Grantee Research Project Results
Final Report: Air Pollution-Exposure-Health Effect Indicators: Mining Massive Geographically-Referenced Environmental Health Data to Identify Risk Factors for Birth Defects
EPA Grant Number: R834790Title: Air Pollution-Exposure-Health Effect Indicators: Mining Massive Geographically-Referenced Environmental Health Data to Identify Risk Factors for Birth Defects
Investigators: Zhan, F. Benjamin , Brender, Jean D. , Langlois, Peter H. , Yang, Jing
Institution: Texas State University , Texas Department of State Health Services , University of North Carolina at Charlotte , Texas A & M University
Current Institution: Texas State University , Texas A & M University , Texas Department of State Health Services , University of North Carolina at Charlotte
EPA Project Officer: Hahn, Intaek
Project Period: February 1, 2011 through January 31, 2014 (Extended to January 31, 2015)
Project Amount: $499,987
RFA: Exploring Linkages Between Health Outcomes and Environmental Hazards, Exposures, and Interventions for Public Health Tracking and Risk Management (2009) RFA Text | Recipients Lists
Research Category: Climate Change , Air Quality and Air Toxics , Human Health
Objective:
In the United States, approximately one in 33 infants is born with a defect. In addition, birth defects are one of the leading causes of infant mortality. The current state of knowledge indicates that researchers have not yet ascertained risk factors for approximately two-thirds of birth defects. Experts in the field suggest that maternal exposure to certain adverse environmental conditions may be partly associated with congenital malformation in offspring. But the process of identifying environmental risk factors associated with some birth defects has remained a challenging task for scientists. Among environmental risk factors, maternal exposure to air pollutants is considered an important risk factor for some types of birth defects. One difficult task in this research area is the development of effective methods that can be used to analyze massive geographically referenced data and identify factors that are most likely to be associated with certain birth defects.
This project aims to develop a three-component computational approach that can be used to identify these risk factors from massive geographically referenced environmental health data. These three components include: (1) air pollution modeling and exposure assessment, (2) visual geospatial data mining, and (3) epidemiological analysis. This project then attempts to achieve two goals using the newly developed method: (1) identify associations between maternal residential proximity to industrial facilities that emit toxic chemicals and select congenital malformations in offspring, and (2) develop air pollution-exposure-health outcome indicators based on identified associations. The environmental data used in this project were Toxic Release Inventory (TRI) data in Texas from 1996 through 2008. The health outcome data used in this project were from the Texas Birth Defects Registry in the same period, including four select groups of birth defects: (1) neural tube defects, (2) heart defects, (3) oral clefts, and (4) limb reduction defects. Overall, for the 13-year period from 1996 through 2008, a total of 449 toxic air pollutants emitted from industrial facilities in Texas, as well as 63,132 cases and 244,927 control births that could be geocoded to street level, were considered in this study.
For air pollution modeling and exposure assessment, the project team used a modified version of the Emission Weighted Proximity Model (EWPM) to calculate the absolute and relative exposure intensities associated with each of the 63,132 cases and 244,927 control births to each of the 449 toxic air pollutants emitted from industrial facilities in Texas over a 13-year period from 1996 through 2008. In the visual geospatial data mining component, the team developed a novel visual analytics tool to select toxic air pollutants that are most likely to be associated with a birth defect. This tool integrates automatic multivariate analysis techniques with interactive visualization methods. In particular, it employs point-biserial correlations (rpb), which measure the correlation between a continuous independent variable (e.g., the concentration of a chemical at a location) and a dichotomous dependent variable (e.g., a birth defect). These correlations are calculated automatically by the tool, visualized on a computer screen, and can be interactively explored by a user.
In the epidemiological analysis component, the team followed a standard case-control study design and used logistic regression as the statistical procedure to examine associations between a toxic chemical and a birth defect and to validate the correlations found through the visual analytic tool. In performing the epidemiological analysis, estimated exposure intensities were categorized using two methods. First, exposure was classified into two groups based on exposure intensities for a given chemical equaling zero or greater than zero. The second categorization was based on four levels that included exposure intensities at zero and greater than zero divided into three equal groups (lowest, middle, highest tertiles) based on the control-mothers distributions of exposure intensity values. Based on the identified associations, the research team developed a set of environmental health indicator maps that can be used to visualize estimated exposure intensity of a given air pollutant in a geographic region of interest.
