Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
Families living on farms live in their work environment and, consequently, have both occupational and environmental opportunities for exposure to pesticides. Sources include pesticides used on the farm and around the home. Depending on the type of pesticide and the exposure pathway, the routes of exposure may be dermal, oral, inhalation, or combinations of these—from direct or indirect contact with the pesticide. Drift of pesticide sprays on farmland and around the home, as well as pesticides brought into the home by the applicator on work boots, clothing, or skin, may contribute to exposure pathways for farm children.1–3 Dermal absorption is considered to be the most frequent route of exposure to pesticides; however, inhalation and ingestion may also contribute to the body burden. Among children, there are important differences in exposure factors related to age such as mobility, behaviors, and activities.4
Biomarkers for a few pesticides have been identified (eg, atrazine, chlorpyrifos, malathion, carbaryl, 2,4-D, dicamba),5 and these markers serve as a useful measure of direct exposure aggregated over all sources and pathways. However, very little biomonitoring information has been collected on children and even less on farm children. What data have been obtained do not reflect short-term exposures but rather general population exposure (eg, NHANES).6 An exception is a study of insecticide residues in the urine of El Salvador farm children that found an association of farm application and parental use with residue levels.7
Epidemiologic studies often rely on self-reported questionnaire data on whether pesticides were used in the time window of interest. We could not locate other studies that compared epidemiologic questionnaire data with biomonitoring data in farm children. The Pesticide Exposure Assessment Study collected both of these types of data. We have previously reported on questionnaire and diary predictors of herbicide residues in body fluids of applicators.8,9 In this article, we outline the questions asked of the father/applicator about herbicide use and potential modifiers of child exposure, and we report herbicide levels in urine of farm children. We then compare the 2 through an analysis of the sensitivity and specificity of questionnaire measures in predicting specific herbicide residues.
Setting and Study Population
The province of Ontario, Canada, has a wide variety of agricultural operations, including livestock, oilseed, grain, fruit, and vegetable farms. The Ontario Farm Family Health Study, initiated in 1991–1992, was used to identify eligible farm families for the Pesticide Exposure Assessment Study. The original research project used telephone interviews of a census of farms in Ontario to identify 2000 reproductive-age couples living on family-run farms, and subsequently collected information on pesticide practices and health outcomes through questionnaires.10–13
We selected the phenoxy herbicides 2,4-D (2,4-dichlorophenoxyacetic acid) and MCPA ([4-chloro-2-methyl]phenoxyacetic acid) as sentinel pesticides. These herbicides are widely used in agricultural and residential settings to control broadleaf weeds and are relatively nonpersistent in the environment, with a biologic half-life in humans that reportedly varies from 12–72 hours.14–17
In early 1996, we recontacted couples that reported using 2,4-D or MCPA in 1991–1992. We asked whether they were planning to use either of these herbicides in the 1996 growing season. Additional eligibility requirements were that their house had to be on the farm property and the husband and wife had to be living there together. Each family was asked to identify any children between the ages of 6 and 15 years; the child whose birth date was closest to the day and month of the interview was invited to participate. For families with no children in this age range, a child who was younger than 6 or older than 15 could also participate in the study. A consent form was signed by the families agreeing to participate, including parental consent for participation by children. Further details on the methods are provided in an earlier paper.8 The study was approved by a committee on research ethics at the University of Guelph in Guelph, Ontario. The field study period was from May through July 1996 when MCPA or 2,4-D was used for the first time during the growing season by the father.
Reports of Pesticide Use and Associated Activities
Throughout the growing season of 1996, the father completed a diary of all pesticides used on field crops. During the evening of the first day of application of the herbicides 2,4-D or MCPA, he completed a day-of-application questionnaire indicating which herbicides had been used on the previous day as well as the current day. Farmers were asked to provide the product name as well as the Pest Control Product Number. This number was linked with a file of products licensed for use in Canada to identify the active ingredients. If the product number was not provided, one was assigned based on the product name.
