Vous êtes sur la page 1sur 8

Research | Mini-Monograph

Air Pollution Exposure Assessment for Epidemiologic Studies of Pregnant

Women and Children: Lessons Learned from the Centers for Children’s
Environmental Health and Disease Prevention Research
Frank Gilliland,1 Ed Avol,1 Patrick Kinney,2 Michael Jerrett,1 Timothy Dvonch,3 Frederick Lurmann,4
Timothy Buckley,5 Patrick Breysse,5 Gerald Keeler,3 Tracy de Villiers,1 and Rob McConnell 1
1Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California, USA;
2Mailman School of Public Health, Columbia University, New York, New York, USA; 3School of Public Health, University of Michigan,
Ann Arbor, Michigan, USA; 4Sonoma Technology, Inc., Petaluma, California, USA; 5Bloomberg School of Public Health, Johns Hopkins
University, Baltimore, Maryland, USA

Lessons Learned in Air

The National Children’s Study is considering a wide spectrum of airborne pollutants that are
hypothesized to potentially influence pregnancy outcomes, neurodevelopment, asthma, atopy,
Pollution Exposure
immune development, obesity, and pubertal development. In this article we summarize six applic- Assessment
able exposure assessment lessons learned from the Centers for Children’s Environmental Health An essential design element of environmental
and Disease Prevention Research that may enhance the National Children’s Study: a) Selecting epidemiologic studies is the a priori considera-
individual study subjects with a wide range of pollution exposure profiles maximizes spatial-scale tion of exposure assessment to ensure that the
exposure contrasts for key pollutants of study interest. b) In studies with large sample sizes, long study exposure range will maximize the ability
duration, and diverse outcomes and exposures, exposure assessment efforts should rely on model- to evaluate key exposure–response relationships
ing to provide estimates for the entire cohort, supported by subject-derived questionnaire data. (Navidi et al. 1994, 1999). Study population
c) Assessment of some exposures of interest requires individual measurements of exposures using selection and exposure assessment design are
snapshots of personal and microenvironmental exposures over short periods and/or in selected linked. Successful selections require considera-
microenvironments. d) Understanding issues of spatial–temporal correlations of air pollutants, the tion of the developmental time frames of inter-
surrogacy of specific pollutants for components of the complex mixture, and the exposure misclas- est and the biologic outcome mechanisms, in
sification inherent in exposure estimates is critical in analysis and interpretation. e) “Usual” tem- addition to understanding the spatial character-
poral, spatial, and physical patterns of activity can be used as modifiers of the exposure/outcome istics of airborne indoor and ambient expo-
relationships. f) Biomarkers of exposure are useful for evaluation of specific exposures that have sures. One potentially successful design strategy
multiple routes of exposure. If these lessons are applied, the National Children’s Study offers a is to maximize the number of contrasting pol-
unique opportunity to assess the adverse effects of air pollution on interrelated health outcomes lution profiles among study subjects by using a
during the critical early life period. Key words: air pollution, airborne, ambient, Centers for quasi-factorial approach to select populations
Children’s Environmental Health and Disease Prevention Research, Children’s Centers, cohort distributed over geographic regions with differ-
study, direct measurement, exposure assessment, modeling, National Children’s Study, personal ent pollution profiles (and/or including homes
measurement. Environ Health Perspect 113:1447–1454 (2005). doi:10.1289/ehp.7673 available with different indoor sources and proximity to
via http://dx.doi.org/ [Online 24 June 2005] specific sources) (Gauderman et al. 2000).
The National Children’s Study proposes to
investigate the relationships between patterns
A major study design challenge for the monthly, yearly, and multiyear metrics may be and histories of exposure during critical peri-
National Children’s Study will be to maxi- most relevant. For these and other outcomes, ods and the development of disease in later life.
mize and characterize exposure contrasts in its time-integrated average levels may capture the This creates an inherent tension because expo-
cohort of 100,000 pregnant women residing effects of chronic exposure during specific peri- sure assessment in large cohort studies requires
in multiple locations across the United States, ods, but more discrete and intense sampling
thereby enhancing the power to estimate frequency or duration may be needed to better This article is part of the mini-monograph “Lessons
exposure–response relationships from child- assess specific exposure–response relationships. Learned from the National Institute of Environmental
hood into adulthood. Multiple outcomes are The purpose of this article is to summa- Health Sciences/U.S. Environmental Protection
of interest, including pregnancy outcomes, rize exposure assessment lessons learned in Agency Centers for Children’s Environmental Health
neurodevelopment, asthma, obesity, and the Centers for Children’s Environmental and Disease Prevention Research for the National
Children’s Study.”
pubertal development. Exposures to a wide Health and Disease Prevention Research Address correspondence to F. Gilliland, Department
spectrum of environmental pollutants are (hereafter Children’s Centers) for air pol- of Preventive Medicine, USC Keck School of
being considered for investigation in the lutants and health outcomes of National Medicine, 1540 Alcazar St., CHP 236, Los Angeles,
study, including air pollutants of indoor and Children’s Study interest. Exposures to aller- CA 90033 USA. Telephone: (323) 442-1309. Fax:
outdoor origin (National Children’s Study gens and bioaerosols are considered elsewhere (323) 442-3272. E-mail: gillilan@usc.edu
2004). in this mini-monograph. Many of the This work was supported by the National Institute
of Environmental Health Sciences (ES009581,
Given the pollutants and health endpoints Children’s Centers have active research pro- ES007048, ES009589, ES009600, ES009142,
currently under consideration, exposure assess- grams involving the assessment of air pollu- ES009089, ES003819, ES009606, ES10688), the
ment for the variable periods during preg- tion in epidemiologic studies (Table 1). On U.S. Environmental Protection Agency (R826708,
nancy, infancy, and childhood will be needed. the basis of experience of investigators from R827027, R826724, and R826710), the National
For asthma-related outcomes, daily, monthly, these centers, we provide recommendations Heart, Lung and Blood Institute (HL61768), the
yearly, and multiyear exposure metrics with for air pollution exposure assessment consid- Hastings Foundation, the Canadian Institutes of
Health Research, and the National Children’s Study.
varying time integration periods may be eration in the study design, population selec- The authors declare they have no competing
required. For pregnancy outcomes, monthly tion, exposure data collection, analysis, and financial interests.
estimates as well as estimates for critical periods interpretation of findings of the National Received 12 October 2004; accepted 24 March
may be needed. For neurodevelopment, Children’s Study. 2005.

