Air pollution exposure assessment methods utilized in epidemiological studies.

Abstract

The assessment of personal exposure to air pollution is a critical component of epidemiological studies associating air pollution and health effects. This paper critically reviewed 157 studies over 29 years that utilized one of five categories of exposure methods (proximity, air dispersion, hybrid, human inhalation, and biomarkers). Proximity models were found to be a questionable technique as they assume that closer proximity equates to greater exposure. Inhalation models and biomarker estimates were the most effective in assessing personal exposure, but are often cost prohibitive for large study populations. This review suggests that: (i) factors such as uncertainty, validity, data availability, and transferability related to exposure assessment methods should be considered when selecting a model; and (ii) although an entirely discreet new class of approach is not necessary, significant progress could be made through the development of a 'hybrid' model utilizing the strengths of several existing methods. Future work should systematically evaluate the performance of hybrid models compared to other individual exposure assessment methods utilizing geospatial information technologies (e.g. geographic information systems (GIS) and remote sensing (RS)) to more robustly refine estimates of ambient exposure and quantify the linkages and differences between outdoor, indoor and personal exposure estimates.

Full text

Page 1

Air pollution exposure assessment methods utilized in epidemiological studies
Bin Zou,*abJ. Gaines Wilson,cF. Benjamin Zhanbdand Yongnian Zenga
Received 11th August 2008, Accepted 9th January 2009
First published as an Advance Article on the web 13th February 2009
DOI: 10.1039/b813889c
The assessment of personal exposure to air pollution is a critical component of epidemiological studies
associating air pollution and health effects. This paper critically reviewed 157 studies over 29 years that
utilized one of five categories of exposure methods (proximity, air dispersion, hybrid, human
inhalation, and biomarkers). Proximity models were found to be a questionable technique as they
assume that closer proximity equates to greater exposure. Inhalation models and biomarker estimates
were the most effective in assessing personal exposure, but are often cost prohibitive for large study
populations. This review suggests that: (i) factors such as uncertainty, validity, data availability, and
transferability related to exposure assessment methods should be considered when selecting a model;
and (ii) although an entirely discreet new class of approach is not necessary, significant progress could
be made through the development of a ‘hybrid’ model utilizing the strengths of several existing
methods. Future work should systematically evaluate the performance of hybrid models compared to
other individual exposure assessment methods utilizing geospatial information technologies (e.g.
geographic information systems (GIS) and remote sensing (RS)) to more robustly refine estimates of
ambient exposure and quantify the linkages and differences between outdoor, indoor and personal
exposure estimates.
1.Introduction
By the mid-20th century, there had been several pronounced air
pollution ‘events’ resulting in substantial adverse effects on
human health. The most well-known were the 1930 event in the
Meuse Valley of Belgium,1the toxic smog in Donora, Pennsyl-
vania in October 1948,2as well as the 1952 London event known
as the ‘killer fog’ which resulted in over 4000 deaths.3As a result
of these events and others, the 1950s saw the first studies exam-
ining the association between human health and air pollution.4It
is now well known and widely accepted that air pollution is
associated with mortality5,6and morbidity, including asthma,7,8
chronic bronchitis and emphysema,7,9myocardial infarction,10
cardiovascular disease,11neurotoxicity,12and lung cancer,13,14
among other ailments and forms of disease.15–22The World
Health Organization (WHO)23estimates that around 1.4 billion
urban residents are living in areas with air above WHO air
quality guidelines (AQG), and consequently, outdoor air pollu-
tion is associated with approximately 200 000 to 570 000 annual
deaths representing about 0.4 to 1.1% of the total annual deaths
in urban areas worldwide. The health effects related to outdoor
air pollution are now recognized as a focus-area for reducing the
global burden of disease, though governments have only begun
to acknowledge this fact through enacting policy.24
Mr Bin Zou is a PhD candidate
at the School of Info-Physics
and Geomatics Engineering in
Central South
China. He was
research scholar at the Texas
Center for Geographic Infor-
mation Science in Texas State
University-San
2007–2009.
University,
a visiting
Marcos from
Dr Jeff Wilson is Assistant Professor of Environmental Science
at the University of Texas at Brownsville.
Dr F. Benjamin Zhan is a professor of Geography and Director
of the Texas Center for Geographic Information Science at Texas
State University-San Marcos.
Dr Yongnian Zeng is a Professor at the School of Info-Physics
and Geomatics Engineering in Central South University, China.
aCentral South University, School of Info-Physics and Geomatics
Engineering, Changsha, Hunan, 410086, China. E-mail: B.ZOU@
txstate.edu; kyzoubin@yahoo.com.cn; Fax: +1 512-245-8353; Tel: +1
512-665-1781
bTexas State University, Texas Center for Geographic Information
Science, Department of Geography, 601 University Drive, San Marcos,
TX, 78666, USA. E-mail: Zhan@txstate.edu
cUniversity of Texas at Brownsville, Department of Chemistry and
Environmental Sciences, Brownsville, TX, 78520, USA. E-mail: jeff.
wilson@utb.edu
dWuhan University, School of Resource and Environmental Science,
Wuhan, Hubei, 430079, China
This journal is ª The Royal Society of Chemistry 2009 J. Environ. Monit., 2009, 11, 475–490 | 475
CRITICAL REVIEW www.rsc.org/jem | Journal of Environmental Monitoring

