Document dQbEQzJ72DR42nkVK2zxx3m2R

. Int. J,cancer: 65. YJ--S(()1996) 1996 Wiley-Lis. lnc. Publicationof the InternationalUnionAgainst Cancer Publicationde I'Union InternationaleContre le Cancer CANCER I N OFFSPRING OF PARENTS ENGAGED IN AGRICULTURAL ACTIVITIES IN NORWAY: INCIDENCE AND RISK FACTORS IN THE FARM ENVIRONMENT petter KRISTENSLN'.A"d, gc ANDERSENLo?r,entz M. IRGENS?,Annc S. BYE^ and Leif S U N D H E I M ~ lNationa1 Institlrli,c!f'~cclcpationHu~ealth. Oslo;'Cancer Registry of N o w a y , Oslo;"edical Birth Registry sf Noway, universiryof Betgiw, R ~ > I X?~YItIa;tisticsNoway, Kongsvinger;.und 5Nowc-panCropResearch Institute, Plunt h t e c t i o n CenM,As, Noway. In this study of cancer in offspring we demonstrate that parental occupation as a farmer poses a risk for leukaemias factors linked to horticulture and use of pesticides are associ- (Van Steensel-Moll et ai., 1985; Lowengart et al., 1987). ated with cancer at an eady age, whereas factors in animal husbandry. in particular poultry farming, are associated with inlater childhoodandyoungadulthood. Incident cancer was investigated in offspring born in 1952-1991 t o parents identified as farm holders in agricultural censuses in Norway in 1969-1989. In the follow-up of 323,292 offspringfor 5.7 million pemn-years, 1,275 incident cancers were identified in the Cancer Registry for 1965- I99I.The standardized incidence for all cancers was equal to the total rural population of Noway, but cohort subjects had an excess incidence of nervous-system Parental exposure to pesticides has also been associated with incrcased occurrence of leukaemia, mainly acute (Lowengart et al., 1987; Laval and Tuyns, 1988; Shu et al., 1988; Buckley et ul., 1989; Buckley, 1992; Leiss and Savitz, 1995). Immunologic and infectious mechanisms have been hypothesized for child- hood leukaemias, and the analogy to malignancies seen in adult farmers is of interest: immunologic mechanisms triggered by pesticides or animal antigens are suspected to be important for the development of leukaemias among adult tumours and testicular cancers in certain regions and strata of farmers, oncogenic zoonotic viruses being among the sus- time that could imply that specific risk factors were of impor- pected agents (Blair et al., 1992). tance. Classification of exposure indicators was based on infor- mation given at the agricultural censuses. Risk factors were found for brain turnours, in particular non-astrocytic neuroepithelial tumours: for all ages, pig farming tripled the risk [rate ratio (RR), 3.1 I; 95% confidence interval (Cl), 1.69-5.131; indicators of pesticide use had an independent effect of the Reports concerning an association between testicular cancer and agricultural exposures are conflicting (Blair et al., 1992). Testicular cancer is suspected to be related to high foetal levels of free oestrogen, and organochlorine pesticides and other oestrogenic compounds have been suggested as causative same magnitude in a dose-response fashion, strongest in chil- agents (Sharpe and Skakkebrek, 1993). dren aged 0to 14 years (RR, 3.37; 95% CI, l.63-6.94). Horticulture and pesticide indicatorswere associatedwith all cancersat ages0 to 4 years, Wilms' tumour, non-Hodgkin's lymphoma, eye cancer and neuroblastoma. Chicken farming was associated with some common cancers of adolescence, and was strongest for osteosarcoma and mixed cellular type of Hodgkin's disease. The main problem in t h i s large cohort study is the crude Pesticide use in the home has also been associated with childhood lymphoma and soft-tissue sarcoma (Leiss and Savitz, 1995). The implication of phenoxy-acid herbicides as a cause of soft-tissue sarcomas and lymphomas in adults (Blair et al., 1992) is also of interest, since these malignancies are relatively common in childhood and adolescence. W u r e indicators available; the resulting misclassificationis To investigate cancer incidence and reproductive outcomes likelyto bias any true associationtowards unity. 8 1996 wiley-Liss, Inc. in the offspring, we established through linkage between several registers a national cohort of families in Norwegian agriculture. Our hypotheses comprised an excess of sarcoma, Reviews on childhood cancer (e.g., Savitz and Chen, 1990) leukaemia, lymphoma, testicular cancer, and particularly ner- have not firmly established parental occupational exposures as vous-system tumours. causal factors. In studies of parental occupation in which all sites in childhood have been the end-points (Zack et a/., 1980; Sanders et al., 1981;Hemminki et al., 1985; Olsen et a/., 1991), the risk estimates for children of farmers or agricultural workers have not deviated much from the null. On the basis of earlier reports, a number of hypotheses were formulated about indicators of exposure and specific groups of malignancies: exposure to animals or pesticides and brain tumours (Gold et al., 1979; West, 1988; Wilkins and Koutras, 1988; Wilkins and Sinks, 1990; Kuijten et al., 1992; Bunin et al., Nevertheless, in several case-control studies brain turnours 1994); exposure to animals (in particular dairy cows and in childhood have been associated with parental farming chickens) or pesticides and acute leukaemia (Lowengart et a[., (West, 1988; Wilkins and Koutras, 1988; Wilkins and Sinks, 1987; Laval and Tuyns, 1988; Shu et al., 1988; Buckley et ai., 1990;Kuijten et al., 1992; Bunin et al., 1994), and environmen- 1989; Buckley, 1992; h i s s and Savitz, 1995); exposure to tal risk factors in farming are considered to be of particular pesticides (in particular organochlorine pesticides commonly Interest (Kuijten and Bunin, 1993). Both contact with animals used in horticulture) and testicular cancer (Sharpe and Skak- and exposure to pesticides have been suspected as risk factors. kebrek, 1993); exposure to pesticides (in particular herbicides Parental exposure to pesticides has been associated with brain in forestry and grain farming) and soft-tissue sarcoma and tumours in several studies (West, 1988; Sinks, 1985; Davis et lymphomas (Blair et af., 1992; h i s s and Savitz, 1995). a[., 1993;Bunin et al., 1994),and Leiss and Savitz (1995) found In general, the potential causative agents were considered to an association with the use of pest strips in the home. Gold et act either during the pre-conceptual period through the a[. (1979) found increased odds ratios for contact with sick pets father's maturing germ cells, during pregnancy via the mother, and use of insecticides in the home, but the results were not or directly in early infancy. If pesticides applied seasonally confirmed in a later study (Howe et al., 1989). In another study, contact of either the mother (McCredie et al., 1 9 9 4 ~ o) r the child (McCredie et al., 19946) with animals was not associated hTowhom correspondence and reprint requests should be sent, at with brain turnours. National Institute of Occupational Health, P.O.B. 8149 Dep, N-0033 , Apart from an increased odds ratio for acute leukaemia seen Oslo, Norway. Fax: +47 22 603276. among the children of mothers in agricultural work in Shang- hai (Shu et al., 1988), there have been few indications that Received: June 14,1995 and in revised form August 29,1995. 40 KRISTENSEN ETAL. cause cancer by acting on the father's maturing germ cells, an uneven distribution by month of birth would be anticipated for the malignancy in question. In particular, we suspected that a TABLE 1- DISTRIBUTION OF CHARACTERISTICS AMONG OFFSPRING (n = 323 292 5 696 606 PERSON-YEARS) BORN IN 1952-1991TO PARENTS IN AGR'ICULkJRAL WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES IN 1969-1989 I 1 i paternally mediated effect of pesticides applied during spring and early summer (as in grain farming and in orchards in Norway) would be strongest for children born during spring. MATERIAL AND METHODS Cohort Statistics Norway have held national agricultural censuses resulting in computerized information files in 1969, 1979, and 1989, and horticultural censuses in 1974and 1985.The criteria for participation have changed slightly over time. The 1969 census included all farms comprising an agricultural area of at least 0.5 hectare (ha); the 1979 and 1989 censuses combined agriculture and forestry and included both farm holdings with a productive forest area of 2.5 ha or more and farms smaller than 0.5 ha which met certain criteria with regard to livestock size, area of greenhouses, fields for vegetables or berries, and number of fruit trees. The inclusion criteria for the horticultural censuses covered area of greenhouses, field-grown vegetables or berries, or number of fruit trees. Participation in the agricultural and horticultural censuses was mandatory, and is considered to be complete, since governmental subsidies to farm holdings were conditional on this information. All personal farm owners at any of the censuses who were born after 1924 were identified by the personal identity number assigned to all residents of Norway. Subtraction of duplicates (5.1 percent; owners of more than one farm at a single census) yielded 149,254 farm holders (125,873 males, 23,381 females). Individuals owning more than one farm were assigned to the place of residence. After linkage of the file of farm owners with the Central Population Register 334,135 offspring born in 1940-1992were identified. The Central Population Register was established in October 1964 and is based on information from all residents of Norway at the time of the 1960 population census and on continuous updating since 1964 (Brunborg and Kravdal, 1986). Linkage of parents and children is possible because each record in this register includes both personal identification numbers. Parental identification was included prospectively at birth for children born after October 1964; for children born earlier, parental identification was included in retrospect at the population census in November 1970. This was complete only for children born after 1951. Records of children born before October 1964 and who died before the 1970 census do not include the parental identification number (Brunborg and Kravdal. 