Environmental Health Perspectives • VOLUME 113 | NUMBER 10 | October 2005 1447

Gilliland et al.

a compromise between the optimal infor- Several modeling frameworks are appli- high-quality input data on the appropriate
mation obtained from individual measure- cable to the National Children’s Study. Basic geographic scale and the need for validation
ments and feasibility constraints related to approaches rely on using questionnaire and calibration studies to enable exposure
sampling methods, respondent burden, and responses as a surrogate for exposure and on uncertainty assignments. There are important
cost. Feasibility considerations likely dictate assigning exposures based on air pollutants limitations of modeling air pollution expo-
that direct measurements will be limited to measured at a central monitor. The latter sures (Sarnat et al. 2001). Studies indicate that
subsets of subjects monitored for short time approach has been successfully employed to for some pollutants, such as particulate matter
periods (“snapshots”) in selected microenviron- detect significant health effects (Dockery et al. (PM) and volatile organic compounds, indoor
ments, whereas exposure metrics used in 1993; Gauderman et al. 2002; Pope et al. sources can predominate (Sax et al. 2004;
chronic effects analyses for the entire cohort 2002; Ritz et al. 2000; Samet et al. 2000). Tonne et al. 2004; Wallace et al. 2004). Any
will be time-integrated over extended periods More refined approaches allow for estimation strategy that relies on ambient modeling
(days to months). The proposed size and dura- within communities using dispersion models should also attempt to assess indoor exposures
tion of the National Children’s Study will and information on transport, land use, and in subsamples of homes and thorough ques-
require the use of modeling to estimate time- meteorology (Brauer et al. 2002; English et al. tionnaire or inspection data that examine
integrated exposures for the entire cohort even 1999; Finkelstein et al. 2003; Hoek et al. important potential sources such as smoking
when direct measurements using snapshots of 2002; Nafstad et al. 2004). Considerations for habits or the presence of an attached garage.
exposure are available for subsets of the cohort. modeled exposures include the availability of This is especially needed for air pollutants for

Table 1. Centers for Children’s Environmental Health and Disease Prevention Research air pollution exposure assessment experience relevant to the National
Children’s Study.
Johns Hopkins University USC Children’s University of
Columbia University University of Michigan Health Study Southern California
Sample 500 pregnant women ~ 250 children with asthma 300 children, moderate to ~ 6,000 public school 202 Los Angeles public
population enrolled in the third in urban Baltimore severe asthma, 7–11 years children, 9–18 years of age school children, 6–16
trimester, and children (ages 2–12) of age at baseline in four specific age cohorts, years of age with asthma
followed from birth from 12 southern California and allergy to house dust
through age 5 communities mite or cockroach
Outcome(s) Asthma and Asthma severity Daily symptom diaries and Pulmonary function (PFTs), Asthma severity
neurodevelopment; pulmonary function symptoms (from annual
follow-up at multiple (PEF, FEV1) medical and residential
time points starting histories for 10 years),
at birth; outcome metrics school-reported absences,
include questionnaires, food-frequency dietary
biomarkers, clinical information, physical
assessments, activity, smoking and ETS,
neurobehavioral GxE interactions
Study Prospective birth cohort Longitudinal intervention Longitudinal intervention Cross-sectional survey Randomized trial
design study with exposures trial (n = 100); longitudinal trial (n ~ 3,600); longitudinal of allergen-reduction
and outcomes measured cohort study (n = 150); cohort study (n ~ 5,600) strategies
at multiple time points cross-sectional
starting during the third case–control study
trimester of pregnancy
Agents Personal PAH and pesticide Indoor/outdoor air pollutants Personal/indoor/outdoor Outdoor air pollutants Settled allergens
assessed exposures of mother in (PM10, PM2.5, O3, nicotine); air pollutants (PM10, PM2.5, [O3, NO2, PM10, PM2.5, (dust mite and cockroach)
third trimester; dust airborne endotoxin and O3, nicotine); PM components acid vapor (HNO3, formic, and endotoxin;
allergens prenatal, mouse allergen; allergens in (trace elements, EC, OC, acetic) EC, OC, PM speciation cockroach counts
12 months, 36 months, reservoir dust (cockroach, endotoxin) (SO4, NO3, NH4, CI)],
and 60 months; mouse, dust mite, cat, dog) PAHs, endotoxin, air toxics,
indoor/outdoor PM2.5, ETS, cigarette smoke
black carbon, and NO2
at 12 months in subset;
biomarkers for ETS,
PAH–DNA adducts,
Other GIS assessment of traffic Home inspection, time– Home inspection, time– Annual residential history by Housing characteristics
exposure proximity; social condition activity data, GIS location, activity data, GIS location, written survey; time–activity and condition, reported
determinants and stress; home meteorology meteorology data, GIS location, traffic and observed behavior,
characteristics density, and proximity humidity and moisture
Assessment Prenatal exposures to PAH Primary exposure assignment Primary exposure assignment Primary exposure assignment Assessment of only indoor
strategy based on personal sampling based on indoor air pollutants, using personal/indoor/outdoor based on community ambient settled dust; no outdoor
and cord blood PAH–DNA and allergens; secondary air pollutants; secondary monitors; secondary exposure assessment
adducts at birth; allergen exposure assignment using exposure assignment using assignment using
exposures based on dust microenvironmental model microenvironmental model microenvironmental model
measures; postnatal air with indoor/outdoor with outdoor air pollution
pollution exposures based air pollution combined with combined with home
on prediction model time–activity information characteristics and time–
developed in subset activity information
Abbreviations: CI, chlorine; EC, elemental carbon; FEV1, forced expiratory volume in 1 sec; GIS, geographic information system; GxE, gene–environment interaction; OC, organic carbon;
PEF, peak expiratory flow; PFT, pulmonary function test.

