Document wr3qQ28rL2rVd9MkBJdk9J7B3

[CANCER RESEARCH (SUPPL.)52.5503a-55095 On& 1.19921 son-Hodgkin's Lymphoma and Occupational Exposure' paul A. %hem,* George B. Hutchison, and Richard S. Neiman3 ~partmentof Epidemiolom, Harvnrd School of Public Health [P.A. S..G.B. H.1, Boston, Mosscrchwenr 02IIS. md D e p v t m a ~of path&gy, Boston IJniwmiQ school of Medicine [R.S. N.], Boston, Massachwenr 02118 \ I* ibstract Nebraska (17) found an odds ratio of 2.4 for those who used 112/189 A case-control study was conducted to p ~ e s sthe effect of occupa- ional exposures on the risk of non-Hodgkin's lymphoma. Interviews ,,ere conducted with 303 persons with non-Hodgkin's lymphomanewly dingnosed from J ~ d . r y1,1980, to May 31,1982, among residents of the Boston, MA, mehopolitan area and 303 age and gender matched mntrols. The study found an increased risk of disease among persons rmployedin the agriculture+forestry, and fishingindustry [relativerisk 2,4-dichlorophenoxyaceticacid more than 20 daydyear and an odds ratio of 2.4 for users of organophosphate insecticides. Leather workers and female textile workers were reported at increased risk of non-Hodgkin's lymphoma due to chemical exposures although no associations with specific chemicals were determined (18, 19). Ross et d.(20) found an excess of male patients with large cell lymphomas of the gastrointestinal (RR)= 3.03; the construction industry [RR = 2.11; and the leather tract and oral cavity associated with occupational exposure to iadustry [RR = 2.11. The particplu job groupingsat ioerrpsedrisk were asbestos. We have conducted a case-control study to identify plant fanners and gardeners (RR unboundedk painters and plasterers occupations and exposures that are associated with increased IRR = 6.0); and carpenters, brick and stone mwonq plumbers, and risk of all non-Hodgkin's lymphomas or of specific histological rwfers (RR= 12.0). Although other exposures may have led to these ,messed risks, the findings in this study are consistent with an in- subtypes of non-Hodgkin's lymphoma. 5/18 4/18 0.7-5.3 0.6-5.4 irmsed risk of non-Hodgkin's lymphoma for workers who may be ex- posed to chlorophenolsor phenoxyaceticacids. Methods 3/18 0.3-3.1 We attempted to identify all cases of non-Hodgkin's lymphoma 6/38 12/38 1/38 119 319 0.4-27 1.2-5.0 0.4-2.5 0.1-5.1 0.6-8.6 =*e Q.* *?? %% :nt clues or chance findings-k t it might be profitable to focus :fforts to evaluate occup ioma. The identification of statistical associations between o m p a [ionsand the occurrence of cancer can be used to identify parricular agents which may lead to the development of cancer. For i!mphomas, most of the increased risk in occupational expo- ,ures appears to be related to chemical exposures. Li et d.(1) ieviewed the causes of death of members of the American ChemicalSociety from 1948 to 1967 and found an observedexpected ratio of 1.8 for lymphoma. Similar results were found in istudy of British chemists and pathologists (2, 3). Studies of workers exposed to benzene, vinyl chloride, styrene. and its polymers suggest increased risk of lymphoma (4-7). Goldsmith md Guidotti (8) reported the occupations recorded on death certificates for males who died of lymphosarcoma in California and Washington. Increased rates were found for people work- newly diagnosed from January 1, 1980, through May 31,1982, among residents of the Boston Standard Metropolitan Statistical Area who were treated in any one of nine participating hospitals. Three hundred seventy-nine cases of all ages were identified in the participating hospitals. With the approval of his or her physician. each livingcase or next of kin for a deceased case was asked to participate in an interview. Parents or guardians wek interviewed for cases ages 17 years or less. Interviewswere completed with 202 cases, 94 proxies, and 7 guardians giving an overall participation rate of 8096 (303 of 379). Refusals to participate were given by 40 cases, 18 physicians for case interviews, I O proxies, and 8 physicians for proxy interviews. A pathology review of the cases was done to confm the diagnosis of non-Hodgkin's lymphoma and to subclassify the type of lymphom,a. Each case was classified according to the Modified Rappaport Classification (21). the Working Formulation(22). nodular or diffuse pattern, and the probable cell type (Bcell, T-cell). For purposes of analysis, we grouped the malignant lymphomas :asing incidence of non-Hodgkin's Rental factors. Cancer Res., S2tSuDd): #* urmeister,L. E., Gibson,R., Everrttqh iluation of risks for non-Hodskin' xposures from amecontrol stud Everett,G. D.. Cantor, K. P, S. F., and Dick, F. Pestrcides for leukemia among men in 6591, 1990. .E.c,auGsreasuomfpdne,aDth.,aMmoonmg , F. X.,v a cohorl t2-168, 1990. A., Herrick, R. F.. and Mahar.H. ral directors. Am. J. Indust. Med. ing in some types of farming, food processing, delivery services, pulp and paper work, electrical work, some building trades, and >ometypes of clerical work. Relative risks of 2.5 (slight expowe) and 9.3 (heavy exposure) were found for workers exposed to chlorophenols or phenoxy acids containing dibenzodioxins 3nd dibenzofurans as impurities. Workers exposed to these 2hemicals were employed in farming, forestry, building, and painting (7, 9-11). Several studies of farming and mortality rrom non-Hodgkin's lymphoma found odds ratios ranging from according to several larger categories. Using the modified Rappaport Classification, we analyzed all cases of histiocytic lymphoma in one category, whether they were nodular or diffuse. In addition, we defined "lymphocytic"as all malignant lymphomas of well differentiated lymphocytic, intermediate differentiated lymphocytic,poorly differentiated lymphocytic, and lymphocyticnot otherwisespecified, as well as those cases that were either nodular, nodular and diffuse, or totally diffuse. Malignant lymphomas of probable T-cell origin were defined on morphological grounds exclusively. Using the Working Formulation, tumors were categorizedinto three 1.2 