Page 2

Exposure, or the event of contact with a pollutant over
a certain period of time, was defined almost concurrently in the
early 1980s by two researchers, Duan25and Ott.26Exposure
assessment can be used to evaluate, at various levels of detail, the
degree and linkage between contaminant sources and concen-
tration of hazards and receptors (e.g. humans) in the environ-
ment bystudying the different exposure pathways (e.g. air,water,
and soil) and routes (e.g. inhalation, ingestion, and dermal
contact) between them while taking into account behavioral
factors.27Air pollution exposure assessment, as one kind of
exposure assessment, indicates human exposure to air pollutants.
Air pollution exposure assessment work, which began in the
early 1990s, has been a critical component in building a case for
the relationship between air pollution and health effects in
epidemiological studies.28Over the past 20 years, a myriad of
methods for assessing human exposure levels to air pollution
have been introduced, with the ultimate goal of estimating
exposure at the individual level within an entire study pop-
ulation.29–32However, most of these exposure assessment
methods place emphasis on case studies and developments of
exposure models. To date, there have been no reviews system-
atically critiquing the literature on methods for air pollution
exposure assessment while evaluating the advantages and
disadvantages of each method in a comparative fashion.
Explicitly, this review addresses the following questions:
(i) what are the most appropriate approaches for assessing
personal exposure to air pollution?
(ii) what factors affect the accuracy and validity of air pollu-
tion exposure assessment? and
(iii) what direction should future studies of air pollution
exposure assessment take?
This article addresses these questions by summarizing the most
current and promising approaches for ambient air pollution
exposure assessment utilized in epidemiological studies. In order
to meet this aim, we first outline the methods and criteria utilized
in understanding the current state of air pollution exposure
assessment. We then provide an overview of concepts and
research frameworks utilized in air pollution exposure assess-
ment (section three). Next, in sections four and five, exposure
models and biomarkers are reviewed in detail by outlining their
developments, applications, and respective merits and weak-
nesses. Finally, based on a comparative review in section six, we
comment on the primary issues of debate and offer our
perspectives about future development of air pollution exposure
assessment methods.
2. Methodology and search criteria
The literature for this review was obtained by systematically
searching the following online databases: the ISI Web of
Knowledge, ScienceDirect, PubMed Central, and the Google
Scholar website. The keywords used to guide our search included
various combinations of ‘exposure assessment’, ‘air pollution’,
‘exposure models’, and ‘biomarkers’ for articles published during
the period 1980–2008. Records on approximately 800 unique
peer-reviewed articles were returned and those publications
deemed to be the most relevant to the research questions of
interest were retained. Since this review will concentrate on the
‘exposure assessment methods of ambient air pollution’, our final
selection consisted of 157 sources (Table 1). Upon selecting
articles, we sorted the literature into one of the following method
categories: (i) exposure modeling fundamentals; (ii) proximity
models; (iii) air dispersion models; (iv) hybrid models; (v) human
inhalation models; (vi) biomarkers; or (vii) emerging geospatial
information technologies. We then critiqued individual studies
while summarizing the literature in a more holistic fashion.
3.
pollution exposure assessment
Fundamental concepts and framework of air
Before summarizing the prominent methods for assessing human
exposure to air pollutants, it is useful to provide an overview of
several key issues related to exposure assessment. These key
concepts include air quality guidelines (AQG), concentration,
human exposure, dose, personal monitoring, questionnaires and
diaries, as well as human activity patterns and microenviron-
ments. We also briefly cover the general conceptual framework
for assessing exposure to air pollution.
3.1Air quality guidelines
Although one of the basic requirements of human health and
well-being is considered to be clean air, air pollution exposure is
increasingly posing a significant threat to human health world-
wide, especially in large urban areas.23In order to provide
a global perspective on assessment of air pollution exposure risk
and consequently reducing the health impacts of air pollution,
WHO began designation of AQG in 1987.33The most recent
update to the guidelines was in 2005,34and isbased on the current
scientific evidence of health effects. Table 2 presents the limits for
the concentrations of six common air pollutants based on various
averaging times. These pollutants are collectively known as
‘criteria pollutants’, and include particulate matter (PM2.5,
PM10), ozone (O3), nitrogen dioxide (NO2), sulfur dioxide (SO2),
carbon monoxide (CO), and lead (Pb), recommended by
WHO.34,35It should be noted that these WHO guidelines are not
universally adopted. Individual governments adopt their own
guidelines and values between levels of government (federal,
state, local) may vary as well. For example, in the United States,
PM2.5guidelines set by the Environmental Protection Agency
(EPA) are (e.g. 1 year: 15 mg m?3; 24 h: 35 mg m?3)36substantially
different from WHO guidelines, while SO2are similar. Table 2 is
illustrative of guidelines and not to be interpreted as universal.
Table 1Composition of literature reviewed
Type of publicationPeer-reviewed papersBooks Government and WHO reportsConference presentations ThesesWebsite pages
No. of publications134312422
476 | J. Environ. Monit., 2009, 11, 475–490 This journal is ª The Royal Society of Chemistry 2009

