Data used to construct exposure estimates or classifications can be drawn from a variety of sources, ranging from general information regarding pesticide use to personal measurements. presents the approaches taken in the studies under review. The first two columns provide source information and environmental measurement methods; the remaining columns categorize various types of exposure samples collected according to age, because different sampling strategies are more or less practical and valuable within these time frames. indicates the analytes measured in five biologic sample matrices collected in these studies.
Exposure data collected in reviewed studies of children’s exposure to pesticides through 2003.
Measured analytes in five biologic sample matrices.a
Pesticide source information.
Virtually all children’s exposure studies collect historical and contemporaneous information regarding pesticide use. In most cases, these data are collected through parental questionnaires or interviews and pertain to pesticides in and around the residence. In general, we have found that parents are best able to provide general information regarding the use of products (e.g., control of particular insects, control of weeds) but may not be able to provide detailed information on specific chemicals (Lu et al. 2004
; Whyatt et al. 2002
). In preliminary analyses of questionnaires administered by the Columbia center, women provided a pesticide product name for only 39% of the pest control methods reported to be used in the home during pregnancy and, in particular, were rarely able to identify the pesticide products used by an exterminator. Further, pesticide products can have the same brand name but contain different active ingredients, further complicating use of questionnaire data in pesticide exposure assessment.
Investigators for most of the reviewed studies have thus gone a step further to visually inspect the pesticide products in the home, sometimes referred to as a pesticide inventory. For example, study staff from the Berkeley center recorded the U.S. EPA registration number and the active ingredients on the label of each home pesticide. The registration number was later entered into a pesticide product database maintained by the California Department of Pesticide Regulation to confirm all active ingredients. Records of commercial pesticide applications can also be accessed during home visits (Berkowitz et al. 2003
; Whyatt et al. 2003
Identification of specific products can be very helpful in determining whether or not a particular class of chemicals has been used in the residence and may inform subsequent sampling plans, but the presence or absence of specific products does not generally enter into the development of an exposure metric for the residents. Frequency of residential pesticide use could be used potentially to sort children into exposure categories, but such an approach has not been fully validated. One study has shown that personal air levels of organophosphate (OP) pesticides were significantly higher among women who reported using exterminator sprays, can sprays, and/or pest bombs during pregnancy compared with those reporting no OP pesticide use (Whyatt et al. 2002
). Another study demonstrated that children whose parents reported garden use of insecticides had higher levels of OP pesticide metabolites than did children whose parents did not use garden insecticides (Lu et al. 2001
Food can be an important source of pesticide exposure for children, but most of the studies reviewed here have not devoted substantial resources to an evaluation of the dietary pathway. The Mount Sinai center obtained maternal prenatal dietary food frequency data during pregnancy only, with specific information about fish consumption. The Berkeley center also obtained a detailed prenatal food frequency questionnaire. Additional information was also obtained on fruit and vegetable consumption for the pregnant women and, later on, for their children. The Berkeley center and the PNASH center have collected duplicate diets from a relatively small number of children (Fenske et al. 2002a
). Such an approach provides very useful quantitative information on exposure but is extremely time-consuming and expensive. A diet diary has also been used to distinguish children whose intake of fresh produce and juices was primarily organic and proved effective in classifying children’s OP pesticide exposure (Curl et al. 2003a
Studies of children of agricultural workers have focused on potential paraoccupational exposure, collecting data on the transmission of pesticides from the workplace to the home by parents or other adult household members, as well as data on residential proximity to pesticide applications (Bradman et al. 1997
; Curl et al. 2002
; Eskenazi et al. 2003
; Koch et al. 2002
; Lu et al. 2000
; Simcox et al. 1995
). Results to date indicate that both of these pathways can contribute to children’s exposures in agricultural communities and would need to be considered in the design of a study that included rural populations. Studies at the Berkeley center have taken advantage of California’s unique Pesticide Use Reporting system, and researchers there are investigating the use of these data as predictors of pesticide exposure in their cohort (Castorina et al. 2003
). The Washington center completed a 2-year intervention to reduce take-home exposure in 2002; the Berkeley center is currently conducting a similar intervention.