For each of the herbicide application-associated activities (ie, mixing and loading the herbicide into the sprayer, application of the herbicide, and clean-up of the equipment), the father identified where the participating child was during the activity according to the following categories: helped with the activity; within 10 feet of the activity; outside but more than 10 feet away; inside a building on the farm; not on the farm property; or do not know where the child was.
Each participating family was provided with instructions and supplies for urine sample collections. Once the father knew when he would be applying the herbicides of interest for the first time that growing season, he contacted study staff who arranged a time to pick up the urine samples. The pesticide applicator was encouraged to follow his normal routine in handling pesticides during the study period.
The index child and both parents were asked to provide one void urine sample immediately before the father's application of the herbicides of interest and then 2 consecutive 24-hour urine samples after handling of the herbicides had begun (day 1 and day 2 24-hour urine samples). Each participant used a cooler bag and ice packs to carry their containers with them throughout the day, but we cannot confirm whether a complete 24-hour urine sample was collected. The first day of monitoring for the study was referred to as the “day of application.”
The pre-exposure sample was collected in a 3-L amber high-density polyethylene 24-hour urine sample container, then transferred to a 250-mL Nalgene bottle and placed in the cooler bag. All urine for each of the next 2 24-hour periods was collected in one of 2 3-L amber containers. The time collection was started and the day (1 or 2) was marked on the bottle. Ice packs were rotated to ensure sufficient cooling of the bag. Field staff revisited the farms soon after the collection period to collect the samples and questionnaires. Samples were delivered to the laboratory and stored at 4°C until analysis.
The following herbicides were extracted from acidified urine samples and derivatized to their methyl esters: 2,4-D, MCPA, MCPB [4-(4-chloro-2-methylphenoxy)butyric acid], MCPP (mecoprop) [(RS)-2-(4-chloro-2-methylphenoxy)propionic acid], 2,4-DP (dichlorprop) [(RS)-2-(2,4-dichlorophenoxy)propionic acid], 2,4-DB [4-(2,4-dichlorophenoxy)butyric acid], and dicamba. Extracts were cleaned up and fractionated on Florisil. Concentrated extracts were examined by gas chromatography-mass spectrometry. Field spikes and laboratory quality control samples were included in the study.18 The detection limit for the acidic herbicides and atrazine in urine was 1 μg/L (1 ppb) and 0.2 ng/mL, respectively, which is comparable to that reported in the literature. The atrazine analysis of urine samples was conducted using an atrazine-directed RaPID Assay kit. Atrazine was included in the herbicides analyzed because of its frequent use in Ontario and its relatively persistent nature particularly in aquatic environments. For descriptive statistics, one half the limit of detection was substituted for nondetects.
Because the focus of this analysis was on concentration rather than total absorbed dose of herbicide, urine results are presented as concentration per unit volume. We did not adjust for creatinine, given concerns that this adjustment could introduce additional sources of error.19–21
Multiple linear regression analysis was used to identify major predictors of herbicide residues in the child's urine. We determined sensitivity and specificity by comparing the applicator's questionnaire reports of herbicide use with whether detectable levels of the herbicide of interest were found in the child's urine samples. The sensitivity (ie, the ability to identify correctly those who have the characteristic of interest) was calculated by dividing the number of children with both questionnaire and urinary data indicating exposure by the total number of children who had detectable levels in their urine. The specificity (the ability of the test instrument to correctly identify those who do not have the characteristic of interest) was calculated by dividing the number of children who were not exposed according to both questionnaire and urinary data by the total number of children who did not have detectable levels of the herbicide of interest in their urine samples. All estimates are expressed as percentages.
We identified 773 farmers from the Ontario Farm Family Health Study as individuals who had used 2,4-D and/or MCPA during the 1991–1992 study. After a telephone screening interview with these farmers in early 1996, we identified 329 families as potentially eligible for the Pesticide Exposure Assessment Study based on the eligibility criteria described previously. Approximately 6% of the farmers refused the telephone screening interview, 17% refused the farm recruitment visit, and 6% of the families refused to participate when visited. Thirteen percent of the families could not be contacted. Although 215 families signed informed consent forms to participate in the study, approximately 40% dropped out, primarily because the spring of 1996 was cool and wet, which hampered the use of the herbicides. A total of 126 families participated, including 92 children ranging in age from 3 to 18 years (mean and median, 11 years). Thirty-four percent of the children were girls; the age and sex of 2 of the children were not available.