Page 3

3.2General definitions and concepts in exposure assessment
3.2.1
sion of the local abundance of a pollutant in ambient air. It can
be defined as a mass (e.g. microgram or nanogram, denoted ‘mg’
and ‘ng’, respectively) or the fraction (mole) of material by
volume (e.g. parts per million or billion, denoted ‘ppm’ and
‘ppb’, respectively) in air.37
Concentration. Concentration is a quantitative expres-
3.2.2
and human exposure is based upon whether or not the interac-
tions between the environment and humans exists. Concentra-
tion only assumes some abundance of pollution but makes no
assumption about whether a human is exposed to the pollution
or not. Exposure is defined by IPCS as ‘concentration, amount or
intensity of a particular environmental agent that reaches the
target population, organism, organ tissue or cell over time and
space’.38Compared with concentration, exposure is a more
accurate measure of human contact with pollution, and there-
fore, estimation of true personal exposure is a critical component
of health risk assessment.
Human exposure. The difference between concentration
3.2.3
human exposure metrics to establish the linkages between some
adverse health effects and air pollutants (e.g. cardiovascular
disease and particulate matter). However, this method cannot
reflect the actual mass/concentration of a pollutant that is actu-
ally inhaled by an individual, given the unique activity patterns,
such as age and sex of individuals. By measuring dose, one is able
to approximate the mass of pollution that crosses into a body’s
theoretical boundaries and reaches the target tissue.39This
approximation can also be stated as the total amount of pollut-
ants that was ‘taken’, or absorbed by an organism, and thus
associated with adverse health effects. Dose is often utilized in
conjunction with biomarkers (reviewed in section five) as a way
of validating cumulative human exposure.40
Dose. Most epidemiological studies currently utilize
3.2.4
approach for measuring the personal exposure to ambient
Personal monitoring.Personal monitoring isan
pollution concentrations with active monitors or passive
samplers, and is considered the most accurate estimate of a per-
son’s ‘true’ exposure.41Personal monitoring is gaining popularity
as a method given recent technological advances that have
reduced the size/weight of personal monitoring devices while
improving accuracy and efficiency. However, the sensitivity and
accuracy of passive samplers used for personal monitoring is
highly dependent upon diffusion factors that influence contact
between the pollutants and instrument sensor, even though these
diffusion factors might not obstruct active sampling. Further-
more, as described by Wallace and Ott,42there are also several
methodological issues for personal monitoring. For instance, the
method is labor-intensive, time consuming, costly, and has
population limitations compared with other methods for
measuring human exposure level.42,43
3.2.5
studies, questionnaires and diaries were used to categorize
subjects into different groups given some potential exposure (e.g.
high exposure or low exposure, exposed or unexposed).44,45
Consequently, diaries were used to aid in interpreting personal
and environmental monitoring results by providing qualitative,
often, retrospective information. However, with the development
of exposure models, the functions of questionnaires and diaries,
in air pollution exposure assessment, have been extended. Apart
from categorizing population exposure qualitatively, question-
naires are now being used to investigate human activity
patterns—a technique which may improve the accuracy of
exposure models.30,32,37
Questionnaires and diaries. In early epidemiological
3.2.6
understanding of human activity patterns is critical as they
strongly influence assessment accuracy of actual human expo-
sure. Human activity patterns may directly indicate the distri-
bution of time among activities and the factors that influence the
degree of media contamination in the activities, and indirectly
reflect the intensity of contact usually measured by breath
volume during the activities. Few studies directly consider
human activity patterns in detail due to a lack of available data,
complexity and high costs of data collection. Instead, researchers
use microenvironments (e.g. the interior of a car, inside a house,
or indoor and outdoor areas) as a proxy of human activity
patterns. Microenvironments are defined as a location where the
concentration of an air pollutant is considered to be spatially
homogenous during the time that individuals are exposed.46–48
Under the microenvironment assumption, the exposure estimate
is defined as the following:37
Human activity patterns and microenvironments. An
Ei¼
X
J
j
Cjtij
(1)
where Eiis the total exposure for person i over the specified
period of time, Cjis the pollutant concentration in microenvi-
ronment j, tijis the time of the person i in microenvironment j,
and J is the number of microenvironments. The units of total
exposure is expressed in mg d m?3. However, we should note that
many of the uncertainties (e.g. additional exposure) are included
in this simple microenvironmental exposure model, as it was
found that the total exposure can be significantly attributed to
Table 2
guidelines (AQG)
Criteria air pollutants suggested by the WHO air quality
PollutantAveraging time AQG value/mg m?3
PM2.5a
1 year
24 h
1 year
24 h
8 h
1 year
1 h
24 h
10 min
15 min
30 min
1 h
8 h
—
10
25
20
50
100
40
200
20
500
100
60
30
10
—
PM10a
O3a
NO2a
SO2a
COb
Leadc
aValues from air quality guidelines—global update 2005: particulate
matter, ozone, nitrogen dioxide and sulfur dioxide.34 bValues from air
quality guidelines for Europe, 2nd edition.35 cA guideline value is not
specified by the WHO.
This journal is ª The Royal Society of Chemistry 2009J. Environ. Monit., 2009, 11, 475–490 | 477

