOR DEVELOPING NON-HODGKIN'S LYMPHOMA. FROM r\ CASE-CONTROL STUDY IN MONTREAL interpret with some unstable estimates due to small numbers I Suhstancc crposurc of data points and large numbers of variables. Conscqucntly. Levcl N o OK' W r i CI wc prefcr to prescnt and interpret the morc stable estimatcs Bronze dust Non-substantial 1 1.2 0.1-10.1 based on adjustment for non-occupational factors only. Substantial 5 2.8 0.8-9.1 Copper dust Alkali. caustic Non-substantial 8 1.1 0.5-2.5 Substantial 8 5.5 1.9-16.0 Non-substan tial 15 1.5 0.8-7. 8 DISCUSSION Many diffcrent classification systems have been uscd to solutions Substantial 7 1.8 0.7-4.5 group lymphoproliferativc cancers. and there is considerable Ammonia Non-substantial 12 0.8 0.4-1.5 Substantial 14 3.4 1.7-6.9 Hvdrogen chloride Nan-substantial I6 1.6 0.9-3.0 Substantial 6 1.3 0.5-3.4 Plastics pyrolysis Non-substantial 8 2.0 0.84.9 products Substantial 6 2.6 1.O-7.1 Fur dust Non-substantial 3 1.7 0.5-6.6 Substantial 5 3.1 1.0-9.3 controversy over which system is most appropriate for epidemiological studies (Weisenburger. 1992; Vineis et al., 1992). There is inevitably a trade-off between specificity of the classification system and numbers of subjects available for statistical analysis. This is the conventional dilemma between "lumping" or "splitting". For our data, we had information on diagnosis as it came from medical records. The data from this Cotton dust Non-substantial 8 1.0 0.4-2.1 source were limited by the prevailing diagnostic technology Substantial 13 2.6 1.3-5.4 and medical opinion during the era in which cases were Plastic dust Formic acid Fluorocarbons Non-substantial 14 1.7 0.9-3.3 Substantial 3 1.0 0.3-3.4 Non-substan tial 2 2.2 0.4-1 1.3 Substantial 2 1.5 0.3-8.0 Non-substantial 5 1.3 0.5-3.6 Substantial 5 2.6 0.8-8.0 accrued (1979-1986) and by the fact that cases were accrued from 19 different hospitals. There is doubt about the validity of classification systems for non-Hodgkin's lymphoma. One promincnt textbook of pathology states that ". . . until the recent past there were morc classifications than expcrts". dcfining recent!\> as after thc timc framc of our study (Cotran er a/.. 'Adjusted for age. proxy status. income (quintiles), ethnicity 1994). It might noncthclcss be uscful to cxaminc histologic (French/Anglo/ot her 1. subgroups of NHL scparatcly, sincc this is a largc and hctcrogcncous catcgory with rcspcct to pathology. clinical TABLE VI - ODDS K,\TIOS ( O K s ) A N D W r ; CONFIDEN('E INTEKVALS (W",('I) FOR DEVELOPIN(; NON-llOD(iKlN'S LYMPIIOMA F K O M I\ (',\SE-CONTKOL STUDY IN MONTREAL fcaturcs. trcatment and actioioLy. The state of the art docs n o t allow us t o do so. On the othcr hand, the classifcation of Hodgkin's disease is sound and has not changed sincc the Ryc Epl\urc (vLcc:vIcrl\, K I I , O K ' '15"; ('I Classification of IYOfi (Lukcs VI al.. 1966).Thcrc is n o overlap bctwccn the two discasc groups. with the possible exception of Occupation?, Salesmen 1-0 23 1.3 0.s-2.2 some follicular (nodular) well-ditfcrcntiatcd NHLs and the rarcst subtype o f Hodgkin's discasc. thc nodular variant of the IO+ 35 1.4 0.Y-2. I lymphocyte-predominant subtype. Patients with Hodgkin's Janitors and cleaners 1-0 7 1.1 0.5-2.0 discasc. cspccially following trcatmcnt. may develop sccond Textile processors Textile makers Rubber and plastic workers IO+ 0 1.6 0.0-4.2 1-9 3 0.9 0.2-3.3 10+ 8 12.2 3.542.2 I 4 3 2.0 0.5-8.0 IO+ 4 15.7 2.9-87.5 1-9 4 3.6 0.')