1448 VOLUME 113 | NUMBER 10 | October 2005 • Environmental Health Perspectives

Lessons learned: air pollution exposure assessment

which indoor sources are often the most sig- and meteorologic processes. Identification of of interest. Because multiple pollutants are of
nificant contributors (Payne-Sturges et al. source contributions within specific geographic interest for the National Children’s Study,
2004). Understanding and assessing the role regions may enhance interpretability of single priorities must be established to allow identi-
of exposure measurement error in health pollutant associations with health outcomes fication of individuals with a wide range of
effects assessment are central issues for the (Laden et al. 2000; Samet et al. 2000). exposure profiles for those key pollutants of
design and implementation of health effect In the following sections, we provide rec- study interest.
cohort studies (Jerrett and Finkelstein 2005). ommendations and issues that may need to be Issues to consider include spatial scale vari-
Finally, interpretation of National Chil- considered in implementing them. These are ations of pollutants, in order to select a study
dren’s Study findings will require information supported by some specific examples from the population that maximizes exposure contrasts
about specific pollutant surrogates because of Children’s Centers listed in Table 1. (Table 2). Table 2 identifies the spatial scales
the complex mixture of covarying pollutants of variability for ambient pollutants to con-
in respirable air (Manchester-Neesvig et al. Specific Recommendations sider in the study design for the National
2003). Pollutants covary because they are National Children’s Study subject selection. Children’s Study. The scales are categorized as
emitted from common sources or are pro- Study populations should be selected to maxi- regional (100–1,000 km), urban (4–50 km),
duced by common atmospheric chemistry mize spatial exposure contrasts for the pollutants neighborhood (50 m to 4 km), and household
(≤ 50 m, including outdoor and indoor
Table 2. Spatial scales of variability for ambient air pollutants. microenvironments). For some exposures,
contrast in exposure can be achieved by con-
Household scale sidering indoor sources and behavior (e.g.,
Regional scale Urban scale Neighborhood scale (≤ 50 m) outdoors
Compound (100–1,000 km) (4–50 km) (50 m to 4 km) and indoor
smoking vs. nonsmoking homes), if indoor-
source pollutant health effects are of interest.
Primary PM2.5 constituents For PM, the spatial scale variability of impor-
EC from combustion x x x
Organics, including PAHs x x tance depends on the constituents of interest.
Metals, including chromium VI, x x x For example, elemental carbon (EC) from
cadmium, lead, beryllium, ambient primary combustion processes varies
nickel, arsenic, iron, on urban and neighborhood scales. Indoor
manganese sources from combustion also contribute to
Other constituents from road x x
personal EC exposure (LaRosa et al. 2002). In
dust, wood smoke,
construction dust, and contrast, particulate sulfates typically vary on a
industrial sources regional scale. To maximize exposure gradi-
Secondary PM2.5 constituents ents to EC, subjects would need to be selected
Sulfate x on a neighborhood scale, such as based on dis-
Nitrate x x tance to busy roadways. Sulfates’ regional
Ammonium x x
nature would be better reflected in a subject
Secondary organics x x
Primary PM2.5–10 constituents selection scheme involving different regions of
Organics, including PAHs x x x the United States.
Metals, including chromium VI, x x To select subjects based on exposure con-
cadmium, lead, beryllium, trasts for ambient pollutants (e.g., ozone, sul-
nickel, arsenic, iron, fate), exposure data on geographic variation in
levels and spatial gradients over time are needed.
Other constituents from road x x
dust, wood smoke, For criteria pollutants, existing data are available
construction dust, and from a national network of monitoring stations.
industrial sources Data for many other pollutants of biologic inter-
Primary PM > 10 constituents est may be sparse or nonexistent (e.g., EC and
Pollen grains x x air toxics). In addition, for other pollutants with
O3 x x
both indoor and outdoor sources (e.g., PM
Nitric oxide x x
NO2 x x mass, nitrogen oxides, volatile organic com-
Sulfur dioxide x x pounds), much of the variability in exposure is
Carbon monoxide x x driven by indoor source activity and/or very
Volatile organic compounds proximate local sources (e.g., traffic). For these
Benzene x x pollutants, levels may need to be measured or
1,3-Butadiene x x
modeled with the appropriate spatial and tem-
Formaldehyde x x
Acetaldehyde x x poral resolution in pilot studies to ascertain the
Acrolein x x appropriate spatial, temporal, and behavioral
Vinyl chloride x x determinants. In addition to variable pollutant
Carbon tetrachloride x x source strengths, subject-specific temporal–spa-
Chloroform x x tial–physical patterns of activity may meaning-
Propylene dichloride x x
fully affect both within and between-group
Methyl chloride x x
Trichloroethylene x x exposure assignments. Capturing this variability
Tetrachloroethylene x x in applicably useful ways for large study popula-
Naphthalene x x tion studies is challenging and often a multi-
Mercury x x faceted approach using self-administered
Bioaerosols, including endotoxin, house dust allergens, fungal spores, and pollen grains, also vary considerably on the questionnaires, walk-through surveys, instru-
household and neighborhood scales; however, they were not included in this analysis. ment deployments, and sentinel monitoring.

Environmental Health Perspectives • VOLUME 113 | NUMBER 10 | October 2005 1449

Gilliland et al.

Because several pollutants of interest for were selected because they would be more expertise, and resulting accuracy and extrapo-
the National Children’s Study are regional likely to exhibit overlapping distributions of lation potential.
in nature, subject selection from areas with confounding risk factors and would allow Modeled estimates can be refined using
contrasting pollution profiles is likely to be adjustments for confounding in the analysis. targeted substudies designed to measure levels
most informative. The national scope of the Replication of exposure profiles was employed at geographic locations over time on the scale
National Children’s Study provides the oppor- to improve the chance of including demo- of spatial and temporal variation of the pollu-
tunity to maximize the number of study pro- graphically comparable communities and to tants under study. The time resolution of the
files. For example, the constituents of PM allow estimation of residual variance within exposure estimate needs to be appropriately
< 2.5 µm in diameter (PM2.5) within a region pollution profiles. Additional details have been matched to outcomes to capture effects of fre-
are highly correlated, but between regions the described previously (Berhane et al. 2004; quency, magnitude, and duration of peak or