Summary/Accomplishments (Outputs/Outcomes):
Results suggest that the three-component computational approach is effective for identifying environmental risk factors from massive geographically referenced environmental health data. When this approach was applied to analyze the data used in this project, 12 chemicals were found to be associated with spina bifida, 10 chemicals with septal heart defect, 5 chemicals with cleft palate alone, and 7 chemicals with limb reduction defects. These 34 chemicals contain two overlapping chemicals (propylene and vanadium) in the lists of chemicals associated with both NTDs and heart defects, giving a list of 32 distinctive chemicals. The identification of these associations is an important step in environmental birth defects research.
Specifically, the associations between maternal residential proximity to emission sources of toxic chemicals and spina bifida in offspring suggest 12 chemicals of significant concern. Odds ratios (OR) were also adjusted (aOR) for birth year and maternal age, education, race/ethnicity, as well as public health region of residence. These chemicals are: 1,3-butadiene (aOR 1.26, 95% CI 1.03, 1.53), 1,4 dioxane (aOR 1.56, 95% CI 1.17, 2.08), 2-ethoxyethanol (aOR 1.63, 95% CI 1.09, 2.45), chloroethane (aOR 1.59, 95% CI 1.18, 2.14), dinitrotoluene (aOR 1.46, 95% CI 1.04, 2.04), propylene (aOR 1.26, 95% CI 1.04, 1.52), tert-butyl alcohol (aOR 1.34, 95% CI 1.06, 1.68), trichlorofluoromethane (aOR 2.44, 95% CI 1.29,4.59), triethylamine (aOR 1.55, 95% CI 1.14, 2.10), vinyl chloride (aOR 1.74, 95% CI 1.27, 2.40), maleic anhydride (aOR 1.30, 1.03, 1.64), and vanadium (aOR 1.44, 95% CI 1.01, 2.06). The strongest associations were noted between trichlorofluoromethane in the middle tertile of exposure and spina bifida (aOR 5.14, 95% CI 2.39, 11.08) and the highest tertile of exposure to 2-ethoxyethanol and spina bifida (aOR 2.49, 95% CI 1.14, 4.40).
For possible associations between maternal residential proximity to emission sources of toxic chemicals and heart defects in offspring, a total of 10 chemicals were identified. These 10 chemicals are: p-xylene (aOR 1.26, 95% CI 1.21, 1.31), m-xylene (aOR 1.23, 95% CI 1.18, 1.27), n-hexane (aOR 1.14, 95% CI 1.12, 1.17), propylene (aOR 1.31, 95% CI 1.27, 1.36), methyl tert-butyl ether (aOR 1.14, 95% CI 1.11, 1.16), benzene (aOR 1.19, 95% CI 1.16, 1.21), vanadium (aOR 1.28, 95% CI 1.20, 1.36), 1,2 4-trimethylbenzene (aOR 1.14, 95% CI 1.11, 1.16), chromium compounds (aOR 1.05, 95% CI 1.03, 1.08), and nickel compounds (aOR 1.22, 95% CI 1.19, 1.26). However, none of the associations was particularly strong. Among these associations, the strongest is the association between vanadium in the highest tertile and septal heart defects (aOR 1.78, 95% CI 1.62, 1.96).
For associations between maternal residential proximity to emission sources of toxic chemicals and oral clefts in offspring, a total of five chemicals were revealed by the analysis. Among these five chemicals, it is clear that three are associated with cleft palate alone. These three chemicals are: 2,4-diaminotoluene (aOR 1.56, 95% CI 1.06, 2.28), iron pentacarbonyl (aOR 1.92; 95% CI 1.19, 3.92), and o-toluidine (aOR 1.50, 95% CI 1.04, 2.17). In addition, two of the five chemicals are associated with cleft lip with or without cleft palate. These two chemicals are: dichloromethane in the middle tertile (aOR 1.16, 95% CI 1.02, 1.33) and n-butyl alcohol in the middle tertile (aOR 1.16, 95% CI 1.04, 1.30).
