Full text of DOI:10.1080/10473289.2000.10464165

This article was downloaded by: [Archives & Bibliothèques de l'ULB]
On: 04 October 2014, At: 11:24
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of the Air & Waste Management Association
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uawm20
Assessing the Relationship between Personal
Particulate and Gaseous Exposures of Senior
Citizens Living in Baltimore, MD
Jeremy A. Sarnat a , Petros Koutrakis a & Helen H. Suh a
a Harvard School of Public Health, Department of Environmental Health , Boston ,
Massachussetts , USA
Published online: 27 Dec 2011.
To cite this article: Jeremy A. Sarnat , Petros Koutrakis & Helen H. Suh (2000) Assessing the Relationship between
Personal Particulate and Gaseous Exposures of Senior Citizens Living in Baltimore, MD, Journal of the Air & Waste
Management Association, 50:7, 1184-1198, DOI: 10.1080/10473289.2000.10464165
To link to this article: http://dx.doi.org/10.1080/10473289.2000.10464165
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose
of the Content. Any opinions and views expressed in this publication are the opinions and views of the
authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should
not be relied upon and should be independently verified with primary sources of information. Taylor
and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses,
damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection
with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

Sarnat, Koutrakis, and Suh
TECHNICAL PAPER
ISSN 1047-3289 J. Air & Waste Manage. Assoc. 50:1184-1198
Copyright 2000 Air & Waste Management Association
Assessing the Relationship between Personal Particulate and
Gaseous Exposures of Senior Citizens Living in Baltimore, MD
Jeremy A. Sarnat, Petros Koutrakis, and Helen H. Suh
Harvard School of Public Health, Department of Environmental Health, Boston, Massachussetts
ABSTRACT
exposures was lowest when individuals spent the major-
ity of their time in well-ventilated indoor environments.
Results also indicate that the potential for confound-
ing by PM2.5 co-pollutants is limited, despite significant
correlations among ambient pollutant concentrations. In
contrast to ambient concentrations, PM2.5 exposures were
not significantly correlated with personal exposures to
PM2.5-10, PM2.5 of non-ambient origin, O3, NO2, and SO2.
Since a confounder must be associated with the exposure
of interest, these results provide evidence that the effects
observed in the PM2.5 epidemiologic studies are unlikely
to be due to confounding by the PM2.5 co-pollutants mea-
sured in this study.
We conducted a multi-pollutant exposure study in Balti-
more, MD, in which 15 non-smoking older adult subjects
(>64 years old) wore a multi-pollutant sampler for 12 days
during the summer of 1998 and the winter of 1999. The
sampler measured simultaneous 24-hr integrated personal
exposures to PM2.5, PM10, SO42-, O3, NO2, SO2, and exhaust-
related VOCs.
Results of this study showed that longitudinal asso-
ciations between ambient PM2.5 concentrations and cor-
responding personal exposures tended to be high in the
summer (median Spearman’s r = 0.74) and low in the win-
ter (median Spearman’s r = 0.25). Indoor ventilation was
an important determinant of personal PM2.5 exposures and
resulting personal–ambient associations. Associations be-
tween personal PM2.5 exposures and corresponding ambi-
ent concentrations were strongest for well-ventilated
indoor environments and decreased with ventilation. This
decrease was attributed to the increasing influence of in-
door PM2.5 sources. Evidence for this was provided by SO42-
measurements, which can be thought of as a tracer for
ambient PM2.5. For SO42-, personal–ambient associations
were strong even in poorly ventilated indoor environ-
ments, suggesting that personal exposures to PM2.5 of
ambient origin are strongly associated with correspond-
ing ambient concentrations. The results also indicated that
the contribution of indoor PM2.5 sources to personal PM2.5
INTRODUCTION
Associations between ambient concentrations of PM and a
variety of adverse health outcomes have been well docu-
mented in numerous epidemiologic studies.1-3 Results from
recent studies indicate that PM2.5 is the particulate size frac-
tion most strongly associated with these observed effects.4
Despite consistent findings from these studies, interpreta-
tion of the results has been questioned, in part due to the
use of ambient concentrations as surrogates of personal
exposure, and also due to concerns about confounding by
PM2.5 co-pollutants, such as PM2.5-10, O3, NO2, and SO2.
Weak correlations between ambient PM concentra-
tions and personal exposures reported in various cross-
sectional exposure studies have been offered by some as
evidence that ambient concentrations are poor surrogates
of personal PM exposures and that findings reported in
epidemiologic studies of PM are inaccurate.5 More recent
observations examining the longitudinal associations be-
tween personal and ambient PM have muted some of this
concern. These studies have found relatively strong asso-
ciations between personal PM2.5, and to a lesser degree
PM10, exposures and ambient concentrations over time.6,7
Although the strength of these associations varied sub-
stantially by individual, results indicated that for certain
individuals ambient PM concentrations are an appropri-
ate surrogate for personal PM2.5 exposures.
IMPLICATIONS
Particulate epidemiologic studies typically use stationary,
ambient concentrations as a surrogate for personal ex-
posure. Questions have been raised concerning the ac-
curacy of ambient PM2.5 measures as a surrogate of ex-
posure due to variability in personal–ambient associations
and confounding by PM2.5 co-pollutants. This paper ex-
amines the personal particulate and gaseous exposures
for a cohort of older adults and investigates their relation-
ship to corresponding ambient concentrations and to one
another. Findings from these analyses will help interpret
results from epidemiological studies.
1184 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
In one of the more recent studies by Rojas-Bracho et
al.,8 this inter-personal variation was attributed to vari-
ability in indoor PM2.5 concentrations, as the association
between personal PM2.5 exposures and outdoor concen-
trations was found to be strongly correlated with corre-
sponding associations between outdoor and indoor PM2.5
concentrations. This finding suggests that indoor sources
and indoor ventilation characteristics are important de-
terminants of personal PM2.5 exposures and their relation-
ship to outdoor levels. (Since the participants spent similar
amounts of time indoors, this was unable to account for
the observed variability in the personal–ambient relation-
ship.) For PM2.5-10, the relationship between ambient con-
centrations and corresponding personal exposures was
shown to be insignificant. Personal exposures to PM2.5-10
and PM2.5 were uncorrelated as well. These insignificant
relationships suggested that PM2.5-10 is unlikely to con-
found observed associations between ambient PM2.5 and
adverse health effects.
The Baltimore multi-pollutant exposure study builds
on the work of these previous longitudinal studies by
1) further examining the influence of indoor ventilation
status upon associations between personal exposures and
ambient concentrations for various particulate and gas-
eous measures and 2) investigating the potential for con-
founding by several PM2.5 co-pollutants. In order to
measure gaseous and particulate pollutants simulta-
neously, a multi-pollutant personal sampler was devel-
oped for this study. Using these simultaneous
measurements, we examined and compared associations
between personal exposures and ambient concentrations
for the various pollutants and determined the relative
importance of indoor sources and ventilation character-
istics to these associations.
collected from June 29 to August 7, 1998. All winter
samples were collected from February 2 to March 13, 1999.
All subjects completed the prescribed 12-day sampling
period with the exception of one individual who dropped
out of the study after the first day and was subsequently
replaced with an additional subject.
Subjects were recruited primarily from senior and com-
munity centers located throughout the Baltimore area.
Subject selection was non-random. All subjects were non-
smokers, retired, physically healthy, and lived in non-smok-
ing private residences (i.e., either single-family houses or
apartments). All residences, except one (SA4), were
equipped with central air conditioning; however, not all
of these residences used air conditioning throughout the
summer. The average age of the subjects was 75 ( 6.8 years).
The subjects were from a range of socio-economic back-
grounds and geographic locations throughout Baltimore.
In addition to wearing the personal sampler, every
subject completed a daily time-activity diary on each
monitoring day. Subjects were given a choice of complet-
ing two types of time-activity diaries during the study.
Twenty-five of the thirty subjects used a closed-form, time-
based diary in which subjects recorded their activities and
location in closed, time-interval spaces that were provided
on the diaries. During the summer, the closed-form diary
was divided into 30-min recording intervals, while in the
winter, the diary consisted of 15-min intervals. The diary
included several columns to allow the subject to indicate
whether a specific pollutant-generating activity was be-
ing performed (i.e., dusting, vacuuming, exposure to en-
vironmental tobacco smoke [ETS], near a busy road). The
remaining five subjects used an open-form, activity-based
diary. Each page of the diary corresponded to a specific
activity with space available on the page to indicate loca-
tion and any special conditions likely to affect exposure.
No formal cross-validation of the two diary formats was
conducted. However, differences between the two formats
were not likely to affect study results, since analyses based
on activity diaries was limited.
This paper reviews the design of the Baltimore study
and provides a characterization of the personal particulate
and gaseous exposures for this sample cohort. Relationships
between the personal exposures and ambient concentrations
for the various pollutants are presented. In addition, factors
influencing these relationships are discussed.
