DEVELOPING STATISTICAL CORRELATION OF VARIOUS TYPES OF BIRTH DEFECTS AND SPECIFIC CONTAMINANTS IN DRINKING WATER

Author

Fan Yang, Southern Illinois University CarbondaleFollow

Date of Award

12-1-2012

Degree Name

Master of Science

Department

Biomedical Engineering

First Advisor

Mohanty, Manoj

Abstract

Our research investigated the correlations of five key contaminants in drinking water in the State of Illinois with the incidence rates of three different categories of negative reproductive outcomes: birth defects, adverse pregnancy outcomes and preterm births. The five water contaminants included three agrichemical-based contaminants (atrazine, nitrate and nitrite) and two disinfectant byproducts (total trihalomethanes and haloacetic acids). We obtained nearly 38,000 measurements of the individual concentrations of these five contaminants from the Illinois Environmental Protection Agency (IEPA). These data, spanning over a five-year period (1998-2002), were collected from nearly 1,800 community water supplies (CWS) located in all 102 Illinois counties. The county-specific incidence rate data for the same time period for preterm births, seven groups of birth defects and seven groups of adverse pregnancy outcomes were collected from the Illinois Department of Public Health (IDPH). The traditional approach of establishing correlation of one explanatory variable at a time indicated that nitrate concentration in drinking water was statistically correlated to the incidence rate of only one category of negative reproductive outcome, i.e., birth defects. However, multiple regression analysis, which took into consideration all water contaminant data simultaneously, established statistical significance of the correlation between nitrate and all three categories of negative reproductive outcomes. Total trihalomethanes (TTHM) and haloacetic acids (HAA5) were also found to be the statistically significant explanatory variables for all three types of negative reproductive outcomes. Nitrite was found to be associated with adverse pregnancy outcomes and preterm births only; its correlation with birth defect rates could not be established at 80% confidence level. Atrazine was found to be among the significant explanatory variables for all three negative reproductive outcomes. However, its correlations with birth defect model developed using censored data and preterm birth model developed based on observed data were not statistically significant. Nearly 21.5%, 35.8% and 16.6% of the variabilities in incidence rates of different types of birth defects, adverse pregnancy outcomes and preterm births, respectively, were correlated to average concentrations of above mentioned five contaminants in community drinking water supplies. The monthly average concentrations of two of three agrichemical-based contaminants, i.e., atrazine and nitrate were significantly higher in SW(surface water)-based than GW(ground water)-based community water supplies. On the other hand, concentrations of both disinfectant byproducts and the other agrichemical-based contaminant, i.e., nitrite were significantly higher in GW-based water supplies. Further correlation analysis based on the actual observed concentration (excluding all 0 detections) was conducted utilizing SW-based atrazine and nitrate data and GW-based nitrite, TTHM and HAA5 data. All categories of negative reproductive outcomes were found significantly correlated with at least one type of water contaminants. The overall negative reproductive outcomes were correlated with nitrate, nitrite and atrazine. Birth defects specifically correlated with nitrate and TTHM, whereas adverse pregnancy outcomes were correlated with nitrate, nitrite and HAA5; preterm birth was correlated with the concentrations of nitrite, atrazine and TTHM. Further investigation indicated significant correlations among the explanatory variables (drinking water contaminants data), which is referred as multicollinearity. In such case, multiple linear regression based analysis may not provide reliable results. Partial least squares (PLS) approach of regression analysis was introduced into our study to overcome the multicollinearity problem. As much as 65.5% of the variability in the county-wide average concentrations of five contaminants in public drinking water supplies was explained by the 4 component PLS regression model, whereas only 7.7% of the incidence rates of different types of birth defects, adverse pregnancy outcomes and preterm births in various Illinois counties, were explained by PLS regression. Although, individual negative reproductive outcome model could be generated, the low R2 values indicated the poor reliability of these models. We attempted to corroborate our statistical analysis findings with the physiological effects of various water contaminants reported in the literature. It is well known that excessive exposure to any of the five key water contaminants may cause malfunction of organism or the immune system, reproduction, nervous and endocrine system and others, which may further result in potential risk of developing adverse effects. Nitrate and nitrite in drinking water associate to a certain extent with congenital malformations, such as central nervous system abnormalities, in human offsprings. Excessive exposure to atrazine is known to increase the risk of potential cardiovascular or reproductive system problems. A certain type of TTHM, i.e., BDCM are known to be associated with an increased risk of spontaneous miscarriage in pregnant women. It also may result in an increased risk of reproductive problem, such as fetal growth restriction.