1986). Descriptive category Gender Male Female Year of birth 1952-1959 1960-1 964 1965-1969 1970-1979 1980-1991 Attained age (years) 0-4 5-9 10-14 15-19 20-29 30-39 Farm location East South West Mid (Tr~ndelag) North Census participation 1969 1974 1979 1985 1989 Census closest to birth date 1969 1974 1979 1985 1989 Productive farming area (hectares) < 0.5 0.5-4.9 5-19.9 2 20 Agricultural activity' Animal husbandry Cattle Dairy cows Pigs Chicken/poultry Sheeplgoats Horticulture . Field vegetables Fruit trees Total person- -time (%) 51.4 48.6 17.6 19.5 26.6 28.9 7.3 18.4 17.8 19.1 18.1 22.0 4.7 30.6 15.2 24.6 12.5 17.2 75.1 3.8 92.9 4.9 93.0 59.7 0.9 33.6 0.4 5.5 19.9 37.3 37.0 5.8 66.7 46.7 39.6 25.1 19.6 34 0 20.5 16.0 3.8 b ~ The present study included all 323,359 offspring born between 1952 and 1991 who were followed up for incident cancer in the Cancer Registry of Norway. Children born between 1952 and 1964were followed up from 1971, as earlier start of follow-up would have led to a negative bias due to lack of identification of children who died before the 1970 popula- tion census. Children born after 1964 were followed up from birth. All children were followed up through 1991, or until death or emigration. Linkage with the Cancer Registry was complete, and only 67 children who either died or emigrated before the start of follow-up in 1971 did not contributc person-time. In total, 102,055childrcn born between 1952 and Berries Green houses Grain farming Potatoes (20.1 hectare) Forestry (2 2.5 hectares)' No farming activity3 Pesticide purchase4 Pesticide spraying equipment' Main income from farm holding Either parent working 2500 hr annual on farm Father working outside farm holding Mother working outside farm holding Either oarent with aericulturaleducation 2.3 1.6 27.2 44.9 63.0 23 0 27 0 26 8 38.6 58.5 5s 1 I8 6 1964 and 221,237 children born between 1965 and 1991 were 'On the farm at the census closest to the year of birth. Children followed up for 5,696,606 person-years (Table I). from farms with several activities at the census contributed person-time in all those activities.-'Information only in 1969,1979 Exposure Information about the farm holding givcn at one or several agricultural censuses was assigned to the records of all members of the holder's family. Since 83% of the population came and 1989 censuses (0.3% missing).-Thildren from holdings with no animal farming, horticultural activity or grain farming at the census closest to the year of birth.-41nformation only in 1969 census (24.9% missing).-'Information only in 1979 census (~7.1% missing). CANCER IN OFFSPRING OF FARMERS 41 from family ho1,iings. information from the Same farm holding was included even in censuses held before the child's parent took Over. This resulted in 1969 CcnSuS information for morc than 75% of the children. Information from the 1979 and 1989 was available for more than 90%, while information from the horticultural censuses i n 1974 and 1985 was available only for a few percent (Tablc I ) . The information included: location (municipality) of the farm holding and whether it was the residence of thc holder; farming arca and area used for cultivation of potatoes, diffcrcnt types of grain, field veg- etables, greenhouses, and forestry; number and species of livestock; number and species of fruit trees and bcrries grown; horses, tractors, sourcc of water (1979), use Of silagc, and use of fertilizers (1979, 1989); amount of money spent on pesticides in the preceding ycar (1969), and pesticide spraying equipment on thc farm (1979): timc spent on agricultural work on the farm and work outside the farm for both parents; fraction of total family income from the farm; agricultural education of holder or spouse: year of conveyance of the farm holding (1979, 1989): and whether a family farm (1979). The exposure indicators and their distributions are given in Table I. The main types of farming in Norway are animal husbandry, horticulture, and grain farming. More than one type of farming was commoner on the holding early in the study period, and specialized production and monoculture became more prevalent in the later*periods. Norwegian farms were mostly small and operated by the family members. Usually, 3 generations lived on the holding, and the children (mostly sons) took over when the parents were too old to operate the farm. The distribution of other explanatory variables in Table I indicates that a large proportion of the holdings were part-time enterprises, and part-time employment outside the holding was more commonly reported in the later censuses. The criteria for participation in the agricultural censuses are loose, and only 58% of the offspring came from holdings where either parent worked for at least 500 hr annually, which is the standard criterion for a farmer in population censuses given by Statistics Norway (Central Bureau of Statistics, 1975). This sub-set of the cohort is termed farmers'offspring. For the majority of holdings, information was available from all 3 agricultural censuses. We considered census information for the pre-conception period, during pregnancy and in infancy most relevant in this study of early cancers and, therefore, defined the census closest in time to the child's birth as the index census, but indicators of pesticide use, water source, and fertilizers were included when available, even for non-index censuses. The index census of the oldest children took place during their adolescence or even in adulthood. For those born between 1952and 1964, the age at the time of the index census was over 10 for 50% and over 20 for 10%. For 81% of children born between 1965 and 1991, the gap between birth and the index census was 5 years or less. Most of the exposure indicators were quantified in categories, but they were analysed as dichotomics (present/absent), 0" the assumption that contact and exposures are as likely to be high for small-scale activities where work practices are more manual. Money spent on pesticides in 1968 was categorized in 4 groups: none; NOK 1-99; NOK 100-499; and NOK 500 or more. Sub-types of horticulture (except field vegetables) were uncommon, so orchards and greenhouse cultivation were combined (7% of offspring exposed). Brain neoplasms and acute leukaemias were each grouped in order to ensure robust categories: neuroepithelial brain tumours werc categorized according to Kleihues et al. (1993) as astrocytomas and non-astrocytic ncuroepithclial tumours; acute leukaemias were grouped as lymphocytic, myelocytic and other types. Hodgkin's disease was divided into nodular sclerosis and the mixed cellular type. Analysis The analyses were performed with the "Epicure" statistical software program (Preston et nl., 1993). Two analytical methods were applied: indirect standardization (observed over expected cases) with an external reference population; and Poisson regression modelling with reference groups within the cohort. Cancer morbidity was investigated by comparing the observed and expected numbers of cases. The expected numbers of cases were estimated from the person-year experience in different strata, and the 5-year age- and gender-specific rural rates in Norway for each year from 1965 through 1991. The rural population comprises residents outside towns, which was 57% of the total population in 1980. Gender-specific standardized incidence ratios (SIR) were calculated for all cancers and for cancers at specific sites for the study population; sub-set