All of the farmers in the study used a boom-type sprayer to apply herbicides during the study period. According to the day-of-application questionnaire completed by the father, only one child helped with the handling of herbicides during the study period. This child had relatively higher urinary levels of 2,4-D (12 μg/L on day 1 and 100 μg/L on day 2). The child with the highest urinary level of MCPA (45 μg/L on day 2), however, was not on the farm during the mixing of MCPA and was reportedly inside a building on the farm during the other pesticide activities. Approximately 30% of the children were outside during the application of 2,4-D or MCPA, with 18% outside during clean up. Approximately half of the children were not on the farm property during mixing and loading (50%), application (43%), or clean-up (36%).
Mean urinary concentrations of 2,4-D and MCPA were higher in those children living on farms where these herbicides were reportedly used (Table 1). The proportion of children who had detectable concentrations of MCPA increased with greater opportunity for exposure (as measured by the location of the child during herbicide handling activities). Although we cannot verify that all the urine samples were correctly collected, some of the children reported missing or spilling part of their urine sample (17% for day 1 and 23% for day 2). There was no apparent relation of the reported completeness of the urine sample with the concentration of 2,4-D and MCPA measured, the volume of the sample, or the age or sex of the child (data not shown). The volume of urine collected in the 24-hour samples ranged from 0.05–3.2 L.
Concentrations of 2,4-D and MCPA measured in the urine of the children in the study are shown in Table 2. Eleven percent to 14% of the 24-hour urine samples exceeded the 95th percentile reference range of 1.8 μg/L for 2,4-D based on a nonoccupationally exposed adult U.S. population sample (NHANES III).22 Urinary concentrations of pesticide residues were highly skewed (data not shown).
Nontarget herbicide residues were detected in 24% of the children's preexposure urine samples, 37% of the day 1 samples and 36% of the day 2 samples. Atrazine, 2,4-DB, 2,4-DP, and MCPB were rarely found (less than or equal to 2% detectable). Combining all 3 urine samples, 22% of the children had detectable levels of dicamba (mean = 0.7; maximum = 3.0 μg/L) and 7% of the children had detectable levels of mecoprop (mean = 0.6; maximum = 1.8 μg/L). Only 2% of the children had detectable levels of at least 3 herbicides.
The mean urinary concentrations of 2,4-D and MCPA in the 24-hour samples were higher in boys, with no apparent differences by age (Table 3). Multiple linear regression models identified child's sex and parent's mean urinary concentration of the herbicide as important predictors of 2,4-D residues in the child's urine (adjusted R2 = 0.45); for MCPA, parents’ mean MCPA urinary levels were the only important predictors identified (adjusted R2 = 0.19).
According to the day-of-application and diary reports of the applicator (the father of the child), approximately 30% of the children were living on a farm where 2,4-D was used on either the day before or 2 days of sampling; the comparable figure for MCPA was 75%. The sensitivity of questionnaire information in predicting whether the child was exposed to this herbicide (as measured by detectable levels in the average of the 2 24-hour samples) was 47% for 2,4-D but 91% for MCPA (Table 4). The specificity of the questionnaire information was 72% for 2,4-D but only 30% for MCPA. If the child was outside during any of the herbicide handling activities, the figures for sensitivity dropped but specificity improved. (Details regarding false-negative reports of herbicide application are available with the electronic version of this article.)
Children with higher mean urinary herbicide residues (arbitrarily chosen as >1.8 μg/L) were more likely to have parents with higher mean urinary concentrations of that herbicide (Table 5). Boys were consistently more prevalent among the children with higher urinary herbicide residues. Two children had high concentrations of both herbicides.
