In a subset of the children for whom all three types of addresses were all available for analysis, we found generally higher point estimates using diagnosis address compared to those estimated using LexisNexis addresses. However, among children for whom we did not have LexisNexis addresses available, we observed the strongest effect estimates based on birth and diagnosis address compared to those with LexisNexis data as a source. The reason for the largely null associations estimated using LexisNexis addresses is that pesticide exposure prevalence increased for both cases and controls compared with the prevalence based on birth and diagnosis addresses . This possibly due to the change in addresses in the early life of children such that more than one address applies and this seems to increase the chance of being exposed at the residential locations. There are several possible scenarios that may lead to such a result. Early life exposures that have been estimated based on birth addresses may represent prenatal exposure levels prior to birth but not early life exposures if the child moved after birth. While it has yet to be determined what the critical windows are for the development of childhood brain tumors, a study of childhood leukemia tried to distinguish between pre-pregnancy, pregnancy and postnatal exposures as critical windows for pesticide exposure, and exposure during pregnancy seemed to have a higher impact than other periods , although the findings do not necessarily transfer to brain tumors. However, in this study,plastic grow bag we primarily focused on early childhood exposures as the susceptible window and did not address prenatal exposures that future research needs to examine.

Diagnosis address of cases possibly represents exposures closer to cancer diagnosis, indicating a different time window. The ‘two-hit’ hypothesis could possibly explain the elevated ORs when exposures were assigned according to diagnosis address: if a second hit is required for cancer initiation, exposures during that period may also have substantial impact on the development of cancer. It is not known whether the ‘two-hit’ hypothesis is relevant for brain tumors, however. Another explanation could be the inaccuracy in reporting of birth and diagnosis address. Although a literature search did not yield validation studies for birth and diagnosis addresses on birth certificates and in Cancer Registry records, it is feasible that families sometimes deliberately misreport their home address at birth due to concerns about legal status or inability to pay for hospital services .Yet, LexisNexis addresses may also have errors, because there are many cases with multiple addresses for the same time period, inaccurate or missing ‘first seen’ and ‘last seen’ dates associated with each address, and inadequate reporting in earlier years. If LexisNexis-based exposures estimates accounting for residential mobility of children’s families are considered to represent exposure levels more accurately and thus would provide effect estimates that are closer to the true associations, then using a single address for each child may underestimate cases and controls’ exposures differentially, for example, if case families tend to move more or closer to the fields more than control families. Of course, besides all other possible explanations, there might be random error in effect estimates because of small numbers of exposed cases in our analysis. Comparing to the subset of children for whom all three addresses were available, the subset with missing LexisNexis information generally showed elevated ORs based on either birth address or diagnosis address. A few scenarios may explain this difference.

According to Table 4.1, the former subset with LexisNexis data represents mothers of older maternal age, higher maternal education and SES, and more Whites, compared to the latter subset. Thus, a likely explanation is that those without LexisNexis records represent a group of individuals with higher pesticide exposure levels in general, insofar they are lower income, lower education, and more Hispanic. Exposure difference exists indeed when we compare exposure prevalence of cases and controls in these two subsets confirming the higher exposure prevalence amongst the subset without LexisNexis. Another possible scenario involves unmeasured variables that may have confounded the associations. In our multivariate adjusted models, we have included most established confounders and confirmed that the other potential confounders such as parity, payment for prenatal care, and neighborhood SES did not change our model results substantially. Thus, uncontrolled confounding would be unlikely in this situation. However, since this subset without LexisNexis addresses has very similar demographic characteristics as those with higher residential mobility shown in earlier studies as well as our own findings, it is also possible that those children who moved into or within rural areas might have greater risks of early life exposures to other rural exposures including other pesticides or to livestock and zoonosis that maybe relevant for childhood brain tumors. As previously mentioned, if diagnosis address reflects true exposures prior to cancer diagnosis, and given that in the group without LexisNexis address exposures prevalence was particularly high in the cases this would explain elevated ORs. However, since controls’ addresses at the reference date are unknown in the current study, we cannot assume that their exposure levels remain the same throughout their early life. To our knowledge, this is the first population-based record-linkage study to examine the impact of residential mobility in early childhood on the associations between agricultural pesticides and childhood brain tumors using a public records database for addresses.