to 1.3 for all ages and higher odds ratios among the elderly groups. Diffuse large cell included all subclassificationsincluding large 12- 14). Cantor et al. (15 ) reported elevated risk among farm- cleaved, large noncleaved, and those with sclerosis. The follicular cat- :IS using the pesticides dichlorodiphenyltrichloroethane RR4 = 1.5), carbofuran (RR = 1.6), and chloramben (RR = 1.2). In a casecontrol study, Hoar et al. (16) found an odds x i 0 of 1.6 for farmers who used herbicides, with frequent users laving an odds ratio of 8.0. A case-control study in eastern egory included all malignant lymphomas classified as follicular with either predominantly small cleaved cells, predominantly large cells, or mixed small cleaved and large cells. We obtained slides for review for 290 of the 303 cases (96%) and all were reviewed by 2 or more members of a 7-member panel of study pathologists. In this report, the pathology classificationsof 286 casesby ' I resented at the National Cancer Institute Workshop, "The Emerging Epi- `mil of Non-Hodgkin's Lymphoma: Current Knowledge Regarding Etiological >clors." October 22-23. 1991. Bethesda MD. Supported by Grants 5 p 0 1 t06373. ROICA38071,and T32-CA09001with the Xational Cancer Institute. *Presentaddress: Aging Studies Branch, National Center for ChronicDisease 'mention and Health Pmmouon. Centers for Disease Control. Atlanta, GA '0?33 ___ .Present address: DepartmentofPathology. Indiana University Schoolof Medqne. Indianawlis IN 46223. a single member of the panel (R. S. N.) are used. In a later report, concordancy among panel members will be reported. For each case, a control of thesamesex and age (within I year), town, and precinct of residence was selectedfrom town residency lists. By law, each town in Massachusettsmust compile lists annually of all residents 17 years of age or older. The list for the town and precinct was entered by a random process and was searched forward to find the first person with the same sex and age as the case. If this potential control subject . . could not be interviewed, the list was searched forward toidentik_-._ -. ~~ __ ~ another potential control subject. For cascs under 18 years of age, the list was searched matching on the age and sex of one parent or guardian. The potential intervieweewas contacted to determinewhether he or she had a child of the same age (within 2 years) and sex as the case. This process was continued until a guardian of the requiredage and sex with a child of the required age and sex was found. Of 423 potential controls, 303 were interviewed, and 120 refused to participate in the study. Each subject (caseor control) was sent an explanatoryletter and was contactedby telephone to arraoge a home interview.If the subject could not be interviewed at home, a mail questionnaire was used. Home interviews were obtained for 301 casea (99%) and 297 controls (98%). The interview included questions on the demographiccharacteristicsOf the subject and on an extensive List of possible risk factors, including occupation. Findings relating to occupation arc reported here, and other exposures will be the topic of later reports. Each subject was asked about his or her cumnt or most recent job, the job held 15 years ago, major occupation, second most major occupation, and up to two occupations in which he or she was exposed to any of the specific agents listed in Table 2. Each occupation was categorized by industry and by occupational title within industry. The industry was coded according to a scheme based on the Standard Industrial Classification Manual of the United States Bureau of the Budget (23.24). The occupational title was coded according to a scheme based on the United States Dictionary of OccupationalTitles (23, 25). The first two digitsof the occupationalcode were used, resultingin generally broad and in some instances rather heterogeneouscategories. For each occupation. information on the dates of employment and exposure to the specific agents in Table 2 was obtained. Industry, occupational title, and exposureto specilkagentsrepresent three interrelatedmethods of classifying a work exposure.An increased risk may be experiencedby anyoneworking in a given industry, such as the textile industry, independentlyof the particular job title held within the industry. Alternatively,an increased nsk could be experiencedby all workers with the same job title (plumbers, for example) independently of the industry in which the person is employed. In this paper, we will consider each of the three methods of cladkation separately. Another goal of the study was to determine whether there was an association between occupation and any of the various histological subtypes of non-Hodgkin's lymphoma. The number of cases of each histological subtype was too small to pcrmit the observation of an association with occupation. Histologic21 subtypes were aggregated into groups with similar histological characteristics in an attempt to form groups suRiciently large for analysi\. The data analysispresented here considered the association of occupation with each of five histological groups and with a sixth group that included all cases of nonHodgkin's lymphoma. Given the large number of exposure measures and the six possible histological categories, a large number of associationswas considered. This led to the problem of distinguishingbetween true associations and associationsthat occurred by chance. T h e approach used was to investigate the pattern of the association within an exposure-histological category as a function of calendar time or of the duration of exposure. Calendar time patterns which showed odds ratios consistently >1 or consistently C1 were considered to favor true associations. Patterns that had a large odds ratio at one point in time and odds ratios near 1 at other timeswere consideredto favorspuriousassociations. Similarly, if odds ratios for at least 10 and at least 20 years of exposurewere both >I, the duration pattern was considered to favor a true association. However, an