analyses were performed for offspring with a parent who fulfilled the occupational criterion for a farmer. Strata of age intervals, period of follow-up, year of birth, and geographical regions were also investigated. Ninety-five percent confidence intervals (CI) were computed under the assumption of a Poisson distribution of the observed cancers under the null hypothesis. The associations between exposure indicators and specific cancers were-assessed in Poisson regression. The main measure of association was the ratio of incidence rates (rate ratio, RR) between subjects with the exposure indicator in question and subjects without the exposure indicator. RR estimates were adjusted for age and calendar year, or birth year and calendar year. Other exposure indicators and explanatory variables were included in the models if they altered the R R estimates by more than 15%.We calculated 95% CI, and those that did not include unity were considered statistically significant. For the pesticide purchase indicator, we performed a test of trend for the association between expenditure level and cancer whenever the number of cases were sufficient. The relevance of the information obtained from the index census was judged on the basis of a priori criteria in order to identify offspring who were most likely to have spent the first years of their life on a farm. Information was considered "relevant" if either parent was the farm holder at the time of the index census and lived on the farm. Information was also considered "relevant" if the farm holder was closely related to the parents and was living at the farm, and either parent took over the farm at a later census. According to this definition, information was "relevant" for 70% of the study subjects. In the analyses of cancer risk at age 0 to 4 years and in analyses of infant tumours (Wilms' tumour, eye cancer, hepatic cancer, neuroblastoma), information was considered "relevant" only if the gap between birth and index census was 5 years or less, in addition to the other criteria. An increased stronger exposureoutcome association for this sub-set or the sub-set of offspring of parents who fulfilled the occupational criterion for a farmer strengthened the assumption of a true association. Cancer identification The outcome classification was based on the information on cancer site and morphology in the Cancer Registry of Norway. Case reporting to the Registry has been mandatory since 1952. RESULTS A total of 1,275incident cases of cancer was identified in the follow-up. The distribution for selected sites are provided in Table 11. Brain tumours ( E D - 7 193.0) counted 182 cases, of 42 KRISTENSEN E T A L TABLE I1 - STANDARDIZED INCIDENCE RATIOS FOR ALL CANCERS AND CANCERS AT SELECTED SITES AMONG OFFSPRING BORN 1952-1991 TO PARENTS IN AGRICULTURAL WORK, ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES 1%9-1989, AND AMONG OFFSPRING OF A PARENT WHO FULFILLED THE STANDARD OCCUPATIONAL CRITERION FOR A FARMER (WORKING 2500 HR ANNUALLY ON A HOLDING) All sites (140-209) Liver (155) Breast (170) Cervix uteri (171) Ovary (175) Testis (178) Kidney'(180) Melanoma (190) Eve (192) NkGous system (193) Thyroid (194) Endocrine glands (195) ~I Bone (196)Connective tissue (197) Hodgkin's disease (201) Non-Hodgkin'slymphoma (200,202) Leukaemia (204) Unsuecified and other sites 1275 14 35 64 40 158 23 133 16 232 46 22 51 27 77 60 181 96 98 92-103 127 69-213 72 50-100 80 62-103 110 78-149 111 95-130 82 52-123 97 82-116 91 52-147 106 92-120 85 62-114 94 59-142 113 84-149 91 60-133 110 87-138 90 69-115 96 8.2~-1~1~1- 94 76-115 739 8 16 31 26 97 14 65 9 141 20 17 32 16 46 33 113 ~~ 55 101 94-109 103 44-202 65 37-106 76 52-108 131 86-193 124 101-152 83 46-140 88 68-113 84 39-160 111 94-131 69 42-106 122 71-196 122 83-172 94 54-152 117 85-156 87 60-122 101 83-122 98 74-128 In = 323,292; 5,696,606person-years.-'n = 188,680;3,323,876 person-years. J which 76 were classified as astrocytomas and 65 as non- The space-time variations in SIR were investigated for , astrocytic neuroepithelial tumours. Nearly all leukaemias were specific major sites. Some of the deviations in Table 111 were acute: 92 lymphocytic, 36 myelocytic, and 35 of other types. due to clustering of nervous-system tumours and testicular Cases of nodular sclerosis numbered 37 and of the mixed cancer. Thus, during follow-up 1985-1991,115nervous-system cellular type of Hodgkin's disease, 34. There were 29 cases of tumours were diagnosed (SIR, 128; 95% CI, 106-154). In this i osteosarcomas, 27 neuroblastomas, and 17 Wilms'-tumour period, 45 cases were diagnosed in the 0- to 14-year age group cases. (SIR, 149; 95% CI, 108-199), of those, 13 cases came from a I SIR analysis The SIR for cancer at all sites was close to that expected, and no site-specific SIR deviated significantly from unity (Table 11). Few breast cancers and cervical cancers were reported in comparison with the numbers expected. Among the major sites, non-significant elevations of the SIR were found for testicular cancer, bone cancer, and Hodgkin's `farm holding in the southern region (SIR, 256; 95% CI, 136-438). The testicular-cancer-incidence excess was confined to the western region, particularly at age 15 and older and for boys born between 1952 and 1969 (50observed cases; SIR, 167; 95% CI, 125-218). Forty of those cases were observed among farmers' sons (either parent working at least 500 hr annually on the farm); the SIR was 216 (95% CI, 156-290). Similar clusters in leukaemias or lymphomas were not apparent. disease. For nervous-system tumours, the SIR was 106 (95% Poisson regression CI 92-120). The incidence of cancer at all sites during the first years of The 188,680 offspring of farmers (either parent worked at life was associated with horticulture and pesticide indicators, least 500 hr annually on a farm) had 739 incident cancer cases whereas cancer later in life was related to contact with animals, and a SIR close to unity (Table 11). This sub-group had SIR in particular chickens (Table IV). For ages 0 to 4 years (total that deviated more markedly from unity for several sites: SIR 281 cancer cases; 1,045,535 person-years), offspring on farms were elevated for testicular cancer, nervous-system tumours, with orchards or greenhouses had a nearly doubled RR; the Hodgkin's disease, bone cancer and ovarian cancer, but association with pesticide purchase showed a dose-response significantly only for testicular cancer. Table 111provides results for all cancers by gender in strata relating to time and geographical location of the farm. The SIR was decreased for females and slightly increased for relationship for different levels of expenditure, but the test for trend was not significant ( p = 0.12). For ages 5 to 19 years (total 403 cancer cases; 3,133,341 person-years) increased RR were found for exposure to chicken farming. males. This gender difference was due mainly to increased SIR Brain tumours for males in the 5- to 14-year age group; the SIR estimates for females were decreased for all ages except 0 to 4 years, strongest during attained age of 15 to 19years. No specific SIR trends were apparent in different periods of follow-up; a moderate increase of borderline significance was found for males during 1985-1991. Moderate increases in the SIR for both genders were found for offspring born during 1980-1991: the SIR for boys and girls combined was 119 (95% CI, 96-148). Brain tumours were the commonest neoplasm (Table V). A RR of 1.59 for brain tumour was found in association with pig ' farming, while the increases were more moderate and non- significant for chicken farming, grain farming. horticulture and pesticide purchase. The increases were restrictcd mainly to the group of non-astrocytic ncuroepithclial tumours, especially for pig farming, pesticide purchase and chicken farming; for pesticide purchase the RR increased for increasing cxpendi- The SIR differences for all sites in males and females in ture levels (test for trend, p = 0.0002). When all exposure adulthood were due to some sex-specific sites (testicular indicators were included simultaneously in a model, the effects cancer, breast cancer, cervical cancer). The gender difference of grain farming and horticulture on the risk for non-astrocytic of SIR in childhood was due to several sites, mainly nervous- neuroepithelial tumours were dependent on pesticide pur- system tumours, bone cancer and leukaemia (data not shown). chase, whereas pig farming (RR, 2.48; 95% C1, 1.37-4.50) and CANCER IN OFFSPRING OF FARMERS TABLE 111 - GENDER-SPECIFIC STANDARDIZED INCIDENCE RATIOS FOR ALL CANCERS FOR SPECIFIED REC;IONSANDTIME PARAMETERS AMONG OFFSPRING BORN 19Q-IY91 TO PAKENTS IN AGRICULTURAL WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES IYhY-1989 Males(n = lhh.2YI: 2.024.670 