correlations may be lower. PM2.5 sulfate is Peters et al. 1999a, 1999b). This design episodic exposure events that may have effects
higher in the eastern United States and lower resulted in contrasting exposure profiles for during windows of vulnerability. Long-term
in the western United States, whereas PM2.5 O 3 and a package of correlated pollutants average exposures, including average peak lev-
nitrate is lower in the eastern United States and (PM10, PM2.5, and NO2) primarily of mobile els or hours above threshold levels, are likely
higher in the western United States. Therefore, source origin. This approach can be extended more important for relationships with chronic
the comparable effect of these PM 2.5 con- to other pollutants, such as ultrafine particles disease, but this assumption needs to be eval-
stituents may be separable by study design. whose concentrations may also vary on a local- uated for specific agents and outcomes of
Replication of pollution profiles in different ized scale of ≤ 50 m. Selecting subjects within focus in the National Children’s Study.
regions is also important to allow for effects of communities based on the distance between Data availability and quality for model
geographic variables such as weather and other the home and the nearest busy roadway or input are critically important. Central-site
confounding variables to be controlled in the other traffic density metric may maximize the monitoring data can be used to assign exposure
analyses (Jerrett et al. 2003a, 2003b; Krewski exposure contrasts of ultrafines within the pro- for outdoor environments, but the utility of
et al. 2000; Peters 1997; Peters et al. 1999a). files of other pollutants such as O3. this assignment will depend on the relative
Exposures within homes with common sources Other potential valuable exposure sam- variability of the pollutant across the sampling
are also highly correlated and may be separated pling designs might consider “matrix” sam- area of interest (intra- vs. intercommunity vari-
by design. pling approaches, which would draw on ability issues). Estimates of indoor concentra-
An example of the integration of these subsets of subjects for specific substudies or tions require individual information on home
approaches is the Southern California Chil- specialty projects. In the larger perspective operating conditions, home source profiles and
dren’s Health Study (CHS), a study performed however, maximizing differences in commu- activity, factors influencing the penetration of
by investigators in the University of Southern nity exposure profiles can provide a rich pop- outdoor pollutants and/or the dilution of pol-
California (USC)/University of California at ulation base from which to develop and lutants of indoor origin (LaRosa et al. 2002;
Los Angeles Children’s Environmental Health inform multiple studies seeking to optimize Navidi et al. 1999). Information about tem-
Center. The USC CHS is a multiyear cohort the National Children’s Study effort. poral, spatial, and physical activity patterns
study of several thousand southern California Exposure metrics. Because of the large size, are also important but are likely to have insuf-
school children (Berhane et al. 2004; Kunzli long duration, and diversity of outcomes and ficient time resolution over the period of
et al. 2003; Peters 1997). The primary USC exposures of interest in the proposed National study interest. Broader categories of “usual”
CHS research question is whether ambient air Children’s Study, the exposure assessment patterns of activity, household operation, and
pollution causes chronic adverse respiratory effort should rely on modeling to provide susceptibility factors can be considered as
health effects during childhood and adolescent estimates for the entire cohort, supported by modifying factors for the exposure–response
growth and development. Almost 12,000 chil- subject-derived questionnaire data. Necessary relationship using available central-site moni-
dren from schools in 13 southern California survey information on temporal–spatial– toring data (Gauderman et al. 2000; Janssen
communities have been recruited into five physical patterns of activity and household et al. 2002).
cohorts since the study began in 1993. characteristics can be collected for the entire An existing national system of central site
Communities were selected to maximize cohort, and targeted exposure substudies can monitors collects continuous data on criteria
differences in outdoor air pollutant concen- be performed in selected subsamples of study air pollutants and more limited data on haz-
trations. To distinguish the effects of different subjects. ardous air pollutants [U.S. Environmental
pollutants, communities were selected to Issues to consider include modeling for Protection Agency (EPA) 2004]. It is possible
minimize the spatial correlations between large-scale investigations over long periods to add additional instruments to monitoring
three priority study pollutants [O3, nitrogen (e.g., the National Children’s Study), which is sites to measure additional pollutants or speci-
dioxide, and PM < 10 µm in diameter currently the only feasible approach for assign- ate PM at reasonable cost. However, the use
(PM 10 )]. However, the full quasi-factorial ing exposure estimates for the entire cohort. of central-site monitoring data for epidemiol-
design could not be fulfilled because all the This is especially true for ambient air pollu- ogy studies requires a quality assurance activ-
potential pollution profiles do not occur in tants that display significant spatial variation ity beyond that which is used for regulatory
nature. Specific community selections were on urban, neighborhood, or household spatial activities as well as methods to address miss-
based on historical air pollution levels for sev- scales. ing data issues. The Health Effects Institute
eral years before study inception, exposure A variety of exposure assessment modeling recently funded a study to compile existing
patterns, and census demographic data. approaches are available, including proximity- estimates of air toxics into a coherent national
Because of differences in the number of loca- based, geostatistical, land-use regression database. When available, these data may
tions at which pollutants were measured and (LUR), dispersion, integrated meteorologic contribute to the National Children’s Study,
the frequency and type of measurements emission, and hybrid approaches involving and selection of the sampling sites for the
made, data available for selecting communi- personal sampling in combination with National Children’s Study should take into
ties were more reliable for O3 than for PM10, one or more of the above methods (Jerrett account the location of existing and upcom-
and more reliable for PM10 than for NO2. et al. 2004). Each model varies by data input ing monitoring data. No similar monitoring
Demographically heterogeneous communities requirements, software/hardware, technical network exists to assess exposure from indoor

