Document M49GQnr5rkNr25QGg1aEaaQEV

' Jnt. J. hncer: 39, 155-161 (1987) @ 1987 Alan R. Liss, Inc. Publication of the International Union Against Cancer Publication de I`Unbon Internationale Contre le Cancer NON-HODGKIN'S LYMPHOMA AND FARMING: AN EXPANDED CASE-CONTROL STUDY Neil E. PEARCE''2'5,Raewyn A . sHEPPARD2, Allan H.SMITH3, and Clinton A. TEAGUE4 `Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC 27514, USA; 2Department of Community Health, Wellington Clinical School of Medicine, Wellington Hospital, Wellington, New Zealand; 3Department of Biomedical and Environmental Health Sciences, School of Public Health, University of California, Berkeley, CA 94720,USA; and 4Departmentof Pathology, Princess Alexandra Hospital, Brisbane, Queensland 4102, Australia. A previously published case-control study of agricultural istered under ICD 200. Second, the data for the previously risk factors involved male cases of non-Hodgkin's lymphoma interviewed ICD 202 cases are re-analysed with additional registered under code 202 of the International Classification controls. Finally, for both case groups, data on additional risk of Diseases (ICD). This study has been expanded with the factors are also presented, including: type of farming; contact inclusion of cases registered under ICD code 200, and additional controls. The expanded study comprises 100 ICD 200 cares and 83 ICD 202 ~asesregistered during the period 1977- with farm animals; and history of various medical conditions and allergies. 81, together with 338 controls selected from other cancer registrationsduring the same period. The largest relative risk METHODS for specific farming types was for orchard workers (odds ra- tio =3.7, 90% confidence limits 1.1-12. I). No elevated risks The expanded target case population (Pearce et a l . , 1985) were observed for exposure to farm animals, nor for potential exposure to phenoxy herbicides (odds ratio = I.O. 90% confidence limits 0.7-1.5). or chlorophenols (odds ratio = 1.4, 90% confidence limits 0.8-2.3). The previous finding of an excess risk associated with fencing work was weakly supported by the expanded study (odds ratio = I.4, 9046 confidence limits 1.0-2.0). However, the previous finding of an excess risk wo- ciated with meat works employment was more strongly sup ported (odds ratio = I.8, 90% confidence limits 1.2-2.6). One comprised male public hospital patients who were registered under ICD codes 200 and 202 during the period 1977-81 and who were under 70 years of age at time of registration. ICD code 200 comprises lymphosarcoma and reticulosarcoma. The cases registered under ICD code 202 had been interviewed previously, and this group comprised all other non-Hodgkin's lymphomas, primarily nodular lymphoma and unspecified nonHodgkin's lymphoma. relevant risk factor is 2.4.6-TCP which is used in the treat- Table I shows the number of exclusions, the target popula- ment of pelts, but the excess risks do not appear to be con- tion, and the response rates. Relevant laboratory reports of fined to pelt department workers. An alternative hypothesis is that meat workers may be exposed to oncogenic viruses. cases coded as non-Hodgkin's lymphoma were examined by one pathologist and 10 cases without a documented tissue diagnosis of non-Hodgkin's lymphoma were excluded. ReaA previous analysis of the occupations of male malignant sons for other exclusions of cases or controls included: private lymphoma patients recorded by the New Zealand Cancer Reg- hospital patients who had been wrongly coded at the Cancer istry during the period 1977-81, compared with other cancer Registry; duplication at the Cancer Registry; wrongly coded controls, found that agricultural workers were at increased age; and persons who had come to New Zealand solely for risk of developing malignant lymphoma (Pearce et a [ . , 1985). medical treatment. Similar excesses have been found in studies in other countries (Blair et al., 1985; Buesching and Wollstadt, 1984; Burmeis- For each of the cases, 4 male cancer patients who had the same year of registration and were born within 2 years of ter, 1981; Burmeister et al., 1983; Cantor, 1982; Goldsmith and Guidotti, 1977), including a Swedish study which found an association between malignant lymphoma and exposure to phenoxy herbicides or chlorophenols (Hardell et al., 1981), both of which have been widely used in New Zealand since the late 1940's (Smith et al., 1983). patients' date of birth had previously been chosen as controls for the study on the basis of Cancer Registry information on occupation (Pearce et al., 1985). For the cases registered under ICD code 202, 2 of the 4 control patients were selected at random for interview, and the same exclusion criteria were applied as had been used for