perbon-yeara) Number of cases SIR 95% CI Females (n = li7.001; 2.77 I .Y30 person-years) Numher ofcases SIR Y i @ h CI Total 684 104 97-112 591 91 84-99 Attained age (years) 0-4 105 89 73-107 96 109 89-133 5-14 143 116 98-136 93 93 75-113 15-19 104 105 86-126 63 78 60-99 20-29 245 I 0 7 94-121 214 YO 79-103 30-39 87 101 81-123 125 88 74-105 Period of follow-up 1965-1 974 88 103 83-126 67 110 86-138 1975-1979 99 99 81-120 56 71 54-91 1980-1984 140 97 82-114 128 87 73-103 I Y85-1991 357 110 99-122 340 94 85-105 Year of birth 1952-1959 195 104 90-120 213 89 77-101 1960-1969 301 104 92-116 227 86 76-98 1970-1979 145 104 88-122 106 97 80-117 1980-1991 43 110' 80-146 45 129 95-171 Location of farm East 206 103 89-117 194 96 84-111 South 101 104 85-125 75 80 63-99 West 195 122 106-140 163 106 90-123 Mid 76 93 74-116 61 75 58-96 North 106 91 75-109 98 84 68-102 TABLE lV - RATE RATIOS FOR ALL CANCERS, 1965-1991.' BY AGE FOR EXPOSURE INDICATORS LINKED TO ANIMAL HUSBANDRY, HORTICULTURE, AND PESTICIDE USE IN OFFSPRING BORN IN 1952-1991 (n = 323,292). TO PARENTS IN AGRICULTURAL WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES IN 1969-1989 Attained age Exposure indicator Number of Persin-years Crude Adusted 95% cI exposedcases instratum RR RR2 All ages Horticulture Field ve etables Orchardgsor greenhouses Pesticide purchase3 Horticulture and pesticide pur- 293 1,164,897 1.16 1.15 1.02-1.32 224 909,762 1.12 1.12 0.97-1.30 113 403,386 1.28 1.21 1.00-1.47 366 1,540,541 1.10 1.16 1.00-1.34 158 586,896 1.26 1.25 1.05-1.50 Orchards or greenhouses and pes- 78 251,042 1.44 1.38 1.09-1.76 ticide p ~ r c h a s e ~ . ~ Animal husbandry 854 3,802,182 1.01 0.98 0.87-1.10 Chicken farming 284 1,117,300 1.18 1.11 0.96-1.27 Pig farming 342 1,429,804 1.09 1.04 0.91-1.18 0-4years Orchards or greenhouses Pesticide purchase3 21 62,502 1.84 1.86 1.18-2.92 Level 1 17 91,180 1.05 1.04 0.62-1.76 Level 2 32 151,068 1.19 1.19 0.79-1.78 Level 3 15 54,293 1.55 1.55 0.89-2.68 5-19 years Animal husbandry 230 1,771,374 1.04 1.08 0.86-1.35 Chicken farming 77 500,584 1.26 1.33 1.03-1.71 Pig farming 89 630,193 1.14 1.21 0.95-1.55 'Sub'ects born 1952-1964 followed up from 1971; subjects born in 1965-1991followed up from birth.-IAdjusted for year of birth and calendar year.JRestricted to subjects covered in the 1969 census (total n = 243,659; 4,280,456 per~on-years).-~References:ubjects from holdings with no horticulture or pesticide purchase.-SReference: subjects from holdings with no orchards, greenhouses, or pesticide purchase. 43 pesticide purchase (RR, 2.11; 95% CI, 1.19-3.75) were indepen- 1.11-2.63) and 22 cases of non-astrocytic neuroepithelial dent risk factors. The independent effect of chicken farming tumours (adjusted RR, 3.37; 95% CI, 1.63-6.94). Pesticide was more moderate (RR, 1.77; 95% CI, 0.99-3.17). purchase was positively associated with non-astrocytic tumours The rates for non-astrocytic neuroepithelial tumours were in all categories of period of follow-up and year of birth. high in all age groups for those exposed to pig farming or However, the association was slightly stronger for children pesticide purchase. The association with pig farming was born between 1980 and 1991, and during follow-up 1985-1991 evenly distributed in all age groups, whereas the association (data not shown). with pesticide purchase was strongest for children under 15 The risk increase for non-astrocytic neuroepithelial tumours years of age. For pesticide purchase, offspring aged 0 to 14 in association with pesticide purchase was not restricted to Years had 41 brain tumour cases (adjusted RR, 1.71; 95% CI, specific types of farming, but particularly high incidence was rq a 44 KRISTENSEN ETAL. TABLEV - R R FOR BRAIN TUMOURS, i 9 h 5 - 1 ~ ~F1O. .R EXPOSURE INDICATORSL I N K E D T A~ F I M A L HUSBANDRY AND PESTICIDE USE IN OFFSPKING BORN 1952-IYYI ( n = 323.292)TO PARENTS IN AGRIC'IILTIIRAI. WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES 1969-1989 Brain Pig farming 61 1,429,804 1.50 1.59 1.16-2.17 50 tumour Chicken 43 1,117,300 1.27 1.32 0.93-1.86 35 (ICD-7 farming 193.0) Grain 59 1351,353 1.28 1.29 0.95-1.77 49 farming Horticulture 44 1,164,897 1.24 1.25 0.89-1.76 39 Pesticide 60 1,540,541 1.39 1.38 0.98-1.94 55 purchase4 Non-astro- Pig farming 30 1,429,805 2.56 3.11 1.89-5.13 25 cytic neu- Chicken 22 1,117,300 2.10 2.42 1.44-4.08 17 roepi- farming thelia1 Grain 25 1,551,353 1.67 1.72 1.05-2.84 19 tumour5 farming Horticulture 18 1,164,897 1.49 1.54 0.89-2.65 17 Pesticide purchase4 Level 1 7 470,523 2.04 2.00 0.854.74 6 Level 2 17 780,646 2.98 2.93 1.54-5.60 16 Level 3 7 289,373 3.31 3.28 1.39-7.76 6 1.42 1.22 1.28 1.39 1.46 3.35 2.32 1.59 2.02 2.11 3.29 3.31 `Subjects born 1952-1964 followed up from 1971; subjects born 1965-1991 followed up from birth.-See "Material and Methods" for definition; adjusted for year of birth and calendar ~ear.-~Adjustedfor year of birth and calendar ~ear.-~Restrictedto subjects covered in the 1969 census (n = 243,659; 4,280,456 erson-years).- Glioma not otherwise specified, choroid plexus papilloma, ependymoma, oligo8endroglioma, medulloblastoma, ganglioglioma, and neuroblastoma. found among offspring aged 0 to 4 years on farms with pesticide purchase and grain farming combined (6 exposed cases; 3.8 per 100,000person-years; RR, 8.01; 95% CI, 1.62-39.7). The associations with pig farming and chicken farming, and pesticide indicators were also investigated for'medulloblastoma which was classified in the non-astrocytic neuroepithelial group. Eight out of a total of 15 medulloblastoma cases were exposed to pigs, and the adjusted R R was even higher than for the whole non-astrocytic neuroepithelial group (RR, 4.92; 95% CI, 1.72-14.1). Chicken farming was also strongly associated with medulloblastoma (7 exposed cases; adjusted RR, 3.91; 95% CI, 1.45-10.5). For pesticide purchase, the RR estimate was doubled,Le., weaker than for the total non-astrocyticgroup. Table V also includes RR estimates for the sub-set of offspring with relevant information from the index census. Compared with the total cohort, the estimates for nonastrocytic neuroepithelial tumours were slightly increased in association with both pig farming and pesticide purchase; the estimate for association with pesticide purchase was increased in this sub-set even for the age group 0 to 14 years (RR, 4.02; 95% CI, 1.78-9.08). Offspring of a parent who worked at least 500 hr annually on the farm also showed stronger associations than the total cohort between non-astrocytic neuroepithelial tumours and exposure to pig farming (RR, 3.99; 95% CI, 2.06-7.73) or pesticide purchase (RR, 4.76; 95% CI, 1.92-11.8). The distribution of month of birth among offspring with brain tumours in association with pesticide purchase was also investigated. Among children aged 0 to 14 years. with the pesticide purchase indicator, 50% of the non-astrocytic neuroepithelial tumours (11 out of 22) and astrocytomas (6 out of 12) were born between April and June. Using children without the pesticide purchase indicator as reference, the RR estimate of non-astrocytic tumours for exposed children born between April and June was 6.15 (95% CI. 2.66-14.2) and only 2.30 (95% CI, 1.00-5.30) for exposed children born between July and March. Acute leukaemias The RR estimates for an association between indicators of exposure to pesticides as well as contact with animals on the one hand and acute lymphocytic or myelocytic leukaemia on the other were close to unity (Table VI). For acute leukaemias there was no trend in the rates for increasing pesticide expenditure (test for trend, p = 0.37). An increased RR of borderline significance was found for other acute leukaemias in association with animal husbandry; the R R was close to unity for pesticide purchase. The R R for other acute leukaemias were increased in asssciation with all animals species, but were unexpectedly strongest for pig farming. Chicken farming and dairy farming were associated only moderately with other acute leukaemias. For most exposures to animals, analyses of the "relevant" information sub-group produced lower RR estimates, except for pig farming. Analyses for the sub-group of offspring to a parent who worked more than 500 hr annually on the holding also produced point estimates closer to unity (for pig farming, RR, 1.78; 95% CI, 0.83-3.81). Other associations in a priori hypotheses There was no association between testicular cancer and exposure to pesticides (Table VII) or the level of pesticide expenditure (test for trend, p = 0.49). An increased risk was found for Hodgkin's disease in association with forcstry, but after adjustment for year of birth and calendar year the estimate was not significant. The associations between grain farming and soft-tissue sarcomas or lymphomas were all close to unity. The RR for non-Hodgkin's lymphoma in association with pesticide purchase was 1.67,with a dose-response relationship for different levels of expenditure (test for trend,!] = 0.04). This association was largely restricted to horticulture, if pesticide purchase was reported. Analysis restricted to the sub-set with "relevant" information yielded an even higher risk estimate for the latter association (10 exposed cases: adjusted RR 2.71). 