Results indicate that maternal residential proximity to emission sources of seven toxic chemicals may be associated with limb reductions in offspring. These seven chemicals are: arsenic, lead, tetrabromobisphenol A, trichloroacetyl chloride, 1,2-dichloropropane, benzyl chloride, and chloroform. Among these seven chemicals, benzyl chloride (aOR 1.54, 95% CI 1.05, 1.26), trichloroacetyl chloride (aOR 2.59, 95% CI 1.10, 6.10), tetrabromobisphenol A (aOR 6.12, 95% CI 2.22, 16.87), and 1,2-dichloropropane, (aOR 2.17, 95% CI 1.11, 4.24) are associated with limb reduction overall. Specifically, benzyl chloride in the highest tertile is associated with longitudinal limb reduction defects (aOR: 3.04; 95% CI: 1.49, 6.20). Trichloroacetyl chloride is associated with limb reduction defects overall (aOR 2.59, 95% CI 1.10, 6.10), as well as with transverse limb reduction defects (aOR 3.90, 95% CI 1.50, 10.16) and upper limb reduction defects (aOR 2.80, 95% CI 1.09, 7.17). In particular, arsenic is associated with transverse limb reduction defects (aOR 1.59, 95% CI 1.14, 2.23). Lead is associated with lower limb reduction (aOR 1.22, 95% CI 1.03, 1.45). Chloroform in the highest tertile is associated with longitudinal limb reduction defects (aOR: 1.99; 95% CI: 1.04, 3.83), upper limb reduction defects (aOR: 1.77; 95% CI: 1.04, 3.02), and lower limb reduction defects (aOR: 2.50; 95% CI: 1.29, 4.81).
It should be reiterated that the findings are restricted to only associations between maternal residential proximity to emission sources of toxic air pollutants from industrial facilities and congenital malformations in offspring and therefore do not infer causality. The study did not take into account synergistic impacts resulting from exposure to multiple chemicals, other sources of exposure to chemicals, or other potentially confounding factors for which data were unavailable, such as folic acid supplementation.
In addition, it is important to consider several limitations of this study when interpreting the associations summarized above. First, the study used maternal residential proximity to point emission sources of toxic air pollutants from industrial facilities as a proxy for exposure assessment. While team members of this study have found the metrics of exposure intensities estimated by the Emission Weighted Proximity Model (EWPM) to be correlated to air monitoring results for a variety of chemicals studied in this project, the estimated exposure based on the EWPM does not represent a mother's actual exposure to a chemical. Second, maternal addresses used in the analysis were addresses at the time of delivery, while the most vulnerable period for the development of the selected birth defects is the first trimester of pregnancy. Previous studies in Texas have indicated that up to one-third of women who are pregnant may change residences between the first trimester and the time of delivery. This limitation adds additional uncertainties in exposure assessment. A third limitation is related to "multiple comparisons" that is evident in the analysis procedures used in the "visual geospatial data mining" component of the study. This third limitation is being examined by members of the research team. Additional results related to this part of the research will be presented elsewhere because this was not within the scope of this project.
Conclusions:
This project has demonstrated that the mining of massive geographically referenced environmental health data in relation to adverse health outcomes, such as birth defects, can help identify potentially harmful environmental exposures and focus further research regarding these exposures and health. This research has refined and further developed the methodology for data analysis of Toxic Release Inventory (TRI) air emissions data for 1996 through 2008 in Texas in comparison with Texas Birth Defects Registry cases of select birth defects and controls. The methodology offers opportunity to evaluate geographic variation in potential exposure risk within the Hazard-Exposure-Health Effect-Intervention structure. The development of the three-component computational approach should be considered as work in progress. Further refinement of this methodology and the availability of new data may allow for assessment of temporal changes in the potential risk of exposure. Findings from this project suggest that maternal residential proximity to air emissions of a number of chemicals might be associated with neural tube, oral cleft, limb deficiency, and congenital heart defects in offspring. These associations can be used to develop environmental health indicator maps. To better understand these relationships and potential public health and policy implications, researchers of this study recommend that future studies focus on populations for which ambient air measurements of these chemicals and residential histories during pregnancy are available.Journal Articles on this Report : 3 Displayed | Download in RIS Format
Other project views: | All 13 publications | 3 publications in selected types | All 3 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Brender JD, Shinde MU, Zhan FB, Gong X, Langlois PH. Maternal residential proximity to chlorinated solvent emissions and birth defects in offspring: a case-control study. Environmental Health 2014;13:96 (16 pp.). |
R834790 (Final) |
Exit Exit |
|
Gong X, Zhan FB, Brender JD, Langlois PH, Lin Y. Validity of the EmissionWeighted Proximity Model in estimating air pollution exposure intensities in large geographic areas. Science of The Total Environment 2016;564:478-485. |
R834790 (Final) |
Exit |
|
Zou B, Wilson J, Gaines Z, Zeng Y, Wu K. Spatial-temporal variations in regional ambient sulfur dioxide concentration and source-contribution analysis:A dispersion modeling approach. ATMOSPHERIC ENVIRONMENT 2011;45(28):4977-4985. |
R834790 (Final) |
Exit Exit |
Supplemental Keywords:
Environmental health, public health, air pollution, exposure assessment, risk assessment, birth defects, health disparities, GIS, data mining, visual analytics;Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.