the cases. The interviews were The New Zealand non-Hodgkin's lymphoma excess oc- conducted concurrently with those of the other cancer controls curred almost entirely among relatively young patients regis- for a case-control study of multiple myeloma and farming tered under code 202 (non-Hodgkin's lymphoma other than (Pearce et al., 19866), yielding a series of 315 interviews from lymphosarcoma and reticulosarcoma) of the International a group of 389 eligible controls. The additional cases regis- Classification of Diseases (ICD) (WHO, 1967, 1977). Accord- tered under ICD code 200 were interviewed towards the end lngly, interviews were conducted with patients from this of the time period during which the series of 315 controls was subgroup and their matched controls with other cancers. The interviewed. An additional group of 30 controls was included interview findings (Pearce et al., 1986a) for exposure to phe- in order to keep the interviewer "blind", and interviews were noxy herbicides or chlorophenols were largely negative. How- conducted with 23 of these. The expanded non-Hodgkin's ever, the interview data suggested that the overall excess risk lymphoma case-control study thus comprised interviews with for farmers was attributable to excesses among farmers who 100 cases registered under ICD code 200, 83 cases registered had carried out fencing work or been employed in a meat works, with particularly high risks for those who had carried both activities. As more resources became available it was also to interview the non-Hodgkin's lymphoma cases refwered under ICD code 200 in order to provide a more comparison with findings from other studies. This thus extends the previously published study in several 5T0 whom reprint requests should be sent, at the University of N. Carolina. First, interview findings for the risk factors previously are presented for the additional group of cases reg- Received: June 18, 1986 and in revised form August 20, 1986 156 PEARCE ET AL. TABLE I - TARGET SAMPLE, ACHIEVED SAMPLE, AND REASONS NOT INTERVIEWED tions selected for questioning were those which had been associated with tumors of the lymphatic and haemopoietic ICD 200 cases ICD 202 CaSeS Cancer controls system in previous epidemiologic studies or case-series reports, and which were expected to have sufficiently high inci- Initially identified in cancer registry 136 106 446 dence in the control population. The Statistical Analysis System logist procedure (Harrell, 1977-81 1983) was used to perform an unconditional maximum likeli- Ineligible: Histology Other reason 4 11 6 6 - 27 hood logistic regression analysis, adjusting for decade of birth and whether the subject or the next-of-kin was interviewed. Target sample 121 94 419 Hospital could not 5 19 RESULTS identify or had no information Medical advice not to contact Refusal Unable to contact Achieved sample 3 2 11 100 Most cases of non-Hodgkin's lymphoma had been report4 2. 15 using Rappaport's classification (1966), and the others were reclassified according to this classification by one pathologist 0 14 when sufficient detail was obtained from the histology repom. 8 43 The remainder were included in unclassified categories (Table 83 338 11). Unfortunately, appropriate information was not available Achieved sample as percentage of 83 88 81 to permit the use of other classifications. The case group included 3 cases of leukaemic reticuloendotheliosis, and 2 target sample (%) cases of Mycosis fungoides, which are not non-Hodgkin's lymphomas despite being registered under ICD code 202. However, excluding this small number of cases made no dif- ference to the findings, and they were included in all analyses TABLE I1 - DISTRIBUTTON OF HISTOLOGICALTYPES OF NON-HODGKIN'S presented here in order to provide greater comparability with LYMPHOMA CASES other studies based on the ICD classification. Cell type Number of cases Table 111 gives the findings for various occupations and activities potentially associated with exposure to phenoxy her- Nodular (42) bicides, and Table IV gives the specific findings relating to Lymphocytic (28) Well differentiated Poorly differentiated Mixed Histiocytic Unclassified Diffuse (1 15) Lymphocytic (44) exposure. The proportions of cases and controls who had 6 22 9 4 worked in the various occupations were very similar. Elevated odds ratios were observed in the ICD 200 case group for persons carrying out spraying for the Ministry of Works, or as 1 forestry or railway workers, but each of these estimates was based on a single exposed case. None of the odds ratios relating to specific phenoxy herbicide exposure were elevated. Well differentiated Poorly differentiated Mixed Histiocytic Unclassified Leukaemic reticuloendotheliosis Mycosis fungoides 12 32 12 