Pesticide Use and Breast Cancer Risk among Farmers Wives in the Agricultural Health Study

Transcription

1 American Journal of Epidemiology Copyright 2005 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 161, No. 2 Printed in U.S.A. DOI: /aje/kwi022 ORIGINAL CONTRIBUTIONS Pesticide Use and Breast Cancer Risk among Farmers Wives in the Agricultural Health Study Lawrence S. Engel 1,2, Deirdre A. Hill 1, Jane A. Hoppin 3, Jay H. Lubin 1, Charles F. Lynch 4, Joy Pierce 5, Claudine Samanic 1, Dale P. Sandler 3, Aaron Blair 1, and Michael C. Alavanja 1 1 Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD. 2 Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, NY. 3 National Institute of Environmental Health Sciences, National Institutes of Health, Research Triangle Park, NC. 4 Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA. 5 Battelle Centers for Public Health Research and Evaluation, Durham, NC. Received for publication December 8, 2003; accepted for publication July 30, The authors examined the association between pesticide use and breast cancer incidence among farmers wives in a large prospective cohort study in Iowa and North Carolina. Participants were 30,454 women with no history of breast cancer prior to cohort enrollment in Information on pesticide use and other information was obtained by self-administered questionnaire at enrollment from the women and their husbands. Through 2000, 309 incident breast cancer cases were identified via population-based cancer registries. Rate ratios were calculated for individual pesticides using Poisson regression, controlling for confounding factors. Breast cancer standardized incidence ratios were 0.87 (95% confidence interval: 0.74, 1.02) for women who reported ever applying pesticides and 1.05 (95% confidence interval: 0.89, 1.24) for women who reported never applying pesticides. There was some evidence of increased risk associated with use of 2,4,5-trichlorophenoxypropionic acid (2,4,5-TP) and possibly use of dieldrin, captan, and 2,4,5-trichlorophenoxyacetic acid (2,4,5-TP), but small numbers of cases among those who had personally used the pesticides precluded firm conclusions. The authors found no clear association of breast cancer risk with farm size or washing of clothes worn during pesticide application, but risk was modestly elevated among women whose homes were closest to areas of pesticide application. Further follow-up of this cohort should help clarify the relation between pesticide exposure and breast cancer risk. agriculture; agrochemicals; breast neoplasms; fungicides, industrial; herbicides; insecticides; pesticides; risk Abbreviations: CI, confidence interval; DDT, dichlorodiphenyltrichloroethane; dichlorvos, 2,2-dichloroethenyl dimethylphosphate; RR, rate ratio; TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin; 2,4,5-T, 2,4,5-trichlorophenoxyacetic acid; 2,4,5-TP, 2,4,5-trichlorophenoxypropionic acid. The estrogenic properties of certain pesticides have focused attention on the possible role of these chemicals in breast cancer etiology. Most of this attention has been directed at the organochlorine insecticides, which persist in the environment and accumulate in body fat. The weight of the evidence to date does not appear to support a relation between exposure to organochlorines and risk of breast cancer (1 24), although unanswered questions remain (25). While several other pesticides cause mammary tumors in rodents (26 29) or exhibit properties in vitro that may be related to breast cancer etiology (30 33), there has been little epidemiologic research on the relation of these compounds Correspondence to Dr. Lawrence Engel, Memorial Sloan-Kettering Cancer Center, Epidemiology Service, 307 East 63rd Street, 3rd Floor, New York, NY ( engell@mskcc.org). 121

2 122 Engel et al. TABLE 1. Selected characteristics of farmers wives at enrollment, Agricultural Health Study, All wives in the cohort Wives who never used pesticides Characteristic (n = 309) (n = 30,145) (n = 152) (n = 13,297) RR, Age (years) , , , , , , , , , , , , , , * 3.9, * 4.3, 20.2 Race White , , Other , , 2.6 State of residence Iowa , , North Carolina , , , , 1.1 Highest educational level Less than high school , High school , , , , 2.1 More than high school , , , , 2.7 Smoking Never smoker , , Former smoker , , , , 1.6 Current smoker , , , , 2.5 First-degree family history of breast cancer , , , , 3.0 Table continues to breast cancer risk. Moreover, although these pesticides generally show only weak hormonal activity and many pesticides appear to have no such activity, the biologic impact of typically combined exposures and exposures at different developmental stages remains largely unknown (32, 