2012 Progress Report: Effects-Based Cumulative Risk Assessment in a Low-Income Urban Community near a Superfund Site
EPA Grant Number:
Effects-Based Cumulative Risk Assessment in a Low-Income Urban Community near a Superfund Site
, Ezzati, Majid
, French, Robert
, Fuentes, Freddie
, Hammitt, James K.
, Korrick, Susan A.
, Rosario, Maria
, Sexton, Ken
, Subramanian, S V
Fabian, Maria Patricia
Korrick, Susan A.
NorthStar Learning Centers
NorthStar Learning Centers
EPA Project Officer:
September 1, 2010 through
February 28, 2014
(Extended to February 28, 2015)
Project Period Covered by this Report:
September 1, 2011 through August 31,2012
Understanding the Role of Nonchemical Stressors and Developing Analytic Methods for Cumulative Risk Assessments (2009)
Human Health Risk Assessment
The study's primary objective is to develop and apply novel statistical and analytical methods for cumulative risk assessment, focusing on a low-income community (New Bedford, MA) living near a Superfund site. We are focusing on specific neurobehavioral and cardiovascular outcomes, ADHD-like behavior and blood pressure and we are considering contributions of multiple chemical stressors to these outcomes in the presence of significant non-chemical stressors. We are leveraging data from public databases and an ongoing birth cohort study in the New Bedford area to predict exposures as a function of individual and neighborhood characteristics. Those exposures are then linked with dose-response functions for key chemical and non-chemical stressors, providing health risk characterization with geospatial and demographic variability. The models will be built and outputs will be communicated to community partners in a manner that informs future studies or intervention strategies.
We have made progress in four major areas over the past year. First, we built predictive exposure models for key chemical and non-chemical stressors influencing ADHD-like behavior, based on data from a New Bedford area cohort. Common predictors of prenatal biomarker measures of lead (Pb), polychlorinated biphenyls (PCBs), dichlorodiphenyldichloroethylene (DDE) and mercury (Hg) included mothers' date of birth, country of birth, and household income. Food consumption patterns (e.g., fish and organ meat consumption) were associated with prenatal exposure to DDE, PCB and Hg, while home attributes (e.g., construction year and assessed value) were associated with postnatal blood Pb concentrations. To make these models applicable to the general population, we also needed to predict food consumption and predictors from the New Bedford cohort as a function of public data. We used structural equation models to predict information only available from the New Bedford cohort as a function of variables available from the U.S. Census and American Community Survey, while at the same time predicting stressor concentrations as a function of cohort data.
For blood pressure, we used data from the National Health and Nutrition Examination Survey (NHANES) 1999-2008 for adults ages 20 years and older to characterize exposures and key predictors. We focused on four chemical exposures previously associated with elevated blood pressure: Pb, Hg, PCBs, and cadmium (Cd). As predictors in linear regression analysis, we examined demographics (age, sex, race/ethnicity, country of birth, education, and poverty status); dietary information (fish and shellfish consumed); and other characteristics (e.g., smoking, housing characteristics). We used structural equation modeling to simultaneously relate these risk factors to systolic blood pressure. Common predictors of Pb, Cd, Hg and PCB included gender, race/ethnicity and smoking. Socioeconomic variables were associated with Pb, Cd and Hg while fish consumption was particular to PCB and Hg.
While these regression models characterize the influence of a number of factors on exposures, applying those models to a large population requires extensive individual-level or cross-tabulated data. However, the data from the census either lacks geographic resolution or extensive individual-level data because of the need to limit potential identifiability. We developed an approach to generate synthetic population data that reasonably represented attributes of New Bedford relevant to chemical and non-chemical stressor exposures. Applying this approach, called probabilistic reweighting using simulated annealing, we were able to reproduce population characteristics at census tract resolution with limited error. We successfully matched our constraint variables with very low error, and we were also able to predict non-constraint variables at the census tract level. We were able to illustrate the value of this approach by linking the synthetic population data with the New Bedford exposure models described above. This helped to identify census tracts at high risk of exposure, although the risk of exposure is spatially patterned due to the prevalence of predictive socioeconomic and food consumption factors rather than spatial patterns of ambient pollutants. Maps can be used to identify neighborhoods at high risk of exposure.
Finally, we implemented a community survey to evaluate whether predictors of exposure-related behaviors in New Bedford have changed from the mid-1990s to the present, given demographic and behavioral shifts over time. We asked questions shown to be predictive of chemical exposures in our regression analyses, and we also asked questions that had been shown to be associated with hypertension outcomes or co-morbidity in the past. The survey was developed in both English and Spanish, and with both paper and online versions. Surveys were implemented at a variety of community settings, with the objective to capture a range of ages, genders, ethnic groups, and neighborhood areas prevalent in New Bedford. Our target was to survey 400 New Bedford residents 18 years and older, which we expected to reach by December 2012.
In Year 3 (9/1/12-8/31/13), we plan for a few key activities. First, we will complete the analyses and related manuscripts for the work described above. Related to these efforts, we will start to develop formal outreach and presentation materials, and we will present study findings at relevant community forums. We will focus new analytical efforts in Year 3 on the development of dose-response functions, utilizing the structural equation modeling outputs but also considering evidence from other epidemiological studies and the toxicological literature. During the second half of Year 3, we will also start integrating the exposure and dose-response information to construct risk maps.
No journal articles submitted with this report: View all 20 publications for this project
ADHD, blood pressure, cadmium, cumulative risk assessment, lead, mercury, non-chemical stressor, PCBs, Superfund
Progress and Final Reports:
2010 Progress Report
2011 Progress Report
2013 Progress Report