odds ratio >1 for those with at least 10 years of exposure and <1 for those with at least 20 years of exposuR were considered more consistent with a spurious association. In this study, odds ratios and x statistics with 1 degree of freedom for matched pairs were calculated by histological group for each of the various industries. occupational titles, and exposum to specificagents for the time periods of 1901-1949, 1950-1959.1960-1969, and 1970 or later. Both early and late exposureswere considered, since there was interest in mechanismsof both tumor induction and tumor promotion. The odds ratios were reviewed to identify odds ratios equal to 2.0 or p$absolute values of 1.73 or greater. If the overall odds ratio or statistic for all time periods together met the above criteria, the ratios and y, statistics were reviewed for each of the four time periodr aIf these data were found to be consistent over the time periodr,(bc association would be included in Curther analysis. olhensise,the ciation was considered to be spurious and would be eliminated further analysis. Similar analysis was done by duration of expoamfor those having specific exposuresfor at least 10 years, at least 20 and at least 30 years. A All of the exposures which showed consistent patterns withii % logical categoriesover calendar time or over duration were c o n s i u as candidate variables for conditional lo@c models retaining (bc matching in the original study design. These models were used to 6 termine whether candidate exposums were sienificpat controlling~ ; ethe efTect of other exposures. The candidate exposurecould entertfr model as an indicator variable for exposure or as an indicator for exposure during 1901-1944,1945-1969, or 1970-1983. For& s u r a to specificagents, the exposure could be expressedas the n u m k of decades of exposure up to a limit of four decades. No demographic variables or interaction terms were included as candidate variables& the models. However, a series of analyses were done which include& religion to control for possible confounding. These models were ob structed for each of the six histological groups within each of the work classifications (industry, occupational title, and occupatioml agent exposures). Models were constructedwhich includedall subjaq additionally models were used which excluded the proxy respow Variables were retained in the final model if their coefficients were significant at the 0.05 level. For odds ratios with point estimates of either zero or infinity, a score statistic (26) was used to estimate the level of significance. The boundary of the 95% confidence interval wu determined by selectinga mal value for the relative risk and adjusting the value until the score statistic had a probability of 0.05. Results A total of 303 cases of non-Hodgkin's lymphoma were inter- viewed for the study, of whom 152 were males and 151 were females. The median age of cases was 65 years. No statistically significant differences between cases and controls were found for education, marital status, current family income, and highest family income. However, religion did differ between cases and controls. Fewer Protestants and more Jews are among cases as compared to the controls. A matched pair analysis of religion yielded a relative risk of 0.4 (P<0.001) for Protestants as compared to all other religions and a relative risk of 3.1 (P< 0.001) for Jews. A review of the numbers by pathological classification of the lymphomas showed a scarcity of many histological subtypes which would provide insufficient cases for analysis by e& subtype. In this analysis, we have combined histological subtypes into the groups listed in Table 1. Since the histological groups are formed using more than one system of classification, the groups listed in Table I are not independent of one another. The follicular group in the Working Formulation includes some Table I GmuDinps ofhistofomcalsubtvms used in the analvsis All histologies Modified Rappapon classification Histiocytic Lymphocytic Cell type T-cell Working formulation Follicular 303 99 94 ~ 24 .- - 5s -- cases that are in the lymphocytic group and some in the histi$ ocytic group in the Rappaport Classification. The diffuse, large cell group in the Working Formulation is largely a subgroup of histiocytic group in the Rappaport C~assification.c u r . rently, there is no known scheme which would group the vari3 ous histological subtypes by common etiologies. The groupings ' ,,sed are based on morphological similarity and are studied in this paper to determine whether morphological groups have ,.ioIogical implications. Table 2 gives the data for the association between exposures certain agents and the occurrence of non-Hodgkin's lym- @ma. Table 2, Column 2, gives the prevalence of the exposure in the total group of subjects, cases, and controls. This may be thought of as an estimate of the prevalence in the population ,,rider the assumption that there is no association between the ,Xposure and disease. The prevalence gives an inclicaiion :,f the probability of observing an association if one exists. In this a relative risk of 2-0r greater would be observed with an 0.80 probability using a significance level of 0.05 for a preva- lence of 4%, with a 0.90 probability for a prevalence of 6%, and with a 0.99 probability for a prevalence of 11Yo. For a relative fiskof 1.5,0.8 and 0.9 probabilities of observingthe association would require prevalences of 15 and 25Y0,respectively. Table 2 shows an increased risk of non-Hodgkin's lymphoma among [hose reporting exposure to particles (RR = 1.4) and those reporting exposure to pesticides or insecticides (RR = 1.8). Other agents listed show lower relative risks. although only one is below 1.0. None of these associations reaches statistical significance. Analyses similar to those reported in Table 2 were done for each of the histological groups listed in Table 1.The analyses were repeated with the elimination of all responses given by proxies. Tests for trends with increasing