34 36). Women engaged in agricultural work or living in agricultural areas may be exposed to higher levels and different types of pesticides than the general population. These exposures may be occupational, such as those incurred through mixing and applying pesticides or working in pesticidetreated fields. Other pesticide exposures may be environmental, resulting from factors such as spray drift, contaminated drinking water, or handling of items that have been contaminated in or near areas of pesticide application. In addition, farm families tend to use more nonagricultural pesticides in and around the home than do nonfarm families (37). Most epidemiologic studies of breast cancer risk among women exposed to pesticides through farming have relied on surrogate measures of pesticide exposure, such as job title, possession of a pesticide application license, or residence on a farm (38 45). Two of these studies reported no association (44, 45), while the remainder reported a slightly decreased risk (39 43). None of these studies had sufficient subjects and exposure information to assess the risks associated with individual pesticides or groups of pesticides. Two casecontrol studies that collected self-reported information on agricultural tasks and pesticide use also found a reduced risk of breast cancer among women employed in agriculture (46, 47), although the findings by Duell et al. (47) suggested an increased risk among the farming women with the greatest likelihood of pesticide exposure; neither study presented results for individual pesticides or chemical classes. We examined breast cancer risk among wives of farmers in a large prospective agricultural cohort study in relation to use of individual pesticides by the women themselves or by their husbands. If exposure to pesticides contributed to the development of breast cancer, one would be more likely to observe this relation in such a population because of their potentially higher exposure than the general population and the availability of specific information on pesticide use and demographic and lifestyle factors collected prior to disease diagnosis. MATERIALS AND METHODS Study population Participants consisted of the wives of private pesticide applicators (largely farmers) from Iowa and North Carolina

3 Pesticides and Female Breast Cancer Risk 123 TABLE 1. Continued Characteristic All wives in the cohort * p for trend < RR, rate ratio; CI, confidence interval. All factors were adjusted for the other factors in the table, except where indicated. Reference category. Missing data (including those for questions for which the participant answered Do not know ) exceeded 10% for highest educational level (9.4% of cases and 10.7% of noncases among all study subjects; 8.5% of cases and 8.1% of noncases who never used pesticides), body mass index (20.1% of cases and 29.7% of noncases among all study subjects; 21.7% of cases and 32.6% of noncases who never used pesticides), age at menarche (7.8% of cases and 20.3% of noncases among all study subjects; 7.2% of cases and 21.6% of noncases who never used pesticides), parity (8.1% of cases and 20.6% of noncases among all study subjects; 7.2% of cases and 21.9% of noncases who never used pesticides), and menopausal status (9.7% of cases and 21.0% of noncases among all study subjects; 8.6% of cases and 22.0% of noncases who never used pesticides). # Weight (kg)/height (m) 2. ** Restricted to postmenopausal women. Restricted to parous women. who were enrolled in the prospective Agricultural Health Study between 1993 and 1997 (48). All 43,475 male private pesticide applicators who indicated that they were married were requested to ask their wives to complete two take-home questionnaires. One questionnaire elicited information on the wives farm exposures and general health ( spouse enrollment questionnaire ), while the other focused on their reproductive health history ( female and family health questionnaire ). Over 32,100 wives (74 percent of eligible wives) were enrolled in the cohort. Of these, 19,578 (61 percent of those enrolled) completed both questionnaires, while 12,549 (39 percent of those enrolled) completed only the spouse enrollment questionnaire. Female licensed pesticide applicators were not included in the analyses reported here, because of their relatively small numbers (n = 1,347; 15 cases) and differences in the nature and extent of their pesticide use in comparison with women who were not themselves licensed but may have applied pesticides through their husband s license. Exposure assessment Wives who never used pesticides (n = 309) (n = 30,145) (n = 152) (n = 13,297) Body mass index,#,** < , , , , , , , , , , , * 1.1, , , 3.2 Age (years) at menarche < , , , , , , 1.2 Parous , , , , 1.9 Age (years) at first birth < , , , , , , , , , 3.5 Postmenopausal , , , , 3.6 Age (years) at menopause** < , , , , , , , , , , , , , , , , 4.3 All pesticide exposure information for this study was obtained at cohort enrollment. Relevant information from the spouse enrollment questionnaire included: 1) ever/never use of 50 specific pesticides (consisting of insecticides, herbicides, fungicides, and fumigants) likely to have been used by this study population; 2) number of years of mixing or applying any pesticides; 3) frequency of mixing or applying any pesticides; 4) number of years the participant had lived or worked on a farm; 5) farm tasks performed; 6) performance of household tasks involving possible pesticide exposure; 7) distance of the participant s house from fields where pesticides were applied; and 8) household and other nonagricultural pesticide use or exposure. The questionnaires also elicited information on a range of demographic, lifestyle, health, and reproductive factors. Questionnaires directed to the farmers elicited similar but more detailed information on lifetime pesticide use, including the duration and frequency of use of specific pesti-

4 124 Engel et al. cides. Information obtained from the farmers was used as a measure of possible indirect pesticide exposure to their wives. Such indirect exposures might result from spray drift, from contaminated drinking water, or from the wives working in pesticide-treated fields or handling items that were in or near areas of pesticide application. (The questionnaires may be viewed at the website of the Agricultural Health Study ( Participant follow-up and case ascertainment Breast cancer cases were identified through populationbased cancer registries in Iowa and North Carolina. Vital status was ascertained through state death registries and the National Death Index. Of the 30,980 participants who were residing in-state at the time of cohort enrollment and provided information on their pesticide use, those with a diagnosis of breast cancer prior to enrollment (n = 455) and/ or incident in-situ breast cancer (n = 74) were excluded from the present analyses, leaving 30,454 eligible participants. consisted of all eligible participants diagnosed with malignant breast cancer (International Classification of Diseases for Oncology, Second Edition, codes C50.0 C50.9) between enrollment and December 31, 2000 (i.e., incident cases; n = 309). Participants who were no longer residing in Iowa or North Carolina were identified through personal contacts with the participants, motor vehicle records, pesticide registration records, and the Internal Revenue Service address database (from which we obtained only current addresses of study participants). Of the 30,454 women included in these analyses, 72 (0.2 percent) moved out of state and 443 (1.5 percent) died during the study period. The average duration of follow-up was 4.8 years, with a total duration of follow-up of 146,653 person-years (99.9 percent of the total possible follow-up person-time). Data analysis Standardized incidence ratios for breast cancer were calculated for study participants who reported any prior pesticide use and those who reported no prior use (49). Expected numbers of cases were estimated using 5-year age and calendar-time, race-specific cancer incidence rates from the cancer registries for each state s population. We used Poisson regression to calculate rate ratios and 95 percent confidence intervals for exposure to various agricultural and nonagricultural risk factors among all 30,454 participants. We calculated person-years at risk for each subject from her date of enrollment to the earliest of the following dates: date of first breast cancer diagnosis, date of moving out of state, date of death, or December 31, We assessed risks associated with reported use of pesticides individually, as well as pesticides grouped into chemical classes, by the women (i.e., potential direct exposures) and by their husbands (i.e., potential indirect exposures). All analyses were adjusted for age (<40, 40 49, 50 59, and 60 years), race (White and other), and state (Iowa and North Carolina). Body mass index, age at menarche, parity, age at first birth, menopausal status, age at menopause, family history of breast cancer, physical activity, smoking, alcohol consumption, fruit and vegetable consumption, and education were examined as potential confounders but were not included in the final models because they did not materially change the estimates. We also performed pesticide-specific exposure-response analyses in which participants were categorized either as nonexposed or by