University of Southern California Dissertations and Theses

Long term effects of air pollution on lung function in the European community respiratory health survey

(USC Thesis Other)

Long term effects of air pollution on lung function in the European community respiratory health survey

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

LONG TERM EFFECTS OF AIR POLLUTION ON LUNG FUNCTION
IN THE EUROPEAN COMMUNITY RESPIRATORY HEALTH SURVEY

by

Thomas Götschi

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EPIDEMIOLOGY)

August 2007

Copyright 2007 Thomas Götschi

ii
Acknowledgements
My sincerest thanks, tremendous appreciation, and deepest respect go to Nino Künzli, for being not only
a brilliant academic advisor, but also a supporter, a challenger, a listener, a motivator, and an inspiration.
Working with Nino has been an invaluable privilege.
I would also like to thank Frank Gilliland for his academic advice, and, most of all, for teaching me the
joy of critical scientific thinking. I would like to thank Jim Gauderman for his countless advice and the
numerous times of significantly increasing the probability of me finding some meaning behind numbers.
My thanks also go to Mike Jerrett who reassured me that space is more than a single location and proved
to me that interdisciplinarity is not an urban myth. I would also like to express my thanks to Costas
Sioutas who as the outside member on my committee provided me with the necessary insight into the
matter of my research.
The European Community Respiratory Health Survey is a huge project made possible thanks to the
collaboration of hundreds of people. I would like to express my sincere thanks to all members of the
working group air pollution. To work with these colleagues from all over Europe was always extremely
pleasant, inspiring and productive. My very special thanks go to Bertil Forsberg, Jordi Sunyer, and
Deborah Jarvis who enabled me to spend time with their local research groups. To visit these
collaborators was definitely one of the highlights of my research activities and not only allowed me to
progress in my work, profit from countless expertise, and experience different work environments, but
also gave me the privilege to befriend many amazing people.
Further, my thanks go to Rob McConnell, Andrea Hricko, and Achim Heinrich for their help, and for
inspiring me with their passionate attitude towards their work. My thanks also go to Ruanne Peters for
her invaluable teaching of the epidemiology basics. I also want to express my thanks to all my
colleagues and friends at USC and elsewhere who inspired, supported, or just simply entertained me in
countless conversations, namely Tracy, Mirna, Talat, Towhid, Made, Nicole, Ketan, Bernie, Danny,
Anne, Benedicte, Žižek, Nuria, and many others. Thousand thanks also go to the staff of the

iii
Environmental Health and the Biostatistics Division at USC, namely to Mary, Celia, Leticia, Wendy,
Jessica, and Helen, among others.
The US Environmental Protection Agency (EPA) generously supported my studies with a STAR fellow-
ship. My gratitude also goes to Stan Azen and John Peters for supporting me as a doctoral student at
USC. I am also grateful for the additional financial support by USC, my parents, and my generous
grandmother.
In contrast to common belief, PhD students do have a life besides their studies. The moral support I
received from friends and family over the past five years was crucial to the success of my studies. Most
of all I owe my sincerest thanks to my parents for being so close despite of the long distance that
separated us. I thank Eva and Natascha for being the reliable friends they were on a path of ups and
occasional downs. Christian, Beat, Lilly, Jan, Kate, Janie, and Federico also deserve credit for helping
me put my research in perspective.
I want to express my thanks and my deepest admiration to all the volunteers at the Bicycle Kitchen.
They taught me how to achieve with passion and reminded me that our minds can only succeed if our
feet stay rooted.
Last but not least, I owe Emily my many thanks for being there during the time when a PhD does
actually not allow for a life besides work. I am truly looking forward to return the favor and be there for
her and share the pleasures and the work that come with a new life.

iv
Table of Contents
Acknowledgements ii
List of Tables vi
List of Figures viii
Abbreviations ix
Abstract xii
Preface: Dissertation Structure xiv
Chapter 1: Introduction 1
Chapter 2: Literature Review on Long Term Effects of Ambient Air Pollution on Lung Function 5
Introduction 5
Results 6
Cross-sectional Studies in Children and Adolescents 7
Studies Using Individually Assigned Exposure 7
Studies Using Traffic Related Exposure Assignment 8
Studies Using Cross-community Exposure Assignment 9
Studies Based on Two or Three Comparison Groups Only 10
Cross-sectional Studies in Adults 11
Studies Using Individually Assigned Exposure 11
Studies Using Traffic Related Exposure Assignment 12
Studies Using Cross-community Exposure Assignment 12
Studies Based on Two or Three Comparison Groups Only 13
Longitudinal Studies in Children and Adolescents 17
Longitudinal Studies in Adults 18
Discussion 25
Interpretation of the Findings within a Biological Framework 26
Comparability of Studies 27
Validity of Studies 31
Exposure Characterization and Assignment 31
Validity of Lung Function Assessment 34
Potential Confounding 36
Selection Bias 38
Measurement Bias 38
Publication Bias 38
Conclusions 39
Chapter 3: Air Pollution Assessment in the European Community Respiratory Health Survey 41
Introduction 41
Methods 45
Results 48
Discussion 57

v
Conclusions 64
Chapter 4: Air Pollution and Lung Function in the European Community Respiratory Health Survey 66
Introduction 66
Methods 67
Statistical analysis 69
Covariate Adjustment 70
Subsample Analyses 71
Sensitivity Analyses for Height Adjustment 71
Sensitivity Analysis for Age Adjustment 72
Loss to Follow-up Analysis 73
Results 73
Results for PM
2.5
and Lung Function Level 78
Results for PM
2.5
and Change in Lung Function 83
Results from Sensitivity Analyses for Height Adjustment 87
Results from Sensitivity Analyses for Age Adjustment 87
Results from Loss to Follow-up Analysis 87
Discussion 90
Lung Function Assessment 95
Characterization of Exposure 101
Central Monitor Based Exposure Assessment 101
Spatial Variability of Exposure Markers 102
Diversity of Study Areas 104
Annual Means as Estimates of Long Term Exposure 106
PM
2.5
as a Marker of Exposure 106
NO
2
at Home Outdoors as an Individual Marker of Exposure 107
Confounding 110
Modeling Approach 115
Conclusions 116
Chapter 5: Synthesis and Outlook 117
References 123
Appendices 140
Appendix 1: Summary Tables of Reviewed Studies of Less Relevance 141
Appendix 2: Historic Air Pollution Data for the ECRHS Cities 150
Appendix 3: Statistical Power 154
Appendix 4: Type of Spirometer and Lung Function 155
Appendix 5: Loss-to-follow-up Effect 156
Appendix 6: ECRHS Lung Function Protocol 157
Appendix 7: Tables of Center Level Correlations 161
Appendix 8: Results of Main Analysis using Sulfur instead of PM
2.5
as Exposure Marker 165
Appendix 9: Sensitivity Analyses using Alternative Exposure Markers. 171
Appendix 10: Sensitivity Analysis using a “SAPALDIA-like” Model 173

vi
List of Tables
Table 1: Reviewed cross-sectional studies of high relevance. 14
Table 2: Reviewed longitudinal studies of high and medium relevance. 20
Table 3: Exposure assignment approaches in relation to quality criteria for studies. 32
Table 4: Characteristics of the ECRHS air pollution monitoring locations. 44
Table 5: Annual means for each air pollution indicator and center. 49
Table 6: Seasonal means for each air pollution indicator and center. 53
Table 7: Within city Pearson correlations between PM
2.5
and all other air pollution indicators. 56
Table 8: Pearson correlations of annual means across centers. 57
Table 9: Spirometers used in ECRHS and changes in equipment and software at follow-up. 68
Table 10: Summary of available lung function data for each ECRHS center. 74
Table 11: Baseline comparison of subjects lost and subjects followed-up. 75
Table 12: Means and proportions of relevant variables for the complete sample. 76
Table 13: Coefficients for lung function level from main models for complete samples. 80
Table 14: PM
2.5
coefficients for lung function level using different adjustment variables. 81
Table 15: PM
2.5
coefficients for lung function level using different subsamples. 82
Table 16: Coefficients for change in lung function from main models for complete samples. 84
Table 17: PM
2.5
coefficients for change in lung function using different adjustment variables. 85
Table 18: PM
2.5
coefficients for change in lung function using different subsamples. 86
Table 19: Effects of loss to follow-up on the association between PM
2.5
and baseline lung function. 89
Table 20: Significantly detectable effects sizes for lung function level and change in lung function. 94
Table 21: Demographic and geographic characteristics of ECRHS centers. 105
Table 22: Country specific socio-economic parameters. 112
Table 23: Annual means of meteorological parameters for ECRHS centers. 114
Table 24: Cross-sectional studies of medium relevance. 141

vii
Table 25: Cross-sectional studies of low relevance. 144
Table 26: Longitudinal studies of lower relevance. 147
Table 27: Availability of historic air pollution data for ECRHS cities. 150
Table 28: Significantly detectable effects sizes for FEV1/FVC ratio, and middle range of PM
2.5
. 154
Table 29: Center level correlations between latitude, PM
2.5
, and various lung function measures. 161
Table 30: Sulfur coefficients for lung function level from main models for complete samples. 165
Table 31: Sulfur coefficients for lung function level using different adjustment variables. 166
Table 32: Sulfur coefficients for lung function level using different subsamples. 167
Table 33: Coefficients for change in lung function from main models for complete samples. 168
Table 34: Sulfur coefficients for change in lung function using different adjustment variables. 169
Table 35: Sulfur coefficients for change in lung function using different subsamples. 170
Table 36: Sensitivity analyses using alternative exposure markers in the main model. 171
Table 37: Results from Sensitivity Analysis using a “SAPALDIA-like” Model. 173