duration of exposure were done. Any association which was significant at the 0.1 level was considered to identify a candidate variable for the conditional logistic models. Variables were retained in the final model if their coefficients were significant at P = 0.05. The results of these analyses are given in Table 3. Table 3 shows the significant associations between non- Hodgkin's lymphoma and exposures to the specific agents. Only the conditional logistic model for all cases of nonHodgkin's lymphoma included a significant term for trend with duration of exposure to particles, expressed in decades of exposure. No other exposures were statistically significant when added to this model. No exposures were significantly related to all cases of non-Hodgkin's lymphoma when proxy responses were excluded from the analysis. The model for cases of histi- ocytic lymphoma showed an association with exposure to pesticides and insecticides only when the exposure occurred be[ween 1945 and 1969.Those with diffuse, large cell lymphoma Table 2 Associations between even being exposed YOspecific agents and non-Hodakin'slymphoma. all roses - Agents - .- - - =-- C isoline or kerosene B nzene C.ilorinated solvents Fumes or gases Particles (dust. sawdust. fibers) Tar Radioactivitv or X-ravs Fertilizers Pestlades or lnsecttcldes Prevalence in population (%) 17 4 24 26 39 5 4 It 7 Relative risk 1.0 1.2 1.2 0.6 1.4 1.3 1.3 1.1 I8 95qo confidence interval 0.6-1.7 0.5-2.6 0.8-1.8 0.6-1.2 0.9-1.8 0.6-2.8 0.6-2.8 0.5-2.3 0.9-3.7 (Working Formulation) also showed an association with pesti- cides and insecticides. However, all the dixordant pairs with the case exposed that led to the association with diffuse large cell lymphomas were the same pairs that led to the association observed for the histiocytic lymphomas. When proxy responses were excluded, the model for diffuse, large cell lymphoma showed an association with exposure to particles. None of the models for the other histological groups, with or without prox- ies had any terms which reached statistical significance. The time dependencies for the associations reported above were selected on the basis of having the largest level of statistical significance. Some of these associations were significant at other time periods or after other durations, but at lower levels of statistical significance. Table 3 dso shows the effect of controlling for religion, a demographic characteristic that differed between cases and controls in this study. This table as well as subsequent tables shows that controlling for religion usually results in small shifts in the relative risks and the confidence intervals. In general, controlling for religion does not alter the conclusions of this study. Associations between employment in specific industries and all non-Hodgkin's lymphomas are given in Table 4. Each occupation entered on the questionnaire was assigned to one of the industries listed in the table; no occupations were excluded from the table. As in Table 2, the prevalence of employment in an industry can be used as a guideline in deciding whether the lack of an association reflects the low prevalence in the population or supports the hypothesis that there is no association. Increased relative risks were found for the agriculture, forestry, and fishing industries (RR = 3.0); the construction industry (RR = 2.1): the paper and wood industry (RR = 1.7); and the leather industry (RR = 2.1). The relative risks were statistically significant at the 0.05 level for the agriculture, forestry, and fishing industries and for the construction industry. A statistically significantly decreased risk was found for those employed in the chemicals, drugs, and paints industries (RR = 0.3). Table 5 lists the associations between employment in industries and histological subgroups which closely approached or achieved statistical significance. To be consistent among industries, Table 5 reports the associations for those who were ever employed in the industry. Ever employment had higher statistical significance than employment in a particular time period for all histological categories except for follicular lymphoma, ,where employment in the construction industry between 1945 and 1969 was more statistically significant (RR unbounded; lower 95% confidence interval, 1.0).Table 5 shows that those employed in the construction industry are at increased risk of the lymphocytic (Rappaport) or the follicular (Working Formulation) lymphoma. Those in the agriculture, forestry, and fishing industries were found to be at increased risk for diffuse, large cell lymphoma (Working Formulation). Table 5 also shows the significant associations identified in a conditional logistic model involving multiple industrial exposure terms as related to the total group of all lymphomas. Both the agriculture and leather industries have significant associations after controlling for other industrial exposures. Table 6 gives associations between employment in an occu- pation and all non-Hodgkin's lymphomas. Table 6 does not include all occupations, since the prevalence of employment in some occupations was quite low. Only occupations with a prevalence of at least 3%or a relative risk significant at the 0.1 level NON-HODGKIN`S LYMPHOMA AND OCCUPAnONN EXPOSURE lime b hf, infinite Excludin~proxies. Table 4 A m knmn emphymenr in sjx&c i d m i a and -Hodgkin's lymphoma. all c a w Indusm Revalence in population (.%.) 95% Relative confidence risk interval Discussion The analysis of the associations of non-Hodgkin's lymp and occupational exposures has considered three aspects3f exposure in an a m m p t to identify risk factors for the d i d Agriculture, foremy. W i g 3 3.0 1.0-8.8 The fmt aspect was an attempt to look at classes of chemic& Construction Paper and wood $5 3 2. I 1.7 1.0-4.6 0.7-4.2 or agents which might increase the risk of disease. This Metal 4 1.2 0.6-2.5 proach would be the most desirable if information were ac&!