Subjects recorded the ventilation status of every vis-
ited indoor location (e.g., windows open, air condition-
ing use). Ventilation status was recorded only during the
summer sampling session, as it was assumed that win-
dows were closed during the winter in Baltimore. For the
summer, personal exposures were classified by the frac-
tion of time the windows were open while a subject was
in an indoor environment (Fv). This metric can be thought
of as a surrogate for air-exchange rate, as several studies
have shown air-exchange rates to be associated with both
open-window status and air conditioning use.8,9 General
information concerning the subject’s residence was also
collected at the beginning of the sampling period. At the
beginning of each 12-day monitoring period, field staff
METHODS
Study Design
During the summer of 1998 and the winter of 1999,
24-hr integrated personal exposures to PM2.5, PM10, SO42-,
O3, NO2, SO2, and select VOCs were measured for 20 older
adults living in the Baltimore metropolitan area. Ten sub-
jects participated in both the summer and winter sam-
pling seasons, while the remaining 10 subjects participated
in either the summer or winter season. Subjects were moni-
tored for 12 days during each of the monitoring seasons.
Four to six subjects were measured concurrently during
each 12-day monitoring period. All summer samples were
Volume 50 July 2000
Journal of the Air & Waste Management Association 1185

Sarnat, Koutrakis, and Suh
recorded household characteristics that could potentially
have influenced personal exposures. Among these char-
acteristics were cooking fuel type, carpeting, garage type,
air-cleaning devices, pets, and number of individuals re-
siding in the residence.
Field staff visited the subjects daily each morning of
monitoring between 7:00 a.m. and 11:00 a.m. to change
sampling equipment, replace pump batteries, adjust flows,
and review the completed time-activity diaries from the
previous 24 hr.
Bellefonte, PA). SO42- concentrations were determined by
extracting the PM2.5 filters and analyzing the aqueous
extract by ion chromatography.
O3, NO2, and SO2 concentrations were measured us-
ing passive badge samplers. Each of the passive badge sam-
plers contained a single cellulose filter, coated with either
nitrite to collect O312 or triethanolamine to collect NO2
and SO2.13 To ensure a constant face velocity across the
filter, each of the passive samplers were placed in the side
of the elutriators, with the face of the passive sampler
exposed to the sampling air stream. Filters were extracted
after sampling and analyzed using ion chromatography.
In addition to personal sampling, field staff operated
a pair of Harvard Impactors (HIs) at a centrally located
monitoring site to measure 24-hr (8:00 a.m.–8:00 a.m.)
integrated ambient PM2.5 and PM10 data. All of the sub-
jects lived within a 20-km radius of this site. O3, NO2,
SO2, and VOC data were obtained from stationary ambi-
ent monitoring (SAM) sites operated by the Maryland
Department of the Environment and located throughout
Baltimore. Ambient concentrations of O3, NO2, and SO2
were measured using a UV photometric analyzer, a
chemiluminiscence monitor, and a pulsed fluorescent
monitor, respectively.
Multi-Pollutant Sampler. All samples were collected using
a specially designed multi-pollutant sampler, which was
attached to a single, BGI model 400 sampling pump. The
entire sampler weighed approximately 5 lb. Sampling in-
lets were placed on the shoulder strap of a backpack to
correspond to the breathing zone of each subject. The
sampling pump and battery pack were carried in a back
or shoulder pack. Subjects were permitted to remove the
pack during activities when the sampler could be dam-
aged or during prolonged periods of inactivity (i.e., sleep-
ing, watching TV). During periods when the sampler was
removed from the subject’s body, subjects were instructed
to keep the sampling inlets as close as possible to their
breathing zone.
The multi-pollutant sampler is a modification of an
earlier, PM-only version described by Rojas-Bracho et al.7
The performance of this modified multi-pollutant sampler
has been discussed in Chang et al.10 The sampler consisted
of active samplers to collect PM2.5, PM10, and VOCs and
passive samplers to collect O3, NO2, and SO2 (Figure 1). To
allow simultaneous PM2.5, PM10, and VOC sampling using
a single pump, the sample air flow was split three ways
using flow restrictors. Two Personal Environmental Moni-
tors (PEMs), one for PM2.5 and the other for PM10, were
placed, nozzle-down, into 10-cm-long PFA Teflon-coated
aluminum elutriators.11 Tests were conducted to examine
the effect of the elutriators on particle collection. Nine pairs
of PM2.5 PEMs and 10 pairs of PM10 PEMs were collocated
on subjects. For each pair, one PEM was placed in an
elutriator, nozzle-down, while the other PEM was also
placed downward without an elutriator. Each PEM sampled
at a flow rate of 3.2 L/min 10%. Each subject wore the
two PEMs for approximately 16 hr. A paired two-sample
t-test was performed to examine whether the elutriator af-
fected the measurement of PM2.5 or PM10 exposures. Re-
sults from this t-test showed no difference between the
elutriator/non-elutriator measurements for either cut size.
Particles were collected by the PEMs on 37-mm
Teflon filters (37-mm Teflo, Gelman Sciences) at a flow
rate of 3.2 and 2.0 L/min for PM2.5 and PM10, respectively.
VOCs were collected at a flow rate of 20 cm3/min using
sorbent tubes filled with CarbopackB (Supelco;
The Teflon filters used to collect PM2.5 and PM10 were
weighed in a temperature- and humidity-controlled weigh-
ing room (temperature = 18–24 °C; relative humidity =
40 5%). Filters were left to equilibrate 24 hr before the
initial weighing and 48 hr prior to post-sampling weigh-
ing. Each filter was weighed in duplicate both before and
after sampling using a using a Mettler Model MT5 micro-
balance (Mettler Toledo International Inc.; Greifensee,
Switzerland). The average of the two weights was used as
the filter weight. When the two filter weights differed by
more than 5 µg, the filter was weighed a third time, with
the final value being the average of the two closest weights.
Quality Assurance. Standard quality assurance/quality con-
trol procedures were followed for this study.14 Collected
data have been assessed for their bias, precision, and com-
pleteness.
Precision and bias of the multi-pollutant sampler meth-
ods were calculated by collocating replicate, fully config-
ured sampling backpacks at SAM sites equipped with
reference sampling methods. The samplers were operated
for 24 hr 10%. For a given pollutant, precision was esti-
mated as the root mean squared difference between the
collocated personal samplers, divided by the square root of
2. Bias for a given method was determined using the mean
relative difference between the multi-pollutant sampler con-
centration and the corresponding mean reference method
concentration. Method limit of detections (LODs) were es-
timated as 3 times the standard deviation of the field blanks.
1186 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
PM10 PEM
Figure 1. Diagram of multi-pollutant sampler.
Completeness was calculated as the number of samples
collected divided by the target number of samples. In to-
tal, 30 12-day sampling sessions were conducted during
the study. At least 9 days of valid samples were collected
for each subject. Precision, bias, LOD, and completeness
results for each method are listed in Table 1.
values ranged from 0.2 to 1.0 parts per billion (ppb) and
0.1 to 1.0 ppb during the summer and winter, respectively.
For NO2, blank correction values ranged from 0.2 to 0.4
ppb and 0.2 to 0.3 ppb during the summer and winter,
respectively. For SO2 badge samplers, blank correction
values ranged from 0.11 to 0.18 ppb.
Filter Quality Assurance. The total number of field blanks
equaled 19% of all personal PM filters and 14% of all out-
door PM filters. Mean blank filter weights, which were
significantly different from 0 (α = 0.05), were used to cor-
rect the particulate mass concentrations. PM masses were
corrected by subtracting the mean field blank weights from
the sample weights. The field blank values for the PEMs
were 2.2 µg (4.0 µg) and 5.3 µg (4.5 µg) for the summer
and winter seasons, respectively. Field blank value for the
HIs were –0.3 µg (3.4 µg) and 1.2 µg (3.5 µg) for the sum-
mer and winter seasons, respectively. Teflon filters were
also corrected for barometric pressure during weighing.
During the winter sampling period, all particle filters were
stored in refrigerated environments post-sampling to re-
duce potential volatilization from the filters.
Data Analysis. Units for PM concentrations and expo-
sures are reported in µg/m3. Coarse particle concentrations,
or PM2.5-10, were calculated as the difference between PM10
and PM2.5. Units for gaseous concentrations and exposures
are reported in ppb. Analysis of VOC data was not in-
cluded in this paper. Negative values for the gaseous pol-
lutants and PM2.5-10, as well as values less than their
respective LODs, were included in the data analyses.
Shapiro-Wilks tests indicated that the data were not
normally distributed. As a result, the median, mean, stan-
dard deviation, and maximum concentrations were used to
describe the distribution of the data. Data were character-
ized using descriptive statistics, graphical displays,
Spearman’s correlation coefficients, linear regression, and
mixed models. Spearman’s correlation coefficient, linear re-
gression, and mixed models were computed using the SAS
system (SAS Institute Inc.; Cary, NC). Statistical significance
is reported at the 0.05 level unless otherwise specified.