viii
List of Figures
Figure 1: Effect estimates from studies on air pollution and lung function in children. 23
Figure 2: Effect estimates from studies on air pollution and lung function in adults. 24
Figure 3: Effect estimates from studies on traffic related air pollution and lung function. 25
Figure 4: Locations of the 21 ECRHS study centers which participated in the air pollution module. 41
Figure 5: Seasonal air pollution patterns 51
Figure 6: Spatial variability of PM
2.5
, reflectance, and elements in Antwerp. 61
Figure 7: Historic NO
2
data available for ECRHS centers. 62
Figure 8 Historic SO
2
data available for ECRHS centers. 63
Figure 9: Distribution of PM
2.5
annual mean concentrations across ECRHS centers. 78
Figure 10: PM
2.5
levels across ECHRS centers by type of spirometer used. 97
Figure 11 Sub-analysis for change in FEV1 in men across centers which used Biomedin spirometers. 99
Figure 12 Sub-analysis for change in FEV1 in men across Southern European centers. 100
Figure 13 Comparison of PM
2.5
and sulfur annual means across ECRHS study centers. 103
Figure 14 Comparison of central monitor based PM
2.5
and NO
2
annual means, and NO
2
2-week
measurements at participants’ homes outdoors. 108
Figure 15: Lung function levels across ECHRS centers by type of spirometer used. 155
Figure 16: Illustration of loss-to-follow-up effect on PM
2.5
FEV1 association at baseline. 156

ix
Abbreviations
Abbreviation Explanation
Abs Light absorption [absorption coefficient /m]
AC Antwerp City, Belgium
AL Albacete, Spain
Al Aluminum
APHEA Air Pollution and Health: A European Approach
AS Antwerp South, Belgium
As Arsenic
ATS American Thoracic Society
BA Barcelona, Spain
Bi Bismuth
BMI Body mass index (weight/height
2
)
Br Bromine
BS Basel, Switzerland
C Carbon
Ca Calcium
Cd Cadmium
CHS Children's Health Study
Cl Chlorine
Co Cobalt
COPD Chronic obstructive pulmonary disease
Cr Chromium
Cu Copper
EC Elemental carbon
ECRHS European Community Respiratory Health Survey
ED-XRF energy dispersive X-ray fluorescence spectrometry
EMEP Co-operative Programme for Monitoring and Evaluation of the Long-range
Transmission of Air pollutants in Europe

x
ER Erfurt, Germany
ERS European Respiratory Society
ETS Environmental tobacco smoke
EXPOLIS Air Pollution Exposure Distribution within Adult urban Populations in Europe
Fe Iron
FEF
25-75%
Forced mid-expiratory flow rate (=MMEF)
FEV1 Forced expiratory volume in 1 second
FVC Forced vital capacity
GA Galdakao, Spain
Ga Gallium
GIS Geographical information system
GN Grenoble, France
GOLD Global Initiative for Chronic Obstructive Lung Disease
H Hydrogen
HU Huelva, Spain
IP Ipswich, United Kingdom
K Potassium
Mg Magnesium
MMEF Maximum mid-expiratory flow rate (=FEF
25-75%
)
Mn Manganese
µg/m
3
micrograms per cubic meter
N Nitrogen
Na Sodium
NHANES National Health and Nutrition Examination Survey
Ni Nickel
NO Norwich, United Kingdom
NO
2
Nitrogen dioxide
O
3
Ozone
OV Oviedo, Spain
P Phosphorus
PA Pavia, Italy
Pb Lead
PEF Peak expiratory flow rate

xi
PM
10
Particles with a median cut off diameter of 10um
PM
2.5
Particles with a median cut off diameter of 2.5um
ppb parts per billion
PS Paris, France
RE Reykjavik, Iceland
SALIA Study on the Influence of Air Pollution on Lung Function, Inflammation, and Aging
SAPALDIA Swiss Study on Air Pollution and Lung Disease in Adults
Se Selenium
SES Socio-economic status
Si Silicon
SO
2
Sulfur dioxide
SO
4
2-
Sulfate
TA Tartu, Estonia
Ti Titanium
TSP Total suspended particles
TU Turin, Italy
UM Umea, Sweden
UP Uppsala, Sweden
V Vanadium
VE Verona, Italy
WHO World Health Organization
Zn Zinc

xii
Abstract
Lung function is an important measure of respiratory health and a predictor of cardio-respiratory
morbidity and mortality. More than fifty publications have addressed long term effects of ambient air
pollution on lung function. The vast majority of studies reported some significant adverse effects on lung
function, including several studies on traffic. There is strong support for air pollution effects on lung
function growth in children, while in adults the evidence is limited to comparisons of lung function
levels and knowledge on effects on lung function decline remains inconclusive. The diversity in study
designs and investigated air pollution and lung function measures is high, limiting the comparability of
studies and the ability to draw quantitative conclusions.
The association of long term air pollution and lung function has not been studied across adult European
multi-national populations. The aim of the European Community Respiratory Health Survey (ECRHS)
was to determine the association between long-term urban background air pollution and lung function
levels, as well as change in lung function among European adults from 21 study centers.
Fine particles (PM
2.5
) were measured using central monitors. Lung function (FEV1, FVC) was tested at
baseline and after 9 years of follow-up. Multilevel linear regression models were used to analyze effects
on lung function level and change in lung function. No significant associations were found between
PM
2.5
and any of the spirometric measures. PM
2.5
, lung function, and numerous covariates were
correlated with latitude.
The observed null-findings may be explained by limitations of ECRHS, in particular various aspects of
exposure misclassification, and the potential for center-level confounding along latitude. These
limitations are inherently associated with the cross-community design, which is a consequence of the
central monitor based exposure assessment. To adequately capture urban air pollution contrasts and to
avoid confounding by center-level factors, future studies should therefore estimate exposures on an
individual or within-community level.

xiii
Besides the question of effects on decline there remain several other unresolved lung function topics that
require further research, such as early life exposures, effects on the plateau phase, susceptibility factors,
and the specific roles of traffic and other pollution sources, among others.

xiv
Preface: Dissertation Structure
The aim of this dissertation is to investigate the long term effects of air pollution on lung function in
adults participating in the European Community Respiratory Health Survey (ECRHS). A comprehensive
review of the state of knowledge on long term effects of air pollution on lung function is provided. The
analysis within ECRHS is discussed in the broader context of epidemiologic research on long term
effects of air pollution on lung function. The two main hypotheses investigated are:
1: Long term urban background air pollution is associated with lung function levels
2: Long term urban background air pollution is associated with change in lung function
Chapter 1 provides a brief introduction on the role of lung function in the investigation of long term
effects of air pollution. Some background on lung function and spirometry is provided. The biological
rationale for air pollution effects on lung function is laid out. A brief introduction of ECRHS is given.
Chapter 2 provides a literature review of previous studies on long term effects of air pollution on lung
function. It is an expanded version of a paper to be submitted for publication. It concludes that further
research on long term effects of air pollution on lung function is needed, in particular for effects on lung
function decline in adults.
Chapter 3 describes the air pollution assessment within ECRHS. Chapter 3 is the subject of a published
paper (Götschi et al. 2005).
Chapter 4 provides the main analysis investigating the association between urban background air
pollution and lung function among participants of ECRHS. Chapter 4 is an extended version of a paper
to be submitted for publication.
Chapter 5 provides a synthesis of this work and previous research on long term effects of air pollution
and lung function. It discusses the relevance of the findings of this work in the broader context of future
air pollution studies and concludes with the main lessons to be learned from this study.

1
Chapter 1: Introduction
Air pollution has been associated with numerous adverse health outcomes. (Brunekreef et al. 2002;
Katsouyanni 2003; Pope et al. 2006; Schlesinger et al. 2006) Acute effects have been shown for
respiratory symptoms and cardiovascular events, hospital admissions and mortality. Long term levels of
air pollution have been associated with chronic bronchitis, markers of atherosclerosis, lung cancer, and
mortality, among others. While many of the short term effects are small and reversible, chronic effects
are of particular public health concern since they may comprise a bigger burden to individuals and
society as a whole (Brunekreef et al. 2002).
Short term variability of lung function may occur over a period of hours or days, and is in most cases
reversible (Dockery et al. 1996). Short term effects of air pollutants on lung function, mainly of gases
such as ozone and sulfur dioxide, are well documented from research in animal models and human
chamber studies (Gong 1992), as well as in observational studies of daily air pollution levels and
pollution episodes (Brunekreef et al. 1995; Schindler et al. 2001; Ward et al. 2004). On the other hand,
steady changes in lung function also occur over much longer periods of time, namely years or even
decades, and may represent chronic loss in lung function (Dockery et al. 1996). As such, lung function is
one of only a few physiological preclinical outcomes suitable for the study of chronic effects of air
pollution. Lung function is further of interest as a well established predictor of morbidity and mortality
(Sin et al. 2005). It is a continuous, objective measure which early on reflects a cumulative health status
which is of relevance for the rest of a subject’s life. The American Thoracic Society (ATS) considers
“any detectable level of permanent lung function loss attributable to air pollution exposure” an adverse
effect (American Thoracic Society 1985; 2000). The literature review and analyses within this
dissertation are exclusively aimed at long term or chronic effects on lung function.
Lung function follows a distinct development from birth to old age which is characterized by three
segments. A sharp growth until early adulthood is followed by a so-called plateau phase of little change
over several years during the mid-twenties. With increasing age lung function is subject to a moderate

2
decline. During the first six years of life lung growth is driven by an increase in the number of alveoli
and airways, and subsequently by an approximately isometric increase in their dimensions. Body size
and elasticity of the lung and thorax are the main determinants of lung function in healthy subjects.
Elasticity of both lung and thorax increases from birth to adolescence when lung function peaks. After
the plateau phase the lung becomes more flaccid, leading to a gradual reduction in lung function.
Adverse effects on lung function are typically distinguished into two categories. Restrictive conditions
diminish the maximum achievable lung volume, while obstructive conditions affect small airways
leading to reduced airflows.
In air pollution studies lung function is typically assessed using spirometry. A number of spirometric
measures are used to assess restriction and obstruction of the lung. The most common ones used in air
pollution studies are measured during a forceful expiration maneuver after maximum inspiration.
Forced vital capacity (FVC) is a measure of lung size which measures the total volume that can be
expired after a maximum inspiration. Forced expiratory volume within one second (FEV1) is the most
commonly used measure of airway obstruction. It measures the maximum volume that can be forcefully
expired within one second, but in fact it can be interpreted as a measure of average flow. To adjust for
the fact that smaller lungs produce smaller flows, FEV1 can be divided by FVC when assessing
obstruction. The GOLD criteria define chronic obstructive pulmonary disease (COPD) as
FEV1/FVC<0.7 (Pauwels et al. 2001). Several studies have also used flow measures, such as peak
expiratory flow (PEF) and maximum mid-expiratory flow (MMEF) or forced expiratory flow between
the 25th and 75th percentile of FVC (FEF25-75), which measure the maximum and average flow during
the middle of the maneuver, respectively.
Age, gender, body size, and ethnicity are strong predictors of lung function. After adjusting for these
variables, approximately 30% (FVC, Becklake 1986) of variability between individuals remain to be
explained by smoking, genetic predisposition, past respiratory diseases, and other risk factors, such as air
pollution.