- Machinery Shipbuilding, auto Food and tobpcco 9 10 0.8 1.0 0.4-1.4 0.6-1.7 rate, since it might lead to the identification of specific ca&f 16 0.9 0.6-1.3 agents. Accurate data on chemical exposures are difficult Textiles Chemicals. drugs, prinn Leather 9 3 4 1.1 0.3 2.1 0.6-1.9 0.1-0.9 0.9-4.8 obtain in home interviews, since the employee may have foe: gotten or never known the specific agents used in the past. The Medicine and OdeDDC Entertainment and -00 Art IO 0.9 0.5-1.5 second approach was to classify occupational exposures by 4 3 0.8 0.6 0.4-1.6 0.3-1.4 industry. This approach would make the best use of the data Job with few ex- 19 0.7 0.5-1.1 when all workers in the industry shared a common exposun, although specific etiological interpretations would be more difficult than in the f i t approach. The third approach was were included in Table 6. Increased r e h i v e risks were found for to categorize occupational exposures by specific occupations, plant farmers and gardeners (RR unbounded); painters, plas- such as secretaries and plumbers, so that workers with sim- terers, and housepainters (RR = 6.0); and carpenters, brick and ilar jobs could be aggregated independently of the industry, stone masons, plumbers, and roofers IRR = 12.0). A further As with industry as the defining exposure, specific occupa- analysis which considered separately each of the building trades tion would be best if occupational titles identified common that make up the latter two high risk groups showed that work- exposures. Although each of these three approaches has its ers in each of the building trades were at increased risk although limitations, a number of potentially useful results have been the number of indbidnals employed in each trade was not large identified. enough to reach staristical significance Decreased relative risks For occupational exposures categorized by exposure to were found for mathematicians and programmers (RR = 0.2) classes of agents. we found an increased risk of disease for those and for those employed in the sewerage, water treatment, and utilities production industries (RR = 0.0). Associations with specific histological subtypes and occupa- tion are reported in Table 7. The associations were reported for those who were mer employed in the occupation, since these associations were more statistically significant than employment in a particular time period. For histiocytic lymphoma (Rappaport), those employed as carpenters, brick and stone masons, plumbers. and roofers were found to have elevated relative risks (RR unbounded); while those employed as students and educaton were at decreased risk (RR = 0.1). Carpenters, brick and stone masons, plumbers, and roofers also had elevated relative risks(RR unbounded) for follicular lymphoma (Working Formulation). Those who were students or educators exposed to particles (dust, sawdust, fibers) and those exposedto pesticides and insecticides. An increased risk of disease with dust exposure has not been generally noted for non-Hodgkin's lymphoma. Increased risk with generalized dust exposure, however, has been reported for Hodgkin's disease (27). For expo- sure to pesticides and insecticides, we found an increased relative risk of 1.8 for all lymphomas and a large relative risk for the diffuse, large cell I!mphomas in particular. An increased risk of 1.9 has been reported for all lymphomas by Everett et af. (2%). Cantor (13) has reported relative risks ranging from 1.53 to 2.4 for farmers under the age of 65. Cantor also noted (15) 811 increased risk of small cell lymphocytic lymphoma among usem of specific pesticides. Hoar ef ai. (16) observed an odds ratio of were at decreased risk (RR = 0.0) for diffuse, large cell lym- I .5 for pesticide e--sure. The odds ratio decreased to 1.1 when phoma (Working Formulation). Table 7 also gives the results of controlled for herbicide exposure suggesting that the increased the conditional logistic model for all histologies of non- risk observed for pesticide exposure may result from the hi@ - Hodgkin's lymphoma. Carpenters, brick and stone masons, correlation between exposure to pesticides and exposum to plumbers, and roofers were at increased risk of non-Hodgkin's herbicides. - lymphoma while mathematicians and programmers were at de- We did not obsene an increased risk for those exposed to creased risk. A model that controlled for religion showed an benzene. The association between lymphoma and benzene_had increased risk for carpenters, brick and stone masons, plumb- been reported by Vianna and Polan (4), where they camps _- ers, roofers, and plant farmers. The association with mathema- the propomonal mortality rate for lymphoma among those.?! ;init andpro- -~ - . however, was _ _ -- not statistically-signif- had worked in an industry that used benzene with the ratefor those who worked in all other industriG.-AX%%TtX- NON-HODCKlNS LYMPHOMA AND OCCLTAllONN. EXPOSURE - Table 5 Association between rmpioynunr in selecied industries and hiadDginl p u p s of non-Hod.kin's I y m m Conuolhg for religion Histology Industry Relative risk 95% omfidewe mornl Relative risk .AlP Lymph& Follicular Diffuse. lvgc a l l Agrieulttm Leather Construction Construction Agriculturr 3.4 2.4 Infinite 6.0 Infinite 1.1-10.8 1.0-5.6 li-infiniu 0.9-38 I .O-bhite 4.4 2.2 Infinite 4.1 Inf~te 0 The mula for the tomi group of all histologies witb multiple exposureswere obtained a d t i o d logistic model. 