Analyses of the ventilation effect on associations be-
tween personal exposures and ambient concentrations
All O3 and NO2/SO2 badge samplers were refrigerated
pre- and post-sampling and were shipped pre- and post-
sampling in refrigerated coolers. Since blank values differed
by batch, blank corrections for the gaseous pollutants were
computed on a per-batch basis. For O3, blank correction
Volume 50 July 2000
Journal of the Air & Waste Management Association 1187

Sarnat, Koutrakis, and Suh
were conducted using mixed-effects models in which
personal exposures were modeled as the dependent vari-
able, ambient concentrations and ventilation status were
modeled as fixed variables, and subjects were modeled
as random variables to account for between-subject varia-
tion. Autocorrelation between pollutant concentrations
over time was modeled using an autoregressive covariance
structure. Since in the mixed model there is no equivalent
single measure of goodness-of-fit, crude R2 values were gen-
erated using simple linear regression techniques and are
presented to give a rough indication of the scatter of the
data around the estimated regression lines.
were conducted both with and without an additional,
single PM2.5 observation from subject SB2, which showed
an extremely high personal PM2.5 exposure. This extreme
value exerted substantial influence on the overall model
results, yet no source for this elevated exposure was iden-
tified in the time-activity diaries.
RESULTS
Summary statistics for the measured ambient PM concen-
trations and personal exposures, stratified by size, season,
and subject, are presented in Tables 2 and 3. Summary
statistics for the gaseous pollutants are presented in Tables
4 and 5.
Exclusion of Data Points. Data points were voided due to
sampling problems (e.g., pump or battery failures, tube
disconnection) or laboratory analysis irregularities. Time-
activity data indicated that one subject, in particular,
(listed as SA4 in the summer and WA3 in the winter) was
heavily exposed to ETS, since this individual spent a large
fraction of time in bingo parlors where smoking was com-
mon. Summary statistics have been presented both in-
cluding and excluding samples for which substantial ETS
exposure was reported. Analysis of associations between
personal exposures and ambient concentrations, however,
did not include this subject, since collected samples did
not typify exposures for a non-smoker or someone living
in a residence with non-smokers.6,15 Similarly, analysis of
the effect of ventilation status also did not include data
from subject SA4, due to ETS exposure, nor did it in-
clude data from subjects SB5 and SC5 due to inadequate
reporting of ventilation status in their activity diaries.
(The analysis of the effect of ventilation status was con-
ducted during the summer only. As such, specific exclu-
sion of data from WA3, a winter subject, was
unnecessary.) In addition, analyses of ventilation effects
During both seasons, personal PM exposures were
comparable, yet significantly lower (p < 0.0001) than
their respective ambient concentrations for all particle
cut-sizes. The sole exception to this finding was winter-
time coarse PM. In all, ambient concentrations of PM2.5,
PM10, PM2.5-10, and SO42- exceeded corresponding personal
exposures 228 of 333 (68%), 206 of 314 (66%), 172 of
310 (55%), and 307 of 314 (98%) of sampled person-
days, respectively. Median ambient concentrations and
median personal exposures for all PM cut-sizes, except
for PM2.5-10, were significantly higher during the sum-
mer than the winter.
During both sampling seasons, personal exposures to
O3, NO2, and SO2 were extremely low. Seventy percent of
the measured personal O3, NO2, and SO2 values were be-
low their respective LOD, even when ambient concentra-
tions were well above their LOD.
Ambient concentrations of O3 and NO2 varied sea-
sonally. As expected, O3 levels tended to be higher in the
summer while NO2 concentrations were higher during the
winter. These ambient seasonal trends were reflected, to a
lesser degree, in personal exposure measurements as well.
Table 1. Method LOD, precision, bias, and completeness.
Summer
Winter
Units
LOD
Precision
Bias
MRD (SD)a Reference
Meanb
Completeness
LOD
Precision
Bias
Completeness
MRD (SD)a Reference
Meanb
PM2.5
PM10
2.6
4.2
0.9c
1.3c
1.5
2.6 (2.6)
0.1 (3.8)
-2.5 (3.3)
N/A
32.5 168/180 (93.3%)
43.2 169/180 (94.0%)
8.6 167/180 (92.8%)
N/A 168/180 (93.3%)
38.4 145/180 (81.0%)
16.0 167/180 (92.8%)
4.0
6.4
7.5c
2.6e
5.5
1.2
2.1
2.7
0.7e
3.7
8.9
2.3
1.7 (2.6)
-1.0 (5.6)
-2.6 (4.9)
N/A
20.1
30.0
9.1
165/180 (91.7%) µg/m3
168/180 (93.3%) µg/m3
165/180 (91.7%) µg/m3
164/180 (91.1%) µg/m3
PM2.5-10 4.9d
SO4
O3
2.6e
6.6
0.7e
3.4
N/A
21.2
23.5
8.1
2-
-0.2 (5.3)
0.0 (2.8)
N/A
-1.0 (5.4)
6.4 (20.5)
-3.5 (1.9)
165/180 (91.7%)
163/180 (90.6%)
160/180 (88.9%)
ppb
ppb
ppb
NO2
SO2
5.7
2.3
11.7
6.5
N/A
N/A
N/A
N/A
Notes: a Corresponds to the mean relative difference (MRD) between the multi-pollutant sampler concentration to the corresponding reference method concentration; b Corresponds to
the mean concentration of the reference method; c Excludes one outlier. When outlier is included, the precision for PM2.5 is 2.9 µg/m3 and 2.2 µg/m3 for PM10; d Square root of the sum
of the squares of the two single sampler LODs; e Estimated LOD and precision from the Harvard-EPA Annular Denuder System.
1188 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
Relationship Between Personal Exposures
and Ambient Concentrations
as significant correlation coefficients were found for 10 of
the 14 summertime subjects, resulting in a median corre-
lation coefficient of 0.76 (Table 6). Despite the observed
high median correlation coefficient, correlation coefficients
for the measured individuals ranged widely (–0.21 < r <
0.95). This finding is consistent with previous studies.6,19-21
As was the case for SO42-, personal–ambient correla-
tions for PM2.5 were lower in the winter as compared with
the summer, with significant positive, correlation coeffi-
cients found for only five of the 14 winter subjects (me-
dian r = 0.25). This seasonal difference was independent of
individual, as the seasonal variation in the personal–ambi-
ent associations was evident even for the 10 subjects par-
ticipating in both sampling seasons. For these 10 subjects,
the median correlation coefficient was 0.70 in the summer
and 0.25 in the winter. This seasonal difference was more
pronounced than that observed for SO42-, suggesting that
both indoor ventilation differences and indoor PM2.5 emis-
sions were responsible for the observed seasonal variation
in the personal–ambient association for PM2.5.
Consistent with previous studies,7 personal PM2.5-10
exposures and ambient concentrations were generally not
significantly associated, as only three of the 28 subjects
(14 subjects per season) had significant and positive cor-
relation coefficients for PM2.5-10. Insignificant personal–
ambient associations for PM2.5-10 probably reflect the
contribution of indoor PM2.5-10 sources, which include
cooking, cleaning, and resuspension, to personal PM2.5-10
exposures and also the reduced effective penetration effi-
ciency of coarse particles indoors.7,8,22,23
In general, the association between ambient concentra-
tions and personal exposures to PM varied by particle
cut-size and by season (Table 6, Figure 2). The strength of
the association increased with decreasing particle size.
The strongest associations between personal exposures
and ambient concentrations were observed for SO42-, which
can be considered a tracer of ambient particles since it is
an important fine particle constituent (da < 1 µm) with
few indoor sources.16,17 When data were analyzed cross-
sectionally and were treated as independent, personal SO42-
exposures and ambient concentrations were strongly as-
sociated, with ambient concentrations explaining 64% of
the variation in personal SO42- exposures (p < 0.0001; N =
127). The personal–ambient associations were even stron-
ger when data were analyzed by individual. Twenty-one
of the 28 subjects (14 subjects per season) had significant
personal–ambient associations. Correlation coefficients for
these subjects were stronger in the summer than the win-
ter. This seasonal difference may be attributed to differ-
ences in indoor ventilation conditions, which were
generally characterized by open windows in the summer
and closed windows in the winter.8,18
Although weaker than that for SO42-, the association
between personal PM2.5 exposures and ambient concen-
trations was also significant. This was not unexpected
given that SO42- [expressed as (NH4)2SO4] was a major com-
ponent of PM2.5 in Baltimore, comprising 40% of the to-
tal ambient PM2.5 mass in the summer and 26% of the
total ambient PM2.5 mass in the winter. When the data
were analyzed cross-sectionally and treated as indepen-
dent, ambient concentrations explained 45% of the vari-
ability in personal PM2.5 exposures (p < 0.0001; N = 133
[R2 = 0.28 including extreme value for SB2]). Again, corre-
lations between personal exposures and ambient concen-
trations improved when data were analyzed by subject,
Similarly, personal–ambient associations were weak for
all of the gaseous pollutants, a result which was not unex-
pected since measured personal gaseous exposures were
frequently below their respective limits of detection (Tables
4 and 5). In addition, spatial variability in ambient gas-
eous concentrations may have contributed to these weak
associations, as spatial variability in outdoor gaseous con-
centrations may be relatively high due to local sources, such
as automobiles for NO2, or to local sinks, such as automo-
biles for O3. Virtually none of the correlation coefficients
for the individual-specific pairwise comparisons were sig-
nificant for O3, NO2, or SO2 (Table 6, Figure 2).