3
The effects of tobacco smoke on lung function may serve as a model for adverse effects of air pollution
on the lung. Abundant studies linked tobacco consumption to such outcomes as slower lung growth,
lower lung function levels attained after the growth period, shortened plateau phase, and accelerated
decline in adults (Dockery et al. 1988; Sherman 1992; Jaakkola 1994; Gold et al. 1996; Anthonisen et al.
2002).
The understanding of the body’s defense mechanisms against air pollutants and the health effects
resulting thereof is growing but still incomplete (Gilliland et al. 1999; Nel 2005; Schlesinger et al. 2006).
Increased levels of oxidative stress induced by oxidative gases and particles in particular affect the redox
balance in the lung lining fluid and can cause direct structural damage to epithelial membranes. This first
line of defense is coupled with a persistent inflammatory response in the lung which leads to tissue
damage and affects lung function. Some inflammation caused effects on lung function will be short term
and reversible, as for example observed after exposure to high concentrations of oxidants, such as ozone,
or during the recovery from a common cold. A chronically increased inflammatory state of the lung,
however, is believed to continuously damage the lung and lead to irreversible loss of lung function.
These chronic pro-inflammatory effects are also considered the main link between air pollution and
cardio-vascular outcomes (Kunzli et al. 2005a; Nel 2005; Pope et al. 2006; Schlesinger et al. 2006) for
which lung function is a marker (Sin et al. 2005). Therefore lung function is an important link in the
investigation of chronic effects of ambient air pollution.
Urban ambient air pollution is a complex mixture of gases and particles, and to date epidemiologic
studies have not been able to identify specific substances responsible for health effects. Although several
gases are irritants it is widely assumed that particles play the crucial role in cardio-respiratory health
effects of air pollution (Dockery 2001; Pope et al. 2006). Various particle characteristics, such as size,
surface, and chemical composition determine the toxicity of particles, although their specific relevance
for long term effects on the human lung remains unclear.
Particles are typically measured as mass concentrations (e.g. µg/m
3
) of particles of a certain size range
(e.g. smaller than 2.5µm). Most particle measures, such as particles smaller than 2.5µm in diameter

4
(PM
2.5
) or total suspended particulates (TSP) are not source-specific, which is a major pitfall from a
regulatory perspective. Besides traffic and industrial emissions, natural dust or sea spray can constitute
significant sources of particles.
The European Community Respiratory Health Survey (ECRHS) is the most comprehensive multi-center
respiratory cohort study in adults in Europe (Burney et al. 1994). The standardized air pollution
assessment during the years 2000 and 2001 aimed at determining long term estimates of PM
2.5
and its
elemental composition across 21 centers participating in the baseline survey and a follow-up survey after
nine years (Hazenkamp-von Arx et al. 2003). Lung function was measured as FEV1 and FVC. The aim
of this project was to test the hypotheses that long-term urban background air pollution was associated
with both lung function level (cross-sectional analysis) and change in lung function (longitudinal
analysis), respectively. As reviewed in the next chapter, evidence of an association is currently
ambiguous.

5
Chapter 2: Literature Review on Long Term Effects
of Ambient Air Pollution on Lung Function
This chapter is subject to a manuscript submitted to a peer reviewed journal.
Introduction
As a consequence of the continuous and cumulative nature of lung function, both cross-sectional and
longitudinal study designs are suitable to assess long term effects of air pollution. Cross-sectional studies
compare lung function levels across exposure contrasts, while longitudinal studies investigate
differences in rates of change in lung function over time across different exposure levels.
This review focuses on the published literature on long term effects of ambient air pollution on lung
function. The OVID MedLine and PubMed databases were searched for the terms “lung function”,
“pulmonary function”, “FEV1” (forced expiratory volume in 1 second), and “FVC” (forced vital
capacity) to find publications on lung function. The search terms “Air pollution”, “particulate matter”
(PM), “PM
2.5
”, “PM
10
” (PM of aerodynamic diameter smaller than 2.5 and 10µm, respectively), “NO
2
”
(Nitrogen dioxide), and “O
3
” (ozone), and were used to identify publications on air pollution. The
relevant publications were selected manually, reviewing titles, abstracts, and reference lists.
This review is comprehensive for studies published in 1990 and later. Older studies are referred to when
of particular relevance. Most of the older studies have been reviewed previously by Sunyer (2001). A
summary of the main findings of the relevant studies is followed by a discussion of their comparability
and validity, and in particular of the characterization of exposure, measurements of lung function, and
potential bias and confounding.

6
Results
The description of studies follows an outcome oriented subdivision into cross-sectional and longitudinal
studies, and into studies in children and adults, respectively. Thereafter, the way exposure was assigned
is considered most influential with regards to the interpretation of effect estimates and the assessment of
their validity. Exposure assignment defines what type of pollution can be assessed, how precisely it can
be assessed, and in addition, has a significant impact on the potential for confounding bias. Therefore, a
further distinction is made between studies that assigned exposure individually and those that assigned
group level exposures based on exposure related clustering of individuals (cross-communities
comparisons). Among the latter, studies conducted across multiple communities, and those conducted
across only two or three communities are distinguished. Studies which based exposure on some marker
of traffic related pollution fall into a middle ground, sometimes assigning truly individual traffic
measures (e.g. residential traffic counts), and sometimes comparing group level exposures across within-
community contrasts (e.g. road buffers).
Overall, 58 publications were reviewed. The vast majority were cross-sectional (41), compared to 17
longitudinal analyses. Two thirds of the studies investigated children (37). There are only two recently
published longitudinal study in adults (Nakai et al. 1999; Sekine et al. 2004), while 4 of the 5 reviewed
cohort studies in children were published in 2002 or later (Neuberger et al. 2002; Gauderman et al. 2004;
Ihorst et al. 2004; Rojas-Martinez et al. 2007).
Community level central monitor based air pollution measurements were the predominant way to
characterize exposure, typically across cities (16), sometimes within cities (6), and sometimes to assign
individual exposure (4). Most of these studies measured several different air pollutants. Results for
multiple pollutants are provided where those are unlikely to be surrogates for the same mixture of air
pollution. Seven studies used some sort of traffic measure as a marker of exposure. Only one study used
residential measurements (Schindler et al. 1998), and one study was based on measurements in a 1km
grid (Hirsch et al. 1999). Among the cross-community studies there are eight studies with seven or more

7
communities and thirteen across only two or three centers. Five studies rely entirely or partly on
descriptive characteristics of the study areas (or the subjects’ profession in a study on policemen (Karita
et al. 2001)) to define exposure contrasts.
Findings of the more relevant studies are summarized below. Smaller studies, in particular those across
two or three communities only are considered less relevant. Summaries of the more relevant studies are
also provided in Table 1 and Table 2, and summary tables for the less important studies are provided in
the annex in Table 24, Table 25, and Table 26, on pages 141 and following. The main findings of the
most relevant studies are also shown in Figure 1, Figure 2, and Figure 3, for studies in children and
adolescents, in adults, and on traffic, respectively.
Cross-sectional Studies in Children and Adolescents
Studies Using Individually Assigned Exposure
The University of California Berkeley Ozone Study (Kunzli et al. 1997a) investigated the effects of O
3
,
PM
10
, and NO
2
on lung function in a convenience sample of 130 college freshmen aged 17-21. They
applied a novel approach of assessing "effective lifetime exposure" to O
3
by combining central monitor
air pollution data, residential history, and information on time-activity from questionnaires and
population based surveys. Despite individually assigned exposure values the exposure contrasts were
strongly driven by the differences in (ecologic) pollution levels between two study regions. The low to
medium life-time O
3
exposures were experienced by participants that grew up in the San Francisco Bay
Area while students with a residential history in Southern California all had medium to high life-time
exposures. They observed significant negative effects of O
3
on small airways (FEF25–75% and
FEF75%), but not on FEV1 or FVC. PM
10
and NO
2
were not associated with lung function and the
associations for ozone were not sensitive to adjustment for those two pollutants.
In a replication of Künzli et al. (1997a) by the same group, Tager et al. (2005) investigated the effects of
lifetime O
3
exposure in a larger sample of 255 Berkeley college freshmen on various measures of lung
function (FEV, FVC, FEF25-75, FEF75). They found significant negative associations between flow