-15.0 primary canccrs at any site. including NHLs. but this is not an issue for newly diagnosed cascs. Scvcral substances. occupations and industrics were includcd in this analysis because prcvious research had indicated a possible incrcascd risk. None of thcsc gave significant rcsults IO+ 1 9 . 0 0.6- I48 in our data. although the numbcrs of cascs cxposcd t o most of Industries these circumstances wcrc quitc small. For solvents. wc had Textiles 1-9 10+ Non-ferrous casting and rolling 1-9 10+ 5 0.6 0.2- I .5 9 3.9 1.6-9.5 2 1.3 0.3-6.4 3 4.1 0.9-18.9 largc numbcrs of cxposcd c a m , but there was no suggcstion of any effect in our data. It is conceivable that some substances are dangerous in onc work environment but not in others, because of the nature of )Adjusted for age, proxy status, income (quintiles), ethnicity the substance or because of cofactors. To evaluate this (French/Anglo/other). possibility, we carried out some analyses looking at joint effects of substances and occupations. Because this requires relatively large numbers, it was only feasible to do it for 2 fairly prevalent substances copper dust, ammonia, and cotton dust and the exposures,_solvents and wood dust. For each of these sub- textile industry, the ORs for substantial exposure in all 6 stances, we assessed the risk among those exposed to the models were significantly raised. substance in particular occupations. There was no evidence T h e results shown in Tables IV to VI are adjusted for a num- that solvent or wood dust exposure in any of the main ber of relevant non-occupational covariates, but the results for occupation groups carried excess risk of NHL. each occupational exposure are not adjusted for other occupa- T h e occupation of sheet metal worker appears to be related tional exposures. However, some pairs of exposure were highly t o myeloma in our data, with high and statistically significant correlated. Approximately half of the cases who had been ORs despite small numbers of cases. A Danish case-control exposed to bronze dust, copper dust, alkalis and caustic study which linked industrial history records with over 1,000 solutions or ammonia had also been exposed to hydrogen cases of myeloma and 4,000 controls (Heineman et d.,1992) chloride. Similarly, about half the cases exposed to alkalis and found raised risks for precision metal working as most recent caustic solutions had also been exposed to ammonia. Cases occupation (OR = 1.4; 95% CI, 1.0-2.1). Among industries, who had been employed as janitors and cleaners were likely to there was a raised risk associated with employment in fabricat- have been exposed to alkali and caustic solutions and to ing structural metal and stationary tanks (OR = 2.0; 95% CI, ammonia. Six of the 7 cases who had been employed as textile 1.0-4.0), and those in this occupation for 5 or more years were makers and 8 of the 14 cases who had worked in the textile significantly more likely to have myeloma than those employed industry had been exposed to cotton dust. Attempts were made for less than 5 years. A number of metal-related occupations to carry out some multivariate analyses with severaI occupa- have been linked to myeloma including smelter and metallurgy L. 502 FR ITSCHI ;\ ND SIEMI,\TY c'xI workers in a study linking cancer registry and occupational data from the ccnsus (McLaughlin et ul.. IY88). as wcll as workers in the composing room of printing plants (Greenc e! aL. 1979). Case-control studies have linkcd mycloma with sclf-reported cxposurc to metals (Morris ct al., 1980) and have linkcd occupation as a metal workcr with mycloprolifcrativc cancers (Giles et al.. 19x4). A study of over 10.000 men with occupation recorded as mctal workcr on thc dcath ccrtificatc (Gallagher and Thrclfall. 