95%c o ~ d e n a interval 1.4-14 0.9-5.4 I .3-infinite 0.54 1S - i i i i t e Table 6 .Irrociafion behwen cmplqynvnr in sela7ed axupations and non-Hodgkin's lymphoma, a l l a t a Oocnpation _- -Mathematician. .o m m m e r Physician Student, educator iccountant auditor Official. manager itmager (am),radio operator, dir traffii controller office w o k secretary Clerk ( o f f i work) salesperson Domestic bourewife, nurse's aide wai- cbet meat cutter Laundering. dry cleaning, leather producrs fabrication Protective setvim (policeman, firrman.military) Senice occupations (janitor) Plant farma and gardener E:ecuical equipment assembler L phoktem. Sewer, stitcher Painter. plasterer, housepainter Carpenter. brick and stone mason. plumber. roofer Truck driver. Pain conductor .airplane or auto mechanic, gas statiotl attendant Packaging warehouseman Working in sewerage or water matmenL or utilities a Inf. intinin?. Prevalence (%) 2 3 11 4 10 3 11 4 t2 25 7 3 4 4 1 3 4 1 2 4 3 5 1 Relative risk 0.2 1.3 0.7 1.6 0.6 0.6 0.8 0.8 0.9 1.3 1.o 1.6 0.7 I .7 Inf" 1.1 1.2 6.0 12.0 1.o 1.4 0.9 0.0 95% confidence interval 0.1-0.9 0.6-3.0 0.4-1.2 0.7-3.5 0.3-1.0 0.2-1.4 0.5-1.4 0.4-1.7 0.5-1.5 0.8-2.2 0.5-1.8 0.6-4.0 0.3-1.6 0.7-4.2 1.3-Inf 0.5-2.7 0.5-2.6 0.9--38 2.0-72 0.4-2.3 0.6-3.6 0.4- I.8 0.0-0.8 showed that the strength of their association was strongly supported by the increased risk of disease presumably due to benzene for woodworkers and farmers. Subsequently, it has &en suggested that the increased risk among woodworkers and farmers may be due to other agents. A later study also found no increased risk with exposure to benzene (29). Analysis of the data by industry found that workers in the agriculture. forestry, fishing industry; the construction indus1': :and the paper and wood industry were at increased risk of non-Hodgkin's lymphoma. Increased risks among agricultural workers have been reported in a number of studies (12-17). The increased risk may, in part, be a result of exposure to pesticides (dichlorodiphenyltrichloroethane,carbofuran, and chloramben) and to herbicides (chlorophenols and phenoxyacids) containing diinzodioxins and dibenzofurans as impurities. The increased risk among workers in the forestry, construction, w o d and paper industry may result from exposure to chlorc ohenols or phenoxyacids being used as herbicides or as wood preservatiws. Workers in forestry would be exposed as they applied the herbicides, and workers using the wood would be exposed when they worked with the wood which was treated with these compounds. In the literature, an increased risk of non-Hodgkin's lym- phoma has been reported for workers in the leather and textile had a relative risk of 2.1 for non-Hodglrin's lymphoma. Rather than being limited to any particular oocopation or group of occupatiom, the increased risk was noted among all persons employed in this industry. This finding is consistent with the study by D e c o d e (I@, who f i t reponed the asochtion in 1979. In that report, Decoufle renewed the various potential carcinogens used in the leather industq but did not single out any specific agents. Chlorophenols, a waterproofingagent used in the leather industry (30), would be a candidate for further study. Ln this study, we found a relathe risk of 1.1 (confidence interval. 0.6-1.9) for employment in the textile indusnies, a value ahich would be consistent with w increased risk among textile workers. Delzell and Gruferman (19), however, reported a proportional mortality ratio oi 1.7 for former female textile m d u s a y employees. The 95woconiidenceinterval for this study indudes their point estimate. Ana&& of the occupational exposure data by occupation group identified three groups of workers at increased risk of non-Hodgkin'slymphoma. The first of thesegroups is the plant farmers and gardeners. This finding is consistent with our o b servation of an increased risk for exposure to insecticides and pesticides exposures that should be highly correlated with plant farmers and gardeners within the same study. Plant farmers and gardeners may also be exposed to herbiddes, but this exposure was not investigated as part of this study. The second and third groups of occupationsat increased risk were (a)painters, plasterers, and housepainters and (b) carpenters, brick and stone masons, plumbers and roofers. This result is consistent with the increased risk found for workers exposed to particles such as dust, sawdust. and fibers. Part of the increased risk could be associated with exposure to wood containing chlorophenols. Also, there has been a report of an in- creased risk of small cell follicular lymphoma among plumbers, presumably resulting from the use of rinyl pipe (31). However, the exposures experienced by workers in these building trades are far too complicated to be attributed to the exposures mentioned above without a more detailed intestigation. It is possible that the results of this 3mdy could be biased by the use of proxy interviews which accounted for approximately one-third of all the responses from cases. Proxy interviews were not used for any of the controls. The use of proxies could have had differing effects on the three categorizations of occupational exposure. The influence of p r o q responses on d a t i v e risks *as probably greatest for the exposures to specific agents. Here it is unlikely that the p r o y respondent would be familiar with the Iifetime history of exposures experienced by the case. Since underreporting of exposures in proxy responses would occur onl? in the case series. there a d d be a decrease in the relati\e risk reported for causal faaors and overestimate of effect of a n y protective factors. In this mdy, since most of the signrficant effects observed relate to increased risk, the use of proxy responses would have tended to underestimate the relatiuedsk-QheLeffects actual11 present in this series may have - ~ -. - , - ~- . NON-HOffiKIN'S LYMPHOMA AKD-QEUE'ATIONAL EXPOSURE Table Arrociations knveen employment in selected o c ~ r d h n f o l o ~ cgraoulps of non-fiodgkm's h'mphoma Controlling for relrgion Hislological group Occupation zril histologies Histiocytic Follicular Diffuse, large cell Carpenter, plumber Mathematics, programmer Plant farming Carpenter, plumber Student, educator Carpenter. plumber Student, educator a Relative risk rstimated from a conditional logistic model. 'Variable not included in h e model. u s not significant;Inf. infmite. Relatin riSP 11.0 0.2 b Inf 0.1 Inf 0.0 95% confidence interval' 1.6-92 0.0s-I.o nsb 1.O-lnf 0.01-0.5 1.3-lnf 0.0-0.4 Relative risP 13.2 lnf n 0.1 Inf 0.0 9590 confidence intena!" 