54 5 Table V gives the findings for various occupations and activities potentially associated with exposure to chlorophenols. The elevated odds ratios for meat works employment and fencing work in the ICD 202 case group were supported 3 by the findings for the ICD 200 case-group, although the 2 fencing work association was relatively weak. The findings Unclassified 21 for meat workers and fencers did not show any clear associa- Total 183 tion with specific subtypes of non-Hodgkin's lymphoma, although this may be due to the small numbers involved in the comparisons (Table 11),or because only the Rappaport classi- under ICD code 202, and 338 controls. The entire control fication was used. group was used in all analyses to provide greater precision for Table VI reports the specific findings relating to chloro- the estimates of control exposure. phenol exposure. The odds ratios for potential chlorophenol Interviews were conducted by telephone with those patients exposure were elevated, but this was largely due to the find- considered to be well enough, or with the patients' next-of- ings for meat workers which included workers potentially kin. All interviews were conducted by one interviewer who exposed to chlorophenols in pelt departments. The odds ratios was not aware of whether the patient had a non-Hodgkin's for all non-Hodgkin's lymphoma for the 3 categories of chlo- lymphoma or was a cancer control. The questionnaire was rophenol exposure were 1.2, 1.1, and 1.0, respectively, when similar to that used in previously published case-control stud- meat workers were excluded from the analysis. ies (Pearce et af.,1986a,b; Smith et af., 1984). As before, Our initial analysis (Pearce et af., 1986a) suggested that stem questions were used to identify whether or not study there was a statistical interaction between the effects of fencing subjects had carried out activities, or worked in particular work and meat works employment, and this is further investi- occupations, which entailed a potential exposure to phenoxy gated in Table VII. In both studies, the odds ratio for exposure herbicides or chlorophenols. If the response to a stem question to both factors was greater than the product of the odds ratios was affirmative, then a series of subsidiary questions were for exposure to each factor alone. asked to clarify the work done and the actual potential for The findings for various types of farming work are pre- exposure, firstly in general terms, and then in specific terms, sented in Table VIII, and those for exposure to various farm seeking the identity of the chemicals used. Additional ques- animals in Table IX. There are few New Zealand farms solely tions were also asked concerning involvement in various types involving cattle, but there are a number of "rnixed/dry stock" of farming, contact with various farm animals, and history of farms with both sheep and cattle. Cropping farms primady various medical conditions and allergies. The medical condi- involve market gardens which grow vegetables and fruit for ~.. NON-HODGKIN'S LYMPHOMA AND FARMING TABLE Ill - ODDS RATIO ESTIMATES FOR NON-HODGKIN'SLYMPHOMA FOR OCCUPA,TIONSAND ACTIVITIES INVOLVING POTENTIAL EXPOSURE TO PHENOXY HERBICIDES Exposure Farming Used any agricultural chemical spray Sprayed gorse, blackberry, pasture, cereal or peas Forestry worker Sprayed chemicals Railway worker Sprayed chemicals Ministry of works Sprayed chemicals Town council worker Sprayed chemicals Chemical sprayer Aerial spray work Ever sprayed an aericultural chemical ICD 200 $2cases 90% Confidence limits ICD 202 $: 90% Confidence limits 41 0.9 0.6-1.3 40 1.2 0.8-1.8 14 0.7 0.4-1.3 19 1.4 0.8-2.3 11 0.7 0.4-1.4 14 1.2 0.7-2.2 81 33 25 8 0.8 0.4-1.5 7 0.9 0.4-1.9 15 1 2.0 0.2-2 1.5 0 - - 1 8 0.6 0.3-1.2 12 1.2 0.7-2.2 20 1 3.2 0.3-37.8 0 - - 1 9 1.2 0.6-2.3 1 6.1 0.6-65.8 6 0 0.9 - 0.4--1.9 15 1 9 0.7 0.4-1.3 9 0.8 0.4-1.6 18 2 0.5 0.1-2.0 1 0.4 0.1-2.4 3 1 1.5 0.2-11.8 1 1.6 0.2-12.2 2 0- - 1 1.4 0.2-11.3 1 55 1.1 0.8-1.7 51 1.4 0.9-2.1 106 All NHL 143 1.o 55 1.o 43 0.9 30 0.8 1 1.1 40 0.9 1 1.4 27 1.o 1 2.1 44 0.7 10 0.5 1 1.7 2 0.6 170 1.2 'Stratified for decade of birth and whether the subject or a relative was interviewed. 