8
measures and O
3
when introducing an interaction term for FEF25-75/FVC*O
3
, indicating higher
susceptibility of subjects with narrower small airways.
In a similarly designed study Galizia et al. (1999) observed significantly lower lung function among
Yale freshmen who grew up in counties with high long term O
3
levels, as compared to their colleagues
who grew up in low ozone counties.
Studies Using Traffic Related Exposure Assignment
Several more recent studies investigated traffic related exposure contrasts within communities. In
Munich, Germany, Nicolai et al. (2003) used traffic counts within 50m of residence and an emission
model which considered stop-and-go traffic characteristics to assign exposure to 2019 children aged 9-
11. They found no associations between traffic and lung function (spirometric measures not specified),
despite significant adverse effects on respiratory symptoms. In contrast, Wjst et al. (1993) showed
significant associations between traffic density in Munich school districts and measures of expiratory
flow (PEF, MEF25) in 4320 9-11 years old children.
Hirsch et al. (1999) used data from an extensive measurement campaign measuring SO
2
, NO
2
, CO,
benzene, and O
3
on a 1km grid and assigned estimated annual means to the home locations of 1256 9-11
year old children in Dresden. They found some associations with respiratory symptoms, but none with
lung function and any of the measured air pollutants. Fritz et al. (2001) provide a descriptive study
indicating a stronger effect of traffic than coal heating related air pollution on lung function of 5 year old
preschoolers in Leipzig, Germany.
Brunekreef and colleagues (1997) studied 877 Dutch schoolchildren who lived within 1000m of
motorways in six different areas of the Netherlands. They found negative effects of truck traffic density
on various lung function indicators (FEV1, PEF, FEF25-75) ranging between 2.5% - 8% per 10,000
trucks. Black smoke, NO
2
, and car traffic density tended to have negative effects as well. In a second
study of that same design now including 24 schools, the Dutch group could not reproduce the earlier

9
findings on lung function, although associations between symptoms and traffic indicators prevailed
(Janssen et al. 2003).
In a novel approach Hogervorst et al. (2006) used oxygen radical formation by particles as a marker for
their potential to cause oxidative stress, a key mechanism in chronic inflammation of the lung and other
diseases. They studied Dutch school children from 6 schools in varying distance to traffic. They found
some significant negative effects of radical formation per particle mass, but less so for radical formation
per volume of air, and some particle mass measures had significant beneficiary effects on lung function.
Studies Using Cross-community Exposure Assignment
Two papers report on repeated cross-sectional assessments conducted across the period of dramatic
decline in air pollution levels between 1991 and 1998 after the German reunion. Sugiri and colleagues
(2006) report improving lung function (Total lung capacity, airway resistance) with decreasing levels of
TSP and SO
2
among 2574 East German 6 year old children, catching up with their western counterparts
by the time of the third survey, eight years after the reunion. The improvement was weaker in children
living within 50m of a busy street, attributed to the dramatic increase in motor vehicles (+50-75%)
during the same period in East Germany. Frye et al. (2003) report improvements of FEV1 and FVC over
three consecutive cross-sectional surveys (1992-97) of 2493 11-14 year olds in three East German
communities. FVC significantly increased by 4.7% for a 50µg/m
3
decrease of TSP and 4.9% for a
decrease of 100µg/m
3
SO
2
, while effects for FEV1 were smaller and not significant.
In the American Harvard 6-Cities Study, neither Ware et al. (1986) nor Dockery et al. (1989) found
significant associations between air pollution and lung function measures in two cross-sectional studies
of preadolescent children enrolled between 1974-79. Ware et al. compared TSP (range: 39-114µg/m
3
) to
FEV1 and FVC in approx. 10,000 children across 6 cities, while Dockery et al. investigated a sample of
5422, limited to 10-12 year olds, and used flow measures (FEV
0.75
, MMEF) and additional particle
measures (PM
15
, PM
2.5
; range: 11.8-36.7µg/m
3
).

10
Schwartz (1989) found highly significant associations between lung function and TSP, NO
2
, and O
3
in a
large cross-sectional study including 4300 6-24 years old subjects from 44 US cities (TSP 10-90%: 43-
95µg/m
3
). The findings suggested threshold effects for TSP and O
3
, and less clearly for NO
2
.
In a comparison across 24 American and Canadian cities Raizenne et al. (1996) found significant
associations between air pollution levels of the previous year (O
3
, NO
2
, PM) and lung function (FEV1,
FVC, FEV
0.75
, FEF
25-75
, PEFR) in 10,251 8-12 year old children. Air pollution levels were also
associated with the proportion of children with low lung function, defined as equal or below 85% of
their predicted value. The observed effects were strongest and most consistent for particle acidity, a
measure driven by the fraction of very small particles.
Peters et al. (1999) reported significant associations between various air pollutants (PM
10
, PM
2.5
, acid
vapor, NO
2
, O
3
) and several measures of lung function (FVC, FEV1, MMEF, PEF) in the cross-sectional
baseline investigation of the Southern Californian Children's Health Study across 12 communities.
Cross-sectional correlations were predominantly observed among the female members of this cohort of
3293 children (age 9-16), and more pronounced in subjects who spent more time outdoors. Results from
the follow-up of this study are described below under longitudinal studies.
Studies Based on Two or Three Comparison Groups Only
Numerous studies report comparisons between only two or three exposure groups, sometimes
distinguished based on air pollution measurements (Xu et al. 1991; He et al. 1993; Schmitzberger et al.
1993; Stern et al. 1994; Goren et al. 1999; Yu et al. 2001; Longhini et al. 2004; Lubinski et al. 2005),
and in other cases based on community characteristics only (Forastiere et al. 1994; Devereux et al. 1996;
Jang et al. 2003). Among those studies with air pollution measurements, negative effects of air pollution
on lung function have been reported by a large Canadian study of more than 3000 children living in ten
rural communities in Saskatchewan and Ontario (Stern et al. 1994), in an Italian comparison between an
urban and a rural area (Longhini et al. 2004), a study of Polish soldiers (age 18-23) (Lubinski et al.
2005), a comparison across three exposure zones in the European Alpine region (Schmitzberger et al.

11
1993), a comparison across two rural communities in Israel (Goren et al. 1999), and in comparisons
across two areas of Hong Kong (Yu et al. 2001), and Wuhan, China (He et al. 1993), respectively. Two
studies based only on community characteristics reported negative or inconsistent findings (Forastiere et
al. 1994; Jang et al. 2003).
Cross-sectional Studies in Adults
Studies Using Individually Assigned Exposure
Schindler et al. (1998) used a sub-sample (N=560) of the Swiss Study on Air Pollution and Lung
Disease in Adults (SAPALDIA) random sample to measure personal and home outdoor NO
2
exposure
using passive samplers. From three repeated measurements they extrapolated individual annual mean
estimates which they used to derive average exposures for residential zones (N=6-13 per center) within
the eight study communities, as well as community averages. They observed consistently negative,
though statistically non-significant associations between NO
2
and lung function, measured as FVC and
FEV1. Associations for between-residential zone comparisons (within community) were weaker than
those for between-communities comparisons. Associations were also stronger for personal NO
2
exposure
than for NO
2
at home outdoors.
Abbey et al. (1998) provide a cross-sectional analysis of adults enrolled in a large cohort study in 1977
(ASHMOG, Adventist Health Study of Smog) who were tested for lung function in 1993. Life time
exposure to PM
10
, O
3
, SO4, and SO
2
were calculated based on subjects’ residential history, back to 1973.
All subjects lived within 20 miles of an air pollution monitor when recruited for the cohort study in
1977. Exceedance frequency indices were calculated as the number of days above a certain pollution
level (40, 60, 80, 100 µg/m
3
for PM
10
). The main findings were negative associations for PM
10
(100) and
FEV1 in males with parental history of airway obstructive disease. They also reported significant
associations between daily variability of self-conducted flow measurements (PEF lability) and
PM
10
(100) in females and never smoking males.

12
Studies Using Traffic Related Exposure Assignment
In a recently published study Kan et al. (2007) observed significantly decreased FEV1 and FVC in
female participants of the Atherosclerosis Risk in Communities (ARIC) study with increased density of
and proximity to traffic. No significant effects were observed in male members of the cohort. Although
this study did conduct follow-up measurements three years after the baseline examination, the follow-up
period was considered too short to detect effects on lung function change. Cross-sectional analysis of the
follow-up data provided similar results as the baseline analysis.
In the German Study on the Influence of Air Pollution on Lung Function, Inflammation, and Aging
(SALIA) Schikowski et al. (2005) included proximity to the nearest busy road (>10’000 vehicles/day) in
an otherwise central monitor based cross-community analysis. They found a significantly increased risk
for COPD measured as FEV1/FVC<0.7 among those women (age ~55) living closer than 100m to a
busy road (OR=1.33, 95% CI 1.03-1.72). FEV1 and FVC were also significantly lowered in proximity to
the nearest road (-1.3%, -1.7%, respectively).
In a comparison based on occupational status of participants, Karita et al. (2001) observed lower lung
function in Thai policemen working in Bangkok traffic, as compared to policemen working in non-
traffic and rural environments.
Studies Using Cross-community Exposure Assignment
In their cross-community analysis, Schikowski et al. (2005) observed strong significant negative
associations between FEV1, FVC, and FEV1/FVC ratio with NO
2
and PM
10
(calculated from TSP)
measurements from fixed site monitors (-4.7%, -3.4%, -1.1%, per 10µg PM
10
, respectively). The cross-
sectional surveys were conducted over several years and the participating communities varied over time.
Strong support for an adverse effect of air pollution on lung function levels in adults is provided by the
Swiss SAPALDIA study. Ackermann-Liebrich et al. (1997) compared air pollution levels (PM
10
, NO
2
,
SO
2
, O
3
) with FVC and FEV1 of 9651 randomly selected participants across the eight study