Fewer than 20% of urine samples from children of farm applicators in this study were positive for 2,4-D or MCPA, and approximately 60% of the urine samples had no detectable level of any of the 8 herbicides measured. These figures are similar to what has been observed in other studies including children6,23–30; however, our study of farm children did observe higher maximum concentrations (100 μg/L for 2,4-D and 45 μg/L for MCPA). Given the high proportion of “nondetects” in our herbicide residue data, the observed differences in urinary herbicide concentrations (eg, arithmetric means) must be interpreted cautiously.
Spot urine samples were collected from family members in the U.S. Agricultural Health Pilot Study and analyzed for 2,4-D and dicamba residues.24 Combining all urinary concentrations for pesticide applicators and their families, the mean value for 2,4-D was higher in the families in whom the pesticide was applied compared with the values for the families from farms where it was not applied (1.1 μg/L in the nonapplication group compared with 8.1 μg/L in the application group). This suggests that these children received more than background exposure to 2,4-D during the application period. The authors speculated that this exposure level may be the result of the children's behavior during outdoor play.5
Reference range concentrations (those concentrations expected in the general population, without occupational exposure) have been estimated for several pesticides based on analyses of urine from 1000 adults who participated in NHANES III.22 No information regarding pesticide use was collected at the time of the urine sampling. For 2,4-D, the mean was <1 μg/L, the median was nondetectable, and maximum value was 37 μg/L. 2,4-D was detected in 12% of the population sampled. No data are available for MCPA.
Although the children in this study were living on farms where herbicides were used, the urinary concentrations of these herbicides indicate that the majority of the children were not exposed either directly or indirectly to the application. Only one child was directly involved in the handling of pesticides and he had the highest measured concentration of 2,4-D in his urine—indicating that proper hygienic handling practices were probably not followed. Because higher herbicide residues were also recorded in some children who were reportedly not on the farm during the pesticide-handling activities, these children are being exposed indirectly, either in other locations or subsequent to the pesticide activity. Children with higher concentrations more often had parents with higher concentrations, which could indicate either transfer of residues to the child (and mother) from the father's activities posthandling (eg, from contaminated surfaces in the home) or possibly genetic similarities among family members in the pharmacokinetics of these herbicides. In either case, these results suggest that parents’ body burdens may be a useful adjunct in assigning exposure likelihood for their children.
The sex differences in urinary herbicide concentrations may be the result of differences in activity patterns, if boys are more likely to be involved in work and play activities that could place them in contact with pesticides. There may also be sex differences in the pharmacokinetics of these herbicides, although this has not been well studied.
In addition to providing data on childhood herbicide exposure, this study also contributes information on the extent of misclassification of exposure in questionnaire-based studies without biomonitoring data. Given the results of our analyses, simply living on a farm (or on a farm where reported use has occurred) is not enough to classify exposure. If sensitivity and specificity are the same for diseased and nondiseased groups, nondifferential misclassification of a dichotomous exposure will bias the true risk toward the null value. Given the low specificity for MCPA measurements based solely on reported use on the farm, there is a high likelihood that someone who is truly unexposed may be classified as exposed (false-positive); however, this figure can be greatly improved by collecting information on whether the child was outside during the pesticide-handling activities. In contrast, sensitivity declines when this additional information is used. The false-negative rate of the questionnaire-based measurement (the probability that someone who truly is exposed will be classified as unexposed) is low for MCPA but may be higher for 2,4-D. Given the smaller number of farms where 2,4-D was used, the differences observed in sensitivity and specificity for 2,4-D and MCPA may be the result of chance or may be the result of real differences in how these 2 herbicides are transported in the environment, or absorbed, distributed, metabolized, or excreted in the body. The formulation of the herbicide product and the composition of inert ingredients may also explain some of the differences observed. The sensitivity and specificity measurements of 2,4-D and MCPA in this analysis of children were similar to those observed for the applicators in this study, except that the specificity of MCPA was considerably higher for the applicators (67% vs. 30% in their children).9
To consider the questionnaire data on reported use as a measure of the exposure to the herbicides, these data should classify a truly exposed person as exposed (sensitivity) with greater probability than they classify a truly unexposed person as exposed (false-positive).31,32 Because this was the intention in this study, the questionnaire items are better than no information at all, although one could argue whether they are sufficiently better to justify their use in epidemiologic studies. There are several issues that should be considered related to sensitivity and specificity. The estimates depend on the cutoff level used (a somewhat arbitrary value), on the distribution of values (with some misclassification as the distribution of values is closer to that of the true-negative, like in this case), on the characteristics of the individuals (such as age), and on the accuracy of the information collected.33
Public concern about the safety of pesticides has been prompted by recent studies reporting associations with cancer in children34–39 and adults.40–44 These studies based their exposure information on questionnaires, and assumed that direct or indirect contact with pesticides implied a dose delivered to the target tissue. As the present analysis has shown, the consequences of this assumption could be a high false-positive rate in classification of exposure. The impact of this kind of error can be profound and has rarely been quantified.45 Until improvements are made in classifying pesticide exposure in epidemiologic studies, results on health effects will be subject to misclassification bias, which makes it difficult to draw valid conclusions about the safety of specific pesticides.