House dust samples have been collected in most of the reviewed studies and have served as a reliable indicator of residential pesticide contamination (studies conducted at the PNASH Center), although not necessarily as a surrogate for children’s exposures (Curl et al. 2002
; Fenske et al. 2002b
; Lu et al. 2000
; Simcox et al. 1995
). A practical problem can arise when insufficient dust is available for analysis, as was the case for the Mount Sinai studies. In the Berkeley center studies, the average mass of 509 dust samples was 9 g/m2. The average of the fraction < 150 μm in diameter used for chemical analyses was 7 g/m2. About 20% of the samples had a fine fraction of < 0.5 g total. Most laboratory methods for pesticides require 0.5–2 g dust. It is likely that only a single chemical analysis will be possible for a significant fraction of homes, thus limiting future tests for other chemicals. The Berkeley, Mount Sinai, and PNASH centers have investigated alternate methods of measuring pesticide concentrations in child environments, such as indoor air and surface wipe sampling (Lu et al. 2004
). A protocol that is currently being validated involves mailing study participants an alcohol wipe with instruction for wiping dust on the top of a specified doorframe. The sample is then placed in a resealable plastic bag and mailed back to the study team. Advantages include low cost of sample collection and low participant burden. However, research is currently ongoing to determine detection limits and detection frequencies using this method.
The Columbia center has conducted extensive indoor air sampling. For chlorpyrifos and diazinon, the correlation between 48-hr personal air samples collected from the mother during the third trimester and average 2-month indoor air levels over the final 2 months of pregnancy were strong (r > 0.7, p < 0.001) (Whyatt et al. 2003
). Air and dust levels were not significantly correlated in a pilot study conducted by the Mount Sinai group; this may have been due to the very small amount of dust collectable in these homes (Markowitz S, personal communication). In addition to the OP pesticides several carbamates and pyrethroids have been measured in personal air samples collected from the mother over 48-hr during pregnancy (Whyatt et al. 2002
Evidence of chemicals in a child’s environment does not necessarily provide the basis for a sound exposure metric. Dust, wipe, and indoor air measurements (including personal air samples) have not shown strong associations with biologic measurements (Curl et al. 2002
; Whyatt et al. 2003
). It is not clear whether the lack of strong associations is due to confounding factors (e.g., dietary exposure), to variability in the biologic measurements (including toxicokinetic considerations (discussed below), or to a relatively weak link between residential contamination and child exposures.
Environmental monitoring in these studies has focused almost exclusively on the home or residential setting and has not yet been extended to child care centers and schools. The Washington studies have included wipe sampling and dust sampling of commuter vehicles of workers to document the movement of agricultural pesticides from the workplace to the home (Curl et al. 2002
; Lu et al. 2000
Hand wipe sampling.
Initial attempts to look at direct child exposures have included the use of hand wipes to collect pesticides from children’s hands. These methods include wiping the child’s hand with sterile gauze dressing pads that have been moistened with isopropanol, or asking the child to place a hand in a bag containing isopropanol (Bradman et al. 1997
). Gordon et al. (1999)
found excellent correlations between chlorpyrifos in indoor air and corresponding dermal wipes but poor correlations between chlorpyrifos in dust and dermal wipes. Another study reported weak associations between OP pesticide concentrations in hand wipes, house dust, and urinary levels of OP metabolites (Shalat et al. 2003
). The Columbia center conducted hand wipes but found all samples to be less than the limit of detection.
Other techniques for assessing children’s dermal exposures include use of clothing dosimeters such as cotton gloves, union suits, and socks (Fenske 1993
; Lewis 2005
). The Berkeley center has experimented with clothing dosimeters in recent studies. Infants (children 6 and 12 months of age) wore precleaned cotton socks and union suits for several hours in their residential environments.
Personal air sampling has been used effectively to monitor maternal exposures during pregnancy by Columbia researchers (Whyatt et al. 2002
). Investigators used motion detectors to determine whether or not the women complied with the request to carry the personal air monitors; motion detectors were installed in the backpacks of randomly selected women. Results were obtained from monitors worn by 113 women for approximately 48 hr each. For the average woman, nearly 95% of the total number of motion detections occurred during waking hours. In addition, 98% of the women self-reported that the air monitor was near them for least 40 of the 48 hr of the personal air monitoring.