157 90% Confidence limits 0.8-1.4 0.7-1.5 0.6-1.5 0.5-1.4 0.1-11.7 0.5-1.4 0.1-15.0 0.6-1.8 0.2-2 1.6 0.5-1.2 0.2-1.5 0.3-7.7 0.1-4.8 0.9-1.7 TABLE N - ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR VARlOUS CATEGORIES OF EXPOSURE TO PHENOXY HERBICIDES' Exposure ICD 200 ICD 202 ""g?zt:C W S Oradtdios 90% confidence limits 90% Confidence limits All NHL 2;: 90% Confidence limits Ever potentially exwsed before cancer registration Potential exposure of more thani day at least 5 years before cancer registration Probable or definite exposure of more than 1day at least 5 years before cancer registration Probable or definite exposure of at least 5 days more than 10 years before cancer registration 23 0.9 0.6-1.5 21 15 0.8 0.4-1.3 17 13 0.7 0.4-1.3 16 9 0.6 0.3-1.2 14 1.1 0.7-1.8 1.2 0.7-2.0 44 32 1.2 0.7-2.1 29 1.3 0.7-2.2 23 72 1.0 0.7-1.5 56 0.9 0.6-1.4 50 1.0 0.6-1.5 41 0.9 0.6-1.5 'Stratified for decade of birth and whether the subject or a relative was interviewed. TABLE V - ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR OCCUPATIONS AND ACTIVITIES INVOLVING POTENTIAL EXWSURETOCHLOROPHENOLS' Exposure ICD 200 $2 90% Confidence CWS limits ICD 202 Oradtidos 90% Confidence limit All NHL F:n$s; 90% Confidence IirniLc Fencing work 34 1.3 0.8-1.9 34 1.6 1.1-2.5 68 Treated own posts 1 0.3 0.1-1.8 1 0.4 0.1-2.6 2 Saw mill or timber 14 1.0 0.6-1.7 10 0.9 0.5-1.6 24 merchant Potential exposure 7 1.1 0.5-2.5 4 0.8 0.3-2.1 11 at saw mill or timber merchant Meat works '24 1.8 1.1-2.9 19 1.7 1.0-2.8 43 Pelt department 6 2.2 0.9-5.3 4 1.7 0.6-4.7 10 in meat works Tannery 0- - 2 1.0 0.3-4.1 2 93 1.4 1.0-2.0 8 0.3 0.1-1.3 45 0.9 0.6-1.5 18 1.0 0.5-2.0 52 1.8 1.2-2.6 10 1.9 0.9-4.0 7 0.5 0.1-1.8 'Stratified for decade of birth and whether the subject or a relative was interviewed. 158 PEARCE ET AL. TABLE VI - ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR VARIOUS CATEGORIES OF EXPOSURE TO CHLOROPHENOU' - ICD 200 ICD 202 All NHL Exposure Excpaosesesd 22 90% Confidence limits :{: 90% Confidence limits -controls Oradtdios 90% Confience limits Ever potentially exposed Potential exposure of more than 1 day at least 5 years before cancer registration 12 1.4 0.8-2.7 9 11 1.3 0.7-2.4 9 1.3 0.7-2.7 1.3 0.7-2.7 21 27 20 . 27 1.4 0.8-2.3 1.3 0.8-2.3 Poofteanttilaelasetx5podsauyrse more than 10 years before cancer registration 7 1.1 0.5-2.4 7 1.4 0.6-2.9 14 21 1.3 0.7-2.3 `Stratified for decade of birth and whether the subject or a relative was interviewed. TABLE VI1 - ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR SEPARATEAND JOINT EFFECTS OF FENCING WORK AND MEAT WORKS EMPLOYMENT' Fencing No Yes No Yes Meat works No No Yes Yes ExcPaOsseusl 55 21 11 13 ICD ZOO Oradtdios 1.o 1.6 1.5 3.4 90% Confidence limits 0.8-3.1 0.8-2.9 0.8-14.3 ICD 202 2;; 90% Confidence limits 42 1.o - 22 1.9 0.9-4.1 7 1.2 0.6-2.6 12 4.0 0.8-19.5 97 43 18 25 All NHL 2; 212 1.o 74 1.7 33 1.4 19 3.8 90% Confidence limits - 1.0-3.0 0.8-2.3 1.2-12.1 `Stratified for farming, decade o f birth, and whether the subject or a relative was interviewed. -TABLE WI ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR VARIOUS TYPES OF FARMING' Exposure ICD 200 ICD 202 %E ""E?cases 90% Confidence limits Sheep farm Daily farm Mixedldry stock farm Cropping farm Poultry farm Orchard 13 0.9 0.5-1.5 19 0.8 0.5-1.4 12 0.9 0.5-1.7 5 1.1 0.4-2.7 1 0.7 0.1-4.9 3 3.5 0.9-14.0 15 1.3 18 0.9 9 0.8 3 0.8 0- 3 4.3 `Stratified for decade of birth and whether the subject or a relative was interviewed. 90% confidence limits 0.7-2.2 0.6-1.5 0.4-1.5 0.3-2.3 - 1.1-17.3 28 37 21 8 1 6 ALL NHL E"," 47 1.1 75 0.9 38 0.9 14 0.9 4 0.4 3 3.7 90% Confidence limits 0.7-1.6 0.6-1.3 0.5-1.4 0.4-2.0 0.1-2.5 1.1-12.1 the consumer market, but some involve the production of wheat and other grains. The only elevated odds ratio was for orchard workers. Finally, elevated risks were observed for persons with reported histories of rheumatoid arthritis, eczema and food allergies (Table X). No patterns were discernible concerning the association with food allergies, which included allergies to flour, dairy products, mushrooms, and seafood. DISCUSSION A major reason for conducting these case-control studies was to ascertain whether the association between exposure to phenoxy herbicides or chlorophenols and malignant lymphoma, which had been observed in the Swedish study (Hardell et al., 1981), existed in New Zedand. The initial analysis (Pearce et al., 1986~)of cases registered under ICD code 202 did not find an association of non-Hodgkin's lymphoma and phenoxy herbicide exposure, but was not strictly comparable to the Swedish study (Hardell et af., 1981) which primarily involved cases of non-Hodgkin's lymphoma coded under ICD 200. The Swedish study also involved cases of Hodgkin's disease (ICD 201), but the findings were similar for the Hodgkin's disease and non-Hodgkin's lymphoma cases. The expanded New Zealand study thus has the advantage of greater comparability with the Swedish study. A 2-tailed test of heterogeneity of the findings in the 2 studies indicates that the differences go well beyond what might be expected by chance (p<O.Ol), and the New Zealand findings suggest that, even if there was a real causal link, it is most unlikely that the true relative risk would be greater than 1.5 for persons exposed to phenoxy herbicides, or 2.3 