tertile/median of nonzero exposure to each pesticide (i.e., either none, low, medium, or high or none, low, or high, depending on the number of exposed subjects) on the basis of their husbands cumulative use of the pesticide (number of years of use of the pesticide average number of days per year on which the pesticide was used). These analyses were restricted to the 13,449 wives who reported no prior pesticide use (i.e., wives with potential indirect exposure only). We assessed exposure-response with linear trend tests, using the mean lifetime number of days of pesticide application in each category. We also conducted the above analyses by state to evaluate the consistency of associations across geographic subcohorts (i.e., Iowa and North Carolina), as well as by menopausal status. Because the women in this cohort provided only information on ever/never use for individual pesticides and our data suggested that they tended to handle fewer and less toxic pesticides for fewer numbers of years and days per year than the men (50), comparisons of the consistency of risk estimates related to the wives use versus the husbands use or Iowa versus North Carolina focused primarily on the direction, not the magnitude, of observed associations. We conducted additional analyses among women with potentially higher exposure to pesticides or greater susceptibility to such exposure by restricting analyses separately to participants who had applied pesticides for at least 10 years (n = 5,563) or at least 40 days in total (n = 6,287), lived on a farm with at least 500 acres (202 hectares) under cultivation (n = 10,638), had grown up on a farm (i.e., some reported exposures may have occurred at a young age) (n = 10,996), or had a history of breast cancer in a first-degree relative (n = 3,434). For brevity, tables presenting the results of subanalyses include only pesticides for which there was an a- priori hypothesis regarding an association with breast cancer (e.g., organochlorines) or for which an altered risk was observed. (More detailed tables can be found at The institutional review boards of participating institutions approved the study protocol and the manner in which informed consent was obtained from participants. RESULTS Differences between cases and noncases with regard to established and suspected breast cancer risk factors generally resembled those identified in previous studies (table 1). Breast cancer risks in relation to these factors were similar among women who never used pesticides and the total cohort. The standardized incidence ratio for breast cancer for all of the women was 0.9 (95 percent confidence interval (CI): 0.8, 1.1). In Iowa, the standardized incidence ratio was 1.0 (95 percent CI: 0.8, 1.2) for women who reported ever applying pesticides and 1.3 (95 percent CI: 1.0, 1.6) for women who

5 Pesticides and Female Breast Cancer Risk 125 TABLE 2. Relation between selected measures of potential cumulative exposure to all pesticides combined and risk of breast cancer among farmers wives, Agricultural Health Study, Characteristic All wives in the cohort Wives who never used pesticides * RR*, RR Wives years of pesticide application , , , , , , , , , 1.2 Wives lifetime days of pesticide application (nonzero quartiles among controls) , , , , , , , , , 1.1 Distance of home (yards ) from areas of pesticide application , , , , , , , , , 1.9 < , , , , 2.9 Acreage of crops on farm < , , , , , , , , , , 1.9 1, , , , , 1.6 Frequency of washing work clothes worn during pesticide application (days/year) < , , , , , , , , , , 2.4 > , , , , 2.7 Grew up on a farm , , , , 1.3 * RR, rate ratio; CI, confidence interval. Each pesticide exposure was examined in a separate model with adjustment for age (<40, 40 49, 50 59, and 60 years), race (White and other), and state of residence (Iowa and North Carolina). Reference category. 1 yard = 0.9 m. 1 acre = 0.4 hectares. reported never applying pesticides, while in North Carolina, the corresponding standardized incidence ratios were 0.7 (95 percent CI: 0.5, 0.9) and 0.8 (95 percent CI: 0.6, 1.1). We found little evidence of association between measures of potential direct and indirect cumulative exposure to all pesticides combined and breast cancer risk (table 2). There was no exposure-response trend in relation to years of pesticide application or lifetime days of pesticide application (i.e., number of years of use of pesticides average number of days per year that pesticides were used) by the women. We also observed no clear trends associated with various measures of possible indirect exposure, although risk appeared modestly elevated among women whose homes were closest to areas of pesticide application. Growing up on a farm was not related to breast cancer risk. Most of the increased risk estimates among these women were linked to insecticide use and were related to use by the women s husbands but not by the women themselves (table 3). These included most of the organochlorine insecticides, with risk estimates ranging from 1.6 (95 percent CI: 1.1, 2.4) for heptachlor to 2.0 (95 percent CI: 1.1, 3.3) for dieldrin when these pesticides were used by the husbands. However,