95th Percentile
90th Percentile
75th Percentile
Ventilation Status and PM2.5 Correlations
The influence of indoor ventilation patterns on the asso-
ciation between personal and ambient PM2.5 was exam-
ined by using mixed models on data stratified into one of
three indoor ventilation categories. Indoor ventilation
conditions were categorized as either well (Fv > 0.72), mod-
erately (0.04 ≤ Fv ≤ 0.72), or poorly (Fv < 0.04) ventilated,
based on the distribution (mean standard deviation) of
the fraction of time indoor environments were character-
ized by open windows (Fv). In the summer, Fv values varied
Median
25th Percentile
10th Percentile
5th Percentile
Sulfate PM2.5 PM10 PM2.5-10
O3
NO3
SO2
Figure 2. Distribution of personal–ambient correlation coefficients.
Volume 50 July 2000
Journal of the Air & Waste Management Association 1189

Sarnat, Koutrakis, and Suh
Table 2. Descriptive statistics for ambient concentrations and personal exposures: PM, summer 1998 (µg/m3).
2-
PM2.5
Na Med Mean SDb Max
PM10
Na Med Mean SDb Max
PM2.5-10
Na Med Mean SDb Max
SO4
Na Med Mean SDb Max
Subject
SA1
SA2
SA4
SA5
A Ambient
SB1
SB2
SB3
SB4
SB5
SB6
B Ambient
SC1
SC2
SC3
SC4
SC5
12 15.3 16.7 5.8 26.2
10 27.4 29.4 8.7 43.8
11 78.4 80.7 41.9 134.8
10 24.6 28.9 12.4 49.6
12 21.7 25.3 12.8 46.6
11 21.0 23.5 11.0 52.3
10 27.4 32.8 17.6 77.9
12 20.5 21.9 7.7 39.3
12 18.6 21.7 6.5 35.1
12 21.2 22.7 8.4 42.1
12 22.8 22.5 6.3 33.7
11 34.4 38.7 12.6 64.2
11 101.0 85.4 39.1 130.9
10 28.0 30.9 11.0 49.6
12 6.1
5.8
6.7
4.7
2.0
7.0
2.9 11.5
3.0 12.4
8.8 22.5
2.4 6.9
1.1 9.0
12 2.7
10 3.7
11 4.8
10 4.8
3.0
5.0
6.4
7.5
9.4
5.0
4.7
6.4
6.3
6.1
7.5
1.6 6.5
3.5 13.8
4.5 14.8
4.9 17.9
7.1 21.2
3.2 13.4
1.8 7.5
3.3 14.2
2.9 11.5
3.4 15.1
2.9 14.9
10 6.4
11 2.8
10 1.4
8
28.0 33.7 15.6 55.6
8
7.1
9
7.0
11 36.1 36.2 10.4 59.8
10 27.3 33.9 18.0 78.1
12 31.1 32.6 9.0 49.0
12 24.5 28.6 9.0 42.4
12 23.4 24.9 8.8 43.1
11 12.8 12.6 3.8 20.0
10 -0.1 1.7 3.0 8.4
12 10.1 10.7 4.1 18.0
12 6.6
12 2.0
11 5.2
10 4.9
12 6.6
12 6.3
12 5.4
6.9
2.2
5.3
9.1
9.3
6.0
4.4 14.7
2.0 5.6
2.8 9.0
2.6 14.5
3.5 14.5
2.7 12.6
9
23.7 23.6 7.5 40.0
9
29.4 28.9 7.4 40.3
9
4.6
9
6.8
12 24.8 26.4 9.5 46.9
12 19.1 20.9 10.3 36.0
12 12.5 12.9 4.2 23.0
12 18.1 19.6 8.1 32.0
11 20.8 20.1 6.8 29.2
12 20.2 18.9 6.4 28.8
13 20.9 23.9 12.7 45.6
12 32.7 35.5 9.9 54.4
12 30.0 30.3 9.6 46.8
12 17.3 18.9 4.9 27.8
12 18.2 30.9 7.7 43.7
11 28.4 30.3 14.0 68.7
12 30.5 33.1 16.1 66.4
13 28.3 32.7 14.2 55.0
12 8.1
12 9.6
12 5.7
12 11.4 11.2 3.3 14.9
11 4.8 10.2 14.0 49.3
12 8.9 14.3 14.1 49.3
12 10.3 11.7 7.8 29.6
12 6.0
12 3.4
12 4.5
11 7.1
12 5.7
6.8
3.6
5.5
6.7
5.8
4.6 16.5
1.8 6.4
3.8 11.3
3.8 14.0
3.1 9.3
C Ambient
13 9.2
8.8
2.1 12.1
13 6.8 10.2 8.8 27.3
Summer Personalc
Summer Personalc,d
Summer Personalc,e
Summer Personalc,d,e
Summer Ambient
37 23.1 26.7 13.7 63.6
37 20.6 22.2 7.5 40.6
37 23.1 26.7 13.7 63.6
37 20.6 22.2 7.5 40.6
37 23.1 25.2 11.5 46.9
37 32.1 33.9 11.7 65.0
37 27.7 29.7 7.5 46.9
34 29.7 32.8 11.5 65.0
34 27.9 29.9 7.6 46.9
34 32.2 34.0 12.8 55.6
37 6.7
37 6.4
34 6.8
34 6.7
34 8.1
7.2
7.3
7.4
7.4
8.4
4.0 24.8
3.8 24.8
4.0 24.8
3.9 24.8
2.3 14.5
37 5.4
37 5.4
35 5.5
35 5.5
5.6
5.6
5.8
5.7
3.1 13.8
3.0 13.8
3.1 13.8
3.1 13.8
35 8.0 10.5 7.8 29.6
a
b
c
Notes: N refers to sample size; SD refers to arithmetic standard deviation; Summer personal values were computed by calculating the mean exposures of all subjects sampled
concurrently; dExcludes personal exposures with extreme ETS exposures; eMatched with ambient sampling days (i.e., personal exposures from days when no ambient monitoring was
available were removed).
Table 3. Descriptive statistics for ambient concentrations and personal exposures: PM, winter 1999 (µg/m3).