13
communities, finding significant negative associations. For PM
10
the predicted effect of 10µg/m
3

increase in annual mean concentration was a 3.4% decrease in FVC and a 1.6% decrease in FEV1.
Chestnut et al. (1991) reported significant associations between TSP and lung function in the adult
sample (25-75 years old, N=6913) of the NHANES I survey (National Health and Nutrition
Examination Survey). The association was non-linear, suggesting a threshold at 60µg/m
3
TSP below
which there was no association. The findings were robust for several subsample analyses (Whites, non-
smokers) and exclusion of the two cities with highest and lowest TSP levels, respectively.
Studies Based on Two or Three Comparison Groups Only
Jammes et al. (1998) observed lower lung function (FEV1, MMEF) in long term residents of the city
center of Marseille, France, compared to residents of outskirts with lower levels of PM
10
and NO
x
. Xu et
al. (1991) conducted a comparison across three neighborhoods in Beijing, one residential, one suburban,
and one industrial, respectively, with large contrasts in air pollution (SO
2
, TSP). After adjustment,
significant effects of air pollution were observed among subjects who did not use coal for domestic
heating. Wang et al. (1999) reported similar results in a comparable study in the city of Chongquing,
China.
In studies without explicit measurements of air pollution, Devereux et al. (1996) observed no significant
differences between lung function in a clean rural (West Cumbria) and a more polluted urban area
(Newcastle) in England among 608 men, while Karita et al. (2001) and Wongsurakiat et al. (1999)
observed lower lung function in traffic exposed policemen in Bangkok, compared to colleagues
employed in more rural areas, and the general population, respectively.

Table 1: Reviewed cross-sectional studies of high relevance. Main study characteristics.
Publication Study Country N
Age
Sex
Health status
FVC, FEV1
FEV1/FVC
PEF, MMEF, FEF()
<0.7/0.8 pred.
Between # of communities
Within community
Between individuals
TSP
PM10, PM2.5
EC, Black smoke, soot
NO2
O3
SO2
Others
Models
Traffic (proximity, density, etc.)
Ackermann 97 SAPALDIA Switzerland 9651 18-61 x 8 x x x x x
Brunekreef 97 Netherlands 877 7-12 x x 6 13 x x x x
Janssen 03 Netherlands 1726 7-12 x x 24 (w) x x x x x
Kan 07 ARIC USA 15792 ~54 x x x x x x x
Nicolai 03 Germany 904 9-11 x (3) x x x x bz x x
Peters 99 CHS USA 3293 9-16 x x 12 x x x H+
Raizenne 96 USA/Canada 10251 8-12 x 22 x x x
Schikowski 05 SALIA Germany 4757 52-56 w na x x x 7 x x (c) x
Schindler 98 SAPALDIA Switzerland 7656 18-60 x 8 <13 x
Other
exposures Study population
Lung function
measures
Exposure
contrast Air pollutants

CHS=Children's Health Study; SAPALDIA=Swiss Study on Air Pollution and Lung Disease in Adults; SALIA=Study on the Influence of Air Pollution on Lung Function,
Inflammation, and Aging; na=non-asthmatics; FVC=forced vital capacity; FEV1=forced expiratory volume in 1s; PEF=peak expiratory flow; MMEF=max. mid expiratory
flow; FEF()=forced expiratory flow (various cutoffs); (w)=consider within community contrasts in some way; (c)=PM
10
calculated from TSP; EC=elemental carbon,
bz=benzene; H+=acidity
14

Table 1 continued: Reviewed cross-sectional studies of high relevance. Summary of statistical analyses.
Publication
Sex, age, height, race
Weight, BMI, obesity
Smoking status
ETS
Asthma, atopy, allergy
Childhood resp. infection
Co-exposures, occupational
SES, income, education
Short term effects, season
Housing characteristics
Sex
Age
Respiratory health
Smoke exposure
Occupational
Income, education
Time of residence
Exposure assessment
Other
Ackermann 97 xxx xx xx x x >10 RE
Brunekreef 97xxxx xx xx
Janssen 03 xxxx x x xx
Kan 07 xxx x x x x x
Nicolai 03xx x x x
Peters 99 xxx xx x x x x
Raizenne 96xxxxxE
Schikowski 05xxxx xxx xx5x
Schindler 98 xxx xx xx >40 x x
Adjusted covariates Sensitivity/ subgroup analyses

BMI = body mass index, ETS = environmental tobacco smoke, SES = socio-economic status, RE = random
effects model, E = exercise prior to test
15

Table 1 continued: Reviewed cross-sectional studies of high relevance. Main results, strengths, and limitations.
Publication Main results Strengths Limitations
Ackermann 97 Sign. neg. assoc. with FVC (all pollutants),
FEV1 (SO2, NO2, O3)
Adjustment for covariates, sensitivity
analyses
Brunekreef 97 Sign. neg. assoc. between truck traffic and
lung function; Stronger effects in girls.
Traffic counts, proximity to freeway,
pollution measures
Short term effects on lung function
measurements?
Janssen 03 Null-findings for lung function, but sign.
assoc. for respiratory symptoms
Sample size, traffic measures Temporal drift in lung function
measurements?
Kan 07 Sign. neg. assoc. and trend with FEV1, FVC
with increased traffic density and proximity in
women, not significant in men.
Traffic exposure assessment; sample size;
adjustment for regional pollution.
Nicolai 03 Null-findings for lung function, but sign.
associations for respiratory symptoms
Model based assessment of exposure to
traffic
Lung function results not reported
Peters 99 Sign. neg. assoc. in females, stronger in
subjects spending time outdoors
Sample size, exposure assess., data analysis
Raizenne 96 Sign. neg. assoc. between PM, acidity, O3 and
FEV1, FVC, FEF
Sample size, exposure assess., data analysis Adjustment for potential confounders?
Schikowski 05 Sign. neg. assoc. FEV1, FVC and FEV1/FVC,
COPD with PM10(TSP), near major road.
Exposure assessment (monitors, traffic, GIS) PM10 based on TSP, partial extrapolation of
exposure?
Schindler 98 Neg. associations of lung function and NO2
within (non-sig.) and across centers (sig.)
Individual passive sampler measurements

COPD = chronic obstructive pulmonary disease defined as FEV1/FVC < 0.7; PM
10
(TSP) = PM
10
based on TSP×0.71; GIS = geographic information system

16

Longitudinal Studies in Children and Adolescents
Six groups have published results from longitudinal studies in children. Except for two sub-studies (Avol
et al. 2001; Gauderman et al. 2007) they are all based on community level exposure assignment. Detels
et al. (1991) included children older than seven and adults in their study.
In the Southern Californian Children's Health Study, Gauderman et al. (2000; 2002; 2004) followed
1759 Children in 12 communities from age 10 to 18 (N=747 at last follow-up). Over the eight years of
follow-up, children living in the most polluted community were predicted to have a growth deficit in
FEV1 of approx. 100ml (~7% for girls, ~4% for boys), as compared to those living in the cleanest
community (exposure range 4-38ppb NO
2
, similar findings for elemental carbon, PM, and acid vapor; no
association with O
3
). Consequently, the proportion of children with clinically low lung function at age
eighteen (FEV1 <80% predicted) was estimated to be almost five times larger in the most polluted
community, as compared to the cleanest community (29µg/m
3
vs. 6µg/m
3
PM
2.5
). In an intervention-like
study of a subsample of 110 children from the same study who moved away after their initial
examination, Avol et al. (2001) observed an improvement in the average lung function growth of those
who moved to areas of better air quality (PM
10
), and slower lung function growth in those who moved to
more polluted areas. In their most recent analysis Gauderman et al. (2007) used distance to freeways as
an exposure metric. Children living within 500m of a freeway had significant deficits in 8-year growth
of FEV
1
(–81 ml, p=0.01) and MMEF (–127 ml/sec, p=0.03), compared to children living at least
1,500m from a freeway, independent of the previously reported effects of regional background pollution.
Rojas-Martinez et al. (2007) followed school children in Mexico City over a period of three years. PM
10
,
NO
2
, and O
3
levels were assigned based on monitors in close proximity (<2km) to the children’s
schools. All three pollutants were associated with significant deficits in lung function growth. Effects for
annual growth in FEV1 per inter quartile range of exposure ranged from -16ml for ozone in boys to
-32ml for NO
2
in girls, with estimates for FVC and FEF
25-75%
(ml/s), and estimates from multi pollutant
models showing effects of similar magnitude.

17

In a series of publications on lung function growth in Austrian school children, Ihorst (2004), Horak
(2002), Kopp (2000), Frischer (1999), and their colleagues reported on seasonal and long term effects of
O
3
and PM
10
. Strong negative effects of ozone and to a lesser degree for PM
10
during summer appeared
to be compensated for during the winter seasons. Over the study period of 3.5 years they could not detect
a significant deficit in overall lung growth among children in higher polluted areas. In a smaller Austrian
cohort study of 200 children drawn from a larger cross-sectional sample, Neuberger and colleagues
(2002) attributed small improvements of lung function to the decrease in NO
2
levels over a five year
period.
In a relatively short longitudinal analysis of 1001 Polish schoolchildren over two years of follow-up,
Jedrychowski et al. (1999) observed significant negative associations between air pollution levels and
lung function growth (FEV1, FVC, proportion of slow growths=lowest quintile) across two zones in
Krakow categorized as low pollution and high pollution, based on black smoke and SO
2
levels.
In a five year follow-up study across two southern Californian communities Detels et al. (1991)
observed faster decline in the subjects who lived in the more polluted community. The most pronounced
differences were observed for measures of small airway impairment, and in particular among women of
the youngest age group (7-10y).
Longitudinal Studies in Adults
Longitudinal studies on lung function in adults are rare. Sekine et al. (2004) provide a cohort study in
Tokyo women which categorized exposure based on traffic proximity, backed up by NO
2
and particle
measurements. Women who provided lung function data (FEV1, FVC) for at least four occasions were
included in the analysis. Cross-sectional annual means of lung function for three exposure groups varied
considerably over the eight study years, as did the number of participating women. The average decline
of lung function for each group was modeled using a two stage regression model. Over 8 years of
follow-up stronger declines in FEV1 were observed among women of the higher exposure groups.