6 126 Engel et al. TABLE 3. Relation between individual pesticide use (ever/never) and risk of breast cancer among farmers wives, Agricultural Health Study, Exposure Wife s pesticide use among all wives in the cohort Husband s pesticide use among wives who never used pesticides (n = 309)* (n = 30,145)* (n = 152)* (n = 13,297)* RR, Any pesticides , , , Chemical classes Carbamates , , , , 1.5 Chloroacetanilides , , , , 1.4 Dinitroanilines , , , , 1.9 Dithiocarbamates , , 2.0 Organochlorines , , , , 2.0 Organophosphates , , , , 4.0 Phenoxy herbicides , , , , 1.8 Pyrethroids , , , , 1.7 Thiocarbamates , , , 1.6 Triazines , , , , 2.0 Insecticides Aldicarb < , , 1.2 Aldrin , , , 2.7 Carbaryl , , , , 2.0 Carbofuran , , , 1.4 Chlordane , , , , 2.5 Chlorpyrifos , , , , 1.8 Coumaphos , , 1.7 DDT , , , , 1.8 Diazinon , , , , 2.0 Dichlorvos , , , 2.5 Dieldrin < , 3.3 Fonofos , , , 1.9 Heptachlor , , , 2.4 Lindane < , , 2.5 Malathion , , , , 2.0 Parathion < , , 2.1 Permethrin (for animals) , , , , 2.1 Permethrin (for crops) , , , 1.7 Phorate , , , 1.6 Terbufos , , , 1.2 Toxaphene < , , 2.1 Table continues the four organochlorines whose use was reported by sufficient numbers of women for analysis were associated with either no risk or a reduced risk when they were used by the women. A similar pattern of increased risk only in relation to the husbands use was observed for the insecticide carbaryl (rate ratio (RR) = 1.4, 95 percent CI: 1.0, 2.0) and the fungicide captan (RR = 2.7, 95 percent CI: 1.7, 4.3). An excess of breast cancer was also observed among women whose husbands used 2,4,5-trichlorophenoxypropionic acid (2,4,5- TP) (RR = 2.0, 95 percent CI: 1.2, 3.2), but no case women reported use of this herbicide. Risk appeared elevated among women whose husbands used any organophosphate (RR = 1.9, 95 percent CI: 0.9, 4.0); however, risk was significantly increased only in relation to malathion (RR = 1.4, 95 percent CI: 1.0, 2.0), and other elevated risks in this class were of similar magnitude. When multiple pesticides were examined in models simultaneously, the associations with 2,4,5-TP and captan were unchanged, while other associations decreased somewhat (data not shown). Patterns of risk were generally inconsistent between Iowa and North Carolina, although few pesticides were reported by sufficient women in North Carolina for a comparison

7 Pesticides and Female Breast Cancer Risk 127 TABLE 3. Continued Exposure Wife s pesticide use among all wives in the cohort Husband s pesticide use among wives who never used pesticides (n = 309)* (n = 30,145)* (n = 152)* (n = 13,297)* Herbicides Alachlor , , , , 1.3 Atrazine , , , , 1.6 Butylate < , , 1.3 Chlorimuron ethyl , , , 1.5 Cyanazine , , , 1.1 Dicamba , , , , 1.5 EPTC < , , 1.6 Glyphosate , , , , 1.9 Imazethapyr , , , 1.8 Metolachlor , , , 1.5 Metribuzin , , , 1.4 Paraquat < , , 2.0 Pendimethalin , , , 1.4 Petroleum oil , , , , 1.9 Trifluralin , , , , 1.4 2,4-D , , , , 1.4 2,4,5-T < , , 1.9 2,4,5-TP < , 3.2 Fungicides Benomyl < , , 2.4 Captan , , , 4.3 Chlorothalonil , , , 2.4 Maneb , , , 1.8 Metalaxyl , , , 1.5 Fumigants Carbon tetrachloride/ carbon disulfide < , 2.4 Ethylene dibromide < , 1.6 Methyl bromide , , , 1.7 * The numbers of cases and noncases included in the analyses differ across pesticides because of missing data. RR, rate ratio; CI, confidence interval; DDT, dichlorodiphenyltrichloroethane; dichlorvos, 2,2-dichloroethenyl dimethylphosphate; EPTC, S-ethyl dipropylthiocarbamate; 2,4-D, 2,4-dichlorophenoxyacetic acid; 2,4,5-T, 2,4,5-trichlorophenoxyacetic acid; 2,4,5-TP, 2,4,5-trichlorophenoxypropionic acid. Each pesticide exposure was examined in a separate model adjusted for age (<40, 40 49, 50 59, and 60 years), race (White and other), and state of residence (Iowa and North Carolina). A rate ratio was estimated only if there were at least three exposed and three unexposed cases and three exposed and three unexposed noncases. Organophosphate insecticides: chlorpyrifos, coumaphos, diazinon, dichlorvos, fonofos, malathion, methyl or ethyl parathion, phorate, terbufos, and trichlorfon; organochlorine insecticides: aldrin, chlordane, DDT, dieldrin, heptachlor, lindane, and toxaphene; carbamate insecticides: aldicarb, carbaryl, and