This study (Whyatt et al. 2003
) also found that levels of several OP and carbamate pesticides measured in the 48-hr personal air samples were significantly correlated with levels in 2-week indoor air samples, indicating that, at least for these pesticides, the 48-hr air samples provided a reasonable estimate of exposure over a longer period during pregnancy. In addition, there was little variability in indoor air levels of the insecticides, and the correlations between each of the insecticides in each of the 2-week air samples were highly significant. In cases where sampling bracketed an application event, it is likely that high levels would be observed initially, increasing temporal variability.
Blood samples have been collected throughout pregnancy to assess body burden of pesticides in the Berkeley, Columbia, and Mount Sinai center studies. No association was seen between insecticide levels in maternal blood collected at delivery and maternal self-reported pesticide use during pregnancy in one study (Whyatt et al. 2003
). Weak correlations were seen between pesticide levels in the maternal personal air samples collected during pregnancy and in blood samples collected at delivery (r = 0.10–0.19). However, the correlations were generally stronger when analyses were restricted to women for whom the personal air sample was collected within a month of collection of the blood samples at delivery (r = 0.13–0.45). Maternal and umbilical blood insecticide levels (chlorpyrifos, diazinon, the propoxur metabolite 2-isopropoxyphenol, and bendiocarb) at delivery were highly correlated, indicating that the pesticides are readily transferred to the fetus during pregnancy. Significant inverse associations were seen between chlorpyrifos in umbilical cord blood and both birth weight and length, whereas no association was seen between chlorpyrifos in maternal personal air samples and the same measures of fetal growth (Whyatt et al. 2004
). These results suggest that the biomarkers may better reflect exposure from all routes, not only the amount of insecticides absorbed by the mother but also the amount of the absorbed dose that has been transferred to the developing fetus (Whyatt et al. 2004
Urine samples have also been collected from women during pregnancy in several studies. Investigators at the Berkeley center found that pesticide metabolites in samples collected in the first and third trimester were not correlated. Within-person variability was approximately two times higher than between-person variability, suggesting that more urine samples collected during pregnancy would improve exposure classification (Eskenazi et al. 2004
). A moving estimate of the coefficient relating dimethyl OP metabolite levels to shorter gestation was used to show that exposures in later pregnancy may be associated with shorter pregnancies. Blood cholinesterase levels were inversely correlated with gestational duration, consistent with findings for dimethyl OP pesticide metabolites, although no significant correlation between blood cholinesterase and urinary metabolite levels was observed.
The Mount Sinai center collected urine samples in the third trimester of pregnancy and found that approximately 70% of the women in the cohort had been exposed to pesticides, but no associations were found between these biologic levels and pesticide questionnaire data (Berkowitz et al. 2003
). In a preliminary analysis of data from the Columbia center, weak but significant correlations were seen between average chlorpyrifos and diazinon levels in indoor air samples collected over the final 2 months of pregnancy and their respective metabolites in urine samples collected biweekly from the mothers over the same time frame.
In summary, it is unlikely that questionnaire data alone can prove adequate for exposure classification of women during pregnancy. However, it appears that systematic monitoring through personal air sampling and biologic monitoring in combination with questionnaire data would yield useful exposure data for epidemiologic investigations.
Several novel sampling methods are under development to determine perinatal exposure levels, including sampling of amniotic fluid, meconium, and cord blood. A pilot study from the Berkeley center of 100 amniotic fluid samples, slated for disposal after amniocentesis, were analyzed for a number of pesticides and their metabolites, including the OP pesticides (Bradman et al. 2003
). Target analytes were detected with frequencies ranging from 5 to 70%. Levels were low compared with levels reported in urine, blood, and meconium. Because of risks to the fetus, amniotic fluid typically can be collected only when medically indicated amniocenteses are conducted, usually around 18–20 weeks of gestation, or during scheduled cesarean sections. Therefore, the population sampled will not necessarily be representative of a larger population of pregnant women. For women already undergoing this procedure, the collection of amniotic fluid for research purposes is noninvasive and causes no additional risk.