2-
PM2.5
Subject
PM
PM2.5-10
SO4
Na Med Mean SDb Ma1x0
Na Med Mean SDb Max
Na Med Mean SDb Max
Na Med Mean SDb Max
WA1
WA2
WA3
WA4
WA5
A Ambient
WB1
WB2
WB3
WB4
B Ambient
WC1
WC2
WC3
WC4
WC5
WC6
12 9.5 10.4 3.4 17.7
11 15.6 17.0 2.5 20.9
10 46.5 53.4 38.7 113.1
12 16.1 19.2 11.4 54.6
11 35.3 34.3 5.1 41.1
10 58.2 66.2 40.3 124.3
12 6.4
11 17.1 17.3 4.9 26.0
10 12.6 12.8 9.4 36.0
8.8 10.7 42.0
12 1.5
11 2.6
10 3.5
1.5
2.3
3.5
2.2
2.6
4.1
2.1
2.4
3.1
2.3
4.8
0.1
1.2
1.2
1.5
1.6
1.3
3.0
0.6 2.5
0.8 3.6
1.1 6.2
0.7 2.8
0.9 4.0
1.0 5.7
0.9 3.8
1.2 4.2
1.4 5.2
1.1 4.1
2.2 8.3
0.4 1.6
0.6 2.3
0.7 2.2
0.6 2.2
0.6 2.4
0.5 2.0
0.9 3.7
9
11.3 15.3 2.9 19.8
9
18.3 20.1 4.1 27.7
9
4.8
4.9
2.3 7.8
4.2 17.0
3.1 11.3
2.6 7.9
9
2.6
11 15.7 16.6 5.5 30.4
12 22.3 22.4 5.8 32.6
11 11.5 12.5 4.5 20.3
12 14.4 21.3 14.8 57.6
11 43.0 46.5 24.4 88.6
12 23.4 26.4 13.0 52.4
12 24.8 25.1 12.0 49.0
11 25.4 26.5 7.7 40.5
11 29.3 28.4 6.3 36.3
11 15.8 16.4 6.3 27.1
12 21.2 35.1 34.4 134.4
11 66.8 71.3 33.8 124.6
12 32.5 37.4 18.3 81.9
12 35.2 33.3 17.1 73.2
12 12.5 12.0 5.9 23.8
11 10.1 9.9
11 5.3
11 4.1
12 5.5 13.8 32.7 117.3
11 27.1 24.8 11.6 43.5
12 9.4 11.0 6.9 29.6
12 5.6
12 6.2
10 2.4
10 3.1
10 3.4
12 0.9
12 6.0
11 3.8
11 2.7
11 4.4
10 2.1
12 2.4
11 3.1
12 2.3
12 5.0
12 0.8
10 1.1
10 1.0
10 1.9
12 1.7
12 1.5
11 3.4
6.9
3.9
8.6
6.1
2.0
3.5
3.6
2.1
7.1
4.8
7.0 24.2
3.5 12.2
1.1 3.4
2.4 6.4
2.5 6.6
2.4 7.3
4.6 18.6
3.8 9.4
12 5.2
10 4.1
10 5.7
10 6.9
12 5.6
5.9
5.5
6.6
7.1
5.8
2.9 11.9
3.8 15.1
3.8 14.6
2.7 11.3
2.4 10.7
11 6.5
11 9.3
7.3
9.8
3.7 17.4
5.2 20.3
11 9.7 10.3 3.0 14.4
12 6.6 7.9 3.8 14.9
12 15.4 18.3 8.4 37.0
11 18.3 18.0 9.4 34.0
12 9.7 11.2 5.0 25.1
12 11.9 14.1 6.8 26.3
C Ambient
Winter Personalc
Winter Personalc,d
Winter Personalc,e
Winter Personalc,d,e
Winter Ambient
36 15.4 18.5 11.2 44.0
36 14.5 16.6 10.2 44.0
36 15.4 18.5 11.2 44.0
36 14.5 16.6 10.2 44.0
36 20.7 5.6 49.0 9.7
36 24.8 28.0 16.5 69.5
36 23.3 26.7 17.6 90.8
34 24.8 28.5 16.7 69.5
34 23.3 27.2 17.9 90.8
34 28.2 7.5 73.2 13.4
36 8.2
9.6
7.9 47.6
36 1.9
36 1.9
34 1.9
34 1.9
34 2.7
2.1
2.0
2.1
2.1
1.0
1.0 4.1
1.0 4.4
1.0 4.1
1.0 4.4
8.3 1.7
36 8.3 10.2 10.9 68.2
34 8.2 9.7 8.1 47.6
34 8.3 10.3 11.2 68.2
34 5.2 -1.3 24.2 5.1
a
b
c
Notes: N refers to sample size; SD refers to arithmetic standard deviation; Winter personal values were computed by calculating the mean exposures of all subjects sampled
concurrently; dExcludes personal exposures with extreme ETS exposures; eMatched with ambient sampling days (i.e., personal exposures from days when no ambient monitoring was
available were removed).
1190 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
Table 4. Descriptive statistics for ambient concentrations and personal exposures: gaseous pollutants, summer 1998 (ppb).
O3
Na
NO2
Na
Subject
Med
Mean
SDb
Max
Med
Mean
SDb
Max
SA1
SA2
SA4
SA5
A Ambient
SB1
SB2
SB3
SB4
SB5
SB6
B Ambient
SC1
SC2
SC3
SC4
SC5
12
10
11
10
12
10
9
11
11
10
8
12
9
10
9
8
0.1
3.7
0.7
6.3
34.5
1.2
0.2
2.7
1.1
0.9
3.1
39.7
2.4
8.5
3.1
9.6
9.7
36.3
0.1
3.0
1.0
6.2
37.5
1.1
0.6
3.2
1.0
0.9
4.0
38.4
4.7
8.5
5.3
9.6
9.7
36.9
1.9
3.0
2.1
3.7
9.4
1.8
0.1
1.8
1.4
1.5
4.0
7.1
6.7
5.6
4.6
2.2
3.5
7.6
3.1
6.5
5.9
11.5
53.6
4.3
2.7
6.9
3.4
4.0
12
12
11
10
12
11
10
12
12
11
9
12
11
12
11
11
12
12
–1.2
3.4
14.8
8.4
18.7
7.4
7.1
5.6
4.9
3.4
–1.0
2.6
15.5
8.3
20.7
8.6
8.2
5.8
5.4
9.8
10.4
21.5
16.7
13.5
10.5
8.5
1.7
4.3
5.7
2.5
5.2
4.4
5.4
3.0
5.1
17.3
13.4
4.5
4.9
16.9
4.5
5.1
6.1
5.2
3.4
8.3
24.2
12.7
30.6
16.2
20.5
11.0
14.4
59.3
39.6
30.8
26.8
65.8
18.1
17.2
22.7
30.5
11.7
3.9
22.7
17.1
8.4
10.0
7.3
21.1
17.1
10.4
14.1
16.9
48.7
7
13
11.3
20.4
11.9
21.8
C Ambient
Summer Personalc
Summer Personalc,d
Summer Ambiente
Summer Ambient
33
33
33
37
2.7
3.3
35.1
36.3
3.5
3.6
37.3
37.6
3.0
3.0
8.3
7.9
9.9
9.9
53.6
53.6
37
37
37
37
8.0
7.4
20.4
20.4
8.7
7.9
21.4
21.4
5.4
5.8
4.9
4.9
28.6
28.6
30.8
30.8
a
b
c
Notes: N refers to sample size; SD refers to arithmetic standard deviation; Summer personal values were computed by calculating the mean exposures of all subjects sampled
concurrently; dExcludes personal exposures with extreme ETS exposures; eMatched with personal sampling days (i.e., ambient concentrations from days when no personal sampling
was available were removed).
Table 5. Descriptive statistics for ambient concentrations and personal exposures: gaseous pollutants, winter 1999 (ppb).
O3
Na
NO2
Na
SO2
Na
Subject
Med Mean SDb
Max
Med Mean SDb
Max
Med Mean SDb
Max
WA1
WA2
WA3
WA4
WA5
A Ambient
WB1
WB2
WB3
WB4
B Ambient
WC1
WC2
WC3
WC4
WC5
WC6
C Ambient
12
11
10
9
0.5
–0.3
0.9
0.7
1.0
11.2
0.6
0.5
0.3
0.8
0.1
0.0
0.9
0.4
0.6
10.6
0.5
0.5
0.9
0.8
15.5
–2.1
–0.8
–0.2
–2.5
–1.3
1.5
1.3
1.7
4.1
3.0
1.8
6.3
2.7
2.5
3.6
3.7
9.6
6.7
3.8
8.5
0.3
2.8
10.6
7.0
1.6
1.0
1.9
1.6
1.0
21.2
1.2
0.9
1.5
1.5
34.5
4.3
3.8
5.0
2.1
3.1
5.7
38.5
12
11
9
3.9
16.3
25.9
15.4
19.2
24.6
5.7
11.3
11.3
6.5
21.7
15.7
9.2
12.3
15.1
25.9
10.0
23.9
4.2
1.9
24.5
41.4
20.0
29.5
4.2
24.4
54.1
52.6
42.1
9.9
94.4
25.9
32.3
36.7
54.5
40.6
5.6
74.0
3.9
8.1
4.5
5.2
11
10
8
0.2
1.1
–0.5
1.3
0.4
5.9
0.5
0.4
0.1
–0.7
9.4
0.2
1.2
–0.7
0.7
0.0
6.6
1.4
5.0
2.3
2.1
2.5
2.6
2.3
3.5
1.7
1.3
3.7
1.8
2.6
1.7
0.7
2.7
1.4
6.3
2.3
1.8
1.5
1.2
1.8
10.4
1.6
1.4
1.7
1.4
17.6
1.5
1.4
1.5
1.3
1.6
1.4
18.8
17.0
25.5
15.2
19.2
24.1
8.6
18.2
18.0
13.6
22.2
20.8
11.8
13.6
18.9
30.1
13.1
21.8
9
9
11
12
11
12
11
12
12
12
12
10
11
12
11
12
11
12
10
11
10
12
12
12
11
11
11
12
11
12
11
12
10
11
10
12
12
12
11
11
11
12
11
12
30.7
8.5
0.3
0.3
17.7
18.6
15.7
39.2
24.8
9.0
9.7
9.2
11.1
10.3
28.8
–0.5
–0.5
10.2
–0.1
–0.3
–0.7
–0.8
0.5
13.0
–2.8
2.3
0.3
–0.3
–0.2
–0.3
0.8
–0.3
8.0
0.0
–2.8
-0.8
0.4
–0.7
9.8
26.5
27.3
Winter Personalc
Winter Personalc,d
Winter Ambient
36
36
0.3
0.3
16.0
0
0
17.8
1.8
1.8
10.3
2.8
2.8
38.5
36
36
36
14.3
13.1
24.0
16.3
15.8
22.7
7.9
8.0
6.9
42.7
42.7
39.2
36
36
36
–0.1
0.0
8.3
0.0
0.0
8.9
0.7
0.7
4.6
1.5
1.5
18.8
a
b
c
Notes: N refers to sample size; SD refers to arithmetic standard deviation; Winter personal values were computed by calculating the mean exposures of all subjects sampled
concurrently; dExcludes personal exposures with extreme ETS exposures.