The results of this study themselves have certain limitations. The study population is a rather small and select group of farm families. They may have been more interested in pesticide safety and, consequently, more likely to handle the pesticides properly, with the result of less exposure for the applicator and family members than for those farm families who did not participate. Alternatively, factors such as the physical–chemical characteristics of the application and the dermal absorption rate may have precluded substantial exposure even for people living on farms. Our results may not be generalizable to other pesticides, types of formulations, application methods, populations, or settings (eg, residential).
This study measured pesticide body burden and an index of the quality of questionnaire data shortly after the pesticide activity. Although we have used the limit of detection as the indicator of exposure for the sensitivity/specificity analysis, what is the appropriate benchmark parameter to determine whether biologically relevant exposure has occurred—the limit of detection or some higher value? Furthermore, we collected information on location of the child during herbicide-handling activities only for the first day of the field study and assumed this was similar for the second day (if the father applied the herbicide on that day as well). In addition, no information was obtained on the amount of time the child was outdoors or on the precise location of the child relative to the activity. We assumed that proximity to the activity would result in higher indirect exposure for the child; however, indirect exposure could also occur from contact with contaminated surfaces or materials (eg, doorknobs, household dust). We have also assumed that the collection period (2 consecutive 24-hour urine samples) was appropriate for children; however, the pharmacokinetics of these herbicides may differ by sex, age, or weight, as well as by the chemicals mixed with the pesticide active ingredient (eg, inerts, adjuvants). For many of the health effects studied, it is assumed that longer-term exposure is critical; however, this study did not collect serial measures of exposure over an extended period of time. Unlike retrospective health effects studies in which recall may be a concern, the applicators in this study reported on their pesticide activities within hours of their occurrence, so one would expect this information to be more accurate.
Given the potential for misclassification of key exposure variables in epidemiologic studies of pesticide health effects among children, we urge incorporation of biomonitoring studies in subsets of children to estimate the extent of such misclassification. All measurements of exposure are imperfect, given the considerable variations in metabolism related to genotype and phenotype, as well as differences in how people answer questionnaires. Hence, in any one study, it is preferable to have multiple measures of exposure and, recognizing the heterogeneity that can occur within exposure “types” (eg, agricultural applications using a boom sprayer), as much detail as possible. Further work is also needed to improve the identification of factors that predict children's exposure, as well as factors related to pesticide pharmacokinetics, to determine whether there are age and sex differences among children.
We gratefully acknowledge the Ontario farm families that participated in the study, Christina Bancej for editing and management of the data, the members of the study consultative and advisory committees for their advice on study design, and the thoughtful comments of Claire Infante-Rivard on an earlier version of the manuscript. The atrazine residue RaPID Assay analyses were done by J. Christopher Hall of the Department of Environmental Biology, University of Guelph.
1.Fenske RA, Lu C, Simcox NJ, et al. Strategies for assessing children's organophosphorus pesticide exposures in agricultural communities. J Expo Anal Environ Epidemiol
. 2000;10(6 Pt 2):662–671.