At the Columbia center, meconium samples were collected from 20 newborns and analyzed for OP pesticide metabolites (Whyatt and Barr 2001
). Detection frequencies were very high for some of these analytes, but others were not detected. Metabolite levels were similar to those seen in adult urine in population-based research. Metabolites were stable at room temperature over 12 hr. These initial results indicate that the measurement of pesticide levels in meconium has promise as a bio-marker of prenatal exposure.
Cord blood has been sampled in three studies. Mount Sinai center investigators collected cord blood for enzyme, lead, and gene analyses. The Mount Sinai group relied on hospital staff for cord blood retrieval, with prenotification of impending delivery and a note on the chart, with the result that 59% of the participants’ cord blood was obtained. Columbia center investigators reported that successful collection of these samples required that a member of the research staff team follow the progress of the labor, go to the labor room before delivery to remind the delivery room staff that the woman is in the study, and assist with the sample collection. Umbilical cord blood was obtained by syringing the blood into heparinized syringes at the point the cord enters the placenta. To date, a cord blood sample has been obtained from 81% of the infants in the study. An average of 29 mL (range, 2–58 mL) was collected per delivery, with > 22 mL collected in 75% of deliveries and ≥30 mL collected in 50% of the deliveries (Whyatt et al. 2003
). The Berkeley center investigators reported a similar proportion of cord blood samples collected and found that successful collection of cord blood required close cooperation with hospital staff to develop procedures that eliminated risks of inadvertent sticks (Eskenazi et al. 2003
In summary, the perinatal sampling procedures described here are in the early stages of development and will need additional study and validation. However, they hold promise for collecting quantitative exposure data at a critical stage of child development.
Infant and toddler exposure.
Traditional urine bags have been used in clinical settings and have proven useful for pesticide-related studies in children (Royster et al. 2002
). The Berkeley center has been successful collecting urine from children 6–24 months of age who were not toilet trained. Urine was collected by applying pediatric urine bags to the children during office or home visits (Eskenazi et al. 2003
). When children were not able to produce a void during scheduled contacts, study staff trained parents to apply the urine bag at home and to then place the urine in a clean cup provided to them. The parent was instructed to call the field office as soon as the void was produced, and study staff then retrieved the sample.
Cotton inserts have also been used to recover urine from diapers (Hu et al. 2000
). However, the most promising development for sampling infants and toddlers who are not yet toilet trained appears to be extracting the metabolites from the diaper gel matrix, although this method still needs to be evaluated for multiple groups of pesticides (Hu et al. 2004
Preschool children’s exposure.
Urine samples have been collected in nearly all studies of pesticide exposure among preschool children. Urine samples have been analyzed for common metabolites, such as the dialkylphosphate (DAP) compounds or for compound-specific metabolites [e.g., 3,5,6-trichloro-2-pyridinol (TCPy) for chlorpyrifos]. Major exposure assessment issues of concern are duration of collection (spot samples vs. 24-hr samples) and frequency of sampling.
Collection of single urine voids, often referred to as spot urine samples, has been selected as a primary sampling strategy for several practical reasons. The burden it places on study participants is relatively low, and sample processing and analysis are manageable and affordable. However, several studies have now determined that pesticide metabolite concentrations in children’s spot urine samples can exhibit high intraindividual (within-child) variability (Adgate et al. 2001
; Koch et al. 2002
). In studies in which it is possible to collect only a single urine sample per day, the first morning void is preferred, because the urine is more concentrated, the collection period is longer (usually > 8 hr), and it appears this sample is most representative of the daily average (Kissel et al. 2005
). Collection of repeated spot urine samples during a single day or over several days is one means of addressing the issue of intraindividual variability. These repeated measures can be averaged to produce a more stable estimate of exposure and would allow evaluation of exposures during specific windows of vulnerability.