1450 VOLUME 113 | NUMBER 10 | October 2005 • Environmental Health Perspectives

Lessons learned: air pollution exposure assessment

sources, which may need to rely on question- housing survey data, residence and school- and their offspring (Miller et al. 2001; Perera
naire information and substudies across level traffic model estimates, and community- et al. 2003, 2004a; Tonne et al. 2004; Whyatt
regions. level air quality measurement data and et al. 2003). In the Columbia PCS, direct air
Modeling of pollutants with large intra- regional transport factors to estimate short- pollution exposure assessment begins in the
community variation requires additional com- term and long-term individual exposures. The third trimester of pregnancy with collection of a
munity measurements. Substudies can be model estimates show the largest amount of 48-hr personal sample of PM2.5 and vapors for
designed to exploit obtainable information for within-community variations in individual each pregnant woman. These samples are ana-
modeling study subject exposures (Jerrett et al. exposures of any of the models; however, vali- lyzed for polycyclic aromatic hydrocarbon
2005). These additional microenvironmental dating these types of models is difficult and (PAH) and pesticide concentrations (i.e., a
measurements can be used for fitting models to resource intensive (Peters 1997). “snapshot” measurement representing “usual”
better estimate exposure, for model validation, Newer modeling strategies such as LUR exposure). In a validation substudy, the investi-
and for assessment of errors in exposure assign- models are promising. LUR employs the pol- gators also collected sequential 2-week inte-
ments. Calibration studies using repeated per- lutant of interest as the dependent variable grated indoor samples, analyzed for the same
sonal monitoring may be designed and and proximate land use, traffic, and physical variables as above, for the entire third trimester
conducted to validate the exposure estimates environmental variables as independent pre- (preferred over the personal snapshot as an
and correct for exposure error in the analysis dictors. The methodology seeks to predict pol- exposure surrogate of third-trimester exposures,
(Berhane et al. 2004; Fraser and Stram 2001; lution concentrations at a given site based on but obviously more intensive laborwise, cost-
Mallick et al. 2002; Stram et al. 1995). surrounding land use and traffic characteris- wise, and subjectwise). A home dust sample was
An illustration of these approaches may be tics. The incorporation of land use variables also collected during the third trimester from
seen in the USC CHS. The USC CHS frame- into the interpolation algorithm detects small- subjects and analyzed for standard allergens rel-
work employed a hierarchical approach for area variations in air pollution more effectively evant to maternal exposures and possible prena-
estimating exposure, ranging from the coarsest than do standard methods of interpolation tal sensitization, based on evidence emerging
spatial estimates based on community pollu- (i.e., kriging) (Briggs et al. 1997, 2000; Lebret from the Columbia PCS (Miller et al. 2001).
tant levels measured at a single central monitor et al. 2000). These methods are promising for Another time interval of study exposure
per community, to the finest spatial-scale esti- the National Children’s Study because they interest was the first 2 years of life, when
mates based on integrated models for individ- can be extrapolated, based on land use cover- infants/toddlers spend substantial amounts of
ual exposure assessment. The framework age, without need for extensive monitoring in time in the home; this may be a critical expo-
involved the following pollutant measurement each location. Most major urban centers sure window for development of allergy and
and modeling levels: a) continuous monitoring maintain land use information, and the U.S. asthma. Columbia PCS homes were visited
of O3, NO2, and PM10, and of PM2.5 mass Census has much of the information needed when the child reached 1 year of age, and a
and composition on a time-integrated 14-day on population density and employment struc- dust sample was collected for allergen analysis.
basis, at a central monitoring station in each tures. The National Children’s Study could Additional sampling was performed in a subset
community; b) measurement of selected pollu- support the monitoring needed to calibrate of 25% of the homes, where 2-week samples of
tants at multiple locations within each com- LUR models that are regionally representative indoor and outdoor air PM2.5, black carbon,
munity; and c) adjustment of the central site of broad land use and emission patterns. and NO2 were collected. These samples are
monitor to the levels around children’s homes Derived coefficients could then be applied to being used to develop and test a spatial LUR
and schools based on a limited number of field other places within the region without need model that will then be used to estimate expo-
measurements. This framework is augmented for extensive monitoring. sures in the full cohort that are representative
by a) modeling of vehicle emissions using Use of limited substudies for exposure of those occurring in early childhood.
geostatistical methods and spatial dispersion refinement. Assessment of some exposures of As a part of its investigations of childhood
models, b) estimating outdoor pollutant con- interest will require individual measurements asthma in Baltimore, Maryland, the Johns
centrations at schools and homes for the entire of exposures using snapshots of personal and Hopkins Center for Asthma in the Urban
study population using spatial statistical mod- microenvironmental exposures over short peri- Environment (JHU Center) has conducted an
els in a hybrid microenvironmental approach, ods and/or in selected microenvironments. intervention trial and a cohort study of asthma
and c) modeling individual exposure estimates Issues to consider include the large number morbidity (Breysse et al. 2005; Swartz et al.
for the entire study population using unified of interrelated factors that are important in 2004). The exposure assessment efforts for
modeling methods that integrated information designing exposure substudies. These include these studies include indoor and outdoor air
with different spatial and temporal resolutions. the substudy’s purpose, the population sample pollution as well as indoor allergens in approx-
These unified methods include community to include, whether personal or microenviron- imately 400 homes. The major focus of these
monitored pollutant levels, studies of indoor mental samples should be collected, the studies was indoor air where investigators
and outdoor levels in homes and schools; step respondent burden, study feasibility, sample assessed 3-day average indoor PM10, PM2.5,
counters; questionnaire-based data on time– collection and analytic costs, temporal varia- NO2, O3, and nicotine at 3-month intervals
activity patterns including commuting pat- tion of exposure, subject activity patterns, (Breysse et al. 2005). In addition, 3-day time
terns, traffic patterns, and housing characteris- household operation by residents, and uses in resolved PM was assessed using a data-logging
tics; and appropriate accounting of uncertainty model validation and calibration. nephalometer. Ambient PM air pollution was
in the exposure estimates. These elements are nicely illustrated in the assessed using a monitoring site centrally
The USC CHS developed a microenvi- Columbia Pregnancy Cohort Study (PCS), a located to the study area.
ronmental exposure model that, in principle, study performed by the Columbia University Results from these studies demonstrate the
can provide estimates of exposures to pollu- Center for Children’s Environmental Health, importance of assessing indoor air. Children,
tants of ambient origin in five microenviron- which has focused on the effects of pre- and particularly young children, spend the great
ments. These include residential outdoors, postnatal exposures to air pollution on birth majority of their time in the home. Others
residential indoors, school outdoors, school outcomes and neurodevelopmental and respira- have noted (Wallace et al. 2004) that indoor
indoors, and inside vehicles. The exposure tory health outcomes in childhood via through PM concentrations are generally higher than
model uses individual-level time–activity and recruitment and follow-up of pregnant women outdoor levels, and cigarette smoking as well