1.i-I (L: n+ 1.54~ nsh 0.01-o.t 1.0-In! 0.0-0.5 been missed because of the high proportion of proxy respon- number of cases of each specific histological subtype was tm, dents. For occupational exposures characterized by industry. small for a meaningful analysis, the only opportunity to obsene the influence of prox! responses is expected to be relative15 any association with histological subtype was obtained by :3x11- small. Here, the respondent is expected to know only about the ing histological groups. cases' recent occupations and major occupations. For the ex- Such a dilution effect appears to be present in the associatioc posures characterized b! specific occupation, there may be a of lymphoma with the student-educator occupation in the tendency for the respondent to "promote" the deceased to a present data. In Table 6, we showed a weak effect in the direc- more prestigious position than the position that was actual11 tion of risk reduction for the student-educator occupation ic held. Such promotions have been observed in studies of OCCUpations reported on death certificates (32). These promotions would tend to remove the case from an occupation of higher exposure to an occupation of lesser exposure, since the exposures identified tend to be associated with less prestigious occupations. This would tend to reduce the estimated magnitude of a -causal" association and thereby reduce the chance of detecting a small true association. For causal associations observed for all three categories of occupational exposures, these considerations would lead one to believe that with the use of proxy responses, the observed relative risk would be lower than would have been found if direct patient responses could have been obtained. We investigated the proxy effect by analyzing our data using all respondents and, separately, using direct respondents onl?. In the study of occupational exposure to dust, sawdust, and fibers, a significant association was found using all respondents. but the risk was not significant when direct respondents were considered alone. On the other hand, this exposure was found to be significantly associated with the histological subtype of diffuse, large cell lymphomas when direct respondents were considered, although it did not reach significance when proxies were included. We do not know of a consistent explanation of these findings, although it seems reasonable to propose that the sisnifcant relation found using direct respondents and the more homogeneous subtpe of diffuse large cell tumors is a true association. In other parts of this study, the inclusion or exclu- sion of proxy reports did not affect conclusions. relation to all histological types taken together. In Table 7 . we found a strong risk reduction for student-educators in relatioc to histiocytic lymphomas (Rappaport) or diffuse large cell tumors (Working Formulation). In fact. the two histologichi sub opes are associated with student-educator occupation in dlj- cordant directions in our data. The subtype of lymphocytic (Rappaport) or follicular (Working Formulation) lymphoma S increased in the student-educator group and tends to obscure the stronger risk reduction effect for histiocytic or diffuse cell types. A less smking example of such a dilution is seen u the construction industry, where the lymphocytic or follicuk subtype is greatly increased (Table 5 ) although the associarioc of this industry with lymphomas generally is of borderline significance (Table 4). In conclusion, we have found an increased risk of lymphom among persons exposed to insecticides, pesticides, and particle such as dust. People employed in the agriculture, forestn. fishing industry; consuuction, paper and wood industry leather industry were also found to be at increased risk. The specific occupations at increased risk were those of plant farmers. gardeners, and people in the building trades such as painters, plasterers, housepainters, carpenters, brick and stone masons, plumbers, and roofers. In each of the occupations- workers had the possibility of exposure to chlorophenols 0: phenoxyacids. The results of this study suggest the need '0' a further investigation of the leather and construction indusme to identify other agents which may confer increased risk. A second factor which could bias the study results is the Acknowledgments selection of those histologies which formed the histological groups for analysis. A broad histological group as used here We are indebted to the administration, record room, and patholm might conbin diseases having different etiologies associated staffs of Beth Israel Hospital, Boston University Medical Cent-. with different specific histologies. For example, the poorly dif- ferentiated lymphocytic lymphoma may have a different etiology from that of the well differentiated lymphocytic lymphoma, Brigham and Women's Hospital, Children's Hospital, D a n a - f m Cancer Institute, Lahey Clinic, Massachusetts General Hospital. England Deaconess Hospital, and TuftsNew England Medical Cent= 10 h.C. Arkin, R.DeLellis, N.Harris,E.Kasdon, G . Pinkus. a d G- both of which we have included in the category lymphocytic Vawter for reviewing the pathology slides; to the physicians who PaVtc r ! lymphoma. If there were a causal or a protective association permission to interview the patients; and to the people who agreed tob between exposure and only one of the types of lymphoma in a interviewed as part of this study. group, the magnitude of the association observed for the group would be less than the magnitude of the true association for that particular histology. In general, the formation of these groups might mean that a true association could be missed. Since the References 1. Li, F. P., Fraumini J. F.,Mantel, N.,etd. Cancer mortalityamone J. Natl. Cancer Inst., 43: 1159-1164,1969. 5508s - - ______ - earle, C ~ % e ~ i study of the mortplityafXIritishchemls Harrington J. M.. and Shannon H S medical labontor) technicians Br Me, Vianna N. J.. and Polan. A. Lymphomas Lancet, I 1394-1395. 