Research staff reviewed the 24-hr sampling record with the parents to ensure accuracy and completeness. Urine samples were stored in the sample refrigerator until daily collection by research staff. Trained, bilingual study staff administered daily questionnaires that assessed the child’s exposure to pesticides, including questions regarding dietary intake of fruits, vegetables, and juices; time spent indoors/outdoors; parental occupational exposures; and residential pesticide use over the previous 24-hr period. Study staff processed the samples at the study field office, recording the weight and volume . On 24-hr sampling days, staff were instructed to select the first FMV sample plus one to three randomly selected additional spot samples for individual analysis. The total volume of the 24-hr composite sample was based on the volume of the individually analyzed samples plus the volume of all samples that were included in the pooled sample. The DAP concentrations were based on volume-weighted averages of concentrations in the individually analyzed samples plus the pooled sample. Samples were stored at −80 °C until shipment on dry ice to the Centers for Disease Control and Prevention for analysis in August and September 2004. Laboratory methods and quality control procedures have previously been described in detail and are available in the Supplementary Materials.
Total dimethyl , total DE,best indoor plant pots and total DAP concentrations were calculated within each sample by summing molar concentrations. We computed metabolite levels in 24-hr samples using the volume-weighted average of concentrations in all samples collected in that 24-hr sampling period . In California, all agricultural pesticide use, including crop, active ingredient, date, pounds applied, and location of use within one square mile sections defined by the Public Lands Survey System are recorded in pesticide use reports by the California Department of Pesticide Regulation . We used the latitude and longitude of the participant’s home, geocoded from their street address, to map pesticide applications. We considered pesticide use within three kilometers of the home in the six months prior to each of the two 24-hr urine sampling days for each study participant, as these are within the range of distances and time periods that have been mostly strongly associated with OP concentrations in samples from this region . We included 11 OPs that devolve into DAPs that are used in the Salinas Valley, which is representative of the most commonly used OPs nationally in the same time period . These 11 OPs include eight DM and three DE pesticides. All estimates were adjusted for the proportion of time the residence was downwind of each pesticide application .
At each study visit, study staff asked parents to report whether their child had consumed fresh fruits or vegetables from a 21-item list since the previous visit. Parents were also asked to report their child’s consumption of any fruits or vegetables that were not on the list; canned, jarred, or frozen fruits and vegetables; and orange, apple, or other 100% fruit juice . Each year since 1991, the United States Department of Agriculture Pesticide Data Program has tested food commodities, including fruits and vegetables, for approximately 450 pesticides and their breakdown products . Using a food consumption-chemical residue approach described previously , we used these publicly available data to calculate the mean concentration of the 11 OPs of interest for each of the food items reported in our study.To estimate dietary OP exposure, we multiplied the estimated concentration of the 11 OPs in each food item by the estimated intake of that food item. Per the US EPA Cumulative Organophosphorus Risk Assessment guidelines, we also included omethoate, the dimethoate oxon, in our dietary assessment, however it was not detected on any of the food commodities of interest in 2004. We made the assumption that each reported consumption of a particular fruit or vegetable was equal to one serving and used data for children ages 3–6 years from the 2003–2004 National Health and Nutrition Examination Survey “What we Eat in America” study linked to Food Commodity Intake Database codes to estimate the weight of each food item. We estimated total exposure for each OP by summing estimated intake across all food items. We included reported food consumption that we were certain had preceded the urine void. For 24-hr samples, we considered the average exposure from all produce reported on the current day and previous day .
For spot samples, we considered all produce reported on the day prior to sampling in order to ensure the produce was consumed before the sample was collected. We used USDA pesticide residue data from 2004 , when available. For commodities not analyzed in 2004, we used data from the most proximate year . PDP samples with values <LOD were set to 0.We used generalized estimating equation models using DAP, DM, and DE dose estimates from each 24-hr composite as the outcome variable and dose estimates from same-day spot as the predictor variable. We also used the combination of each same-day FMV and non-FMV spot sample as a predictor variable by computing the arithmetic average of the dose estimate from the individual samples. Missing voids from 24-hr samples were excluded from the analysis, as both the volume of the sample and DAP concentrations were unknown. Analyses were conducted for volume- and creatinine-adjusted dose estimates. All dose estimates were log10-transformed. We assessed the performance of the models for each predictor variable using the predictive power of the model defined as the coefficient of determination ; the root mean squared error , which is a measure of both precision and accuracy of the model; and the intraclass correlation , which measures agreement between the dose estimates. In this study of 25 children living in an agricultural region, we found that volume- and creatinine-adjusted non-FMV spot urine samples had relatively weak ability to predict 24-hr cumulative OP dose. Moreover, our results indicate that reliance on non-FMV spot samples may underestimate daily cumulative OP dose and the percentage of samples exceeding regulatory guidelines, regardless of the method used to account for expected daily urinary excretion. Models including the average of an FMV and non-FMV spot had the greatest ability to predict 24-hr dose, however models containing just an FMV sample were often similarly predictive of daily dose. Our findings are consistent with previous analyses in this population in which we found that spot urine samples had relatively weak ability to predict cumulative exposure over one week and that reliance on spot samples to reflect chronic OP pesticide exposure may result in exposure misclassification that could bias effect estimates towards null findings . Because 24-hr sampling, considered the “gold standard”, or the collection of multiple daily spot samples is infeasible in most epidemiologic studies, we recommend that future studies prioritize the collection of FMV samples to most accurately characterize OP dose. To our knowledge, only two other studies have examined the ability of same-day spot urine samples to predict 24-hr OP pesticide exposure or dose . In a study of 13 2–5 year old children, Kissel et al., analyzed OP metabolite concentrations from urine samples collected during each of two 24-hr sampling cycles in two different seasons and found that FMV samples were the best predictor of weighted average daily metabolite concentration in both creatinine-adjusted and unadjusted analyses . They also observed high intra-child variability in metabolite levels from urine samples collected on the same day . Their findings indicate that full 24-hr sampling may reduce measurement error due to within-person variability, however if spot sampling is to be conducted,blueberry container size collection of FMV samples are preferable for analytes with short half-lives . In another analysis of 20 farmers and their children, Scher et al., analyzed agreement between two OP parent compounds/metabolites and 3,5,6-trichloro-2-pyridinol in morning void samples with 24-hr composite exposure and dose estimates from urine collected between 24 h before through 96 h after pesticide application .