Environmental Health Perspectives • VOLUME 113 | NUMBER 10 | October 2005 1451

Gilliland et al.

as other household activities are responsible correlations of air pollutants, the surrogacy of Exposure misclassification is a critical issue
for this increase. In some cases, the PM contri- specific pollutants for components of the for exposure assessment efforts, especially
bution from cigarette smoking to indoor PM complex mixture, and the exposure misclassi- modeled exposures. In most large cohort stud-
is greater than that penetrating from outdoor fication inherent in exposure estimates will be ies, it is not possible to accurately measure the
air. The JHU Center results indicate, for critical in analyzing and interpreting National true personal exposure of individuals over the
example, that a single cigarette contributes Children’s Study findings. time interval that is most relevant for the out-
between 1 and 2 µg/m 3 to indoor PM. In Issues to consider include the fact that air comes of interest. Thus, virtually all exposure
addition, a strategy that uses repeat measures pollutants occur as complex mixtures of gases assessments provide at best estimates of true
allows larger time frame variability to be and particles, but coexisting constituents may exposures, with some error. Errors may arise
assessed (e.g., seasonal). covary, based on their common sources or because of temporal factors (e.g., the exposure
Results from the Michigan Center for the photochemical pathways. The ambient level of metric captures only a snapshot of the relevant
Environment and Children’s Health demon- one pollutant may therefore be a surrogate for time interval) or spatial factors (e.g., the expo-
strate the importance of focusing on the other pollutants arising from the same source, sure metric is collected at a location different
home as an important microenvironment for so interpretation of findings for individual from where the subject lives and breathes).
children’s exposure (Keeler et al. 2002; Yip pollutants must account for this surrogacy Additionally, inherent imprecision in the spe-
et al. 2004). An important lesson from these (Manchester-Neesvig et al. 2003; Sarnat et al. cific method selected for study application
studies is that home-based exposure assess- 2001). Identification of pollutant sources may also result in some measurement error.
ments are feasible for studies involving hun- therefore provides a potentially important For the results of the study to ultimately be
dreds of children and need to be considered mechanism to evaluate source-specific health interpretable, it is important in designing the
in the National Children’s Study. This con- effects and can ultimately lead to effective study for investigators to analyze the nature of
clusion is particularly true for newborn chil- strategies for reducing population exposure. the exposure misclassification errors that are
dren who spend essentially all of their time in Substudies among subjects in differing likely to be present. Quantitative estimates of
the home. The microenvironments of impor- geographic locations may be useful for defin- exposure errors can be obtained by carrying
tance include the indoor environment in a ing pollutant relationships. For example, in out calibration substudies where results from
range of housing types, because there is a assessing PM, chemical tracers have been iden- more complete exposure metrics are com-
growing recognition that housing quality is an tified that can serve as “fingerprints” for indi- pared with results from the modeled metrics
important predictor of indoor air pollution vidual sources, or source types, of air pollution (Berhane et al. 2004; Fraser and Stram 2001;
and can affect outdoor pollution penetration (Laden et al. 2000; Manchester-Neesvig et al. Mallick et al. 2002; Sarnat et al. 2001; Stram
rates as well as being a general risk factor for 2003; Sarnat et al. 2002). This type of infor- et al. 1995). Bayesian statistical frameworks
poor health (Kingsley 2003). mation can be used to apportion contributions may assist with assessing the impact of meas-
As described above, the USC CHS experi- to the measured PM mass on a per sample urement error on the exposure–response rela-
ence suggests that exposure assignment accu- basis, along with providing data critical to the tionships (Berhane et al. 2004).
racy can be improved by conducting substudies assessment and interpretation of health effects Modifiers of exposure–outcome relation-
with a limited number of measurements associated with individual chemical compo- ships. “Usual” temporal, spatial, and physical
extended temporally and spatially. In evaluat- nents of PM. Quantitative assessments of patterns of activity can be used as modifiers of
ing the minimal sampling needed to success- source contributions for large data sets are the exposure–outcome relationships. Highly
fully predict long-term exposures in study often determined using a statistical receptor time-resolved activity information over the
communities, USC CHS investigators found modeling approach. This type of data analysis study period of interest may not be necessary,
that the intraclass correlation between esti- is best suited to longitudinal study designs and and is not likely to be available, for all National
mated annual average of pollutants, based on can be limiting because it may require collec- Children’s Study participants throughout the
2-week subset measurements, and the true tion of a large number of samples to obtain study. Personal exposure estimates, based on
annual average was greater than 0.9 for O3, robust results. time in microenvironments, are likely to be
NO2, and nitric oxide in southern California, The recent successful development and associated with large uncertainties. “Usual” pat-
if two winter, two summer, and one spring deployment of several types of continuous terns of activity, such as time usually spent out-
sample were obtained. Greater numbers of portable PM mass and number monitors offer doors, can be collected by questionnaire and
samples did not appreciably improve the corre- the potential for producing real-time (< 5-min used as modifiers of exposure–outcome rela-
lation. These results indicate that accurate esti- interval) data. The continuous data collection tionships (Gauderman et al. 2002). Activity-
mates of the pollutant annual average levels can format of these samplers allows a better under- level assignments may be important in moving
be obtained at homes, schools, and other cen- standing of source emission patterns and expo- from exposure to delivered dose of an airborne
tral site locations with a limited number of sures, especially in urban environments, and pollutant to the lung. For example, for asthma
samples. Local measurements can then be can be used to enhance investigations of short- prevalence and incidence, USC CHS investiga-
combined with concurrent central site meas- term peak exposures. These highly time- tors saw little association with community levels
urements to estimate neighborhood and house- resolved exposure data can be coupled with of exposure. However, when physical activity
hold scale concentrations for the entire cohort. personal time–activity pattern data to quanti- was considered, O3 was strongly associated with
Although the optimum number of samples tatively identify exposures from specific emis- asthma incidence (where variation entered from
may differ by region of the country or in differ- sion sources. To date, real-time PM samplers increased ventilation rates associated with exer-
ent neighborhoods within communities, do not yet offer the ability to determine PM cise and likely increased dose to the lung). An
depending on the pollutants of interest and chemical speciation. A combination of important challenge for the National Children’s
geographic and temporal variation in the methodologic approaches (employing chemi- Study is assessing activity patterns among
processes driving air pollution, this general cal tracers and continuous PM number and mothers, infants, and young children.
strategy may be of use in planning efficient mass count information) may improve the For extremely large study populations
National Children’s Study substudies. ability to identify specific sources and source for which individual questionnaires may be
Analytic and interpretation issues. types contributing to the measured exposure impractical to administer and/or collect, ran-
Understanding issues of spatial/temporal to PM and other pollutants. domized sampling schemes or oversampling in