18

The remaining longitudinal studies in adults are limited to a few exposure comparison groups only. In an
earlier Japanese study Nakai et al. (1999) compared lung function measurements of 444 30-59 years old
women who lived in three different zones of Tokyo, one in close proximity to a busy road. Up to ten
spirometry tests were taken over a period of two and a half years. No significant differences in lung
function level or change in lung function associated with air pollution contrasts could be detected.
In a 5 year follow-up study of 2625 Southern Californian adults living in three communities, Tashkin et
al. (1994) observed greater declines in FEV1 of up to 26ml/yr (men living in Long Beach) between the
most polluted and the cleanest area (Lancaster). Effects in men were independent of smoking status and
of similar magnitude as active smoking. In women, significant effects were only observed in never
smokers.
Jedrychowski et al. (1989) followed a cohort of 1414 subjects over 13 years in Krakow, Poland.
Exposure was assigned based on residence in three districts of the city varying in SO
2
and PM
concentration, and sulfur transformation ratio (STR, a measure of acidity). Decline in lung function was
significantly associated with STR, but not with SO
2
or PM levels.
In a study of 3337 Dutch adults over 12 years of follow-up Van der Lende (1981) found a stronger
decline in VC (8.9ml/yr) and FEV1 (9.7ml/yr) among subjects living in an urban area, as compared to
subjects living in a rural area.

19

Table 2: Reviewed longitudinal studies of high and medium relevance.
Publication Country Relevance N
Age
Sex
Health status
Smoking status
FVC, FEV1
PEF, MMEF, FEF()
<0.7/0.8 pred.
# of measurements
Years of follow-up
Between # of communities
Within community
Between individuals
Avol 01 USA High 110 10-15 xx 2+ 5 x
Gauderman 04,02,00 USA High 1759 10-18 xxx 8 8 12
Gauderman 07 USA High 3677 10-18 xxx 8 8 12 x
Goss 04 USA Medium 11484 6~40 cf x 8 2 x
Horak 02 Austria Medium 975 6-9 xx 6 3 8
Ihorst 04 Austria Medium 2153 6-10 x73.53/15
Kopp 00 Austria Medium 797 6-9 x x 4 2 3/10
Neuberger 02 Austria Medium 3451 ES xx 2-8 5 (2)
Rojas-Martinez 07 Mexico Medium 3170 8-12 xx 3 10
Sekine 04 Japan Medium 406 30-59 wnr ns x 8 8 3
Study population
Lung function
measures
Exposure
contrast

cf = cystic fibrosis, w = women only, nr = no respiratory disease, ns = non-smokers, 3/# = 3 exposure groups across total # of
communities. (2) = comparison of communities with declining vs. increasing NO
2
.
20

Table 2 continued: Reviewed longitudinal studies of high and medium relevance.
Publication
TSP
PM10, PM2.5
EC, Black smoke, soot
NO2
O3
SO2
Others
Residential history
Traffic (proximity, density, etc.)
Sex, age, height, race
Weight, BMI, obesity
Smoking status
ETS
Asthma, atopy, allergy
Occupational
SES, income, education
Short term effects
Season
Housing characteristics
Baseline lung function
Sex
Respiratory health
Smoke exposure
Time of residence
Exclude extreme centers
Exposure assessment
Avol 01 x x x x xx x
Gauderman 04,02,00 x x x x H+ xxxxx xx x xxx x
Gauderman 07 x x x x x xxxxx xx x x xx
Goss 04 x x x x xx x x x
Horak 02 x x x x x xxxx x xxx x
Ihorst 04 x x x x xx xx xxx x
Kopp 00 (x) (x) x xx
Neuberger 02 x dx x x x
Rojas-Martinez x x x x xx x x x
Sekine 04 x x xxxx xx x
Air pollutants
Other
exposures Adjusted covariates
Sensitivity/ subgroup
analyses

(x) = no quantitative use of pollution contrasts, H+ = acidity, BMI = body mass index, ETS = environmental tobacco smoke;
21

Table 2 continued: Reviewed longitudinal studies of high and medium relevance.
Publication Main results Strengths Limitations
Avol 01 Lung function growth lowered by move to high
pollution area, up when moved to low air
pollution area
Exposure intervention Sample size
Gauderman
04,02,00
Sign. reduced lung function growth in children
from higher polluted communities
Length of follow-up, data analysis
Gauderman
07
Lung function growth independently assoc.
with freeway distance and regional pollution
Exposure assessment, study design
Goss 04 Pulmonary. exacerb. sign. assoc. with PM, O3;
FEV1 decline assoc. with PM2.5
Sample size, individual exposure assignment Specific population (Cystic fibrosis patients)
Horak 02 Sign. reduced lung function growth in summer
in children from higher polluted communities
Semi annual lung function assessments Effect size? Confounding? Analytical
approach?
Ihorst 04 O3 decreased lung function growth in summer,
opposite pattern in winter
Semi annual lung function assessments Exposure categories? Focus not long term
effects
Kopp 00 O3 decreased lung function growth in summer,
opposite pattern in winter
Short follow-up period. Exposure
categorization?
Neuberger 02 Faster growth in MMEF in districts with
declining NO2
Potential confounding, loss to follow-up?
Rojas-
Martinez 07
Sign. reduced lung function growth in children
from areas with higher PM10, NO, O3
Data analysis; subjects' proximity to monitors. Short follow-up period.
Sekine 04 Indication of faster lung function decline in
proximity to traffic
Exposure measure Loss to follow-up, unstable lung function
estimates

22

Figure 1: Effect estimates from some of the most relevant studies on long term effects of air pollution
on lung function in children.
Publications are listed alphabetically and by study in the case of multiple publications from the same
study. Effect estimates are selected based on relevance, and, where available, a measure of lung
volume (preferably FEV1) and a measure of flow (preferably mid expiratory flow; MMEF, FEF
25-
75%
) is provided. Scales on the x-axis, units and interpretation of effect estimates are not directly
comparable across studies and only shown for the purpose of qualitative comparisons. Dashed lines
indicate the null effect level. For exact interpretation of effect estimates refer to the text or original
publications.

23

Figure 2: Effect estimates from some of the most relevant studies on long term effects of air pollution
on lung function in adults.

a)
among men without a parental history of asthma, bronchitis, emphysema, or hay fever;
b)
among men with a parental history of asthma, bronchitis, emphysema, or hay fever;

24

Figure 3: Effect estimates from some of the most relevant studies on long term effects of air pollution
on lung function using traffic indicators as exposure measures.

Discussion
The vast majority (49) of the reviewed publications reported at least one statistically significant adverse
effect of air pollution on lung function. Only nine reported no association. The reviewed studies differ
considerably in their size and settings, their applied methods, in their results, and ultimately in their
overall quality. Summaries of the main characteristics of most relevant studies are provided in Table 1
and Table 2.
The structure of this discussion section is as follows: First, the findings of the studies of the highest
quality are interpreted in the context of a biological framework for the various phases of lung function
development throughout life, and for different types of exposures. Then the considerable challenges of

25

comparisons across the reviewed studies are addressed. Figure 1, Figure 2, and Figure 3 illustrate the
challenges of comparing the findings of the arguably most relevant studies. The next section addresses
several issues that affect the validity of the studies, with a particular focus on three main determinants of
the quality of a study, namely the way exposure was assessed, the lung function assessment, and the
potential for confounding. Finally, the assessment of potential selection bias concludes the discussion.
Interpretation of the Findings within a Biological Framework
Strong evidence for adverse long term effects of air pollution on lung function comes from the
longitudinal Southern California Children’s Health Study which found reduced lung function growth in
children, resulting in significant deficits at the end of adolescence (Gauderman et al. 2004). In adults,
SAPALDIA provides the strongest evidence for cross-sectional differences in lung function of subjects
exposed to different levels of air pollution (Ackermann-Liebrich et al. 1997). These findings have been
confirmed in SALIA, which included women only (Schikowski et al. 2005). Numerous other studies
support these findings in one or the other way. The diversity in methods and quality, however, precludes
conclusions on several more specific aspects of air pollution effects on lung function.
For example, no study has ever followed adolescents until they reached the plateau phase. It therefore
remains unknown whether air pollution related growth deficits merely reflect a delayed lung growth that
may be compensated for by a prolonged growth phase, or whether these subjects will later as young
adults enter the lung function decline phase starting with a reduced lung function. Further, the scientific
basis is not sufficient to determine whether air pollution only affects lung function in children and cross-
sectional differences in adults merely reflect the growth deficits obtained during childhood, or whether
adults’ lungs are similarly affected by air pollution, resulting in accelerated decline in lung function, as
suggested by four longitudinal studies (van der Lende et al. 1981; Jedrychowski et al. 1989; Tashkin et
al. 1994; Sekine et al. 2004).
With the exception of one study (Avol et al. 2001), the reviewed studies also do not permit to address
whether reduced lung function growth rates observed in teenagers were due to concurrent air pollution