We quantitatively assessed the magnitude of potential bias introduced by exposure misclassification using a real datasets, while the only previous study that examined the degree of bias resulting from misclassification due to residential mobility in pregnancy used simulated data . Another contribution of this study is that we examined potential bias introduced by restricting to children with available LexisNexis addresses and provided some cautions to conducting environmental epidemiological studies that rely on LexisNexis records only. Nevertheless, the coverage of LexisNexis records has been improving substantially over the years as shown in Table 4.1, providing future large record-linkage studies with a possibly valuable source of residential addresses besides the ones routinely collected on birth certifies and cases diagnosis records. This study has a few limitations. A major issue with our current data is that we were not able to obtain a full residential history as a “gold standard” for everyone in this study, limiting our ability to examine the full scope of the impact of residential mobility on effect estimates for the entire study population. Although LexisNexis records have a promising potential for the future research because they became more complete over time, current studies trying to track down child-bearing age women may still face some challenges. On top of that, reconstructed residential histories from LexisNexis records are by no means the “gold standard”. They have several intrinsic limitations including providing multiple addresses for the same time period,PE grow bag inaccurate or missing ‘first seen’ and ‘last seen’ dates associated with certain addresses , and time-varying sources of addresses over years, therefore limiting its usefulness for large-scale record-based epidemiological studies. In addition to concerns regarding the quality of reconstructed residential histories from LexisNexis records, our approach assumes that infants or young children spent most of their time at home during the day when pesticides are sprayed on nearby fields, while they could be in the day care centers or kindergartens. Although childcare centers are usually near homes, this could still introduce exposure misclassification when a smaller buffer such as 500m or 1km is used. Using a single source of address to assess children’s early life exposure to pesticides does not account for their residential mobility and may lead to exposure misclassification that may even be differential as well as non-differential. Databases with public records could be used to augment routinely collected residential information from birth certificates and cancer registries but researchers should be cautious about the potential selection bias introduced by the availability and accuracy of such data sources.

This dissertation first examined prenatal exposures to pesticides known or suspected to be reproductive toxicants in relation to adverse birth outcomes among women living near agricultural fields in California, and found that first and second trimester exposures to most selected pesticides were associated with preterm delivery but the evidence for term low birthweight was limited. Our findings corroborate previous evidence suggesting early or mid-pregnancy is the critical period of fetal development and underscore the importance of protecting pregnant women from exposures to pesticides. Synthetic pesticides have been a double-edged sword in the past century. While they control the harm caused by pests and improve productivity of agricultural crops, many compounds are only tested for its environmental safety in the laboratory or in field trials, but their impact on human health is unknown. Earlier pesticides have been constantly linked to chronic adverse effects in later research and eventually phased out or replaced with less toxic and more selective agrochemicals. Because of the trade-off between the health and environmental effects and the need to produce food , it is a long-term process for national and international agencies to collect evidence for the harmful effect of pesticides and requires a comprehensive search for literature to conduct scientific review to ban toxic pesticides. Our study makes a significant contribution by providing knowledge of susceptibility of fetal growth to maternal pesticide exposure, for the references of public awareness, future research, and health policy making. In the latter part of this dissertation, we then assessed patterns of residential mobility and examined the impact of mobility on the first year of life exposure measures for agricultural pesticides. We also estimated and compared the effect estimates for pesticide exposure duringchildren’s early life based on birth residence and/or diagnosis address with those estimated using reconstructed residential history based on a public records database LexisNexis, using childhood brain tumors as an example. We highlight the importance of accounting for residential mobility in estimating environmental exposures during children’s early life and provide new information on the application of LexisNexis records in augmenting existing address information and estimating environmental exposures for large record-linkage epidemiological studies of childhood health outcomes. However, researchers should be aware of the potential selection bias introduced by the availability and accuracy of such data sources, because individuals with lower socioeconomic status or underrepresented are likely those with sub-optimal living environment and having higher risk of being exposed to agricultural pesticides and therefore subject to chronic adverse outcomes more than other groups. Recent literature has been increasingly focused on the health effects of pesticide exposure including pregnancy outcomes, neurodevelopment, and cancer in lower SES population, or among Latino migrant farm workers and their offspring throughout the US and suggests that they may also be more vulnerable to pesticide exposure because of poor housing condition and insufficient protection from occupational hazards . Furthermore, our findings suggest that some population subgroups living in agricultural regions, whether or not they are actively farming, may represent potential high-risk populations. Future studies should pay more attention to these vulnerable population groups, develop better strategies to obtain accurate residential information or synthesize data from different sources to provide more accurate exposure and effect measures for agricultural pesticides.California has the largest and most complex agricultural labor market in the United States, reflecting seasonal employment demands, the predominance of immigrant workers and the significant role of labor contractors in matching workers and jobs. Whether measured in sales, production or acres, California agriculture expanded in the 1990s . Farm sales reached $27 billion in 2000, with about 77 million tons of crops produced on 8.8 million acres. More than half of these sales were in fruits and nuts, vegetables and melons, and horticultural specialties , such as flowers and nursery products. Rising yields gally, and that those Mexicans who come into the country do so with proper documents. Regularization does not mean rewarding those who break the law. Regularization means that we give legal rights to people who are already contributing to this great nation.” President George Bush agreed: “When we find willing employer and willing employee, we ought to match the two. We ought to make it easier for people who want to employ somebody, who are looking for workers, to be able to hire people who want to work” .