2.Eskanazi B, Bradman A, Castorina R. Exposures of children to organophosphate pesticides and their potential adverse health effects. Environ Health Perspect
. 1999;107(suppl 3):409–419.
3.Gladen BC, Sandler DP, Zahm SH, et al. Exposure opportunities of families of farmer pesticide applicators. Am J Ind Med
4.Adgate JL, Sexton K. Emerging issues: children's exposure to pesticides in residential settings. In: Krieger RI, ed. Handbook of Pesticide Toxicology: Principles
, 2nd ed, vol 1. New York: Academic Press; 2001.
5.Barr DB, Barr JR, Driskell WJ, et al. Strategies for biological monitoring of exposure for contemporary-use pesticides. Toxicol Ind Health
6.Kutz FW, Cook BT, Carter-Pokras OD, et al. Selected pesticide residues and metabolites in urine from a survey of the US general population. J Toxicol Environ Health
7.Azaroff LS. Biomarkers of exposure to organophosphorus insecticides among farmers’ families in rural El Salvador: factors associated with exposure. Environ Res
. 1999;80(2 Pt 1):138–147.
8.Arbuckle TE, Schrader SM, Cole D, et al. 2,4-Dichlorophenoxyacetic acid residues in semen of Ontario farmers. Reprod Toxicol
9.Arbuckle TE, Burnett R, Cole D, et al. Predictors of herbicide exposure in farm applicators. Int Arch Occup Environ Health
10.Savitz DA, Arbuckle T, Kaczor D, et al. Male pesticide exposure and pregnancy outcome. Am J Epidemiol
11.Curtis KM, Savitz DA, Weinberg CR, et al. The effect of pesticide exposure on time to pregnancy. Epidemiology
12.Arbuckle TE, Savitz DA, Mery LS, et al. Exposure to phenoxy herbicides and the risk of spontaneous abortion. Epidemiology
13.Arbuckle TE, Lin Z, Mery LS. An exploratory analysis of the effect of pesticide exposure on the risk of spontaneous abortion in an Ontario farm population. Environ Health Perspect
14.Sauerhoff MW, Braun WH, Blau GE, et al. The fate of 2,4-dichlorophenoxyacetic acid (2,4-D) following oral administration to man. Toxicology
15.Kolmodin-Hedman B, Höglund S, Swensson Å, et al. Studies of phenoxy acid herbicides: II: oral and dermal uptake and elimination in urine of MCPA in humans. Arch Toxicol
16.Kohli JD, Khanna RN, Gupta BN, et al. Absorption and excretion of 2, 4-dichlorophenoxyacetic acid in man. Xenobiotica
17.Knopp D, Glass S. Biological monitoring of 2,4-dichlorophenoxyacetic acid-exposed workers in agriculture and forestry. Int Arch Occup Environ Health
18.Ripley BD, Mena F, Lyttle C. Pesticide Exposure Assessment Study: Pesticides in Urine Part 1—Analytical Results, Methodology and Quality Control.
Pesticide and Trace Contaminants Laboratory, Ontario Ministry of Agriculture, Food and Rural Affairs; February 1997.