26

exposures or whether they were caused by exposures experienced much earlier in life. A unique study
presented by Avol et al. made use of individual changes in air pollution over time by following up on
kids that moved away after the baseline investigation of the Children’s Health Study (Avol et al. 2001).
This intervention-like study provides some of the strongest support for effects of concurrent air pollution
exposures and suggests that improvements in air quality may allow children’s lungs to recover from
previously experienced adverse effects.
Besides a few suggestive results from some studies (e.g. Raizenne et al. 1996; Kunzli et al. 1997a;
Gauderman et al. 2004), the overall health relevance of specific pollutants cannot be judged beyond the
conclusion that all studies investigated predominantly combustion related pollutants. This is despite the
fact that some studies were specifically designed to address ozone (i.e. the UC Berkeley studies), or
varying mixtures of pollutants (i.e. the Children’s Health Study). Findings by several studies specifically
addressing traffic as a major source of air pollution have been mixed. On the other hand, the Children’s
Health Study found reduced lung growth independently related to higher levels of air regional air
pollution (Gauderman et al. 2000; 2002; 2004) and to proximity to traffic (Gauderman et al. 2007).
Adverse effects of both regional air pollution and proximity to traffic have also been found in SALIA
(Schikowski et al. 2005). Also Kan et al. (2007) reported adverse effects for proximity to traffic after
adjusting for regional air pollution, although only significant in women, but not in men.
Comparability of Studies
Besides the distinction of studies in children and adults, and cross-sectional and longitudinal studies,
several additional design aspects limit the direct comparability of the reported effects. Figure 1, Figure 2,
and Figure 3 provide an overview of the various effect measures used in the reviewed studies.
Most noteworthy are the differences in exposure assessment, namely between community level
(ecological) exposure assignment based on only one or a few monitors (Kunzli et al. 1997b), various
approaches to assign exposure individually (e.g. Kunzli et al. 1997a; Abbey et al. 1998; Schindler et al.
1998; Galizia et al. 1999), and exposure measures based on various markers of traffic-related pollution

27

(i.e. density, proximity, etc.) (e.g. Brunekreef et al. 1997; Wongsurakiat et al. 1999; Schikowski et al.
2005; Gauderman et al. 2007; Kan et al. 2007).
Community level exposure assignment can be considered the standard approach since it has been applied
by most studies, the majority of which reported some significant effect of air pollution on lung function
(e.g. Schwartz 1989; Ackermann-Liebrich et al. 1997; Gauderman et al. 2004; Schikowski et al. 2005).
The approach is limited to regional background pollution and cannot capture within-community
(individual) exposure contrasts due to local sources, such as for example traffic. Although most of these
studies measured various pollutants, pollutant specific interpretation of effects is not possible due to a
number of reasons. The correlation of long term averages is often high, the differences in within-
community spatial characteristics lead to different precision and accuracy of estimates, further, methods
and settings of the studies vary, and some studies used unique measures which have not been
investigated by others. In general, however, the investigated pollutants all reflect various aspects of
combustion related pollution.
Individual level exposure assignment has only been used by a few studies (Kunzli et al. 1997a; Abbey et
al. 1998; Schindler et al. 1998; Galizia et al. 1999; Avol et al. 2001; Nicolai et al. 2003; Schikowski et
al. 2005; Tager et al. 2005; Gauderman et al. 2007; Kan et al. 2007), some of which attempted to
improve exposure estimates from central site monitors by taking individuals’ mobility over time or
residential history into account (Kunzli et al. 1997a; Abbey et al. 1998; Tager et al. 2005).
Measurements of individual exposures to actual pollutants are difficult and were only conducted in a
sub-study of SAPALDIA (Schindler et al. 1998). This analysis may serve as an example to illustrate
why findings based on individual and community level exposure assignment may not be easily
compared. Effect estimates for NO
2
varied considerably within the SAPALDIA study when estimated in
a cross-community analysis (Ackermann-Liebrich et al. 1997), as compared to a within-community
analysis (Schindler et al. 1998). When interpreting such results it has to be kept in mind that NO
2
serves
as a marker of a pollution mix, and when measured on the individual level does not represent the same
type of pollution as when measured at a regional background monitor. Residential NO
2
measurements

28

(outdoors) capture the large spatial variability in NO
2
which predominantly results from traffic
emissions. When measured at a central background monitor, however, NO
2
levels are assumed to reflect
the average background of combustion related pollution of the entire area represented by the monitor.
From a chemical and physical, as well as from a health perspective “NO
2
” measured at a regional
background monitor is therefore a marker of a considerably different pollution mix than when measured
at the residential level, and consequently these cannot directly be compared with each other.
Comparisons across studies are further complicated by a considerable variation in the concentrations
investigated which may affect their ability to detect air pollution effects. For example, in the Six Cities
Study no effects of TSP (range 34-80µg/m
3
) (Ware et al. 1986) or PM
2.5
(12-37µg/m
3
) (Dockery et al.
1989) could be found on lung function levels of children. In contrast, the NHANES study reported
strong effects above relatively high threshold levels for TSP (>85µg/m
3
), NO
2
(>0.04ppm), and ozone
(>0.04ppm), in both children (age 6-24y) (Schwartz 1989) and adults (24-75y) (Chestnut et al. 1991).
A rather specific complication in the comparison of effect sizes can arise from converting TSP into
PM
10
, as done in SALIA. PM
10
levels were calculated from TSP measurements using an average
conversion factor derived at several locations in the study region (×0.71; range 0.62 – 0.84), however,
the true ratios in the study communities were unknown. The assumption of a ubiquitous PM
10
/TSP ratio
could influence the PM
10
effect size considerably, if the true PM
10
/TSP ratio would in fact depend on
pollution concentrations. Such a scenario could arise if particles in highly polluted areas would primarily
stem from combustion and therefore be smaller, whereas in the lower polluted areas they might consist
more of windblown dust and therefore be larger. If for example in SALIA the true PM
10
/TSP ratios had
been smallest (0.62) in the low polluted communities and largest (0.84) in the high polluted communities
the effect size for PM
10
would have been only half of the reported estimates (13ml/µgm
-3
vs.
27ml/µgm
-3
).
Ultimately, the sometimes contradictive findings for specific pollutants, such as for example ozone (e.g.
compare Kunzli et al. 1997a; vs. Gauderman et al. 2004), may serve as the strongest caveat against
interpreting pollutant specific effect estimates at face value. Exposure measures that may in one study

29

very well reflect a specific pollutant may somewhere else be a marker of a broader mixture of pollution,
or vary in precision, or be confounded by unmeasured factors, and therefore lead to presumably
contradictive findings.
Several studies aimed at traffic as a major source of urban air pollution, without measuring actual
pollutants (e.g. Wjst et al. 1993; Janssen et al. 2003; Nicolai et al. 2003). Even more so than studies
based on pollution measurements, comparisons of traffic studies are challenged by widely differing
exposure measures and contrasts (see Figure 3). Although exposure contrasts and categories strongly
depend on local study settings, further research would profit from some standardization of traffic
exposure measures.
Further limitations on comparisons across studies arise from the populations studied, the way data was
analyzed, and the lung function measures used.
The vast majority of the reviewed studies either only measured lung volumes (FEV1, FVC), or reported
consistent effects for both volume and flow measures. Only a few studies observed significant
associations for flow measures but not for volumes (Wjst et al. 1993; Kunzli et al. 1997a; Neuberger et
al. 2002; Tager et al. 2005). Although there is biological support for stronger effects on smaller airways,
(Frank et al. 2001; Churg et al. 2003) the reviewed epidemiologic studies do not provide enough of a
basis for such a conclusion. Some studies reported effects for unique measures not used by any other
study, such as pulmonary exacerbation among cystic fibrosis patients (Goss et al. 2004), lability of peak
expiratory flow (Abbey et al. 1998), or total lung capacity and airway resistance (Sugiri et al. 2006).
Based on these few studies there is no conclusion to be made for the relevance of these specific lung
function measures.
Subgroup analyses are an important tool to detect small effects of air pollution which may only affect a
group of susceptible individuals. Identification of effect modifiers can enhance the understanding of the
underlying biology of air pollution effects. Stratified analyses are also a valuable tool to exclude the
possibility of (residual) confounding (by the stratifying variable) and to reduce random misclassification

30

by limiting analyses to subjects with more precise estimates, such as for example long term residents
(Ware et al. 1986; Raizenne et al. 1996; Ackermann-Liebrich et al. 1997; Schikowski et al. 2005). Some
studies observed differences in effects between sexes, but the findings are not consistent. Abbey et al.
(1998) and Galizia et al. (1999) observed stronger effects on lung function levels in males, Jedrychowski
et al. reported a stronger effect on lung function growth in boys (1999) and on decline in men (1989), as
did Tashkin et al. (1994). On the other hand, Peters et al. (1999), Gauderman et al. (2007), Brunekreef et
al. (1997), and Kan et al. (2007) all observed stronger effects of air pollution either in girls or women.
The interpretation of these sex differences must take into account the different potentials for exposure
misclassification and confounding between sexes. Based on the published literature the question of a
biological difference in effect mechanisms between sexes cannot be answered. Two studies used
interaction terms to identify significant effects in unique subgroups which have not been analyzed by
others. Abbey et al. (1998) found their strongest air pollution effects in males who’s parents had asthma,
bronchitis, emphysema, or hay fever, which they interpret as a suggestion of a genetic component of
susceptibility towards air pollution. Tager et al. (2005) present significant interaction terms for ozone
and the ratio between FEF
25-75
and FVC (O
3
*FEF
25-75
/FVC), indicating stronger effects among subjects
with small airways. Both studies do not include the main effects of the interaction variable in their
models, which limits the interpretability of these findings.
Validity of Studies
Exposure Characterization and Assignment
The way exposure is assigned has not only implications on the interpretation and comparability of
results, but also on the overall quality of a study, in particular its feasibility to adequately estimate
exposure and address the study question in general. As pointed out earlier, studies can be divided into
three main groups based on how they assigned exposure, namely on an individual level, on a community
level, and more recently using traffic related markers of either individual or within-community exposure
contrasts.