for persons exposed to chlorophenols. The discrepancies in the findings of the two studies are puzzling, particularly in the light of the recent case-control study of non-Hodgkin's lymphoma in Kansas (Hoar et al., 1986) which found an overall relative risk of 2.2 for phenoxy herbicide exposure. The effect corresponded with the use of 2,4-dichloro~henoxyaceticacid (2,4-D), since usage of other phenoxy herbicides was small. The New Zealand study differed from those in Sweden and Kansas in that the control group consisted of other cancers selected from the National Cancer Registry. The main advantages of using other registrants as controls lie in the minimi- NON-HODGKIN'S LYMPHOMA AND FARMING TABLE IX - ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR CONTACT WITH VARIOUS FARM ANIMALS' Exposure Exposed cases ICD 200 Odds ratio con9fi0d%ence limits EXPOS4 cases ICD 202 Odds rat10 Con9f0id8ence limits cases ALL NHL Odds controls ratio Sheep cows Cattle Poultry Pigs 36 0.9 0.6-1.4 37 1.3 0.9-2.0 73 38 0.9 0.6-1.3 33 0.9 0.6-1.4 71 27 0.8 0.5-1.3 30 1.3 0.8-2.0 57 41 29 11 ..o5 1.0-2.2 0.6-1.5 28 27 1.o 0.7-1.6 1.o 0.7-1.6 69 56 122 1 . 1 134 96 0.9 1 .o 106 97 11 ..o2 `Stratified for decade of birth and whether the subject or a relative was interviewed Con9f0id%ence limits 0.8-1.5 0.7-1.2 0.7-1.4 0.9-1.7 0.7-1.4 TABLE X - ODDS RATIO ESTIMATES FOR NON-HODGKIN'S LYMPHOMA FOR VARIOUS MEDICAL CONDITIONS AND ALLERGIES` Exposure cases ICD 200 Oradtdios 90% Confidence limits ICD 202 Oradtdios 90% Confidence limits ALL NHL F:r2 z: 90% Confidence limits Rheumatoid arthritis Eczema Asthma Asthma medication Hay fever Food allergies Drug allergies 3 1.5 0.4-4.7 4 2.2 0.7-6.3 7 14 2.5 1.3-4.6 9 1.6 0.8-3.2 23 6 0.6 0.3-1.3 9 1.2 0.6-2.3 15 5 0.8 0.3-1.9 5 1.0 0.4-2.3 10 14 0.8 0.4-1.3 1 1 0.7 0.4-1.2 25 8 4.8 1.9-11.9 5 3.3 1.2-9.2 13 2 0.4 0.1-1.3 10 2.1 1 . 1 4 . 3 12 7 1.7 0.7-4.2 21 2.0 1.2-3.4 29 0.9 0.5-1.5 19 0.9 0.4-1.7 52 0.8 0.5-1.2 7 3.8 1.7-8.5 20 1.2 0.6-2.2 `Stratified for decade of birth and whether the subject or a relative was interviewed. zation of information bias since cases and controls are drawn from the same registry. This study design also minimizes selection bias due to geographical differences in registration, while any bias due to other cancers being associated with farming is likely to be small, since the overall cancer mortality in New Zealand farmers is identical to national mortality rates (Pearce and Howard, 1985). Furthermore, the previously published analysis (Pearce et al., 1986~)also included a second control group from the general population as an additional check, and this gave very similar findings to those obtained with the main control group of other cancer patients. Lack of validation of the exposure data based on personal recall is also a potential problem in the present study. This might be exW t d to aPPlY equally to cases and Controls (Since both groups involve Cancer Patients), and hence to reduce the obServed relative risk (Copeland et al., 1977). However, in Hoar's SmdY (1986) there was little change in the findings when exposure was based on the records of pesticide Suppliers, rather than on personal recall. Hence, the phenoxy herbicide findings in the expanded study appear unlikely to be due to bias, although the possibility of unidentified confounding factors always exists. A further possibility (Pearce et al., 1986u), is that phenoxy herbicide and chlorophenol exposure could be carcinogenic only when occurring jointly with other exposures present in Sweden and Kansas, but not in New Zealand. However, this study provides no evidence that can be used to assess this speculation. The major positive findings from the expanded study relate to fencing work and meat works employment. We have previously suggested (Pearce et al., 1986a) that the excess risk associated with fencing work could be due to exposure to pentachlorophenol (PCp) which is occasionally used for timber preservation, or to sodium pentachlorophenate (NaPCP) which is widely used for anti-sapstain treatment. On the other hand, it has been estimated that 99% of fencing timber produced in New Zealand since 1955 has been treated by vacuum-pressure impregnation with a copper chrome arsenate (CCA) solution (Department of Statistics, 1983), of which arsenic comprises 3650% of the active ingredients (Timber Preservation Authority, 1980). Hence, another risk factor of interest is arsenic which has been associated with an increased risk of lymphatic and haemopoietic malignancies in several studies (Axelson et al., 1978; Baetjer et al., 1975; Fergusson, 1976; Ott et al., 1974). It has been suggested that arsenic may not generally act as a direct carcinogen, but may primarily act as a co-carcinogen (Pershagen, 1983), and the finding of a strong excess risk associated with joint exposure to fencing work and meat works employment is of interest in this regard. The findings for orchard workers may also be relevant due to the past use of lead arsenate spray in orchards (Nelson et al., 1973), although many other chemicals may also have been used. The previous finding of an association of non-Hodgkin's lymphoma with meat works employment has been supported by the expanded study. We have previously suggested (Smith et al., 1984) that one potential risk factor was exposure to 2,4,6-trichlorophenol (2,4,6-TCP) or other chemicals used in the treatment of pelts. 