Volume 50 July 2000
Journal of the Air & Waste Management Association 1191

Sarnat, Koutrakis, and Suh
Table 6. Correlation between personal and ambient pollutant level by subject and season.a
SUMMER
WINTER
2-
2-
SO4
PM2.5
PM10 PM2.5-10
O3
NO2
SO4
PM2.5 PM10 PM2.5-10
O3
NO2
SO2
SA1
SA2
SA5
SB1
SB2
SB3
SB4
SB5
SB6
SC1
SC2
SC3
SC4
SC5
Median
Mean
Stdev
95% CI
0.95b
0.79b
0.9b
0.55
0.85b
0.89b
0.65b
-0.21
0.82b
0.73b
0.73b
0.53
0.5
0.32
0.21
-0.39
0
0.36
0.12
0.49
0.21
-0.03
-0.11
0.47
0.77b
0.12
0.33
0.01
0.27
0.05
-0.29
0.17
0.20
0.28
0.14
0.72b
0.27
-0.07
-0.31
-0.13
0.06
WA1
WA2
WA4
WA5
WB1
WB2
WB3
WB4
WC1
WC2
WC3
WC4
WC5
0.4
0.22
0.2
0.25
0.26
-0.48
0.68b
0.36
0.53
0.48
-0.07
0.32
0.1
0.55
0.57
0.32
0.25
0.32
0.29
0.30
0.16
-0.27
-0.28
0.52
0.3
0.35
0.29
-0.43
0.18
0.3
-0.43
-0.11
0.54
-0.05
-0.11
0.07
0.06
0.34
0.18
0.74b
-0.01
-0.18
0.13
0.3
0.05
-0.75b
-0.02
0.65b
-0.21
-0.39
0.36
-0.26
-0.22
-0.28
0.14
0.35
0.06
0.29
0.02
-0.02
0.37
0.19
0.86b
0.79b
0.63b
0.08
0.75b
0.59b
0.71b
0.4
0.54
0.33
0.49
0.93b
0.88b
0.95b
0.81b
0.75b
0.73b
0.71b
0.47
0.83b
0.55
0.72
0.67
0.20
0.11
-0.38 -0.36
-0.18 -0.79b
0.66b
0.82b
0.86b
0.89b
0.89b
0.53
0.99b
0.92b
0.9b
0.22
0.8b
0.3
0.69b
0.54
0.51
0.1
0.59
0.89b
0.79b
0.81b
0.69b
0.35
0.53
0.38
0.47
0.25
0.26
-0.11
0.58b
0.29
0.37
-0.6b
-0.34
-0.33
-0.12
0.64b
0.11
0.06
0.38
0.20
0.62
0.55
-0.12
0.74b
0.79b
0.28
0.19
0.81b
0.01
0.25
0.33
0.40
0.21
0
0.14
-0.64b
-0.04
0.3
-0.35
-0.63
0.75b
-0.08
0.4
-0.16
0.36
-0.01
0.95b
0.78b
0.85b
0.78b
0.55
0.79b
0.29
0.71
0.27
0.64b
0.64
0.57
0.23
0.12
-0.25
-0.29
0.17
-0.01
0.43
-0.01
0.05
0.34
0.18
0.87b
0.71b
0.88
0.83
0.12
0.07
WC6
0.76
0.68
0.29
0.15
Median
Mean
Stdev
95% CI
0.07
0.40
0.21
Notes: aCorrelations represent Spearman’s r values; bSignificant at the α= 0.05 level.
0.24, p < 0.02, crude R2 = 0.05 including extreme value from
SB2]). The weaker personal–ambient associations for the
poorly ventilated category is reflected by the increased vari-
ability in the ratio of personal to ambient PM2.5 for the
poorly as compared to the well- and moderately ventilated
groups (Figure 5). Slopes for the personal on ambient re-
gressions (slope = 0.83, well; slope = 0.59, moderate; slope
= 0.46, poor) and median personal/ambient ratios also de-
creased with ventilation. This may be due to lower effec-
tive penetration efficiencies for poorly ventilated indoor
environments. In the winter, personal–ambient associations
for PM2.5 (both crude R2 and slope) were similar to those for
the summertime poorly ventilated category (slope = 0.47,
p < 0.0001, crude R2 = 0.19).
Ventilation status also affected the association be-
tween personal exposures and ambient concentrations for
SO42-; however, the effect was not as pronounced as that
for PM2.5. In the summer, ambient SO42-concentrations
were strongly associated with personal exposures for all
three ventilation categories (crude R2 = 0.88, 0.73, and 0.72
for the well, moderately, and poorly ventilated indoor
environments, respectively) (Figure 6). As was the case
with PM2.5, the slopes of personal ambient level decreased
with ventilation status (0.70, 0.40, and 0.39 for well,
moderately, and poorly ventilated categories, respectively),
which again may be due to the fact that the effective pen-
etration efficiency for SO42- is lower for these environ-
widely by individual for most subjects spanning ventila-
tion categories (Figure 3). In the winter, Fv values for all
subjects were assumed to fall into the poorly ventilated
category. Since ventilation conditions during the summer
were quantified (and not assumed) using activity data,
analyses for summer and winter data were performed sepa-
rately.
When summertime personal PM2.5 exposures were
measured for individuals spending time in well-venti-
lated indoor environments (e.g., Fv > 0.72), ambient con-
centrations were excellent predictors of personal PM2.5
exposures (slope = 0.83, p < 0.0001, crude R2 = 0.80) (Fig-
ure 4). The association between personal PM2.5 exposures
and corresponding ambient concentrations was weaker for
the moderately ventilated category (slope = 0.59, p< 0.0001,
crude R2 = 0.57), and still weaker for the poorly ventilated
category (slope = 0.46, p < 0.0001, crude R2 = 0.25 [slope =
2-
ments. In winter, even though ambient SO4
concentrations were relatively low, ambient concentra-
tions remained relatively strong predictors of personal
SO42- concentrations. Wintertime personal–ambient PM2.5
associations (slope = 0.50, p < 0.0001, crude R2 = 0.71)
Figure 3. Range of Fv values by subject.
1192 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
Ambient Concentration (µg/m3)
Ambient Concentration (µg/m3)
Ambient Concentration (µg/m3)
Figure 4. Personal PM2.5 exposures vs ambient PM2.5 concentrations: Summer 1998.*
were similar to those for the moderately and poorly ven-
tilated categories in the summer. These results suggest that
the weak personal–ambient associations observed for PM2.5
for the poorly ventilated category are attributable prima-
rily to the influence of non-ambient PM2.5 sources.
of a 24-hr sampling period) and that the variation in
this time was minimal.
Correlation between Ambient PM2.5 and
Personal Exposure to PM2.5 of Ambient Origin
Personal exposure to PM2.5 of ambient origin was estimated
using the expression
The fraction of time subjects spent outdoors was
added to Fv to examine whether consideration of time
spent outdoors improves our ability to explain variabil-
ity in the personal–ambient associations. Results from
these analyses differ little from analyses involving Fv
alone, suggesting that information about time spent
outdoors explains little of the variability in the
personal–ambient relationship. The unimportance of
time spent outdoors may be attributed to the fact that
subjects spent little time outdoors (approximately 5 6%
where Personalij represents the personal exposure to SO42-
for subject i on day j and Ambientj represents the ambient
concentration measured at the stationary site on day j.
The “effective penetration” of ambient PM2.5 to personal
Volume 50 July 2000
Journal of the Air & Waste Management Association 1193

Sarnat, Koutrakis, and Suh
PM2.5
SO42-
Figure 5. Personal–ambient ratios for PM2.5 and SO42- by ventilation status: Summer 1998.
exposures for all fine particles was assumed to equal that
for SO42-.
were significantly associated with ambient PM2.5 concen-
trations during the summer (r = 0.34, p < 0.04; r = 0.63, p
< 0.0001; and r = 0.43, p < 0.008, respectively). Significant
correlations were also found between wintertime PM2.5-10
concentrations and ambient PM2.5 concentrations (r = 0.57,
p < 0.0004). Ambient concentrations of PM2.5-10, O3, and
NO2 were also significantly associated with personal PM2.5
exposures and personal exposure to PM2.5 of ambient ori-
gin for some subjects. This was especially true for NO2,
where the personal PM2.5 exposures for 10 of the 28 sub-
jects (5 subjects per season) were significantly associated
with ambient NO2 concentrations (Table 7).