1 i I I CANCER IN OFFSPRING OF FARMERS TABLEVI - RR FOR ACUTE LEUKAEMIAS, 1Yh5-1YYl.' FOR EXPOSURE INDICATORS LINKED TO ANIMAL HUSBANDRY AND PESTICIDE USE IN OFFSPRING BORN 1Y52-1YYl ( n = 32.3.292)TO PARENTS IN AGRICIJI TURAL WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES IY6Y-198Y Cancer diagnosh Exposure Total population Exposed Person-veiirs Crude Adjusted cases instratum RR RR' y5c,8 cI "Relevant" information' Exposed Adjusted cases RR' Acute leu- Pesticide kaemia purchase4 Animal hus- bandry Acute lym- Pesticide phocytic purchaseJ leukaemia Animal hus- bandry Acute Pesticide myelocytic purchase4 leukaemia Animal hus- handry Other acute Pesticide leukae- purchase4 mias Animal hus- bandry Cattle farming Dairy farming Chicken farming Pig farming 52 1,540,541 1.09 112 3,802,182 1.09 29 1,540,541 1.10 57 3,802,182 0.81 12 1,540,541 1.33 26 3,802,182 1.30 11 1,540,541 0.89 29 3,802,182 2.41 21 2,661,395 1.71 16 2,254,687 1.29 10 1,117,300 1.64 15 1,429,805 2.16 1.06 0.75-1.49 1.19 0.85-1.65 1.03 0.65-1.64 0.97 0.63-1.49 1.35 0.64-2.85 1.28 0.62-2.66 0.90 0.44-1.86 2.28 0.95-5.51 1.64 0.83-3.22 1.22 0.63-2.37 1.54 0.74-3.22 2.10 1.07-4.12 44 92 24 48 11 21 9 23 17 14 7 14 1.13 0.95 0.99 0.92 1.38 0.84 0.86 1.90 1.21 1.04 1.08 2.26 'Sub'ects born 1952-1964 followed up from 1971; subjects born 1965-1991 followed u from birth.-lAdjusted for year of birth and calendar ~ear.-~Se"eMaterial and Methods" for deznition, adjusted for year of birth and calendar ~ear.-~Restrictedto subjects covered in the 1969 census (n = 243,659;4,280,456 person-years). TABLEVII - A PRIORI SUSPECTED ASSOCIATIONS BETWEEN CANCERDIAGNOSED IN 1965-1991.1 (EXCEPT BRAIN TUMOURS AND LEUKAEMIAS AND EXPOSURE INDICATORS IN OFFSPRING BORN IN 1952-1991 (n = 323,292) TO PARENTS IN A G R I C U L hRAL WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES IN 1969-1989 RESULTS OF POISSON REGRESSION ANALYSIS ~ ~~~~~~ ~~ Cancer diagnosis indicator Number of Person-years Crude Adjusted 95% cI exposed cases in stratum RR RR2 Testicular cancer Pesticide purchase3 36 790,254 0.84 0.89 0.60-1.32 Horticulture 10 208,056 0.88 0.79 0.41-1.49 Soft-tissue sarcoma Grain farming 6 1,551,352 0.76 0.76 0.31-1.88 Forestry4 13 3,589,579 0.54 0.53 0.25-1.14 Pesticide spraying 8 1,527,954 1.23 1.25 0.53-2.92 equipment5 Hodgkin's disease Grain farming 17 1,551,352 0.76 0.75 0.44-1.29 Forestry4 56 3,589,579 1.67 1.56 0.94-2.60 Pesticide spraying 22 1,527,954 1.18 1.27 0.76-2.11 equipmentS Non-Hodgkin's Grain farming 14 1,551,352 0.81 0.81 0.45-1.47 lymphoma Forestry4 39 3,589,579 1.09 1.04 0.61-1.77 Pesticide purchase6 Level 1 5 470,523 1.27 1.30 0.49-3.42 Level 2 10 780,646 1.53 1.57 0.75-3.30 Level 3 Horticulture and Res- 6 11 289,373 2.47 2.50 1.02-6.15 586,896 2.10 2.08 1.00-4.32 ticide purchase 'Sub'ects born 1952-1964 followed up from 1971; subjects born 1965-1991 followed up from birth.-IAdjusted for year of birth and calendar ~ear.-~Restrictedto males covered in the 1969 census (n = 125,571;2,202,325 per~on-years).-~Restrictedto subjects covered in the 1969, 1979 or 1989 censuses (n = 322,270; 5,678,035 per~on-years).-~Restrictedto subjects covered in the 1979 census (n = 300,555; 5,292,372 person-years).-6Restricted to subjects covered in the 1969 census (n = 243,659;4,280,456 person-years). 45 Associations not included in the a priori hypotheses The associations between horticulture or exposure to pestitides and cancer at an early age, and between chicken farming and cancer at later ages were also reflected in other specific groups of neoplasms (Table VIII). Some of the associations between indicators of pesticide use in horticulture and the commoner malignancies in infancy were strong. Although there were few exposed cases, significantly increased RR were found between exposure to orchards or greenhouses and Wilms' tumour, especially when use of pesticide spraying equipment was reported. Field vegetable 46 KRISTENSEN E T A L TABLE VI11 - POSITIVE ASSOCIATIONS NOT SUSPECTEDA PRIORI BETWEEN CANCER AND EXPOSURE INDICATORS FOR OFFSPRING BORN 1952-1991 (n = 323,292) TO PARENTS IN AGRICULTURAL WORK ACCORDING TO THE NORWEGIAN AGRICULTURAL CENSUSES 1969-1989 RESULTS OF POISSON REGRESSION ANALYSIS - Cancer diagnosis Exposure indicator Total population Number of Person-years Crude Ad usted exposed cases in stratum RR RR' 95% cI "Relevant" information` R R ~Exposed Ad usted cases Wilms' tumour Orchards or greenhouses Wilms' tumour Pesticide spraying equipment4 Wilms' tumour Orchards or greenhouses and pesticide spraying equipment4.5 Eye cancer Field vegetables and pesticide Neuroblastoma Field vegetables (age 0-4 years) Osteosarcoma Chicken Hodgkin's disease, Chicken mixed cellular type 4 403,386 4.04 4.78 1.56-14.7 2 4.42 9 1,527,954 2.77 2.54 0.98-6.58 6 4.88 4 227,982 7.95 8.87 2.67-29.5 2 12.5 4 442,289 3.12 3.17 0.93-10.9 4 4.91 7 162,642 2.38 2.51 1.03-6.13 6 3.00 12 1,117,300 2.89 2.90 1.38-6.11 8 2.70 14 1,117,300 2.87 2.68 1.35-5.30 13 3.61 `See "Material and Methods" for definition, adjusted for year of birth and calendar year.-?Adjusted for year of birth and calendar year.-3Restricted to subjects covered in the 1969 census (n = 243,659; 4,280,456 per~on-years).-~Restrictedto subjects covered in the 1979 census (n = 300,555; 5,292,372 person-years).-SReference: subjects from holdings with no orchards, greenhouses, or pesticide ' purchase.-hReference: subjects from holdings with no field vegetables or pesticide purchase. farming was a risk factor both for eye cancer and for neuroblastoma at young age. However, only 2 of the 4 exposed eye-cancer cases were retinoblastomas. A positive association was found between chicken farming and osteosarcoma, whereas an increased RR of borderline significance was seen for Hodgkin's disease; the latter association was confined to the mixed cellular type (Table VIII). The associations reported in Table VI11were stronger when the analyses were restricted to the sub-set of children with "relevant" information from the index census. The exception was the association between chicken farming and osteosarcoma, but the slightly decreased RR was still significant. Analysis of the association for the sub-set of the cohort with a parent who worked at least 500 hr annually on the farm yielded a RR of 4.15 (95% CI, 1.61-10.7). DISCUSSION The total cancer experience among the offspring of farm holders was similar to that of the rural population of Norway, in accordance with case-control (Hemminki et al., 1985; OIsen et al., 1991) and proportionate mortality (Sanders et nl., 1981) studies in other countries. The incidence of cancer at specific sites did not deviate significantly from reference values in the whole study population. Offspring of a parent who worked 500 hr or more annually on a farm had moderate increases in the SIR for testicular cancer, nervous-system tumours, Hodgkin's disease and bone cancer. However, the increase was significant only for testicular cancer. For cancer at all sites, male offspring had a SIR slightly above unity, whereas the female SIR was below unity. This difference can be explained only partially by low rates for cancers of the breast and cervix and high rates of testicular canccr. Since the diffcrence encompasses several sites, chancc is a plausible explanation. The gendcr differcnce at the ages of 5 to 19 years could also be due to environmental factors that act after birth. Children of farmers are unique in that they live at their parents' workplace and are potentially exposcd to occupational agents. Most fatal accidents among farmers' children are linked to farm activities, and affect boys far more often than girls, even in the 0 to 4 year age group (Directorate of Labour Inspection, Oslo. personal communication). It is thereforc probable that the risk of accidental childhood exposure is higher for boys than for girls. The Poisson regression results indicate that pesticides poses 2 risk for some of the commonest malignancies of infancy and carly childhood, whereas contact with animals, in particular poultry and pigs, constitutes a risk for some of the malignancies most typical of late childhood and adolescence. Brain tumours The most interesting confirmation of an a priori hypothesis was the clear relationship between brain turnours and indicators of exposure to pesticides and contact with animals. Children aged 0 to 14 years had a