Page 4

the exposure in certain microenvironments with high PM levels
where less time is spent.49
3.3Framework of air pollution exposure assessment
The means by which individuals come into contact with envi-
ronmental contaminants include inhalation, ingestion, and/or
skin (dermal) absorption, which often depends on the pollutant
exposed to. Air pollution exposure can be defined as a process of
hazardous substances in air entering an organism through
exposure inhalation. Based on an early framework proposed by
Nieuwenhuijsen et al.,50the fundamental process of air pollution
exposure is illustrated in Fig. 1.
As indicated in Fig. 1, air pollutants usually derive from both
outdoor and indoor pollution sources. Both ambient (outdoor
background) exposure and indoor exposure should be consid-
ered if we attempt to assess personal exposure accurately. The
methods used to assess personal exposure include personal
monitoring, biological monitoring (e.g. biomarkers), and envi-
ronmental monitoring/modeling. The former two are often
utilized to assess indoor exposure and ambient exposure while
the latter one has more extensively been applied to ambient
exposure alone. In this review, we place emphasis on environ-
mental modeling and biomarkers.
4. Exposure modeling
According to the WHO, exposure modeling is a ‘logical or
empirical construct which allows estimation of individual or
population exposure parameters from available input data’.51
Thus, air pollution exposure models can be developed to estimate
exposures and doses of individuals, defined population sub-
groups or entire populations. The aim of these models is to reflect
‘real-world’ human exposure to air pollutants over time and
consequently assess the health outcomes of air pollution expo-
sure.52Moreover, compared with direct air pollution exposure
assessment methods (personal monitoring and biological moni-
toring), the advantages of exposure modeling are as follows:53
? exposure models are typically used to supplement the
monitoring data where direct measurements are not available or
where these techniques are not appropriate for measuring air
pollution concentration;
? they can predict future exposures, as well as reconstruct
historical exposure by utilizing existing data from different types
and sources;
? the degree of complexity adopted by the model can be set
according to the needs of the assessment;
? contributions of different chemicals can be clearly separated
in exposure assessment; and
? source-intensive monitoring programs and expenses can be
reduced, allowing for economicaly substantial extension of air
pollution exposure assessment.
As early as 1991, various models for estimating exposure to air
pollutants were been developed. These models are generally
classified into three groups: (i) statistical, (ii) mathematical
(deterministic), and (iii) mathematical–stochastic models.53The
development of models to estimate air pollution exposure has
recently received enhanced attention from the research commu-
nity. Jerrett et al.54categorized the exposure models for air
pollution as proximity models, dispersion models, land use
regression models, interpolation models, integrated meteorolog-
ical-emission models, and hybrid models. The authors noted that
proximitymodels are relatively crudeestimators,because noneof
the parameters influencing the dispersion process of pollutants
are generally considered. Hybrid models were recommended as
having the greatest potential and reliability. Fryer et al.55
reviewed various environmental models, including air pollution
exposure models as both environmental concentration models
and ‘human intake’ models. Environmental concentration
models were described as simulating environmental processes to
generate chemical concentrations in particular media to which
humans may come into contact. Human intake models went one-
Fig. 1 A framework of air pollution exposure and its possible assessment approaches (modified from Nieuwenhuijsen et al.50).
478 | J. Environ. Monit., 2009, 11, 475–490This journal is ª The Royal Society of Chemistry 2009