1983) found no incrcasc in PMR from myeloma for all mctal workers combined but a raised risk ( P M R = 209; 95% CI. 108-366) for machinists (based on 12 deaths). T h e risk of myeloma for sheet metal workers was not raised in a largc study which linked data from the Canadian Labour Forcc 10% Samplc Study with death certificate information (Lindsay et al.. 1993). Our findings suggest a raised risk in occupations rclating to the manufacture of textiles for both Hodgkin's and nonHodgkin's lymphoma. For Hodgkin's lymphoma. risks werc raised for exposurc to thc substances fabric dust and cellulose. while for NHL. thcrc were high ORs with exposurc to cotton dust and work in thc tcxtilc industry. especially in thc occupation of textile making. Thc tcxtilc industry has bccn examined in a number of previous studies. Textile workers were found to be a t increascd risks of death from N H L in one study using dcath ccrtificatc information (Delzcll and Grutfcrman. 10x3). but not in othcrs (O'Bricn and Dccouflc. 1988: Dubrow and Gutc. 1988: Schumachcr and Dclzcll. 1988). A rctrospcctivc cohort o f synthetic tcstilcs workers (Goldberg and Thcriault. 1904) found an increased risk of onc type of N H L only (reticulum cell sarcoma). Case-control studies h a w shown no increased risk of NHL in tcstilc workers (Schcrr ct d.. 1YO2) and a non-significant incrcascd risk in workers in tcxtilc mill products and apparel (Blair CI ti/.. lOY3). I t is also worth noting that studics havc shown an incrcasc in risk of multiplc myeloma in textile workers (Linct c[ 01.. 1094) and in the manufacture of synthetic yarns (Hcincman ct o/.. IYY2). A rclatcd tinding in o u r study was the incrcascd risk of NHL with cxposurc t o fur dust. No incrcascs in S M R for lymphoma wcrc t'ound in 7 retrospective cohort studies of fur workers (Sweeney et al., 1985: Guay and Sicmiatycki. 1987) but numbers of dcaths were vcry small in both studies. T h e occupation of leather worker was linked t o Hodgkin's lymphoma in our data. A rcview of the leather industry hy I A R C did not find convincing cvidcncc of a link hctwccn the industry and lymphoma (IARC. 1981). Sincc then thcrc have been scvcral rcports on cohorts of leather workcrs. cspccially shoe manufacturers (Garabrant and Wcgman. 1984: Walrath er al.. 1987: Stern et ul., 1987: Seniori-Constantini et al.. 1989: Walker et al.. 1903; Mikoczy et al.. 1YY4). Thcsc studies havc not demonstratcd any raised risk for lcathcr workers. but they have bccn based on small numbers of dcaths from lymphoma and usually do not subdivide the group further. A registry study in Swedish womcn found no increase in thc incidcncc of N H L in leather workcrs (Linct et al.. 1994), but a case-control study in Boston (Schcrr et al., 1992) found a slightly incrcascd risk of N H L in workers in the leather industry (OR = 2.1: 95% CI. 0.948). Other substances consistently and significantly associatcd with N H L in o u r data wcrc copper dust and ammonia. No incrcasc in thc incidencc of blood. bonc marrow and Ivmphatic system canccrs was found in coppcr smcltcry workers in Swedcn (Sandstrom