1452 VOLUME 113 | NUMBER 10 | October 2005 • Environmental Health Perspectives

Lessons learned: air pollution exposure assessment

certain nested subsamples of possible increased cord blood levels in many cases (Whyatt et al. interest will require individual measure-
interest may be worth careful consideration. 2003). For some but not all pesticides, corre- ments of exposures using snapshots of per-
Use of biomarkers. Biomarkers of exposure lations also were demonstrated between sonal and microenvironmental exposures
offer utility for evaluation of specific exposures plasma levels at birth (either cord blood or over short periods and/or in selected micro-
that have multiple routes of exposure. For spe- maternal) and air measurements collected environments.
cific airborne pollutants, exposure assessments during the third trimester of pregnancy. Cord • Analytic and interpretation issues. Under-
may need to consider multiple routes of plasma, but not air, levels of the insecticide standing issues of spatial–temporal correla-
human exposure. In addition to inhalation, chlorpyrifos and diazinon were significantly tions of air pollutants, the surrogacy of
dermal absorption and oral ingestion may be associated with decreased birth weight and specific pollutants for components of the
important pathways of exposure for pollutants length (Whyatt et al. 2004). Of particular sig- complex mixture, and the exposure misclassi-
of interest with regard to young children, nificance, levels of several pesticides in both fication inherent in exposure estimates will be
infants, and pregnant or lactating mothers. air and plasma showed significant declines critical in analyzing and interpreting findings
The use of exposure biomarkers is one poten- across women enrolled before and after the from the National Children’s Study.
tially valuable approach in this area (Weaver U.S. EPA insecticide phase-out (Whyatt et al. • Modifiers of exposure–outcome relationships.
et al. 1998). Interpreting the relationship 2003). Furthermore, associations with adverse “Usual” temporal, spatial, and physical pat-
between these markers and exposures, how- birth outcomes were significant only for terns of activity can be used as modifiers of
ever, is a complex function of the timing and infants born before the phase-out (Whyatt the exposure/outcome relationships.
routes of exposure, and of the pollutant toxi- et al. 2004). These findings illustrate the util- • Use of biomarkers. Biomarkers of exposure
cokinetics. As discussed above, temporal–spa- ity of well-targeted biomarker measurements, may be required for evaluation of specific
tial–physical patterns of activity will almost in conjunction with health and external expo- exposures that have multiple routes of
surely affect this dynamic in important ways, sure measures, for birth cohort studies. exposure.
from modification of ventilation rates to facili- Cotinine and nicotine as markers for ETS, We have learned that there are many chal-
tated dermal absorption during periods of ele- an important source of PM exposure, has a lenges to assessing air pollution exposures to
vated, increased, or extended activities. As long history of use in biomonitoring. Hair children. To overcome these challenges, the
exposure assessment tools, biomarkers offer nicotine has the potential to provide estimates National Children’s Study will need to commit
the potential advantage of integrating the net of ETS exposure over a 2–3 month period or extensive resources to exposure assessment
effect of all of these factors in producing a longer (Jaakkola and Jaakkola 1997), and activities. With optimal subject selection, expo-
given internal dose for a given individual. other nicotine metabolites (e.g. cotinine) may sure estimates can be modeled for the entire
Such measurements may better represent true be useful indicators of both exposure and cohort, supported by direct measurement of
health-relevant exposures for an individual bioavailability. selected pollutants in a subset of the study pop-
than any external measure of exposure can. ulation. Biomonitoring is likely to be a valu-
Biomarker measurements are substantially Summary able adjunct to the exposure assessment design,
integrated into the exposure and health assess- The National Children’s Study offers a unique helping to trace the mechanistic linkages
ment designs of the Columbia PCS. From an opportunity to understand the adverse effects between exposures and effects. Prioritization of
exposure perspective, biomarkers focus on of air pollution on a broad range of interre- pollutants of study interest and developmental
DNA-bound PAHs (Perera et al. 2004a, lated outcomes during the critical period of periods of study focus would allow optimiza-
2004b), pesticides in blood plasma and meco- early life development and growth. Six recom- tion of the study design for the National
nium (Perera et al. 2003; Whyatt et al. 2001, mendations for air pollution exposure assess- Children’s Study to maximize contrasting pol-
2003, 2004), and the environmental tobacco ment are proposed from lessons learned in the lution profiles and enhance the ability to assess
smoke (ETS) metabolite cotinine in urine Children’s Centers. exposure–response relationships.
(Perera et al. 2004b), beginning with mater- • National Children’s Study subject selection.
nal and infant cord blood samples at birth, Study populations should be selected to REFERENCES
and continuing with follow-up assessments in maximize spatial-scale exposure contrasts for
the child at 2 and 5 years of age. PAH-DNA the pollutants of interest. Because multiple Berhane K, Gauderman WJ, Stram DO, Thomas DC. 2004.
Statistical issues in studies of the long term effects of air
adducts also can be viewed as early measures pollutants are of interest for the National pollution: the Southern California Children’s Health Study.
of procarcinogenic health effects (Perera et al. Children’s Study, priorities must be estab- Stat Sci 19(3):414–449.
2004b). Other effect-related biomarkers focus lished to allow identification of individuals Brauer M, Hoek G, Van Vliet P, Meliefste K, Fischer PH,
Wijga A, et al. 2002. Air pollution from traffic and the
on the time course of sensitization to environ- with a wide range of exposure profiles for development of respiratory infections and asthmatic and
mental allergens, including measurements of those key pollutants of study interest. allergic symptoms in children. Am J Respir Crit Care Med
maternal, cord-blood, and child IgE, and pro- • Exposure metrics. Because of the large size, 166(8):1092–1098.
Breysse P, Buckley T, Williams D, Beck C, Kanchanaraksa S,
duction of proinflammatory cytokines or pro- long duration, and diversity of outcomes Swartz L, et al. 2005. Indoor exposures to air pollutants
liferation of mononuclear cells in response to and exposures of interest in the proposed and allergens in the homes of asthmatic children in inner-
specific allergens (Miller et al. 2001). National Children’s Study, the exposure city Baltimore. Environ Res 98(2):167–176.
Briggs D, Collins S, Elliott P, Kingham S, Lebret E, Pryl K, et al.
The integration of newly developed pesti- assessment effort should rely on modeling to 1997. Urban air pollution GIS: a regression-based approach.
cide biomarkers within the epidemiologic provide estimates for the entire cohort, sup- Int J Geogr Inf Sci 11:699–718.
design of the Columbia PCS has made possi- ported by subject-derived questionnaire Briggs D, de Hoogh C, Gulliver J, Wills J, Elliott P, Kingham S,
et al. 2000. A regression-based method for mapping traffic-
ble significant new advances in our under- data. Necessary survey information on tem- related air pollution: application and testing in four con-
standing of the health effects and patterns of poral–spatial–physical patterns of activity trasting urban environments. Sci Total Environ 253:151–167.
exposures to pesticides among urban women and household characteristics can be col- Dockery D, Pope CA, Xu X, Spengler J, Ware J, Fay M, et al.
and children (Perera et al. 2003; Whyatt et al. lected for the entire cohort, and targeted 1993. An association between air pollution and mortality in
six U.S. cities. N Engl J Med 329:1753–1759.
2001, 2003, 2004). A wide range of pesticides exposure substudies can be performed in a English P, Neutra R, Scalf R, Sullivan M, Waller L, Zhu L. 1999.
have been shown to be quantifiable in the selected subsample of study subjects. Examining associations between childhood asthma and
plasma of women and their newborns, with • Use of limited substudies for exposure refine- traffic flow using a geographic information system.
Environ Health Perspect 107:761–767.
significant correlations between maternal and ment. Assessment of some exposures of Finkelstein M, Jerrett M, DeLuca P, Finkelstein N, Verma DK,