Estimates of personal exposure to PM2.5 from ambi-
ent sources were plotted against total personal PM2.5 ex-
posures (Figure 7), where deviations from the 1:1 line
reflect non-ambient source contributions to personal PM2.5
exposures. As with previous analyses, the association be-
tween personal exposures to PM2.5 of ambient origin and
personal exposures to total PM2.5 varied by ventilation sta-
tus, with the scatter around the regression line greatest
for the poorly ventilated category. Similarly, the contri-
bution of personal PM2.5 of ambient origin to personal
PM2.5 exposures varied by ventilation category. PM2.5 of
ambient origin comprised an average 75% ( 13%) of per-
sonal PM2.5 exposures for the well-ventilated category, 70%
( 16%) for the moderately ventilated category, and 55%
( 13%) for the poorly ventilated category. These differ-
ences are reflected in the intercepts of the regression of
personal exposures to PM2.5 on that for personal PM2.5 of
ambient origin, which equaled 2.7, 3.3, and 5.9 µg/m3 for
the well-, moderately, and poorly ventilated categories,
respectively. The slopes for the three ventilation catego-
ries were comparable (1.2, 1.3, and 1.3 for high, medium,
and low Fv categories, respectively).
For personal exposures, significant correlations
among the pollutants were not observed. Personal PM2.5
exposures were not significantly associated with PM2.5-10
,
O3, NO2, or SO2 when analyzed either cross-sectionally,
by individual, or by ventilation status. Similarly, associa-
tions between personal exposures to PM2.5 of ambient ori-
gin and of non-ambient origin were not significant, with
only one subject having a significant correlation coeffi-
cient. Personal exposures to PM2.5 of ambient origin were
also not significantly associated with personal exposures
to the gaseous pollutants.
DISCUSSION AND CONCLUSIONS
Relationship between PM2.5
and Other Pollutants
Correlations among the pollutants were examined to inves-
tigate the potential for confounding in epidemiologic stud-
ies of PM2.5. Ambient PM2.5-10, O3, and NO2 concentrations
Personal exposures to all of the measured pollutants
tended to be lower than corresponding ambient concen-
trations. For PM, this finding is contrary to the results from
several studies of healthy individuals, which attributed
higher personal particulate exposures to the presence of a
1194 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
Ambient Concentration (µg/m3)
Ambient Concentration (µg/m3)
Ambient Concentration (µg/m3)
Figure 6. Personal SO42- exposures vs ambient SO42- concentrations: Summer 1998.*
exposures to indoor and personal particulate sources may
also have contributed to the observed lower personal par-
ticulate exposures for our Baltimore cohort.
personal particulate cloud and indoor and personal par-
ticulate sources.7,15,20 Lower personal particulate exposures
as compared with ambient concentrations, however, have
been found in two studies of individuals with limited ex-
posure to indoor particulate sources. Bahadori et al., for
example, found mean personal PM2.5 concentrations of
individuals with chronic obstructive pulmonary disease
to be lower than corresponding outdoor concentrations
and attributed these lower exposures to the cohort’s low
activity level.24 Similarly, Tamura et al. found mean per-
sonal PM10 exposures to be lower than corresponding out-
door levels for a cohort of older adults with limited exposure
to indoor PM sources.21 It is possible that similarly reduced
From the current analyses, it is clear that indoor ven-
tilation was an important determinant of personal PM2.5
exposures through its competing effects on effective
penetration efficiency and indoor source contributions.
These effects were clearly illustrated by the relationship
between ventilation status and the personal–ambient as-
sociations for SO42- and PM2.5. This relationship was inde-
pendent of subject, as evidenced by the fact that most
individuals had exposures that spanned ventilation cat-
egories and by the results from the mixed models that
Volume 50 July 2000
Journal of the Air & Waste Management Association 1195

Sarnat, Koutrakis, and Suh
Personal exposure to PM2.5 of Ambient Origin (µg/m3)
Personal exposure to PM2.5 of Ambient Origin (µg/m3)
Personal exposure to PM2.5 of Ambient Origin (µg/m3)
Figure 7. Personal PM2.5 exposures vs personal exposure to PM2.5 of ambient origin: Summer 1998.*
controlled for subject. For SO42-, both the slopes of the
personal-on-ambient regression lines and the median per-
sonal/ambient ratios decreased significantly with venti-
lation status, where the slopes and ratios were highest for
the well-ventilated category and were lowest for the poorly
ventilated category. This decrease, however, was not ac-
companied by a concomitant decrease in the strength of
the personal–ambient relationship for SO42-, as shown by
a similar degree of scatter around the regression line for
all three ventilation categories. Since SO42- has relatively
few indoor or personal sources, the observed decreases in
personal–ambient slopes and ratios for SO42- with ventila-
tion may be attributed primarily to corresponding reduc-
tion in the effective penetration efficiency of particles from
outdoor to personal environments. This reduction in the
effective penetration efficiency with ventilation affected
the magnitude of the personal–ambient association but,
importantly, not the strength of this relationship.
The slopes and median ratios also decreased for PM2.5,
which may also be attributed to reduced effective pen-
etration efficiencies in poorly ventilated environments.
These reductions, however, were less substantial than
those observed for SO42-, probably as a result of the influ-
ence of indoor PM2.5 sources, which have been shown in
several studies to be greatest in poorly ventilated indoor
environments.7,8,23
Evidence of the influence of indoor ventilation con-
ditions on indoor PM2.5 source contributions is provided
by estimates of the fraction of personal PM2.5 that is of
ambient origin. As reflected by both the intercepts of the
regression lines and by the ratios of personal PM2.5 of
ambient origin to total personal PM2.5, this fraction was
1196 Journal of the Air & Waste Management Association
Volume 50 July 2000

Sarnat, Koutrakis, and Suh
Table 7. Correlations between personal PM2.5 and ambient pollutant concentrations.a
Personal PM2.5 vs Ambient:
Personal PM2.5 of Ambient Origin vs Ambient:
Subject
PM2.5
O3
NO2
PM2.5-10
O3
NO2
PM2.5-10
SUMMER
SA1
SA2
0.55
0.85b
0.89b
0.65b
–0.21
0.82b
0.73b
0.73b
0.53
0.95b
0.78b
0.85b
0.78b
0.55
0.15
0.31
0.38
0.66b
0.82b
–0.15
0.81b
–0.14
–0.34
–0.42
–0.38
0.66b
0.36
–0.12
0.57
0.27
0.21
0.71b
0.64
0.81b
–0.74b
0.08
0.15
0.68
0.79
–0.03
0.33
0.00
–0.14
0.33
0.32
0.57b
0.76b
0.80b
0.51
0.27
–0.24
0.02
0.00
0.40
0.66b
0.59b
0.60
0.48
0.69
0.71b
–0.45
0.67
0.42
–0.45
0.33
0.45
SA5
0.18
0.64
0.33
SB1
0.40
0.38
0.89b
0.26
SB2
–0.62
0.55
0.62b
0.45
0.15
SB3
–0.04
–0.12
0.23
0.52
–0.20
–0.29
–0.48
–0.57
0.63b
0.65b
0.71b
0.50
SB4
0.45
SB5
0.36
SB6
0.15
0.12
–0.03
0.83b
0.66b
0.69b
0.50
SC1
0.78b
0.68b
0.78b
0.66b
0.51
0.65b
0.51
SC2
SC3
0.73b
0.59
0.68b
0.70b
0.43
SC4
SC5
0.32
0.34
0.33
WINTER
WA1
WA2
WA4
WA5
WB1
WB2
WB3
WB4
WC1
WC2
WC3
WC4
WC5
WC6
Summer
Winter
0.22
–0.18
–0.07
0.67b
–0.43
–0.84
–0.32
–0.45
–0.01
–0.62
–0.88
–0.42
–0.84
–0.62
–0.03
0.48
–0.26
–0.36
–0.22
0.61b
0.77b
0.59b
0.62b
0.34
–0.05
–0.70
–0.29
0.50
–0.78b
–0.15
–0.33
–0.72b
–0.87b
–0.76b
–0.77b
–0.80b
–0.64b
–0.87b
–0.77b
–0.72b
–0.76b
–0.75b
0.41
–0.04
–0.15
0.20
–0.38
–0.18
0.22
0.80b
0.62b
0.55
–0.09
0.53
0.41
0.09
0.59b
0.56
0.04
–0.12
0.74b
0.79b
0.28
–0.10
0.44
0.77b
0.57
0.68b
0.02
–0.15
0.17
0.25
0.03
0.30
0.19
0.50
0.45
0.22
0.81b
0.01
0.08
0.81b
0.37
0.05
0.65b
0.37
0.19
Median
Median
0.76
0.41
0.42
0.25
–0.43
0.26
0.39
–0.76
0.21
b
Notes: aCorrelations represent Spearman’s r values; Significance at the α = 0.05 level.
lowest for the poorly ventilated as compared with the well-
and moderately ventilated categories. These results indi-
cate that the contribution of indoor PM2.5 sources to per-
sonal PM2.5 exposures was highest when individuals spent
little to no time in well-ventilated—or high air-exchange
rate—indoor environments.
compared with personal exposures for particles of non-
ambient origin. As a result, personal PM2.5 exposures of
individuals who spent more time in these well-ventilated
indoor environments are more strongly associated with
corresponding ambient concentrations.