Collection of complete 24-hr urine samples has become a standard part of many occupational exposure studies but has generally been viewed as impractical for small children. Several studies reviewed here have attempted to collect 24-hr samples but have been only partially successful. A recent study (Kissel et al. 2005
) of 25 children in a low-income, low-literacy population by the Berkeley center provided intensive training of participants, detailed record keeping by participants, use of small refrigerators, and daily contact by research staff to improve compliance; it was estimated that 28% of participants provided complete samples, an additional 12% were likely complete, 52% missed one or two voids, and 8% likely missed more than two voids.
Several of the centers have also collected blood samples from children postnatally. The Columbia center has employed a pediatric phlebotomist to draw blood when children came to the center for the developmental assessment. Samples were collected from 98% of the children that were seen. However, volumes were generally low (an average of 6.8 mL collected at 24 months and 6.2 at 36 months). The Berkeley center hired a pediatric phlebotomist to collect blood for both state-required lead screening and the CHAMACOS (Center for Health Analysis of Mothers and Children of Salinas) study, increasing the rate of blood collection. Repeat blood samples can be collected from young children but are more difficult to obtain than are urine samples.
Children’s activities are an important variable in assessing pesticide exposure. The Berkeley center has used a visually based, low-literacy child activity time line for parents to record child activity and location. The University of Washington center and the PNASH center have employed miniaturized global positioning system (GPS) units to produce detailed documentation of children’s time–location patterns (Elgethun et al. 2003
). Recent studies have found that time–location diaries kept by parents produce relatively poor agreement with the GPS measurements, suggesting that such diary data would result in substantial misclassification. The GPS analysis has also shown that transient peak exposures can occur both temporally and spatially and that such exposures are not adequately captured within the resolution of most microenvironmental analysis studies.
School-age children exposure.
Sampling procedures for school-age children are similar to those described above for preschool children. However, as children reach school age, they are more likely to be able to participate more actively in studies. They may be able to assent to study procedures, wear personal sampling devices, collect more complete urine samples, and provide helpful information regarding pesticide sources and their own activities. Here we would stress greater emphasis on personal sampling devices to improve the quality of exposure data for this age group.
The PNASH center has explored the feasibility of saliva sampling for pesticides in both workers and children (Denovan et al. 2000
; Lu et al. 2003
). Current saliva sample collection methods require that children chew on a cotton or synthetic plug for approximately 2 min. The plug, containing up to 2 mL of saliva, is then placed in a vial for storage. The plug is similar in size to a dental sponge and could pose a choking hazard to children < 3 years of age. The Berkeley center has experimented with pipettes to directly transfer saliva from a child’s mouth into a collection container. Sample volumes, however, have been < 1 mL. In rare cases, children have spit directly into a beaker. It is not clear that these techniques provide an adequate or appropriate saliva sample for pesticide analysis.
Participation of cohort members in environmental and biologic sampling.
Collection of an array of biologic and environmental samples from women during pregnancy and soon after birth places a burden on study participants and may lead to attrition regarding participation in the exposure assessment component of these studies. and provide data from the birth cohort studies reviewed here to indicate what might be anticipated in the National Children’s Study. Sample sizes are presented for each study and for each relevant time category; the percentage of enrolled study members is then provided for each of the biologic or environmental samples. It is important to recognize that not all of the rates in and are directly comparable. For example, the Berkeley study accepted all eligible enrollees with no condition that they participate in every exposure assessment event; in contrast, enrollment criteria for the Columbia study included collection of a cord blood sample from each participant at delivery. Participation in environmental and biologic sampling tends to drop over time and can be relatively low for certain types of samples. Factors contributing to low participation include reliance on delivery staff, emergency deliveries, inability to schedule appointments that include both parents, mobile populations that are hard to track, and the absence of children from the home at the time of visits by study staff. Participation can also be enhanced; for example, the Berkeley center saw an increase from 64 to 81% between 12 and 24 months for child blood samples because of the hiring of a child phlebotomist who went to each home.
Percent participation of cohort members in biologic sampling procedures in four birth cohort studies.a
Percent participation of cohort members in biologic sampling procedures in three birth cohort studies.a