| This dissertation presents a combination of three approaches designed to minimize exposure misclassification and improve risk estimation for studies of disinfection by-products (DBPs) and prenatal development. First, as a means to reduce misclassification in exposure assessment, we conducted a feasibility study to identify simple methods for a priori selection of sites with low spatial variability. These methods were tested upon quarterly DBP concentrations from the distribution system of 198 facilities, listed in USEPA's Information Collection Rule. We classified spatial variation of trihalomethanes (THMs) within these distribution systems using two methods: (1) based upon a two-way analysis of variance, and (2) based upon cutpoints deemed biologically important in epidemiologic literature. Second, we evaluated confounding in studies of DBP exposures over the variable-length third trimester of gestation in relation to time-dependent outcomes. Using examples from a hypothetical population modeled after the U.S. distribution of term and preterm births, we evaluated how selection of participants and cutpoint use affected risk estimation. We found that failure to adjust for the variable length of third trimesters and use of cutpoints potentially leads to significant bias, both away from and towards the null, and further complicates estimation of true risk in epidemiologic studies. Lastly, we conducted a retrospective cohort study, evaluating exposure over specific windows of gestation, to determine whether a relationship exists between exposure to trihalomethanes (THMs) or the five most prevalent individual haloacetic acids (HAA5), including trichloroacetic acid (TCAA), and cleft palate. Birth certificate data (n = 51,717) were obtained and logistic regression techniques were used to estimate the relationship between mean exposures to DBPs over the first trimester, second month, and 6th, 7th and 8th weeks of pregnancy and cleft palate. We found no positive associations for categorical levels of exposure to THMs during critical time windows. Results for average 2 nd tertile exposure to TCAA (4–5.25 ug/l), as compared to the referent (<4 ug/l), were slightly increased for individual weeks and the second month. Overall, the findings in this dissertation will provide practical and theoretical guidance to future epidemiologic investigations of DBPs and birth outcomes. |