carbofuran; pyrethroid insecticides: permethrin or pyrethroid products; triazine herbicides: atrazine, cyanazine, and metribuzin; chloroacetanilide herbicides: alachlor and metolachlor; thiocarbamate herbicides: butylate and EPTC; dinitroaniline herbicides: pendimethalin and trifluralin; phenoxy herbicides: 2,4-D, 2,4,5-T, and 2,4,5-TP; dithiocarbamate fungicides: mancozeb, maneb, and ziram. across states (table 4). Risk appeared to be increased in relation to dinitroanilines, but only among women who used them in North Carolina (RR = 2.3, 95 percent CI: 0.9, 5.6). A husband s use of any organochlorine was associated with a rate ratio of in Iowa but was unrelated to risk in North Carolina. Moreover, risk was not increased among women who used organochlorines in either state. In fact, risk was reduced among Iowa women who used the organochlorines chlordane or dichlorodiphenyltrichloroethane (DDT). A similar pattern was seen for carbaryl and for several of the organophosphate insecticides, which appeared to be associated with increased risk only in relation to the husband s use in Iowa. Interestingly, there was a comparable excess of breast cancer for the husband s use of 2,4,5-TP in both states (RR = 1.9 (95 percent CI: 1.0, 3.4) in Iowa and RR = 2.1 (95 percent CI: 0.9, 4.6) in North Carolina); there were too few wives who used this herbicide in either state for comparison. Similarly, a husband s use of captan was associated with an

8 128 Engel et al. TABLE 4. Relation between individual pesticide use (ever/never) and risk of breast cancer among farmers wives, by state, Agricultural Health Study, Exposure Wife s pesticide use among all wives in the cohort Husband s pesticide use among wives who never used pesticides * * * * RR, Iowa (n = 207) (n = 20,592) (n = 93) (n = 8,221) Any pesticides , , , Chemical classes Carbamates , , , , 1.7 Chloroacetanilides , , , , 2.0 Dinitroanilines , , , , 2.4 Dithiocarbamates < < Organochlorines , , , , 4.0 Organophosphates , , , , 2.5 Phenoxy herbicides , , , , 2.3 Pyrethroids , , , , 2.1 Thiocarbamates , , , 2.0 Triazines , , , , 2.7 Insecticides Aldrin , , , 3.5 Carbaryl , , , , 2.5 Chlordane , , , 3.6 Chlorpyrifos , , , 2.1 Coumaphos , , 2.0 DDT , , , 2.6 Diazinon , , , , 2.6 Dichlorvos , , 2.6 Dieldrin < , 3.4 Fonofos , , , 2.2 Heptachlor , , , 2.8 Lindane < , , 2.8 Malathion , , , , 2.2 Parathion < , 4.1 Phorate , , , 1.8 Terbufos , , , 1.2 Toxaphene < , 3.2 Herbicides Atrazine , , , , 1.7 Cyanazine , , , 1.2 Metribuzin < , , 1.4 2,4-D , , , , 1.5 2,4,5-T < , , 1.9 2,4,5-TP < , 3.4 Fungicides Captan , , 5.6 Fumigants Methyl bromide < < Table continues increased risk in both states (RR = 3.2 (95 percent CI: 1.8, 5.6) in Iowa and RR = 1.9 (95 percent CI: 0.9, 4.3) in North Carolina). In analyses stratified by menopausal status, all of the increased risks related to the women s pesticide use occurred among the premenopausal women, while all of the decreased

9 Pesticides and Female Breast Cancer Risk 129 TABLE 4. Continued Exposure Wife s pesticide use among all wives in the cohort Husband s pesticide use among wives who never used pesticides * * * * North Carolina (n = 102) (n = 9,553) (n = 59) (n = 5,076) Any pesticides , , , Chemical classes Carbamates , , , , 2.3 Chloroacetanilides , , , 1.3 Dinitroanilines , , , 2.1 Dithiocarbamates < , 2.0 Organochlorines , , , 1.2 Organophosphates , , , Phenoxy herbicides , , , 2.1 Pyrethroids < , 1.7 Thiocarbamates < , 1.8 Triazines < , , 2.4 Insecticides Aldrin < , 2.2 Carbaryl , , , , 1.7 Chlordane , , , 1.9 Chlorpyrifos , , , 1.9 DDT < , , 1.5 Diazinon , , , , 2.0 Dichlorvos < < Dieldrin < < Fonofos < < Heptachlor < < Lindane < , 3.3 Malathion , , , , 2.7 Parathion < , 1.8 Phorate < , 1.9 Terbufos < , 2.1 Toxaphene < , 1.7 Herbicides Atrazine < , , 2.1 Cyanazine < , 1.8 Metribuzin < , 2.2 2,4-D , , , 1.3 2,4,5-T < , 2.4 2,4,5-TP < , 4.6 Fungicides Captan < , 4.3 Maneb < , , 1.9 Fumigants Methyl bromide , , , 1.8 * The numbers of cases and noncases included in the analyses differ across pesticides because of missing data. RR, rate ratio; CI, confidence interval; DDT, dichlorodiphenyltrichloroethane; dichlorvos, 2,2-dichloroethenyl dimethylphosphate; 2,4-D, 2,4-dichlorophenoxyacetic acid; 2,4,5-T, 2,4,5-trichlorophenoxyacetic acid; 2,4,5-TP, 2,4,5-trichlorophenoxypropionic acid. Each pesticide exposure was examined in a separate model adjusted for age (<40, 40 49, 50 59, and 60 years) and race (White and other). A rate ratio was estimated only if there were at least three exposed and three unexposed cases and three exposed and three unexposed noncases. Organophosphate insecticides: chlorpyrifos, coumaphos, diazinon, dichlorvos, fonofos, malathion, methyl or ethyl parathion, phorate, terbufos, and trichlorfon; organochlorine insecticides: aldrin, chlordane, DDT, dieldrin, heptachlor, lindane, and toxaphene; carbamate insecticides: aldicarb, carbaryl, and carbofuran; pyrethroid insecticides: permethrin or pyrethroid products; triazine herbicides: atrazine, cyanazine, and metribuzin; chloroacetanilide herbicides: alachlor and metolachlor; thiocarbamate herbicides: butylate and S-ethyl dipropylthiocarbamate; dinitroaniline herbicides: pendimethalin and trifluralin; phenoxy herbicides: 2,4- D, 2,4,5-T, and 2,4,5-TP; dithiocarbamate fungicides: mancozeb, maneb, and ziram.