1979 Nicholson. H.J.. Selikoff, 1. J.. and S styrene-polyrtwne polymerization wor Environ. Health. I(Supp1.2).247-252 6. Tabershaw, 1 R. and Gaffey, H R manufactureofnn!l chloride and its pol 7. Hardell. L., E r i h n , M.. Lenner. P.. ei sure to chemials especially organic sc acids a casceonuol study Br. J. Cancc 8. Goldsmith. J. R. and Guidotti, T. L E, . ology of Hardell, lymphosarcoma. Pathol. Annu L., Enlcrson, M.. and Lenner, P kemiske subsun~r,sprskitt organiska 1 . isyror. Laklnidnmgen, 77. 208-210. 15 Hardell, L M a b a n t lymphoma of hist . acetlc acids or cblorophenols Lancet, I Hardell, L. Relrnon of soft-tissue sarco cancer to phenox) acids, chlorophenol* Envvon Health.7: 119-130, 1981. Schumacher, M.C. Farmingofcupation . lymphoma in Utah.J Oceup. Med..2; Cantor, K. P. Farmlng and mortalit! case-conuol stud!. Int. J Cancer, 29. 2 Burmeister, L. F.. Everett G D., Van mortality end farm practices rn Iowa AI Cantor, K., E v a e ~G.,Blur, A,, CI QI phoma AbsUacrs of 18th Annual Meet] Research, p. 39. 1985. Hoar, S., B h . A, Holmes, F., et a/. A lymphoma and soft nssue sarcoma. JAh Zahm, S. H.. Wesenburger, D. D., Babt of non-Hodglmt lymphoma and the t acid (2,4-D) m eastern Nebraska Epide Decoufle, P Cancer nrks assoclated H NON-HODGKIN'S LYMPHOMA AND OccUpATIONAL U(POSURE - -i 2. S e d % C E, Waterhouse, J. A. H., Henman, B. A., et d.Epidemiologic study of tbe mortality of British chemists. Br. J. Cancer. 38: 192-193,1978. 3. Hadogtom J. M.,and Shannon H. S. Mortality otudy of pathologists and medial L;lbontory technicians. Br. Med. J., 4: 329-332.1975. d. V i m N. J- and Polan. A. Lymnhomas and occunatiolul benzene exwsure. - b a t . I: 1394-1395. 1979,5. Nicholrols W. J., Selikoff, 1. J.. and Seidman, H. M d t y experience of en-s polymerization workers. Initial lindings. Sand. J. Work En- Haltb. 4(Suppl. 2): 247-252, 1978. 6. Tabenbrr, I. R,and Gdey, W. R. Mortality nudy of workers in the manuhawe ofvinyl chlorideand its polymers. J. Oeeop. M d . I& 509418, 1974. 7. HPrddL L,~ h nM.,, Lenner, P., et d.Malignant lymphoma and expo- - SUR to dumuk, especially organic solvents, chlmphenoh and phenoxy acids a -nml study. Br. J. Cancer, 43: 169-1741981. 8. Goldrmim J. R,and Guidotti, T. L. Environmental factors in the epidemi- ology dlymphos~rcoma.Pathol. Annu., 22: 411424, 1977. 9. Hard& IEriluJon, M., and Lenncr, P. Maligna lymfom och expositionfor kemhke -, mkin organnka losingsmedalklorfenoler och fenoxy- 10 isymr. LpLunmungen. 77: 208-210, 1980. \e IO. Harden, L Malignant lymphoma ofhutiofytic typeand exposureto phenoxy acetic acids or chlorophenols. Lancet, 1: 55-56, 1979. 1- I I,Hardell L Relation of soft-tissue sarcoma, malignant lymphoma and colon c a n e r to phenoxy acids, chlornphenols and other agents. Sand. J. Work ,n Enviroh Hdtb, 7: 119-130.1981. I 2. Schumrber, M. C. Fanning occupationsand mortality fmm non-Hodgkin's le lymphoma in Utah.J. Occup. Med., 27: 580-584, 1985. c- in 13. Cantor, K. P. Fanning and mortality from non-Hodgkin's lymphoma: a casxoufml study. lnt. J. Cancer, 29: 239-247. 1982. 14., -B L F., Everett G. D., Van Ljcr S. F., a aL Selected m c e r re mortality and farm practices in Iowa. Am. J. Epidemiol., 118: 72-77,1983. in 15. Cantor, K.Everett, C., Blair, A., et al. Fanning and non-Hodgkin's lym- phoma A b n c t s of 18th Annual Meeting of the Society for Epidemiologic U- Rew.rcb, p. 39. 1985. b- 16. Hoar. S, Blair, A., Holmes, F., et d.Agricultural herbicide use and risk of 1s- lymphoma and soft tissue sarcoma. JAMA, 256: 1141-1147, 1986. 17. Zahm. S H.,Weisenburger, D. D.,Babbitt, P. A, a aL A ase-control study ic of non-Had@n's lymphoma and the herbicide 2,4-dichlorophenoxyatic is acid (2.4-D)in eastern Nebraska. Epidemiology, I: 349-356. 1990. re 18. Deco~dk.P.Cancer risks associated with employment in the leather and e: in ar Jn 4- *uba products industry. Arch. Environ. Halth. 34: 33-39.1979. 19. Dekk E., and Gruffman, S. Cancer and abs Q ~ LoIf death among fermk textile workers, 1976-1978. J. Natl. c.oeSlnn.7I:735-740,1983. 20. Rorr R.Nichols.P., Wright, W., et aI. Asbauaerp~arraud lymphomas ofthe gasmintestinaltract and oraI cavity. I- 1118-1120.1982 21. Rappaport, H. Tumors of tbe Hematopoietic Sysem Wrrhigton, DC .Armed Forces Institute of Pathology, 1966. 22. The non-Hodgkin's lymphoma pathologic d. *- poject. National 1MtitutC SpODsorrd study Of d of r!oa-Hod&in's 1ymphom.s-summary and description of a Gmmdmton for clinical usage Cancer (Phila).49: 2112-2135, 1982 23. H o u , S . , M o r r i s o n , k S . , C o l ~ P . , e t d . A q ~ d a p o s u r r L i n k . g e s-vs,te-m-f-o~ r-th-e sNdv of ocmoationd mummsEa J. Oecuo. Med.. 12: 722-726, 1980. 24. United S&ta Department of Commmc B u ~aaflkbeccmtp 1970C c n w of Population, clysified Index of Industries .nd Libnry of Colgpepa 74-179488. 2s. Diaionary of Oceupation Titles, Vols. 1 and 2 W-n. Dc: United States Department of Labor. United Stata G~SDEPUrinUting Office. I%? 26. Cox. D. R.. and Hinkley. D.V. Theoretical StntktnLoodmChapmanand H.U.1972. 27. Cutensohn. N. M. The epidemiologyof Hodgkia's bduesto etiology. University of Texas M. D.Andemon clia. C o d cancer, 27: 3-9, 1984. 28. Everett. C., Blair, A., Cantor, K., et d.E n v i r o d e b c d cxposurrs as risk factors for leukemia and non-Hodglrio'slplphomr AbsPacts of 18th Annual Meeting of the Society for EpidemiologkR e , p. 39, 1985. 29. Baonrd S.. Canwright. R,Bud C.,et ai. A a d o s k haon b lymphoid mdignancies: a -nml epidemiologial samly. LmL Rer. 8: 681-689, 1984. 30. HYdelL L.. and Sandsmom,A. Caoe-control Smtr: soft tksue sarcoma and exposure to phenoxyaatk acids or chloropbmok R. J. Caocer, 39: 711- 717. 1979. 3I. Dolam B. P.. Levine. A. M.. and Dolan, D. C S d dadfollicular Center cdI lymphonu: seven cases in California ptmnbcrr J. Oecup. Med., 25: 613-615, 1983. 32. The Registrar General`s Decennial Supplement J k g h d aod Wales. 1951, Occupational Mortality,,Part 2, Vol. 1. LondoE H a \(aiatY's Stationery O S u . 1958. na les hnd he n- nt- 1 ne ns, or ra ies JgY ter, ber -ew iter, IG. ?us- Jbc -- -. .-