19.Alessio L, Berlin A, Dell'Orto A, et al. Reliability of urinary creatinine as a parameter used to adjust values of urinary biological indicators. Int Arch Occup Environ Health
20.Harris SA, Purdham JT, Corey PN, et al. An evaluation of 24-hour urinary creatinine excretion for use in identification of incomplete urine collections and adjustment of absorbed dose of pesticides. Am Ind Hyg Assoc J
21.O'Rourke MK, Lizardi PS, Rogan SP, et al. Pesticide exposure and creatinine variation among young children. J Expo Anal Environ Epidemiol
22.Hill RHJr, Head SL, Baker S, et al. Pesticide residues in urine of adults living in the United States: reference range concentrations. Environ Res
23.Hill RHJr, To T, Holler JS, et al. Residues of chlorinated phenols and phenoxy acid herbicides in the urine of Arkansas children. Arch Environ Contam Toxicol
24.Shealy DB, Bonin MA, Wooten JV, et al. Application of an improved method for the analysis of pesticides and their metabolites in the urine of farmer applicators and their families. Environ Int
25.Aprea C, Sciarra G, Bozzi N. Analytical methods for the determination of urinary 2,4-dichlorophenoxyacetic acid and 2-methyl-4-chlorophenoxyacetic acid in occupationally exposed subjects and in the general population. J Anal Toxicol
26.Aprea C, Colosio C, Mammone T, et al. Biological monitoring of pesticide methods: a review of analytical methods. J Chromatogr B Analyt Technol Biomed Life Sci
27.Lioy PJ, Edwards RD, Freeman N, et al. House dust levels of selected insecticides and a herbicide measured by the EL and LWW samplers and comparisons to hand rinses and urine metabolites. J Expo Anal Environ Epidemiol
28.Adgate JL, Barr DB, Clayton A, et al. Measurement of children's exposure to pesticides: analysis of urinary metabolite levels in a probability-based sample. Environ Health Perspect
29.Minnesota Department of Health. Comparative Risks of Multiple Chemical Exposures.
Final report for the Legislative Commission on Minnesota Resources. Minnesota Department of Health; July 2000.
30.Baker SE, Barr DB, Driskell WJ, et al. Quantification of selected pesticide metabolites in human urine using isotope dilution high-performance liquid chromatography/tandem mass spectrometry. J Expo Anal Environ Epidemiol
. 2000;10(6 Pt 2):789–798.
31.Armstrong BK, White E, Saracci R. Principles of Exposure Measurement in Epidemiology
. New York: Oxford University Press; 1992.
32.Kelsey JL, Whittemore AS, Evans AS, et al. Methods in Observational Epidemiology
, 2nd ed. New York: Oxford University Press; 1996.
33.Szklo M, Nieto FJ. Epidemiology: Beyond the Basics
. Gaithersburg, MD: Aspen Publishers; 2000.
34.Valery PC, McWhirter W, Sleigh A, et al. Farm exposures, parental occupation, and risk of Ewing's sarcoma in Australia: a national case-control study. Cancer Causes Control
35.Daniels JL, Olshan AF, Teschke K, et al. Residential pesticide exposure and neuroblastoma. Epidemiology
36.Meinert R, Schuz J, Kaletsch U, et al. Leukemia and non-Hodgkin's lymphoma in childhood and exposure to pesticides: results of a register-based case-control study in Germany. Am J Epidemiol
37.Buckley JD, Meadows AT, Kadin ME, et al. Pesticide exposures in children with non-Hodgkin lymphoma. Cancer
38.Infante-Rivard C, Labuda D, Krajinovic M, et al. Risk of childhood leukemia associated with exposure to pesticides and with gene polymorphisms. Epidemiology
39.Kristensen P, Andersen A, Irgens LM, et al. Cancer in offspring of parents engaged in agricultural activities in Norway: incidence and risk factors in the farm environment. Int J Cancer
40.Waddell BL, Zahm SH, Baris D, et al. Agricultural use of organophosphate pesticides and the risk of non-Hodgkin's lymphoma among male farmers (United States). Cancer Causes Control
41.McDuffie HH, Pahwa P, McLaughlin JR, et al. Non-Hodgkin's lymphoma and specific pesticide exposures in men: cross-Canada study of pesticides and health. Cancer Epidemiol Biomarkers Prev
42.Settimi L, Comba P, Bosia S, et al. Cancer risk among male farmers: a multi-site case-control study. Int J Occup Med Environ Health
43.Sharpe CR, Siemiatycki J, Parent ME. Activities and exposures during leisure and prostate cancer risk. Cancer Epidemiol Biomarkers Prev
44.Mao Y, Hu J, Ugnat AM, et al. Non-Hodgkin's lymphoma and occupational exposures to chemicals in Canada. Canadian Cancer Registries Epidemiology Research Group. Ann Oncol
. 2000;11(suppl 1):69–73.
45.Greenland S. Basic methods for sensitivity analysis of biases. Int J Epidemiol