2,4,6-TCP is highly skin-permeating (Roberts et al., 1977) and the treatment of pelts has involved the workers' manually removing pelts from the treatment SOlutions and handling wet skins without protection. Some com- panies now use other systems involving less direct contact. Studies in rats and mice have shown that 2,4,6-TCP can cause lymphomas, leukaemias and hepatocellular carcinomas (Na- tional Cancer Institute, 1979). However, we also noted that there was an overall excess of soft-tissue sarcomas among meat workers, and that focusing on the pelt departments could be misleading. This observation is supported by the findings presented here, since *e risks for pelt-department workers are not markedly different from those for other meat workers. Furthermore, the propor- tion of meat workers F l t - d e P m e n t work was small, and there is no association with other types of chloroPhenol exposure in the expanded study- In our previous report (Pearce et al., 1986a) we noted these problems with the 2,4,6-TCP hypothesis, and suggested an alternative hypothesis that meat workers may be exposed to zoonotic oncogenic viruses. In a recent report, Johnson et d. 160 PEARCE ET AL (1986) also proposed this hypothesis and presented a comprehensive review of relevant evidence. In particular, they observed that the greatest risk for Hodgkin's disease in their cohort study of meat workers was associated with slaughtering, particularly of cattle, pigs and sheep, supporting the hypothesis of an infective origin for these cases. The increased interest in recent years in a potential role of zoonotic oncogenic viruses stems from the generally increased interest in RNA tumor viruses, or retroviruses, since the discovery that sequences closely related to viral oncogenes were present in human cells (Hall, 1984). These cellular oncogenes, or proto-oncogenes, have been discovered in the genomes of all vertebrate cells examined so far and for each of the viral oncogenes discovered to date. The major interest in the agricultural field centres on the possible role of the bovine leukaemia virus which causes lymphosarcoma in cattle (Heath et al., 1975), and can induce antibody production in other species (Olson et al., 1972). This is a C-type RNA virus, one of a group of viruses which represent the aetiologic agents of tumours of the lymphatic and haemopoietic system in all the higher non-human mammalian species studied to date (Armenian and Hamaden, 1983). Bovine leukaemia virus has been found in New Zealand herds, although the prevalence of infection was low (Parrish et al., 1981). The virus has been found in meat and unpasteurized milk in other countries (Ferrer et al., 1981; Miller and Van der Maaten, 1979). Preliminary human studies have shown no evidence of infection (Donham et al., 1977), but this may be because current laboratory techniques or knowledge of the mechanism of infection are not sufficiently developed (Johnson et al., 1986). Epidemiologic evidence for the role of zoonotic oncogenic viruses is currently weak (Blair et al., 1985), although some studies have found associations between various haematologic malignancies and dairy (Bartsch et al., 1975; Heath, 1970; Henricson et al., 1968), cattle (Blair and Thomas, 1979) or poultry (Priester and Mason, 1974) farming. Most studies have been ecologic in nature and have not included serological tests for evidence of viral infection. However, it is not clear whether such tests would be informative without more adequate knowledge of the possible mechanism of infection. Furthermore, it appears that viruses can cause tumours in animals without &serological evidence of viral replication (Francis et al., 1981. Witter et al., 1970). The picture is further complicated by evidence suggesting that chemical leukaemogens activate leukaemogenic viruses, rather than inducing a somatic mutation (Armenian and Hamaden, 1983), and that some viruses may be oncogenic only when combined with other factors (Fraumeni, 1979). Hence, although the evidence is currently weak, the zoonotic oncogenic virus hypothesis clearly warrants further investigation. Finally, although