It should be added that the above associations be-
tween personal exposure and ambient concentrations were
unique to PM2.5 and were not observed for any of the other
measured pollutants. As a result, the potential for con-
founding appears to be limited, despite significant corre-
lations that were observed among ambient pollutant
concentrations. In contrast to ambient concentrations,
neither personal exposures to total PM2.5 nor PM2.5 of
ambient origin were significantly correlated with personal
exposures to the co-pollutants (i.e., PM2.5-10, PM2.5 of non-
ambient origin, O3, NO2, and SO2). Not surprisingly,
personal–ambient associations for PM2.5-10, O3, NO2, and SO2
were similarly weak and insignificant. Since a confounder
Given these results, it was not surprising that the per-
sonal–ambient association for PM2.5 also declined with ven-
tilation status and that this decline was more pronounced
than that for SO42-. As the contribution of particles of in-
door origin to personal PM2.5 exposures increased, the cor-
responding association between personal exposures and
ambient concentrations became weaker. In contrast, the
personal SO42- exposures were strongly correlated with
ambient concentrations for all three ventilation categories,
suggesting that personal–ambient associations involving
personal exposures to particles of ambient origin are strong
and are less affected by indoor ventilation conditions as
Volume 50 July 2000
Journal of the Air & Waste Management Association 1197

Sarnat, Koutrakis, and Suh
4. Schwartz, J.; Dockery, D.W.; Neas, L.M. J. Air & Waste Manage. Assoc.
1996, 46, 2-12.
5. Vedal, S. J. Air & Waste Manage. Assoc. 1997, 47, 551-581.
6. Janssen, N.A.H.; Hoek, G.; Brunekreef, B.; Harssema, H.; Mensink, I.;
Zuidhof, A. Am. J. Epidemiol. 1998, 147(6), 537-547.
7. Rojas-Bracho, L.; Suh, H.H.; Koutrakis, P. J. Expos. Anal. Environ.
Epidemiol., in press.
8. Abt, E.; Suh, H.; Allen, G.; Koutrakis, P. Environ. Health Perspect. 1999,
108(1), 35-44.
9. Suh, H.H.; Spengler, J.D.; Koutrakis, P. Environ. Sci. Tech. 1992, 26,
2507-2516.
10. Chang, L.T.; Sarnat, J.A.; Wolfson, J.M.; Rojas-Bracho, L.; Suh, H.H.;
Koutrakis, P. J. Atmos. Pollution 1999, 40th Anniversary Issue, 31-39.
11. Thomas, K.W.; Pellizzari, E.D.; Clayton, C.A.; Whitaker, D.A.; Shores,
R.C.; Spengler, J.D.; Özkaynak, H.; Froehlich, S.E.; Wallace, L.A.
J. Expos. Anal. Environ. Epidemiol. 1993, 3(2), 203-226.
12. Koutrakis, P.; Wolfson, J.M.; Bunyaviroch, A.; Froehlich, S.E.; Hirano,
K.; Mulik, J.D. Anal. Chem. 1993, 65, 209-214.
must be associated with the exposure of interest, these
results provide among the first evidence from an expo-
sure study that the effects observed in the PM epidemio-
logic studies are probably not due to confounding by
the measured PM co-pollutants. Given the strong corre-
lations among the ambient pollutants and the strong
personal–ambient associations for PM2.5, however, am-
bient co-pollutant concentrations may be appropriate sur-
rogates for personal exposures to PM2.5 or PM2.5 of ambient
origin for some individuals. It should be noted that other,
unmeasured co-pollutants, such as carbon monoxide and
specific PM components, may also act as surrogates for
personal PM2.5 or, alternatively, may confound the ob-
served associations between ambient PM2.5 and health
effects. The importance of these unmeasured co-pollut-
ants should be evaluated in future studies.
It is also important to note that inter-pollutant vari-
ability in method precision between PM and the gaseous
pollutants and the general low level of personal gaseous
exposures may account for some of this lack of correla-
tion among personal PM2.5 and its co-pollutants. An analy-
sis employing more precise methods of co-pollutant
sampling would improve the strength of these findings.
A detailed analysis of subject time-activity diaries is ex-
pected to yield more information of the relationships
outlined in this paper and the corresponding strength
between ambient concentrations and personal exposures.
13. Ogawa & Company, USA, NO, NO2, NOx, and SO2 Sampling Protocol
Using the Ogawa Sampler; 1998, Version 3.
14. Quality Assurance Project Plan; Harvard School of Public Health, De-
partment of Environmental Science and Engineering, 1999.
15. Özkaynak, H.; Xue, J.; Spengler, J.; Wallace, L.; Pellizzari, E.; Jenkins,
P. J. Expos. Anal. Environ. Epidemiol. 1996, 6(1), 57-78.
16. Wilson, W.E.; Suh, H.H. J. Air & Waste Manage. Assoc. 1997, 47, 1238-
1249.
17. Mage, D.; Wilson, W.; Hasselblad, V.; Grant, L. J. Air & Waste Manage.
Assoc. 1999, 49, 1280-1291.
18. Suh, H.H.; Koutrakis, P.; Spengler, J.D. J. Air & Waste Manage. Assoc.
1993, 43, 845-850.
19. Wallace, L.A. J. Air & Waste Manage. Assoc. 1996, 46, 98-126.
20. Lioy, P.J.; Waldman, J.M.; Buckley, T.; Butler, J.; Pietarinen, C. Atmos.
Environ. 1990, 24B(1), 57-66.
21. Tamura, J.K.; Ando, M.; Sagai, M.; Matsumoto, Y. Environ. Sci. 1996,
4, 37-51.
22. Clayton, C.A.; Perritt, R.L.; Pellizzari, E.D.; Thomas, K.W.; Whitmore,
R.W.; Wallace, L.A.; Özkaynak, H.; Spengler, J.D. J. Expos. Anal. Environ.
Epidemiol. 1993, 3(2), 227-250.
23. Long, C.M.; Suh, H.H.; Koutrakis, P. J. Air & Waste Manage. Assoc., in
press.
24. Bahadori, T.; Suh, H.H.; Rojas-Bracho, L.; Koutrakis, P. Dissertation,
Harvard University, Boston, MA, 1998.
ACKNOWLEDGMENTS
The authors wish to thank all of the participants of the
study as well as M. Wolfson, G. Allen, S. Ferguson, L.-T.
Chang, P. Catalano, J. Evans, A. Wheeler, J. Baker, C. Jang,
D. Millman, J. Sullivan, A. Joshi, F. Plucienik, R. Simon,
M. Erdman, C. Howard-Reed, G. Lau, J. Sekula, D.
Belliveau, K. Misra, N. Alba, and R. Cheung. Fixed-site
monitoring data were provided, in part, by the Mary-
land Department of the Environment. This study was
funded by the Electric Power Research Institute and the
American Petroleum Institute. Additional funding for this
analysis was provided by the Harvard-EPA Center on Par-
ticle Health Effects (grant #R827353-01-0).
About the Authors
Jeremy A. Sarnat is a doctoral student at the Harvard School
of Public Health, Environmental Science and Engineering
Program. Dr. Petros Koutrakis is a professor of environmental
sciences at the Harvard School of Public Health. Dr. Helen
H. Suh is an assistant professor of environmental chemis-
try and exposure assessment at the Harvard School of Public
Health. Correspondence should be addressed to Jeremy
A. Sarnat, 665 Huntington Avenue, Building 1, Room 1308a,
Boston, MA 02115, jsarnat@hsph.harvard.edu.
REFERENCES
1. Dockery, D.W.; Pope, C.A.; Xu, X.; Spengler, J.D.; Ware, J.H.; Fay, M.E.;
Ferris, B.G.J.; Speizer, F.E. N. Engl. J. Med. 1993, 329, 1753-1759.
2. Schwartz, J. Am. J. Epidemiol. 1994, 139, 589-599.
3. Pope, C.A.; Thun, M.J.; Namboodiri, M.M.; Dockery, D.W.; Evans,
J.S.; Speizer, F.E.; Heath, C.W., Jr. J. Respir. Crit. Care Med. 1995, 151,
669-674.
1198 Journal of the Air & Waste Management Association
Volume 50 July 2000