Page 5

step further by quantifying human chemical intake from contact
with the relevant environmental media. The paper cautioned that
results from exposure modeling may not always be reliable,
especially since air pollution exposure is often affected by many
factors that are difficult to quantify accurately, as well as extra
confounding factors. In addition, data availability, model val-
idity, variability and uncertainty are also the current key issues
that need to be resolved in future exposure modeling.54,55
The aim of the following section is not to exhaustively
summarize and classify the universe of research on exposure
models for air pollution. Instead, it attempts to examine the
ambient air pollution exposure assessment literature with focus
on historical and recent developments in air pollution exposure
models for epidemiological studies. The types of models reviewed
below include proximity models, air dispersion models, hybrid
models, and human inhalation models. Recent application
progress of these models in epidemiological studies will be pre-
sented first followed by a critique of each model design.
4.1Proximity models
Proximity models are the most basic approach for assessing
human exposure to air pollutants. Proximity models measure the
proximity of a receptor to a pollution source in epidemiological
studies. These models are based on the assumption that exposure
at locations nearer to an emission source are higher relative to
positions further from the source. Although there have been
many recent studies on the relationships between air pollution
and health outcomes using proximity as a proxy for exposure, the
measurement data are not always available and/or the causalities
of disease and pollutants are not always clear. Examples of
proximity studies refer to the associations between residential
proximity to point, line, and area emission sources and respira-
tory disease,56lung cancer,57,58stroke mortality,59,60and birth
defects.61,62
For point and area sources of air pollution, the literature is
lending increasing attention to adverse health outcomes of waste
sites and industrial facilities by using simple proximity models.
Taking birth defects as an example, Orr et al.63suggested an
increased risk of chromosomal anomalies in offspring of women
who lived near hazardous waste sites. Conversely, Cordier et al.64
found no association between such anomalies and maternal
residence in communities with municipal waste incinerators. In
order to examine whether maternal residential proximity to
hazardous waste sites or industrial facilities with chemical air
emissions were associated with offspring chromosomal anoma-
lies, Brender et al.19designed a case-control study in Texas. The
study found that women 35 years or older who lived within one
mile of industries with emissions of heavy metals were two times
more likely (odd ratio (OR), 2.4; 95% confidence interval (CI)
1.1–4.1) than women living farther away to have offspring with
chromosomal anomalies. Maternal residences within one mile of
hazardous waste sites (relative to farther away) were not asso-
ciated with chromosomal anomalies in offspring except for
Hispanic births after the adjustment for maternal age, race/
ethnicity, and education. Suarez et al.22examined the relation-
ship between maternal proximity to National Priorities List
(NPL) or state superfund sites and neural tube defect (NTD)
risk, resulting in similar findings.
Residential proximity to roads is the primary domain in air
pollution exposure modeling-based epidemiological studies in