Compared to estimates based on 24-hr samples, investigators found that single morning void urine samples tended to overestimate daily exposure and dose estimates of 2,4-D and chlorpyrifos . More specifically, four children had chlorpyrifos dose estimates above the acute population adjusted dose regulatory level of 0.5 μg/kg/day based on morning void samples, whereas no 24-hr dose estimates exceeded EPA safety thresholds . Taken together with our results, these findings suggest that reliance solely on non-FMV spot samples may underestimate OP dose, whereas analysis of FMV samples alone may overestimate dose. Previous epidemiologic analyses among children living in the Salinas Valley have found DMs to drive associations between urinary DAPs and adverse child neurodevelopment . We observed that DMs had a substantial influence on OP dose estimates and ability of spot samples to predict 24-hr dose. There are a few possible explanations for this. First, of the 11 OPs examined in this analysis, 8 are DMs and only 3 are DEs. These eight DMs had a much higher total molar mass than the three DEs . Second, oxydemeton methyl, a highly toxic DM with a large RPF , increased in use in the Salinas Valley shortly after our study started ) and may be influencing the associations observed in our study and previous epidemiologic analyses from this region. Pesticide use trends have shifted drastically since we conducted this study and some of the most toxic OPs have largely been phased out of agricultural use in the Salinas Valley and across the United States. Additional investigations are needed to examine cumulative OP dose estimates and potential contributions from DEs and DMs with the current mixture of OPs being applied. In addition to the potential influence of specific OPs, it’s possible that DEs are chemically less stable and have higher intrinsic variability than DMs . We found that estimates adjusted for expected 24-hr creatinine had similar ability to predict daily OP dose as estimates adjusted for observed 24-hr creatinine excretion or urine volume. Conversely, in a study of 109 children living in an agricultural area in Washington State, investigators found that creatinine-adjusted doses tended to be lower than those calculated with daily urine volume . Previous studies have found that creatinine concentrations may be highly variable due to factors such as age, sex, BMI, diet, and fluid intake and that correcting for specific gravity may introduce less variability and may be a more robust method in studies focusing on children . Additional research may be needed to evaluate the validity of creatinine correction in children. Furthermore, we recommend that future studies collect urine specific gravity information, particularly given the ease of measuring this metric . This study has multiple strengths and implications for future risk assessments and epidemiologic studies. We extended previous examinations that estimated cumulative OP dose from diet and nearby agricultural pesticide use separately by considering these exposures in conjunction. Additionally, this is one of only a few studies to examine cumulative OP pesticide dose among children living in an agricultural area and to examine the ability of spot samples to predict 24-hr dose. These results have important implications for risk assessments and could be applied to other non-persistent environmental chemicals. This study also has limitations. We did not have specific gravity measurements and could not compare adjustment for urinary dilution using specific gravity. Notably, while DAPs represent exposure to approximately 80% of the OPs used in the Salinas Valley , children may have been exposed to other OPs that do not devolve into DAPs. While California’s unique and comprehensive PUR database allowed us to estimate the mix of pesticides participants may have been exposed to from nearby agricultural pesticide use, relying solely on these data to estimate all non-dietary exposures may not adequately account for all sources and pathways of exposure. We examined agricultural pesticide applications near participants’ residences in the six months prior to each 24-hr sampling in order to tiy to account for exposures from multiple sources, including agricultural drift and accumulation of pesticides in the home , however participants may have also been exposed to pesticides via the take-home exposure pathway, particularly if they lived with farm workers . However, because the dose calculations incorporate the proportion of potential exposure to each pesticide in relation to total DEs and DMs applied, rather than a sum of each pesticide, and because we anticipate that children living with farm workers were likely exposed to a similar mixture of OPs from para-occupational exposures, we do not believe that this impacted our results substantially.