and Wall. 1993). In conclusion. we havc found a number of possible risk factors for lymphoma and myeloma. However. for some of them the risks arc not strong or the issue has previously hccn examined and no supporting cvidcncc has heen found. The rclationships which seem most worthy of further study arc those bctwccn myeloma and shcct mctal workers. and bctwccn lymphoma and the rcxtilc industry. ,\('KN( )W LEIX; EMEN'I'S This study was funded hy grants trorn the Institut clc rcchcrchc en s;intC et cii sccuritc du travail du Qutihcc. thc Fonds de rcchcrchc en santc du QuChcc. the N;itional Health Research and Development Progrmi. and the Nation:il Canccr Institute of C:inada. J.S. is ;I Nation;il Health Scholar. The ticldwork was supervised hy Ms. L. Richardson, and the chemical coding was carried out by Dr. M. G r i n . Dr. L. Nadon, Mr. D. Begin and Mr. R. Lakhani. Dr. B. Case providcd helpful d v i c c regarding pathologic classification. REFERENCES BLAIR. A.. LINOS. A.. STEWAKI. P.A.. BUKMEISTER. L.F.. GIBSON. R.. EVEREIT.G.. SCHUMAN. L. and CANTOKK..P., Evaluation of risks for non-Hodgkin's lymphoma by occupation and industry exposures from a case-control study. Amer. J . ind. Med.. 23,301-312 (1993). COTKANRS, KUMAR. V.. ROeBlNs. S.L.. eds.. Rohhins puthulugic hasis ofdireuse (5th ed.). W.B. Saunders, Philadelphia (1994). DELZELL, E. and GRUFFERMAN. S.. Cancer and other causes of death among female textile workers, 197678.J. m.Cancer Insf.,71.735-740 (1983). DUBROWR. . and CUTE.D.M.. Cause-specific mortality among male textile workers in Rhode Island. Amer. J. ind. Med., 13,439-454 (1988). GALLAGHER, R.P. and THRELFALL. W.J., Cancer mortality in metal workers. Can. med. Assoc. J., 129,1191-1194 (1983). GARABRANDT.,H. and WEGMAN, D.H., Cancer mortality among shoe and leather workers in Massachusetts. Amer. 3. ind. Med., 5, 303-314 (1984). GILES, G.G., LICKISS,J.N., BAIKIE,M.J., LOWETHAL. R.M. and PANTONJ,., Myeloproliferative and lymphoproliferative disorders in Tasmania, 1972-80: occupational and familial aspects. J. nut. Cancer Insf.,72,1233-1240 (1984). GOLDBERG, M.S. and THERIAULT, G., Retrospective cohort study of workers of a synthetic textiles plant in Quebec. 1. General mortality. Amer. J. ind. Med., 25,889-907 (1994). GREENE, M.H., HOOVER. R.N., ECK, R.L. and FRAUMENI,J.F., Cancer mortality among printing plant workers. Environ Res.. 2 0 , 6 6 7 3 (1979). GUAY, D. and SIEMIATYCJ.K, HI,istoric cohort study in Montreal's fur industry.Amer. J. ind. Med., 12. 181-193 (1987). HAKrCIi. P. and DEVESAS..S.. ~ U a n t i ~ C J t i Oonf the imp;icl ol known risk factors o n time trends in non-Hodgkin's lymphoma incidrncc. Cuncer Rev.. 52. S556h-S5569 (1992). HARTCE.P.. DEVESAS..S. and FKAUMENJ.FI.. Hodgkin's and n o n Hodgkin's lymphomas. Cancer Srrnr.. 20.423453 (1994). HEINEMAN. E.F., OISEN. J.H.. POITERNL. .M., GOMEZM, .. RAWN. E. and BLAIR. A., Occupational risk factors for multiple myeloma among Danish men. Cancer Causes Control, 3,555-568 (1992). [ARC MONOGRAPHSON THE EVALUATION OF THE CARCINOGENIC RISK OF CHEMICALS TO HUMANS, Vol. 25. Wood, leather and some ussociafed industries. IARC, Lyon, (1981). JARRETTR,.F., Viral involvement in Hodgkin's disease. fnr. J. Cell Cloning, 10,315-322 (1992). LINDSAY, J.P., STAVRAKKY.,M. and HOWE,G.R., The Canadian Labour Force IO-Percent Sample Study-cancer mortality among men, 1965-1979.3. occ. Med., 35,408-414 (1993). LINET, M.S., MCLAUGHLINJ.K, ., MALKER,H.S.R., CHOW, W.H.. WEINER, J.A., STONE, B.J., ERICSSONJ.,L.E. and FRAUMENI, J.F., Occupation and hematopoietic and lymphoproliferative malignancies among women: a linked registry study. J. occ. Med., 36, 1187-1198 (1994). LUES, R.J., CRAVERL,.F., HALL,T.C., RAPPAPORT, H. and RUBEN, P., Report of the nomenclature committee. Cancer Res., 26, 131 1-1311 (1966). MCLAUGHLIJN.K,., LINETM, .S., STONE, B.J., BLOT, W.J., FRAUMENI. JR. J.F., MALKER, H.S.R., WEINER, J.A. and ERICSSONJ.L, .E.. Multiple myeloma and occupation in Sweden. Arch. environ. Health, 43, 7-10 (1988). I I I 1 M i ~ o c z vZ. .. S.tit.1-Z. A. and ~ H X ; M A K . L.. Cancer incidence and TEI.I.I. C. and LENZIS... Cancer mortality among workers in thc TuscLin mort;ility among Swedish leather tannerb. Occ. c t i i v o t i . Mecl.. 51. tanning industry. Bnr. J. t i i d Mcd..46. 384-388 ( IY8lJ). 530-535 (IYW). SIEMIATYCJK..I.Risk $tcfor.y jiJr cancer tti rlic workp/ucc. CRC Press. Moiwis. P.D..KOEPS~LTL..D.. DAI.IN(;J..R.. TAYI.OKJ.W. .. LYON. Bocn Raton ( 109I ). J.L.. SWANSONG,.M.. CHILDM. . and WEISS.N.S..Toxic siihst;incc SIEMIATYJ(.-. KDIE.WAR. R.. Fu'moh. L. and G ~ R I MN.. Occupational exposure and multiple myeloma: ;I case-control study. J. IIUI. Curiccr risk factors for hladder canccr: results from ;I case-control study in / t l , ~ f .7, 6. YX7-9Y4 (1986). Montreal. Quehec. Canada..-ttm~rJ.. Eprcl~witol.1. 40. 1061-IO80 ( 1994). O'BKIENT. .R. mid DL-(OI 1-1.~P... Cancer mortality among northern Georgia carpet and textile workers. ..ltm*r. J. i d . Mcd. 14, 15-24 (i9xx). P ~ A R C LN.. ; i d BETHWAITCP.. Increasing incidence of nonHodgkin's lymphoma-occupational and environmental factors. Curicer Res.; 52. SS496-SSjlK) (IW2). SI'EKNF..B.. B~AIJMOJN.JI... HAII'EIUN. W.E.. MUKI'HY1..1.. H1l.t~. B.W. and FAJEN.J.M.. Mortality of chrome leather tannery workers and chemical exposures in tanneries. .\`cutid. J. Work fm*irori.Heulth. 13, 108-1 17 (1987). SWEENEY. M.H.. WXKATH.J. and WAXWEILERK.J... Mortality among retired fur workers. Scoiid. 1. Workenviron. Hcdth. 11.957-264 (1985). RICDEI..D.A. and P O ~ T E RL~.M. .. The epidemiology ol multiple myeloma. Hemulo/. Oticol. C h . N . .-lrnc,r..6. 225-247 ( 1902). SANDSTROMA. and WALL. S.. Cancer incidence among male salaried employees at a smeltery in northern Sweden. .4cru Oticol.. 32, Y-14 (1993). SCHERK. P.A.. HUTCHISONG.B. and NEIMAN,R.S.. Non-Hodgkin`s lymphoma and occup:itronal exposurc. Cuitccv Res.. 52. SS503-S5509 (1992). SCHUMACHEMK.C. . and DCI:LEI.I.. E.. A death-certificate case-control studv o f non-Hodrkin's Ivmphoma and occupation in men in North C;irc;lina. .4tni.r. J. bid. ,Wed.. 13, 317-330 ( 1988). S ~ N I O K I - C ~ N S ~ . A NAI,. I ~PM~ ~'I.~ .E.. MILIGIL. .. BUIA.TI'1. E.. MAH- VINEISP.., D'AMOREF. . and WORKING GROUP ON THE EPIDEMIOLOGY o t HEMATOLYMPHOPOIEMTAILTIGICNANCIES IN ITALY. The role of occupational exposure and immunodeficiency in B-cell malignancies. EpJdLmiOl~~3~.T2,66-270 ( 1992). WALKEK. J.T.. BLOOM. T.F.. STERN. F.B.. OKUN. A.H.. FINGERHUT. M.A. and HAI.PERIW~.E.. Mortality of workers employed in shoe manufacturing. Scotid. J. Work ewirotl. Hculrli. 19. 89-95 ( I 903). WALKATH..I..DL(o( 1-I.L. P. and THOMAS. T.L.. Mortality among workers in ii shoe manufacturing company. Atner. J. Cid. Mcd.. 12, 615-023 (1987). WEISENHURGD~.DK... Pathological classiticalion of non-Hodgkin's lymphoma for epidemiological studies. Cutiwr Rer.. 52. SS4.56-SS4M (IYY2).