recent years, as exposure to air pollution is mostly dominated by
traffic emissions while industrial emissions are now measured to
a greater extent.65There is growing evidence that proximity to
traffic is an accurate surrogate for exposure to traffic-related air
pollution.66–68In early studies, researchers simply used the resi-
dential proximity distance to the closest main road as proxy for
traffic air pollution exposure. Miyake et al.66reported that
proximity of residence, not school, to major roads may be
associated with an increased prevalence of allergic disorders,
headache, stomach ache, and fatigue among Japanese adoles-
cents. Maheswaran and Elliott59utilized the distance from the
centroid of each district population to the nearest main road as
a proxy for exposure. The study found that stroke mortality was
lower by 7% (95% CI, 4–9%) for men living $ 1000 m away from
a main road compared with men living within 200 m away and
these risks increased with decreasing distance from main roads.
Roosbroeck et al.69measured personal exposure to traffic-related
pollutants in children to validate exposure classification based on
school location, and confirmed that the school’s proximity to
a freeway can be used as a valid estimate of exposure in epide-
miological studies on the effects of the traffic-related air pollut-
ants: soot and NOx, in children.
The most recent proximity-model studies of air pollution
exposure and health combine proximity measures with other
factors such as road type, traffic flow and land use. Wilhelm and
Ritz70used distance-weighted traffic density (DWTD) as a proxy
for accounting different exposure to air pollutants. They
concluded that exposure to traffic-related pollutants is important
for low birth weight (LBW) (OR, 1.39; 95% CI, 1.16–1.67) and
preterm birth (OR, 1.15; 95% CI, 1.05–1.26) during fall/winter
months. Hochadel et al.71examined the relationship between
daily traffic flow and maximum traffic intensity of buffers with
radii from 50 to 10 000 m, as well as distances to main roads
and highways, and ‘truly’ traffic-related pollutants by using
a geographic information system (GIS). The study found that
traffic-based predictors were strongly correlated with NO2
concentration (r2¼ 0.81) and PM2.5absorbance (r2¼ 0.81).
Sahsuvaroglu et al.72established a land use regression model
(LUR) including variables such as traffic density, industrial land
use, open space, location downwind from the Highway 403, and
distances from a highway and Lake Ontario in Canada. The
results confirmed that a LUR can provide more accurate esti-
mates of traffic pollution in areas with large industrial sources
and complex meteorological and topographic conditions than
those based on distance to road calculations. Brauer et al.73also
confirmed the linkages between birth outcomes and traffic-
related air pollution recently through the study for observing the
population-based cohort with low outdoor air pollution expo-
sure.
In addition, GIS are increasingly used in estimating proximity
exposure to air pollution, and have been recently recognized as
a potential method to minimize the exposure misclassifications,
which is vital to the accuracy of exposure assessment produced
by proximity models.74,75In many of the early epidemiological
studies, GIS were utilized to define study population, and to
establish the exposure surrogate on the basis of distance of study
population to the nearest contaminant sources.76More recently,
This journal is ª The Royal Society of Chemistry 2009 J. Environ. Monit., 2009, 11, 475–490 | 479