10 130 Engel et al. TABLE 5. Relation between individual pesticide use (ever/never) and risk of breast cancer among farmers wives, by menopausal status at enrollment, Agricultural Health Study, Exposure Wife s pesticide use among all wives in the cohort Husband s pesticide use among wives who never used pesticides * * * * RR, Premenopausal women (n = 87) (n = 13,087) (n = 36) (n = 5,674) Any pesticides , , , Chemical classes Carbamates , , , , 3.2 Chloroacetanilides , , , 1.9 Dinitroanilines , , , 1.9 Dithiocarbamates < < Organochlorines , , , 2.4 Organophosphates , , , Phenoxy herbicides , , , , 2.3 Pyrethroids , , , 2.2 Thiocarbamates < , , 1.6 Triazines , , , 4.3 Insecticides Aldrin < , 3.8 Carbaryl , , , , 1.7 Chlordane , , 3.2 Chlorpyrifos , , , 1.6 DDT < , 3.2 Diazinon , , , , 3.2 Dichlorvos , , 3.3 Dieldrin < < Fonofos < , , 2.5 Heptachlor < , 3.3 Lindane < , 4.6 Malathion , , , , 3.0 Parathion < , 3.0 Phorate , , , 2.5 Terbufos , , , 2.1 Toxaphene < , 3.1 Herbicides Atrazine , , , 2.1 Cyanazine , , , 1.2 Metribuzin < , , 2.1 2,4-D , , , , 1.5 2,4,5-T < , 2.4 2,4,5-TP < , 5.2 Fungicides Captan < , 3.5 Fumigants Methyl bromide , , 2.5 Table continues risks associated with the women s use and the increased risks associated with the husbands use were found among the postmenopausal women (table 5). All three pesticides associated with elevated risks when used by the women were organophosphates (chlorpyrifos: RR = 2.2, 95 percent CI: 1.0, 4.9; 2,2-dichloroethenyl dimethylphosphate (dichlorvos): RR = 2.3, 95 percent CI: 1.0, 5.3; terbufos: RR = 2.6, 95 percent CI: 1.1, 5.9). Reduced risks among postmeno-

11 Pesticides and Female Breast Cancer Risk 131 TABLE 5. Continued Exposure Wife s pesticide use among all wives in the cohort Husband s pesticide use among wives who never used pesticides * * * * Postmenopausal women (n = 192) (n = 10,736) (n = 103) (n = 4,700) Any pesticides , , , Chemical classes Carbamates , , , , 1.5 Chloroacetanilides , , , 1.6 Dinitroanilines , , , 2.4 Dithiocarbamates < , 2.0 Organochlorines , , , , 2.2 Organophosphates , , , , 3.7 Phenoxy herbicides , , , , 1.9 Pyrethroids , , 2.2 Thiocarbamates < , , 1.9 Triazines , , , 2.2 Insecticides Aldrin , , , 2.6 Carbaryl , , , , 2.2 Chlordane , , , 2.3 Chlorpyrifos , , , 2.4 DDT , , , 1.7 Diazinon , , , , 2.3 Dichlorvos , , 2.6 Dieldrin < , 3.0 Fonofos < , 2.1 Heptachlor , , 2.7 Lindane < , 2.7 Malathion , , , , 2.3 Parathion < , 2.5 Phorate , , , 1.6 Terbufos , , , 1.4 Toxaphene < , 2.2 Herbicides Atrazine , , , 1.7 Cyanazine < , , 1.4 Metribuzin < , , 1.4 2,4-D , , , , 1.5 2,4,5-T < , , 2.2 2,4,5-TP < , 3.9 Fungicides Captan < , 6.1 Fumigants Methyl bromide , , 1.9 * The numbers of cases and noncases included in the analyses differ across pesticides because of missing data. RR, rate ratio; CI, confidence interval; DDT, dichlorodiphenyltrichloroethane; dichlorvos, 2,2-dichloroethenyl dimethylphosphate; 2,4-D, 2,4-dichlorophenoxyacetic acid; 2,4,5-T, 2,4,5-trichlorophenoxyacetic acid; 2,4,5-TP, 2,4,5-trichlorophenoxypropionic acid. Each pesticide exposure was examined in a separate model adjusted for age (<40, 40 49, 50 59, and 60 years), race (White and other), and state of residence (Iowa and North Carolina). A rate ratio was estimated only if there were at least three exposed and three unexposed cases and three exposed and three unexposed noncases. Organophosphate insecticides: chlorpyrifos, coumaphos, diazinon, dichlorvos, fonofos, malathion, methyl or ethyl parathion, phorate, terbufos, and trichlorfon; organochlorine insecticides: aldrin, chlordane, DDT, dieldrin, heptachlor, lindane, and toxaphene; carbamate insecticides: aldicarb, carbaryl, and carbofuran; pyrethroid insecticides: permethrin or pyrethroid products; triazine herbicides: atrazine, cyanazine, and metribuzin; chloroacetanilide herbicides: alachlor and metolachlor; thiocarbamate herbicides: butylate and S-ethyl dipropylthiocarbamate; dinitroaniline herbicides: pendimethalin and trifluralin; phenoxy herbicides: 2,4- D, 2,4,5-T, and 2,4,5-TP; dithiocarbamate fungicides: mancozeb, maneb, and ziram.