nearly doubled risk for brain tumours and a more than tripled risk for neuroepithelial tumours other than astrocytomas in association with pesticide purchase on the farm; RRs also increased by increasing expenditure levels. Even stronger associations were seen for sub-groups of offspring considered to have grown up on the farm or whose parents had a large work input on the farm, I 1 ! , strengthening the evidence for a true relationship. The results are in accordance with those of several case-control studies in which indicators of parental or childhood exposure to pesticides or insecticides were associated with brain tumours (Preston-Martin et al., 1982; Sinks, 1985; West, 1988; Wilkins and Koutras, 1988; Davis et al., 1993; Bunin et al., 1994). In those studies, the odds ratios were 1.5 to 2.0 and above, although several of the results were non-significant because of low exposure prevalence. Contrasting our results, West (1988) found that pesticides were associated more closely with astro- cytoma. Use of insecticides was also reported in all case homes in a-report of a cluster (Wilkinset al., 1991). No increased risks were found, however, in other studies in which contact with pesticides (Howeetal., 1989)or use of insecticides in the home before diagnosis (McCredie et al.. 1994u, b ) were investigated. . : In a study of gliomas in adults, several groups of pesticides were found to be risk factors (Musicco et al., 1988). The credibility of a true relationship between pesticides and specific brain tumours in our study is strengthened by the space-time pattern of nervous-system tumour incidence in the cohort with an increasing SIR during the last years of follow-up and in offspring born 1980 to 199I . The same strata had a stronger association betwecn pcsticidc purchase and nonastrocytic neuroepithelial tumours. Also, the highest SIR was found in the south region. where pesticides are more exten- sivcly used. Clustering of neuroepithelial brain tumours in offspring born April to June to parents reporting pesticide use may indicate that paternal exposure 0 to 3 months before conception is crucial. Wilkins and Sinks (1990) found that the association between glioma and having a father in farming was strongest when he had held this occupation during the pre- CANCER IN OFFSPRING OF FARMERS 47 conception peribd. West (1988) found that indicators of paternal exposclre to pesticides constituted a risk for several histologic sub-types of brain tumours. The association betwccn pig farming and brain tumours was clear. RR was significantly increascd by more than 50 percent for all brain tumours and 5-fold for medulloblastoma. Chicken farming also was associated with medulloblastoma. Our results are in accordance with Bunin er a/. (1994), who reported testicular cancer among tarmers' sons, especially at the ages of 15 to 19 years, is at variance with the only other report on parcntal occupation and testicular cancer (Kardaun ef a/., 1991). Our data suggest that testicular cancer is associated with farm practices that were common in the western part of Norway. Risk factors for testicular cancer other than exposure to pesticides in the study population arc reported elsewhere (Kristensen et a/., 19%). elevated odds ratios for thc tumour category including medulloblastoma in association with pigs and poultry. Toxoplasma gondii infection is a potential cause of brain tumours that is worth considering, sincc it infects domestic animals including chicken and pigs, and is cven a suspccted causc of glioma in animals (Schuman et a/., 1967). Pig farmers in Finland had an increased prevalence of antibodies to this micro-organism (Seuri and Koskela, 1992). In a case-control study of children R i . ~ nkot included in the study hypotheses Several positive associations not included in our study hypotheses were found, although they should be interpreted with caution, since many associations were investigated. The credibility of several of the results is strengthened, however, as they followed a specific pattern: tumours of infancy were associated with indicators of exposure to pesticides and horti- and adults with tumours of the central ncrvous system, a culture, and malignancies that are commoner in later child- significant association was found between antibodies to Toxo- hood were associated with indicators of contact with animals. plasma and gliomas (Schuman et al., 1967). In another study, a Wilms' tumour was related to living on a farm with orchards relationship was found with meningiomas but not with gliomas, or greenhouses and pesticide spraying equipment. A signifi- but the cases included only adults (Ryan et al., 1993). cant 9-fold risk was found for offspring with both these exposure indicators, but this was based on only 4 exposed Other hypotheses cases. This result is in agreement with those of a study from The SIR analysis showed that leukaemia incidence was not Brazil (Sharpe et al., 1995), reporting paternal and maternal increased in the study population, and the Poisson regression pesticide exposure as strong risk factors. It is also in agreement analysis did not indicate a relationship between animal hus- with results from the National Wilms' Tumor Study in the US, bandry or exposure to pesticides and acute lymphocytic or in which insecticide application during the last 3 years before myelocytic leukaemia. The RR for other acute leukaemias diagnosis constituted a significant risk factor (Olshan et al., were increased in association with contact with several animal 1990, 1993). Parental farming or rural farming residence has species: the association was strongest for pig farming but was not been identified as risk factors of Wilms' tumour, however of borderline significance. Similar results have been reported (Kantor et al., 1979; Wilkins and Sinks, 1984; Olshan et al., for adult farmers (Blair et al., 1992). The evidence for a true 1990).A relationship between paternal farming and childhood relationship is not convincing, however, since most of the death from renal tumour has been reported (McDowall, 1985). associations were weaker for the sub-set of the study popula- Growing field vegetables combined with pesticide purchase tion for which presumably more relevant census information constituted risk factors of borderline significance for eye was available and for offspring from holdings where a parent cancer and neuroblastoma. Eye cancer has been associated worked at least 500 hr annually. with the rural environment (Doll, 1991), but neuroblastoma Interpretation of the results for leukaemia are difficult,since has not been related to parental farming (Spitz and Johnson, classification of sub-types in the Cancer Registry changed 1985; Davis et al., 1987; Wilkins and Hudley, 1990), except in during the study period. Most of the lymphocytic and myelo- one study (Bunin et al., 1990). Detailed exposures were not cytic sub-types were diagnosed later than 1975, and several assessed in those studies. other acute leukaemias would probably be classified as lympho- Chicken farming was associated with some of the common cytic or myelocytic today. The hypothesis of an association cancers of later childhood. The strongest associations were between exposure to pesticides and acute myelocytic leukae- found for osteosarcoma and the mixed cellular type of mia at a young age (Buckley et al., 1989; Buckley, 1992) is not Hodgkin's disease. To our knowledge, similar findingshave not supported by our results. been reported, except for a suggestion of an association Moreover, the suspected,association between soft-tissue between Hodgkin's disease at young age and contact with sarcoma and lymphoma and exposure to phenoxy acids in grain rabbits (Dorken, 1975). It is interesting that chicken farming farming and forestry (Blair et al., 1992) was not supported by was a stronger risk factor than farming of other animal species. our data. An increased R R for Hodgkin's disease was found Chicken have been suspected of inducing malignancies in adult for offspring from farm holdings with forestry, but the adjusted farmers through poultry viruses or chronic antigenic stimuli estimate was not significant. Our data suggest that parental use (Blair et al., 1992). Exposure to organic dust containing of pesticides in horticulture is a risk factor for non-Hodgkin's micro-organisms may be very high in hen houses (Clark et al., lymphoma in childhood. The R R for pesticide purchase follow 1983). a dose-response pattern. As the increase is largely restricted to horticulture, a relationship is suggested with insecticides Study validity rather than phenoxy herbicides, which are prevalent in grain The study has some design features that ensure its validity. farming. Support for our result is provided by the results of Use of the personal