the main findings of this expanded study relate to agricultural risk factors, the finding for rheumatoid arthritis is of interest, since several other clinical reports and epidemiological studies have suggested that patients with rheumatoid arthritis are at increased risk of non-Hodgkin's lymphoma (Pearce and Porta, 1986). In summary, this expanded study has supported the previous finding of an absence of association between non-Hodgkin's lymphoma and exposure to phenoxy herbicides in New Zealand. It also supports a similar absence of association with exposure to chlorophenols, although this finding is more open to question, due to the marginally elevated risks for overall chlorophenol exposure, and the major positive findings relating to fencing work and meat works employment. One relevant risk factor in meat works is 2,4,6-TCP which is used in the treatment of pelts, but the excess risks do not appear to be confined to pelt-department workers. An alternative hypothesis is that meat workers may be exposed to oncogenic viruses. ACKNOWLEDGEMENTS This work was completed while the first author was funded by an Overseas Research Fellowship of the Medical Research Council.of New Zealand, and was supported by grants from the War Pensions Medical Research Trust Board, the Medical Research Council, and the Northern California Occupational Health Centre. The authors thank Dr. K. Howard and Mrs. H. Giles for their help with this study. They also thank Mr. J. Findlay and the staff of the New Zealand Cancer Registry, the collaborating hospitals throughout the country, the consultants and general practitioners involved, and in particular the patients and families who participated. REFERENCES ARMENIAN, H.K., and HAMADENR,.R., Epidemiology of non-Hodgkin's lymphoma. In: A.M. Lilienfeld (ed.), Reviews in cancer epidemiology, Vol. 2, Elsevier, New York (1983). AXELSON,O., DAHLGREN, E., JANSSONC, .D., and REHNLUND, S.O., Arsenic exposure and mortality: a case-referent study from a Swedish copper smelter. Brit. J. industr. Med., 35, 8-15 (1978). BAETJER, A.M., LILIENFELD, A.M., and LEVINM, .L., Cancer and occupational exposure to inorganic arsenic. Abstracrs ofthe XVIII Internarional Congress on Occuparional Health, Permanent Commission and International Association on Occupational Health, London (1975). BARTSCH, D.C., SPRINGHER, F., and FALK, H., Acute nonlymphocytic leukemia. An adult cluster. J. Amer. med. Ass., 232, 1333-1336 (1975). BLAIR, A., MALKER, H., CANTOR, K.P., BURMEISTER, L., and WIKLUND, K., Cancer mortality among farmers: a review. Scand. J. Work environ. Hlth, 11, 397-407 (1985). BLAIR, A., and THOMAS, T.L., Leukemia among Nebraska farmers: a death certificate study. Amer. J. Epidemiol., 110, 264-273 (1979). BUESCHING, D.P., and WOLLSTADT, L., Cancer mortality among farmers. J. nat. Cancer Inst., 72, 503 (1984). BURMEISTER, L.F., Cancer mortality in Iowa farmers, 1971-8. J. nar. Cancer Inst.. 66,461-4&1(1981). BURMEISTER, L.F., EVERETT, G.D., VANLIERS, .F., and ISACSON, P., Selected cancer mortality and farm practices in Iowa. Amer. J. Epidemiol., 118, 72-77 (1983). CANTOR, K.P., Farming and mortality from non-Hodgkin's lymphoma: a case-control study. Inr. J. Cancer, 29, 239-247 (1982). COPELAND, K.T., CHECKOWAY, H., MCMICHAELA, .J., and HOLBROOK, R.H., Bias due to misclassification in the estimation of relative risk. h e r . J. Epidemiol., 105,488-495 (1977). DEPARTMENT OF STATISTICS. Monthly Abstract of Statistics, October, 1983, Department of Statistics, Wellington (1983). DONHAMK,.J., VANDERMAATEMN.,J., MILLER, J.M., KRUSEB, .C., and RUBINOM, .J., Seroepidemiologic studies on the possible relationships of human and bovine leukemia: brief communication. J. nat. Cancer Inst., 59, 851-853 (1977). FERGUSSON, W., Epidemiology of arsenic. In: B.W. Carnow (ed.), Health effects of occupational lead and arsenic exposure: a symposium, Department of Health, Education and Welfare, Washington, DC (1976). FERRER, J.F., KENYONS,.J., and GUPTAP, ., Milk of dairy cows frequently contains a leukemogenic virus. Science, 213, 1014-1016 (1981). FRANCIS, D.P., ESSEX,M., COTTER, S.M., GUTENSOHN,., JOKOWSKL R.. and HARDY, W.D., Epidemiological association between virus-ne@tive feline leukaemia and the horizontally transmitted feline leukaemia Virus. Cancer Lett., U ,37-42 (1981). FRAUMENJI.F, ., Epidemiological studies of cancer. I n : A.C. Griffin and C.R. Shaw (eds.), Carcinogens: identification and mechanisms of action, Raven Press, New York (1979). GOLDSMITH, J.R., GUIDOTITT, .L., Environmental factors in the epidemiology of lymphosarcoma. Path. Ann., 12,411-425 (1977). HALL, A., Oncogenes-implications for human cancer: a review. J. roY. Soc. Med., 77,410-416 (1984). HARDELLL,.. ERIKSSON, M., LENNER, P., and