12 132 Engel et al. pausal women were linked to their use of the organochlorines chlordane (RR = 0.5, 95 percent CI: 0.2, 1.0) and DDT (RR = 0.5, 95 percent CI: 0.2, 0.9) and to their use of atrazine (RR = 0.4, 95 percent CI: 0.1, 1.0). In contrast, risk was increased by percent among postmenopausal women whose husbands used the organochlorines aldrin, chlordane, dieldrin, and heptachlor and the organophosphates chlorpyrifos, diazinon, and malathion; increased risks were also related to the husbands use of 2,4,5-TP (RR = 2.2, 95 percent CI: 1.3, 3.9) and captan (RR = 3.6, 95 percent CI: 2.1, 6.1). Lindane use by the husbands was associated with an increased risk among both pre- and postmenopausal women. We found evidence of exposure-response trends in relation to cumulative use of certain pesticides by the husbands. Compared with women whose husbands did not use dieldrin, the rate ratios associated with low and high cumulative dieldrin use were 1.4 (95 percent CI: 0.6, 3.5) and 3.2 (95 percent CI: 1.3, 8.0), respectively (p for trend = 0.002). The corresponding risk estimates for 2,4,5-TP use were 1.5 (95 percent CI: 0.7, 3.5) and 4.7 (95 percent CI: 2.2, 9.6) (p for trend < 0.001), which were similar in both states. An apparent trend was seen for use of 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), with rate ratios of 1.2 (95 percent CI: 0.7, 2.1) for low cumulative use, 2.0 (95 percent CI: 1.0, 3.9) for medium cumulative use, and 2.2 (95 percent CI: 1.2, 4.3) for high cumulative use (p for trend = 0.009). Too few husbands provided the information necessary to evaluate trends for captan. There were no apparent trends in relation to the husbands cumulative use of other pesticides. Other evidence that suggested increasing risk with increasing exposure was the higher risk estimate associated with methyl bromide use by the women when analyses were restricted to women who had applied pesticides for at least 10 years (RR = 3.2, 95 percent CI: 1.2, 8.7 (six exposed cases)) or at least 40 days in total (RR = 2.3, 95 percent CI: 0.9, 5.8 (six exposed cases)). In addition, when analyses were restricted to women on farms with at least 500 acres planted, risk estimates increased in relation to the husbands use of heptachlor (RR = 2.7, 95 percent CI: 1.3, 5.7 (17 exposed cases)) and phorate (RR = 1.9, 95 percent CI: 1.0, 3.8 (21 exposed cases)). The risk of breast cancer related to a woman s use of diazinon was significantly higher among women with a family history of breast cancer (RR = 1.7, 95 percent CI: 0.9, 3.2 (13 exposed cases)) than among women without a family history (RR = 0.8, 95 percent CI: 0.5, 1.2 (17 exposed cases)), with a p for interaction of A similar pattern was seen for the husbands use of parathion (RR = 4.2, 95 percent CI: 1.6, 10.6 (seven exposed cases) and RR = 0.9, 95 percent CI: 0.5, 1.8 (11 exposed cases), respectively; p for interaction = 0.04) and paraquat (RR = 3.9, 95 percent CI: 1.7, 8.9 (11 exposed cases) and RR = 0.9, 95 percent CI: 0.5, 1.6 (18 exposed cases), respectively; p for interaction = 0.03). We observed no other multiplicative interactions with family history of breast cancer that were consistent between states (data not shown). Results did not materially change when we restricted analyses to women who had grown up on a farm, whose houses were within 100 yards (91.4 m) of fields where pesticides were applied, or who reported frequently washing work clothes used during pesticide application; nor did the results change when we excluded cases diagnosed within the first year after enrollment in the cohort (data not shown). DISCUSSION Although breast cancer risk was not related to overall pesticide use in this cohort, it was increased in relation to the use of several specific pesticides. The strongest evidence of an increased breast cancer risk was seen for the husbands use of 2,4,5-TP, which showed consistent results, including exposure-response trends, between states; however, too few women used this herbicide to estimate the risk associated with personal use. Weaker evidence, including either consistency between states or an exposure-response trend, was found for the husbands use of dieldrin, captan, and 2,4,5-T. Other observed associations lacked consistency and exposure-response relations, and many were of borderline significance. There was no clear association of breast cancer risk with any measure of potential cumulative exposure to all pesticides combined, although risk appeared to be modestly elevated among women with homes closest to areas of pesticide application. In evaluating our findings, the strength, consistency, and dose-response of the observed associations, as well as the biologic plausibility conferred by other human and animal studies, need to be taken into account. Among the organochlorines, the most well-studied pesticide group, pesticidespecific results were inconsistent between states and between the wives and husbands use, although insufficient numbers of exposed subjects precluded some subgroup comparisons. Dieldrin was the only organochlorine to show an exposure-response relation. A Danish study also reported a dose-response trend for dieldrin (51), but other studies found no association (1, 15, 52). The elevated risks associated with certain organochlorine exposures in our study may be due in part to residual confounding by age (since bans or phaseouts of most of these pesticides decreased or eliminated their use by the 1970s or early 1980s), although different methods of adjusting for age in our analyses produced similar results. Several organochlorine pesticides, such as DDT, chlordane, dieldrin, and toxaphene, show hormonal activity in vivo or in vitro (32, 53 55). However, while a number of these chemicals induce tumors in various organ systems in animals, they have not been linked to mammary tumors (56 59). In addition, most epidemiologic studies of organochlorines (primarily DDT and dichlorodiphenylethane (DDE)) have reported no association with breast cancer risk (1 24), although some studies have reported elevated risks (52, 60 69). Results have been inconsistent but largely null in subgroups defined by race (16), body mass index (1, 7, 8, 16), menopausal status (1, 5 8, 10, 11, 13, 14, 16 18, 20 24), family history of breast cancer (7, 8), and estrogen-receptor status of the tumor (1, 4 7, 11, 13, 15 17, 20 24) and have shown no clear pattern by type of specimen used or timing of specimen collection. Sample size and chance may have contributed to the discrepant findings. We are unaware of any epidemiologic studies that have examined specifically the relation between 2,4,5-TP or