identification number made possible several studies of non-Hodgkin's lymphoma in adults, in which complete identification, linkage and follow-up, and cancer insecticides (Cantor, 1982; Woods et al., 1987; Zahm et al., registration in the Cancer Registry of Norway is considered to 1990,1993; Cantor et al., 1992) and orchard farming (Pearce et be complete. Failure of inclusion of children born before 1965 a/., 1985, 1987) constituted risk factors. who died before the 1970 population census would introduce a Testicular cancer was associated neither with the available indicators of exposure to pesticides nor with horticulture, where organochlorine pestjcides with oestrogenic effects are used. The theory that intra-uterine exposure to such pesticides selective loss and a negative bias if a large proportion died of cancer. This bias was avoided by starting follow-up after the 1970 population census for children born between 1952 and 1964. is a cause of testicular cancer (Sharpe and Skakkebaek, 1993)is Agriculture in Norway occurs mainly on small and medium- therefore not supported by our results. The increased SIR for sized farms that are owned and operated by the family. Only 48 KRISTENSEN ETAL II 10% of the total agricultural workforce in the 1970 population at all censuses, such as that on pesticide use; pesticide census were employees (Central Bureau of Statistics, 1975), purchase information was available at the 1969 census, but the who did not participate in the agricultural censuses. The association with brain tumours was strongest for children born national agricultural and horticultural censuses therefore offer 10 to 20 years later. The cross-sectional information on an opportunity to identify most people engaged in agricultural exposure offers little or no opportunity to make Inferences work. We did not include offspring of farm owners born before about the potential mechanisms of tumourogenesis, except for 1925. On the basis of the age distribution of farm owners and the birth-month clustering of cases associated with pesticide the fertility of Norwegian farmers, we estimated that about purchase, which may be compatible with a paternalmediated 15% of all farmers' offspring born between 1952and 1991were mechanism of pesticides. children of farm owners born before 1925. The limitations of the quality of the information on exposure Inclusion criteria in the agricultural censuses were loose, and its timing are not, however, related differently to the and a large proportion of the parents had a low agricultural outcomes under study. The result of such non-differential work input. The inclusion of subjectswith a peripheral connection misclassification for dichotomous categories will be a bias to farming may constitute a problem in the interpretation of the towards unity for any true association. We should therefore results of both the SIR and the Poisson regression analyses. We regard the results for the unconfirmed hypotheses as ``non- therefore performed separate analyses confined to offspring of positive" rather than negative. farmersdefined accordingto the standard occupational code. Also the SIR estimates could be biased towards unity. This Cohort subjects could also have parents who were not could be due to inclusion of offspring whose parents were not farmers at the time of conception or birth, since the cohort was living or working on the farm at the time of conception or birth. based on information obtained at the agricultural censuses, Also, 89% of the study population lived on farms in rural 1969-1989, whereas the offspring under study were born municipalities, and contributed to the person-time experience between 1952 and 1991. We were unable to quantify offspring of the total rural population. The study cohort had a person- born before the first census participation, whose parents time experience that was 0.20 to 0.25 that of the total rural started in farming after the birth of the child. However, this population from 1965 to 1991 for people under 20 years. This problem is not likely to be substantial: less than one thousand ratio declined rapidly to 0.11 for ages 25 to 29 years and 0.01 new farms were settled in the 1950s and even fewer in the for ages 35 to 39 years. The contribution of the study 1960s. During the same period, farms were almost entirely population to both the numerator and the denominator will taken over by children (mainly sons) who lived on the farm. lead to SIR estimates biased towards the null value, although The main structural change in Norwegian agriculture during the impact of this bias is limited for SIR close to unity. If the study the study period has been a downward trend in the number of population'sperson-time experience is subtracted from that of the holdings and the number of people engaged in agricultural reference population, providing a "worst-case" scenario, none of production. Therefore, farm owners whose children were born the reported SIR would changeby more than a few percent. 1 after the census year could have left the farm at the time the child was conceived or born. However, the inclusion of offspring of former farmers into the cohort would not be a major problem: less than 10,000 offspring, including only 8 Our results could be influenced by potential confounders. We had no information on tobacco smoking and alcohol consumption among the parents, but the cancer profile among the adult farmers strongly indicates low prevalences of both ' cancer cases, were born after the last census participation of smoking and alcohol consumption compared with the total the parents (the majority in 1990-1991). rural population (Kristensen et aZ., 199%). If parental smoking The main problem of our study was the use of crude proxies and alcohol consumption were risk factors, confounding could for true exposures and resulting exposure misclassification in be the consequence, especially for the SIR results. However, the Poisson regression analysis. This may be most relevant for parental lifestyle factors are not firmly established as risk pesticides, as we had only information on money spent on factors for cancer in their offspring. Also, the selection of pesticides or spraying equipment. We have assumed that younger parents (born after J924) could introduce confound- associations with these indicators and with grain farming and ing with regard to cancers related to parental age or parity; forestry reflect exposures to phenoxy herbicides, and that the however, stratification on parental age in the analyses did not indicators of exposure to pesticides and work in orchards and influence point estimates. greenhouses are related to exposure to organochlorine, organophosphate, and carbamate compounds. Although the approval, uses and application of pesticides were firmly regulated in the study period, our assumptions can be questioned, and inferences about types of pesticides used can hardly be made on the basis of our data. Several potential exposures to animals on individual farms are also unknown, such as the presence of Although our study lacks precise exposure characterization, several results are in agreement with other reports, and farmers should consider, when handling pesticides, that even their offspring could be at risk of cancer. Further studies should address specific groups of pesticides and potential mechanisms for cancer development caused by animal contact. zoonotic viruses or antigenic provocation by protein dust. Equally problcmatic is the cross-sectional nature of our ACKNOWLEDGEMENTS information on cxposure. We have to assumc that activity on a Dr. T'. Bjerkedal, Dr. P. Laake and Dr. T. Norseth have farm during the index census year reflects the activity before offered valuable supervision and advice at all stages of the conception, during pregnancy or in infancy. This assumption study. We thank Mr. S. Hansen and Mr. A. Johansen for file introduces rnisclassification,which is most serious for cancer linking and preparation in the Cancer Registry, Mr. 0. occurring after the first ycars of life, to which subjects born Sandvin for designing reference rate files, and Dr. E. Heseltine before the first census in 1969 make a considerable contribu- for rcvision of the manuscript. This work was supported by the tion. The problem affects in particular information unavailable Research Council of Norway (grant 103542/1IO). REFERENCES BI-AIRA... ZAHMS.H.. PEAKCEN. .E.. HEINEMAN, E.F. and FRAUMENI. BRUNBOROH, . and KRAVDAL. 0..Fertiliy bv birth order iri Notway: 0. J.F. J R . . Clues to cancer etiology from studies of farmers. Scnnd. J. register-based utra/ysis. Report X6127. Central Bureau of Statistics Work Owiron. Health. 