|
|
|
From Sources to Biomarkers
|
|
|
From Sources to Biomarkers:
A Hierarchical Bayesian Approach for Human
Exposure Modeling
|
|
|
|
Project Overview
|
|
|
In September, 2004, a
team of SSES researchers from OSU (Catherine Calder, Peter Craigmile,
Noel Cressie, and Tom Santner) and Battelle (Bruce Buxton, Nancy
McMillan, and Michele Morara) was awarded a three-year contract funded
by the American Chemistry Council's (ACC)
Long-Range Research Initiative for the research project "From
Sources to Biomarkers: A Hierarchical Bayesian Approach for Human
Exposure Modeling". The funding was a result of a submission to the EPA's FY2003 STAR Grant program, in response to an RFA titled, "Environmental Statistics Research:
Novel Analyses of Human Exposure Related Data", which was jointly funded by the EPA's National Center for Environmental
Research (NCER) and the ACC. Noel Cressie is PI.
STB Research Team (Summer 2005)
Front row, from left: Noel Cressie,
Crystal Dong, Nancy McMillan, Kate Calder, Jian Zhang, Bruce Buxton
Back row, from left: Greg Young,
Peter Craigmile, Tom Santner, Ke Wang, Michele Morara
Project Description
The objective of this study is to characterize multi-pollutant
(arsenic, lead, cadmium, and chromium) human exposures by linking
sources to biomarkers using a multi-scale (areal, residential, and
personal) hierarchical Bayesian model (HBM) that describes how
multi-media pathways contribute to direct routes of exposure
(inhalation, ingestion, dermal). We hypothesize that by incorporating
several different sources of data that inform about pollutant
pathways, our model will discern patterns in human exposures and allow
us to draw more informed conclusions about current and future
distributions of biomarkers. Our approach is to use a hierarchical
Bayesian multivariate statistical modeling framework to coherently
manage and account for variability and uncertainty. The model has
explicit stages for sources, areal environmental levels, indoor
(residential) environmental levels, personal exposures, and
biomarkers. The primary data sources are the National Human Exposure
Assessment Survey (NHEXAS) Phase I data from EPA Region 5 and Arizona,
supplemented by census data, ambient air monitoring data, and
emissions data. NHEXAS data provide information to stages of the
model addressing areal and indoor environmental conditions, personal
exposures, and biomarkers. These stages combine this information in a
manner akin to structural-equation modeling to discern pathways and
routes of exposure. Ambient monitoring and emissions data provide
information to stages of the model addressing source rates and areal
environmental conditions. At these stages, we discern geographical
patterns in environmental contamination using spatial statistical
models. The expected results of this study include, but are not
limited to, characterizations of [a] the distribution of biomarkers
across the population as a whole and within subpopulations (defined
geographically or by personal, behavioral, or housing conditions); [b]
the impact of source emissions reductions on personal exposure levels;
[c] the impact of personal, behavioral, and housing conditions on the
distribution of biomarkers; and [d] the relative contribution to
biomarkers from various pathways and routes of exposure. The
multi-scale, multi-media, and multi-pollutant exposure models
developed in this study will also serve as examples, illustrating how
the HBM approach can be used to aid in the risk-assessment process
for other pollutants and in other settings.
|
|
|
Hierarchical Bayesian Modeling of Arsenic Exposure
Pathways
Tutorial on Hierarchical Bayesian Modeling for Exposure to Arsenic
The STB research group has developed a hierarchical Bayesian model
(HBM) that describes the pathways of arsenic exposure:
Global Environment (G) => Local Environment (L) => Personal Exposure
(X) => Biomarker (B)
The figure below illustrates the structure of
the model. Solid arrows from variable A to variable B indicate that
a priori B is assumed to be linearly related to variable A. The
dotted arrows from variable A to variable B imply that A is assumed to
be the baseline level of variable B. [Source: Cressie et al., 2007]
Note: While each of the local variables is assumed to have an
associated global variable, only the baseline link for global
topsoil to local soil is shown since topsoil is the only global variable that
depends on another global variable (stream sediment) in the
preliminary model.
The pathways model was fitted using NHEXAS exposure data from 249
individuals in EPA's Region 5 (Illinois, Indiana,
Michigan, Minnesota, Ohio, and Wisconsin). In the figure above, the
posterior median and 95 percent credible intervals for each of the
pathway links are given above the corresponding arrow. Bold
solid arrows indicate that the posterior probability that the link is different
from zero is at least 95 percent.
In addition to the NHEXAS measurements, data on the
concentration of arsenic in topsoil (obtained from the USDA-NRCS's Soil
Geochemistry Spatial Database) and in stream sediment (obtained
from the USGS's National
Geochemical Survey Database) were used to provide information about
the background, or global level, of arsenic in soil. The figure below
shows the extensive spatial coverage of the global stream-sediment
measurements (locations are denoted by circles).
The map coloring corresponds to the posterior mean of the the global
levels of arsenic in stream sediment across Region 5
watersheds. This figure illustrates the potential for understanding
the mechanisms for the spatial variation in arsenic exposure across a
large geographic area by supplementing NHEXAS data with information on
background arsenic levels in various media.
Hierarchical Bayesian Approach for Human
Exposure Modeling
A more detailed description of the HBM for arsenic exposure pathways
can be found in: Tutorial on Hierarchical Bayesian Modeling for Exposure to Arsenic
, and in the following article:
Cressie, N., Buxton, B.E., Calder,
C.A., Craigmile, P.F., Dong, C., McMillan, N.J., Morara, M., Santner,
T.J., Wang, K., Young, G., and Zhang, J. (2007). From sources to
biomarkers: A Bayesian approach to human exposure modeling.
Journal of Statistical Planning and Inference, 137,
3361-3379.
Subpopulation-Specific Arsenic Exposure
Pathways
An extension of the hierarchical Bayesian
model for human-exposure pathways described in Cressie et al. (2007) that
considers subpopulation-specific routes of exposure to arsenic can be
found in the following manuscript, which can be obtained from the SSES Preprints
webpage.
Santner, T.J., Craigmile, P.F., Calder,
C.A., and Paul, R. (2007). Demographic and behavioral modifiers of
arsenic exposure pathways: A Bayesian hierarchical analysis of NHEXAS
data. Department of Statistics Preprint No. 801, The
Ohio State University, Columbus, OH.
Supplemental information for this manuscript is available here.
|
|
Links
|
|
|
|
|