Environmental Health Perspectives • VOLUME 113 | NUMBER 10 | October 2005 1453

Gilliland et al.

Chapman K, et al. 2003. A cohort study of income, air pollu- Lebret E, Briggs D, van Reeuwijk H, Fischer P, Smallbone K, Pope CA, Burnett RT, Thun MJ, Calle EC, Krewski D, Ito K, et al.
tion and mortality. Can Med Assoc J 169:397–402. Harssema H, et al. 2000. Small area variations in ambient 2002. Lung cancer, cardiopulmonary mortality, and long-
Fraser GE, Stram DO. 2001. Regression calibration in studies with NO2 concentrations in four European areas. Atmos Environ term exposure to fine particulate air pollution. JAMA
correlated variables measured with error. Am J Epidemiol 34(2):177–185. 287(9):1132–1141.
154(9):836–844. Mallick R, Fung K, Krewski D. 2002. Adjusting for measurement Ritz B, Yu F, Chapa G, Fruin S. 2000. Effect of air pollution on
Gauderman WJ, Gilliland GF, Vora H, Avol E, Stram D, McConnell error in the Cox proportional hazards regression model. preterm birth among children born in southern California
R, et al. 2002. Association between air pollution and lung J Cancer Epidemiol Prev 7(4):155–164. between 1989 and 1993. Epidemiology 11(5):502–511.
function growth in Southern California children: results from Manchester-Neesvig JB, Schauer JJ, Cass GR. 2003. The distri- Samet J, Dominici F, Curriero F, Coursac I, Zeger S. 2000. Fine
a second cohort. Am J Respir Crit Care Med 166(1):76–84. bution of particle-phase organic compounds in the atmos- particulate air pollution and mortality in 20 U.S. cities,
Gauderman WJ, McConnell R, Gilliland F, London S, Thomas D, phere and their use for source apportionment during the 1987–1994. N Engl J Med 343(24):1742–1749.
Avol E, et al. 2000. Association between air pollution and Southern California Children’s Health Study. J Air Waste Sarnat JA, Long CM, Koutrakis P, Coull BA, Schwartz J, Suh HH.
lung function growth in southern California children. Am J Manag Assoc 53(9):1065–1079. 2002. Using sulfur as a tracer of outdoor fine particulate
Respir Crit Care Med 162(4 Pt 1):1383–1390. Miller RL, Chew GL, Bell CA, Biedermann SA, Aggarwal M, Kinney matter. Environ Sci Technol 36(24):5305–5314.
Hoek G, Brunekreef B, Goldbohm S, Fischer P, Brandt PA. 2002. PL, et al. 2001. Prenatal exposure, maternal sensitization, and Sarnat JA, Schwartz J, Catalano PJ, Suh HH. 2001. Gaseous
Association between mortality and indicators of traffic- sensitization in utero to indoor allergens in an inner-city pollutants in particulate matter epidemiology: confounders
related air pollution in the Netherlands: a cohort study. cohort. Am J Respir Crit Care Med 164(6):995–1001. or surrogates? Environ Health Perspect 109:1053–1061.
Lancet 360(9341):1203–1209. Nafstad P, Håheim L, Wisløff T, Gram F, Oftedal B, Holme I, Sax SN, Bennett DH, Chillrud SN, Kinney PL, Spengler JD. 2004.
Jaakkola M, Jaakkola J. 1997. Assessment of exposure to envi- et al. 2004. Urban air pollution and mortality in a cohort of Differences in source emission rates of volatile organic
ronmental tobacco smoke. Eur Respir J 10(10):2384–2397. Norwegian men. Environ Health Perspect 112:610–615. compounds in inner-city residences of New York City and
Janssen NA, Schwartz J, Zanobetti A, Suh HH. 2002. Air condi- National Children’s Study. 2004. Study Plan. Rockville, Los Angeles. J Expo Anal Environ Epidemiol 14(suppl
tioning and source-specific particles as modifiers of the MD:National Children’s Study. Available: http://national 1):S95–S109.
effect of PM(10) on hospital admissions for heart and lung childrensstudy.gov [accessed 3 June 2004]. Stram DO, Longnecker MP, Shames L, Kolonel LN, Wilkens LR,
disease. Environ Health Perspect 110:43–49. Navidi W, Thomas D, Langholz B, Stram D. 1999. Statistical Pike MC, et al. 1995. Cost-efficient design of a diet validation
Jerrett M, Arain A, Kanaroglou P, Beckerman B, Potoglou D, methods for epidemiologic studies of the health effects of study. Am J Epidemiol 142(3):353–362.
Sahsuvaroglu T, et al. 2005. A review and evaluation of air pollution. Res Rep Health Eff Inst 86:1–56. Swartz L, Callahan K, Butz A, Rand C, Kanchanaraksa S, Diette
intra-urban air pollution exposure models. J Expo Anal Navidi W, Thomas D, Stram D, Peters J. 1994. Design and G, et al. 2004. Partnering with an inner city community to
Environ Epidemiol 15(2):185–204. analysis of multilevel analytic studies with applications to conduct an environmental randomized clinical trial in
Jerrett M, Burnett RT, Goldberg MS, Sears M, Krewski D, a study of air pollution. Environ Health Perspect 102(suppl asthma. Environ Res 95(2):156–165.
Catalan R, et al. 2003a. Spatial analysis for environmental 8):25–32. Tonne CC, Whyatt RM, Camann DE, Perera FP, Kinney PL. 2004.
health research: concepts, methods, and examples. Payne-Sturges DC, Burke TA, Breysse PN, Diener-West M, Predictors of personal polycyclic aromatic hydrocarbon
J Toxicol Environ Health A 66(16–19):1783–1810. Buckley T. 2004. Personal exposure meets risk assess- exposures among pregnant minority women in New York
Jerrett M, Burnett RT, Willis A, Krewski D, Goldberg MS, ment: a comparison of measured and modeled exposures City. Environ Health Perspect 112:754–759.
DeLuca P, et al. 2003b. Spatial analysis of the air pollution- and risks in an urban community. Environ Health Perspect U.S. EPA. 2004. About AirData. Research Triangle Park, NC:U.S.
mortality relationship in the context of ecologic con- 112:589–598. Environmental Protection Agency. Available: http://www.
founders. J Toxicol Environ Health A 66(16–19):1735–1777. Perera FP, Rauh V, Tsai WY, Kinney P, Camann D, Barr D, et al. epa.gov/air/data/info.html [accessed 18 May 2004].
Jerrett M, Finkelstein M. 2005. Geographies of risk in studies 2003. Effects of transplacental exposure to environmental Wallace L, Mitchell H, O’Connor G, Liu L, Neas L, Lippmann M,
linking chronic air pollution exposure to health outcomes. pollutants on birth outcomes in a multiethnic population. et al. 2004. Particle concentrations in inner-city homes of
J Toxicol Environ Health 68(13–14):1207–1242. Environ Health Perspect 111:201–205. children with asthma: the effect of smoking, cooking, and
Keeler G, Dvonch T, Yip F, Parker E, Israel B, Marsik F, et al. 2002. Perera FP, Rauh V, Whyatt RM, Tsai WY, Bernert JT, Tu YH, et al. outdoor pollution. Environ Health Perspect 111:1265–1272.
Assessment of personal and community-level exposures to 2004a. Molecular evidence of an interaction between pre- Weaver VM, Buckley TJ, Groopman JD. 1998. Approaches to
particulate matter among children with asthma in Detroit, natal environmental exposures and birth outcomes in a mul- environmental exposure assessment in children. Environ
Michigan, as part of Community Action Against Asthma tiethnic population. Environ Health Perspect 112:626–630. Health Perspect 106(suppl 3):827–832.
(CAAA). Environ Health Perspect 110(suppl 2):173–181. Perera FP, Tang D, Tu YH, Cruz LA, Borjas M, Bernert T, et al. Whyatt RM, Barr DB, Camann DE, Kinney PL, Barr JR, Andrews
Kingsley G. 2003. Housing, health, and the neighborhood con- 2004b. Biomarkers in maternal and newborn blood indi- HF, et al. 2003. Contemporary-use pesticides in personal
text. Am J Prev Med 24(3 suppl):6–7. cate heightened fetal susceptibility to procarcinogenic air samples during pregnancy and blood samples at deliv-
Krewski D, Burnett R, Goldberg M, Hoover K, Siemiatycki J, DNA damage. Environ Health Perspect 112:1133–1136. ery among urban minority mothers and newborns. Environ
Jerrett M, et al. 2000. Reanalysis of the Harvard Six Cities Peters JM. 1997. Epidemiologic Investigation to Identify Chronic Health Perspect 111:749–756.
and the American Cancer Society Study of Particulate Air Health Effects of Ambient Air Pollutants in Southern Whyatt RM, Jedrychowski W, Hemminki K, Santella RM, Tsai
Pollution and Mortality: Phase II. Sensitivity Analysis. California: Phase II Final Report. Contract No. A033-186. WY, Yang K, et al. 2001. Biomarkers of polycyclic aromatic
Cambridge, MA:Health Effects Institute. Los Angeles, CA:University of Southern California School of hydrocarbon-DNA damage and cigarette smoke expo-
Kunzli N, McConnell R, Bates D, Bastain T, Hricko A, Lurmann F, Medicine, Department of Preventive Medicine. sures in paired maternal and newborn blood samples as a
et al. 2003. Breathless in Los Angeles: the exhausting Peters JM, Avol E, Gauderman WJ, Linn WS, Navidi W, London measure of differential susceptibility. Cancer Epidemiol
search for clean air. Am J Public Health 93(9):1494–1499. SJ, et al. 1999a. A study of twelve southern California com- Biomarkers Prev 10(6):581–588.
Laden F, Neas LM, Dockery DW, Schwartz J. 2000. Association munities with differing levels and types of air pollution. II. Whyatt RM, Rauh V, Barr DB, Camann DE, Andrews HF,
of fine particulate matter from different sources with Effects on pulmonary function. Am J Respir Crit Care Med Garfinkel R, et al. 2004. Prenatal insecticide exposures and
daily mortality in six U.S. cities. Environ Health Perspect 159(3):768–775. birth weight and length among an urban minority cohort.
108:941–947. Peters JM, Avol E, Navidi W, London SJ, Gauderman WJ, Environ Health Perspect 112:1125–1132.
LaRosa LE, Buckley TJ, Wallace LA. 2002. Real-time indoor and Lurmann F, et al. 1999b. A study of twelve southern Yip FY, Keeler GJ, Dvonch JT, Robins TG, Parker EA, Israel BA,
outdoor measurements of black carbon in an occupied California communities with differing levels and types of et al. 2004. Personal exposures to particulate matter among
house: an examination of sources. J Air Waste Manag air pollution. I. Prevalence of respiratory morbidity. Am J children with asthma in Detroit, Michigan. Atmos Environ
Assoc 52:41–49. Respir Crit Care Med 159(3):760–767. 38(31):5227–5236.

1454 VOLUME 113 | NUMBER 10 | October 2005 • Environmental Health Perspectives