LUNDGREN E., Malignant I i i I ,I 4 -/ I. ~ NON-HODGKIN'S LYMPHOMA AND FARMING 161 lymphoma and exposure to chemicals, especially organic solvents, chlorophenols and phenoxy acids: a case-control study. Brit. J. Cancer, 43, 169-176 (1981). HARRELL, F., The logist procedure. In: SAS supplemental library user's guide, SAS Institute, Cary, NC (1983). HEATH, C.W., Human leukemia: genetic and environmental clusters. Bibl. Haemat., 36, 649-653 (1970). HEATH, C.W., CLADWELL, G.G., and FEORINO, P.C., Viruses and other microbes. In: J.F. Fraumeni (ed.), Persons at high risk of cancer, Academic Press, New York (1975). HENRICSONB, ., and RINGERTZN, ., Comparative statistical studies on the geographical distribution of human and bovine leukemia in Sweden. Bibl. Haemat., 31, 330-332 (1968). HOAR, S.K., BLAIR, A., HOLMES, F.F., BOYSENC, ., ROBEL,R.J., HooVER,R., and FRAUMENI, J.F., Agricultural herbicide use and risk of lymphoma and soft tissue sarcoma. J. Amer. med. Ass., 256, 1141-1147 (1986). Zealand Cancer Registry based study. Amer. J. Epidemiol, U1,225-237 (1985). PEARCE, N.E., and PORTA, M., Association of non-Hodgkin's lymphoma with rheumatoid arthritis (letter). Amer. J. Med., in press (1986). PEARCE, N.E., SMITH, A.H., HOWARD, J.K., SHEPPARD, R.A., GZLES, H.J., and TEAGUE,C.A., Non-Hodgkin's lymphoma and exposure to phenoxy herbicides, chlorophenols, fencing work and meat works employment: a case-control study. Brit. J. industr. Med., 43,75-83 (1986~). PEARCE, N.E., SMITH, A.H., HOWARD, J.K., SHEPPARD, R.A., GZLES, H.J., and TEAGUE, C.A., Case-control study of multiple myeloma and fanning. Brit J. Cancer, in press (1986b). PERSHAGENG,., The epidemiology of human arsenic exposure. In: B.A. Fowler (ed.), Biological and environmental effects of arsenic, Elsevier, Amsterdam (1983). PRIESTERW, .A., and MASON, T.J., Human cancer mortality in relation to poultry population, by county, in 10 southeastern states. J. nut. Cancer Inst., 5 3 , 4 5 4 9 (1974). JOHNSON, E.S., FISCHMAN, H.R., MATANOWSKI, G.M., and DIAMOND, RAPPAPORT, H., Tumors of the haemopoietic system. In: Atlas of Tumor E,., Cancer mortality among white males in the meat industry. J. occup. Pathology, Fascicle 8, Armed Forces Institute of Pathology, Washington, Med., 28, 23-32 (1986). DC (1966). MILLER, J.M., and VANDERMAATEMN.,J., Infectivity tests of secretions ROBERTS, M.S., ANDERSON, R.A., and SWARBRICK, J., Permeability of and excretions from cattle infected with bovine leukemia virus. J. nut. human epidermis to phenolic compounds. J. Pharm. Pharmacal., 29, Cancer Insr., 62,425-428 (1979). 677-683 (1977). I NATIONAL CANCER INSTITUTE. Bioassay of 2,4,c%trichlorophenolfor pos- SMITH, A.H., FISHER, D.O., GILES, H.J., and PEARCEN, .E., The New . sible carcinogenicity, Technical Report Series 155, Department of Health, Zealand soft tissue sarcoma case-control study: interview findings con- Education and Welfare, Washington, DC (1979). cerning phenoxylacetic acid exposure. Chemosphere,U ,565-571 (1983). NELSON, W.C., LYKINS,M.H., MACKEYJ, ., NEWILL,V.A., FINKLEA, SMITH, A.H., PEARcE, N.E., FISHER, D.O., GILES, H.J., TEAGUE, C.A., J.F., and HAMMER, D.I.,Mortality among orchard workers exposed to and HOWARD, J.K., Soft tissue sarcoma and exposure to phenoxyherbi- lead arsenate sprays: a cohort study. J. chon. Dis., 26, 105-118 (1973). cides and chlorophenols in New Zealand. J. nut. Cancer Inst., 73, 1111- OLSON, C., MILLER, L.D., MILLER, J.M., and Hoss, H.E., Transmission 1117 (1984). of lymphosarcoma from cattle to sheep. J. nut. Cancer Inst., 49, 1463- TIMBER PRESERVATION AUTHORITY. Timber preservation in New Zealand: 1467 (1972). specifications, Amendment No. 4, New Zealand Forest Service, Rotorua Om, M.G., HOLDER, B.B., and GORDON, H.L., Respiratory cancer and (1980). occupational exposure to arsenicals. Arch. environ. Hlrh, 29, 250-255 WITTER, R.L., NAZWAN,K., PURCHASE, H.G., and BUROOYNEG,.H., (1974). Isolation from turkeys of a cell-associated herpesvirus antigenically re- PARRISH, C.R., OLIVER, R.E., WEDDELLW, ., and Ross, A.D., Bovine lated to Marek's disease virus. Amer. J. vet. Res., 31, 525-538 (1970). leukemia virus infection in New Zealand cattle. N. 2. vet. J . , 30, 56-58 WORLD HEALTH ORGANISATION. Manual of the international statistical (1981). classification of diseases, injuries and causes of death, 8th revision, PEARCEN,.E., and HOWARD, J.K.,Occupational mortality in New Zea- WHO, Geneva (1967). land males 1974-78. Commun. Hlth Srud., 9,212-219 (1985). WORLDHEALTH ORGANISATION. Manual of the international statistical clas- PEARCE,N.E., SMITH, A.H., and FISHER, D.O., Malignant lymphoma and multiple myeloma linked with agricultural occupations in a New sification of diseases, injuries and causes of death, 9th revision, WHO, Geneva (1977).