18, 209-215 (1992). Norway. Oslo-Kongsvinger (1986). BUCKLE\. J.D.. The aetiology of cancer in the very young. Brit. J. LEISS.J.K. and SAVITZ. D.A., Home pesticide use and childhood Cancer; 66, SUPP'X.VIII, SI(-S12 (1992). cancer: a case-control study.Amen J. prrbl. Health, 85.249-252 (1995). BUCKLEYJ.D.. ROBISON, L.L.. SWOTINSKI, R.. GARABKANT. D.H.. LEBEAVM. ., MANCHESTER, P.. NESBI.I.M.E.. ODOM, L., PETERS. J.M., WOODS, W.G. and HAMMOND. D.G., Occupational exposures of parents of children with acute non-lymphocytic leukemia: a report from the Children's Cancer Study Group. Cuncer Res.. 49. 4030-4037 (1989). BUNIN, G.R., BLKUEYJ,.D., BOESELC.,.P., RORKE, L.B. and MEADOWS, A.T., Risk factors for astrocytic glioma and primitive neuroectodermal tumor o f the brain in young children: a report from the Children's Cancer Group. Ctrncer Epidemiol. Biornarker Pretl., 3 , 197-204 (1994). BUNING,.R.. WAKIIE,.. KRAMEKS, ., RHELC, .A. and MEADOWS.A.T.. Neuroblastoma and parental occupation. Amer. J. Epirlemiol., 131, 776-780 (1990). CANTOR, K.P.. Farming and mortality from non-Hodgkin's lymphoma: a case-control study. Int. J. Cancer. 29,239-247 (1982). CANTOR, K.P., BLAIR, A.. EVERETT,G., GIBSON. R., BURMEISTER,L.F., BROWN, L.M., SCHUMALN., and DICK, F.R., Pesticides and other agricultural risk factors for non-Hodgkin's lymphoma among men in Iowa and Minnesota. Cancer Res., 52.2447-2455 (1992). LOWENGAKT, R.A., PETERS. J.M.. CICIONI, C., BUCKLEY. J., BERNSTEIN.L., PRESTON-MARTIN, S. and RAPPAPORT, E., Childhood leukemia and parents' occupational and home exposures. J. nar. Cancer Inst., 79, 39-46 (1987). MCCKEDIEM,.. MAISONNEUVE. P. and BOYLLP, ., Antenatal risk factors for malignant brain tumours in New South Wales children. Imf. J. Curice,; 56, 6 1 0 (1904~). MCCREDIEM,., MAISONNEUVE. P. and BOYLE, P.. Perinatal and early postnatal risk factors for malignant brain tumours in New South Wales children. lift.J. Cancer, 56, 11-15 (19946). MCDOWALML,.E., Occupational re roductive epidemiologv. The UFC of routinely collected srutistics in Engfand orid Wales 1980-82. HMSO, Office of Population Censuses and Surveys,London (1985). Musicco, M., SANI,M., MOLINARI, S.. FILIPPINI, G., GATTA,G. and BERRINO, F., A case-control study of brain gliomas and occupational exposure to chemical carcinogens: the risk to farmers. Amer. J. Epidemiol., 128,778-785 (1988). OLSEN, J.H., DE NULLY BROWN, P., SCHULGEN, G. and JENSEN, O.M., Parental employment at time of conception and risk of cancer in offspring.Europ. J. Cancer, 27,958-965 (1991). CENTRAL BUREAU OF STATISTICS, Population and housing census 1970. OLSHAN, A.F., BRESLOW,N.E., DALING, J.R.. F A L L E ~J.AM,., GRUFF- Vol. II: Industry occupation, working hours, etc. Central Bureau -of ERMAN, S., ROBISONL.L., WASKERWITZ, M. and HAMMOND, C.D., Statisticsof Norway,Oslo (1975). Wilms' tumor and paternal occupation. Cancer Res., 50, 3212-3217 CLARK, S., RYLANDER, R. and LARSSON, L., Airborne bacteria, (1990). endotoxin and fungi in dust in poultry and swine confinement OLSHAN, A.F., BRESLOW, N.E., FALLETTA, J.M., GRUFFERMAN, S., buildings.Amen i d . Hyg.Ass. J., 44,537-541 (1983). PENDERGRASS, T., ROBISONL,.L., WASKERWITZ, M., WOODS, W.G., DAVIS, J.R., BROWNSON, R.C., GARCIA: R., BENTZ, B.J. and TURNER, A., Family pesticide use and childhood brain cancer. Arch. environ. VIETTI,T.J. and HAMMOND, G.D., Risk factors for Wilms tumor. Cancer, 72,938-944 (1993). Contam. Toxicol., 24,87-92 (1993). PEARCE, N.E., SHEPPARD, R.A., SMITH, A.H. and TEAGUEC,.A., DAVIS, S., ROGERS, M.A.M. and PENDERGRASS, T.W., The incidence and epidemiologic characteristics of neuroblastoma in the United Non-Hodgkin's lymphoma and farming: an expanded case-control study. Int. J. Cancer, 39,155-161 (1987). States.Amer.J. Epiderniol., 126,1063-1074 (1987). PEARCE, N.E., SMITH, A.H. and FISHER, D.O., Malignant lymphoma DOLL, R., Urban and rural factors in the aetiology of cancer. Int. J. Cancer,47,803-810 (1991). DORKENI,-I., Hodgkin's disease: an epidemiologic study among 140 children-urbanirural relations, parental occupation, contact with domestic animals (in German). Arch. Geschwulsfforsch., 45, 283-298 and multiple myeloma linked with agricultural occupations in a New Zealand cancer-registry-based study.Amer. J. Epidemiol., 121,225-237 (1985). . PRESTON, D.L., LUBINJ,.H., PIERCE, D.A. and MCCONNEYM, .E., Epicure. Hirosoft International Corporation, Seattle (1993). (1975). PRESTON-MARTIN, S., Yu, M.C., BENTONB, . and HENDERSON, B.E., GOLD, E., GORDIS, L., TONASCIA, J. and SZKLOM, ., Risk factors for brain tumors in children.Amer. J. Epidemiol., 109,309-319 (1979). N-nitroso compounds and childhood brain tumors: a case-control study. Cancer Res., 42,5240-5245 (1982). HEMMINKKI,., SALONIEMI, I., SALONENT,., PARTANEN, T. and VAINIO, H., Childhood cancer and parental occupation in Finland. J. Epidemiol. Commun.Health, 35,ll-15 (1985). RYAN, P., HURLEY, S.F., JOHNSON, A.M., SALZBERGM,., LEE, M.W., NORTH, J.B., MCNEILLJ,.J. and MCMICHAELA, .J., Tumours of the brain and presence of antibodies to Toxoplasma gondii. Int. J. Epidemiol., 22,412419 (1993). HOWE, G.R., BURCH, J.D., CHIARELLI, A.M., RISCH, H.A. and CHOI, B.C.K., An exploratory case-control study of brain tumors in children. CancerRes., 49,4349-4352 (1989). SANDERS, B.M., WHITE, G.C. and DRAPER, G.J., Occupations of fathers of children dying from neoplasms. J. Epidemiol. Commun. Health, 35,245-250 (1981). KANTORA, .F., CURNEN, M.G.M., MEIGS, J.W. and FLANNERY, J.T., Occupations of fathers of patients with Wilms' tumour. J. Epidemiol. Commun. Health, 33,253-256 (1979). SAVITZD,.A. and CHEN,J., Parental occupation and childhood cancer: review of epidemiologic studies. Environ. Health Perspect., 88,325-337 (1990). ~ R D A UJ.NW,.P.F., HAYES, R.B., POTTERNL,.M., BROWN, L.M. and HOOVER, R.N., Testicular cancer in young men and arental occupational exposure.Amer. J. ind. Med., 20,219-227 (19917. SCHUMALN.,M., CHOI,N.W. and GULLENW, .H., Relationship of central-nervous-system neoplasms to Toxoplasma gondii infection. Amer. J. Publ. Health, 57,848-856 (1967). ~ E I H U EPS.,, BURGER, P.C. and SCHEITHAUER, B.W., Histological typing of rumours of the central nervous system. World Health Otganization.Internationalhistological classification of rumours. Springer,Berlin, Heidelberg (1993). SEURI, M. and KOSKELA,P., Contact with pigs and cats associated with high prevalence of Toxoplasma antibodies among farmers. Brit. J. ind. Med., 49,845-849 (1992). KRISTENSEPN.,. ANDERSEN, A,, IRGENSL,.M., BYE, AS. and VAGSTAD, N., Testicular cancer and parental use of fertilizers in agriculture. Cancer Epidemiol. Biomarkers Prev., (1995a). (In press). SHARPE, C.R., FRANCOE,.L., DE CAMARGO, B., LOPES, L.F., BARRETO, J.H., JOHNSONR, .R. and MAUAD, M.A., Parental exposures to pesticides and risk of Wilms' tumor in BraziLAmer.J. Epidemiol., 141, 210-217 (1995). KRISTENSEPN.,, ANDERSEN, A., IRGENS, L.M., LAAKE, P. and BYE, A.S., Male and female cancer in Norwegian agriculture:incidence and risk factors in the farm environment. Scand. J. Work environ. Health, (19956). (In press). KUIJTEN.R.R. and BUNING, .R., Risk factors for childhood brain tumors. Cancer Epidemiol. Biomarkers Prev., 2,277-288 (1993). SHARPE, R.M. and SKAKKEBIENK.,E., Are oestrogens involved in falling sperm counts and disorders of the male reproductive tract? Lancet, 341,1392-1395 (1993). SHU, X.O., GAO, Y.T., BRINTON, L.A., LINET, M.S., TU, J.T., ZHENG, W. and FRAUMENI, J.F. JR., A population-based case-control study of childhood leukemia in Shanghai. Cancer, 62,635-644 (1988). KUIJTEN.R.R., BUNING,.R., NASS,C.C. and MEADOWS,A.T., Parental SINKS, T.H. JR., N-nitroso compounds, pesticides, and parental expo- occupation and childhood astrocytoma: results of a case-control study. sures in the work-place as risk factors for childhood brain cancer: a CancerRes., 52,782-786 (1992). case-control study. Diss. Abstr. Int., 46, 1888-B (1985). LAVALG. . and TUYNSA, .J., Environmental factors in childhood SPITZ, M.R. and JOHNSON, C.C., Neuroblastoma and paternal occupa- leukaemia. Brit. J. ind. Med., 45,843-844 (1988). tion.Amer. J. Epidemiol., 121,924-929 (1985). KRISTENSEN ETAL. WILKINS, J.R. and KOUTRAS, R.A., Paternal occupation and brain cancer in offspring: a mortality-based case-control study.Amer. J. ind. Med., 14,299-318(1988). WILKINS, J.R. 111, MCLAUGHLIJN.,A., SINKS, T.H.and KOSNIK, E.J., Parental occupation and intracranial neoplasms of childhood: anecdotal evidence from a unique occupational cancer cluster.Amer. J. ind. Med., 19,643-653 (1991). parents exposed to hydrocarbon-related industries and occupations. J, Epidemjol., 329-336 (1980). ZAHM,S.H., WEISENBURGER, D.D., BABBITT, P.A., SAAL, R.C., VAUGHTJ,.B., CANTOR, K.P. and BLAIR, A., A case-control study of non-Hodgkin's lymphoma and the herbicide 2,4-dichlorophenoxyace. tic acid (2,4-D) in eastern Nebraska. Eptdemidow, 1,349-356 (1990). i \> 1 I t 1 I I I 1 I k