31

The exposure aggregation level determines a study’s capability to keep exposure misclassification at a
minimum, capture a sufficient exposure range, measure the biologically relevant exposure(s), and limit
the potential for confounding to a minimum. In Table 3 the three exposure assignment approaches
distinguished within this review are loosely compared with regards to the most relevant quality criteria.
Table 3: Exposure assignment approaches in relation to quality criteria for studies.
For example, community level exposure assignment allows for air pollution measurements over long
time periods leading to estimates of very high temporal precision, whereas the precision of exposure
estimates on the individual level is much more prone to be affected by the temporal variability of the
true exposure. (+ = advantageous, o = somewhat limited, - = disadvantageous).
Quality criterion
Community
level
Traffic
indicators
Individual
level
Temporal precision of exposure estimates + o-
Spatial precision of exposure estimates - o+
Exposure range (Stat. power due to exposure) - o+
Sample size (Stat. power due to outcome) + o-
Potential health relevance of exposure - +o
Diversity of measurable pollutants + --
Potential for confounding - o+
Exposure assignment approach

Community Level Exposure Assignment
Community level exposure assignment has been applied by most studies. Continuous measurements at
central monitoring sites often conducted over several years allow determining precise long term
estimates of multiple pollutants. Using central monitors, the community level approach allows for large
study samples, thereby accepting a considerable amount of (random) misclassification of individual’s
exposures by assigning all individuals of the same community the same exposure value. The validity of
exposure estimates measured at a central monitor is therefore mainly determined by the spatial
homogeneity of a pollutant. For a single monitor to reflect the pollution of a wide area the measured
pollutant’s spatial variability needs to be low. The approach therefore is limited to so-called background
pollution, a term referring to a pollution mix that is not immediately affected by local sources, such as
traffic. In fact, central monitors in close proximity to traffic (or other) sources will produce biased
estimates of average community exposures, most likely for such pollutants as NO
2
, fine particles, or O
3

32

due to scavenging effects by nitrogen oxides (NO →NO
2
). Further, the same logic as for the location of
central monitors applies to the location of subjects (homes). Central monitors will poorly predict
exposures at subjects’ homes if those are strongly affected by local sources, again most likely for NO
2
,
fine particles, and O
3
. Estimates from central monitors are also likely to be less precise with increasing
distance between subjects and the monitor. Community level exposure estimates for ozone are likely to
be less precise than those of other pollutants since indoor concentrations for O
3
are much lower than
outdoors. Consistent with this, some authors reported significant associations between O
3
and FVC and
PEFR for children who spent more time outdoors, but not for those who spent less time outdoors
(Gauderman et al. 2002).
Individual Level Exposure Assignment
Several studies attempted to improve exposure estimates from fixed site monitors by taking individuals’
mobility over time or residential history into account (Kunzli et al. 1997a; Abbey et al. 1998; Galizia et
al. 1999; Tager et al. 2005). For example, the study by Künzli et al. (1997a), and the later replication of
their study by Tager et al. (2005) used a sophisticated calculation of individual long term exposures
taking time spent outdoors and residential history into account. The overall exposure contrasts remained
dominated by the cross-community differences between the San Francisco Bay area and the Los Angeles
Basin, where the participants came from. A stratified evaluation of their exposure data, however,
revealed that in the high ozone area of Los Angeles there were considerable differences between purely
central monitor based exposure estimates and estimates taking individuals’ time spent outdoors into
account (Pearson r=0.56), whereas in the low ozone region of San Francisco the differences were small
(r=0.75).
In yet another individualized use of central monitor data, Avol et al. (2001) assigned pollution data from
the nearest monitors to movers of the Southern California Children’s Health Study. Subjects’ lung
functions were measured before and after the move into the new environment. This fact enabled the
authors to use relatively large exposure contrasts directly resulting from the children’s move in their
analysis.

33

Since these studies do rely on central monitors their strength lays not actually in the precision of the
exposure assessment, but rather that through assigning exposure to each study subject individually they
overcome the problem of confounding by contextual variables (community level confounding).
Actual individual measurements as conducted by Schindler et al. (1998) profit from directly measuring
considerable exposure contrasts that are found between subjects within communities, but they are
limited by increased measurement error due to more rudimentary sampling technology (i.e. passive
samplers) and limited sampling periods.
Traffic Indicators as Exposure Metrics
In contrast to the results from numerous studies of traffic effects on respiratory symptoms and other
outcomes, (Venn et al. 2000; Wyler et al. 2000; Brauer et al. 2002; Hoek et al. 2002a; Janssen et al.
2003; Nicolai et al. 2003; Wilhelm et al. 2003; Kim et al. 2004; Bayer-Oglesby et al. 2006; McConnell
et al. 2006) the overall picture seems somewhat less clear for the fewer lung function studies. Based on
the published studies it seems premature to speculate on the validity of the various exposure metrics
used. They all have in common that they reflect exposure to tailpipe emissions in one or the other way.
Factors that influence the precision of traffic measures, however, such as the quality of traffic density
data and prediction models, wind direction, and topography are rarely addressed in the reviewed
publications. For example, Brunekreef et al. (1997) speculate in their critical discussion section whether
the significant associations between truck traffic counts and lung function measures in school children
could have been biased by acute effects on lung function in the schools with the highest traffic
exposures, because during lung function measurements at both schools winds were blowing in direction
from the highways.
Validity of Lung Function Assessment
Spirometry is a well established technique to measure lung function, and all major air pollution studies
are in compliance with the guidelines of the American Thoracic Society (ATS) (1995) or the European
Respiratory Society (ERS) (1993). Hence in most cases, study design and implementation likely had a

34

bigger influence on the quality of lung function measurements than technology related variations.
Nevertheless, it cannot be excluded that different devices across study centers introduced error in one or
the other study, as shown in a comparison study by Kuenzli et al (2005b), and the significance of the
highest degree of standardization possible, including equipment, procedures, and personnel must
therefore not be underestimated.
While in cross-sectional studies it is standard to only conduct one lung function assessment, the quality
of cohort studies is clearly improved with increasing number of measurements. The Children’s Health
Study is a good example for a thorough approach to measure the growth curve of children’s lungs over
more than eight years, with spirometry conducted every year.
Janssen et al. (2003) pointed out that under certain circumstances, seasonal timing of lung function
measurements may be of relevance. Since they conducted spirometric measurements at 24 schools
sequentially over a period of several months, a systematic drift of the measured values, possibly due to
changing temperatures, could not be excluded and may have biased their results. Such and similar
concerns apply in particular to multi-center studies, where systematic misclassifications may affect a
whole cluster of subjects and considerably influence the cross-community association. Another example
that might be affected by seasonal effects on the precision of lung function measurements is the Austrian
study presented by Ihorst et al. (2004). Adverse effects of O
3
on lung function occurring over the
summer appeared to be reversed over the winter, leading the authors to conclude that O
3
effects might be
reversible. An alternative explanation, however, was pointed out by Brunekreef (2002) who suggested
that spring measurements of lung function could be confounded by winter PM
10
effects, respiratory
infections or exposure to pollen. If spring measurements of lung function were systematically
underestimated due to one of these factors this could have caused the observed pattern without there
being any true effect of ozone at all.
Despite the potential for bias the above mentioned issues appear to be of limited concern in most of the
larger studies, since they most likely occur at random (i.e. independent of air pollution) and the number
of subjects investigated may partly compensate for misclassification.

35

Potential Confounding
Potential confounding is a main determinant of the validity of the reviewed studies. Where exposure
contrasts result from geographical clusters (i.e. cross-community comparisons), contextual factors are of
particular interest. (Diez Roux 2004) Studies across two (or very few) communities are of particular
concern. Any differences in lung function predictors across the communities, either of truly contextual
factors (e.g. climate) or unadjusted for individual factors (e.g. occupational exposures) will distort the
estimated effects of air pollution on lung function and therefore preclude any strong conclusions. In
contrast, in multi-center studies such confounders would need to be associated with air pollution levels
to bias the estimates. The large multi-center studies on lung function and air pollution are NHANES
(Schwartz 1989; Chestnut et al. 1991), 6-Cities (Ware et al. 1986; Dockery et al. 1989), SAPALDIA
(Ackermann-Liebrich et al. 1997), SALIA (Schikowski et al. 2005), the Southern California Children’s
Health Study (CHS) (Peters et al. 1999; Gauderman et al. 2004), a study across 24 Canadian and US
communities (Raizenne et al. 1996), and a study across 15 Austrian and German communities (Ihorst et
al. 2004). Community level confounding is mostly determined by study design and populations, and
hard to control for retrospectively. The most efficient way to overcome community level confounding is
assigning exposure on an individual, or within-community basis. Studies which use some sort of traffic
measures as exposure are confronted with the challenge that socio-economic factors often follow the
same within city contrasts and therefore could confound association between lung function and traffic
proximity (e.g. Gauderman et al. 2007). Such studies usually attempt to adjust for socio-economic status
(SES) using various crude measures for a multi-factorial phenomenon, namely the impact of SES on
lung function, which, however, is not understood in its entirety. Also, individual level exposure
assignment may not entirely prevent contextual confounding if subjects (and either their exposure levels
or outcomes) cluster within more than one geographical area, such as in the University of California
Berkeley studies (Kunzli et al. 1997a; Tager et al. 2005).
Most studies adjusted for potential individual level confounders, such as smoking, although quality of
available data varied. The Children’s Health Study provides a good example for thorough adjustment for

36

confounders using a hierarchical regression model (Gauderman et al. 2000; 2002; 2004). Also helpful in
assessing potential confounding is a transparent presentation of sensitivity analyses, such as in the
SAPALDIA paper which demonstrates the robustness of the estimates towards various potential
confounders (Ackermann-Liebrich et al. 1997). Difficult to interpret are results from models which
include interaction terms but do not provide coefficients for the interacting variables (Abbey et al. 1998;
Tager et al. 2005). Several studies analyzed subgroups based on smoking status (e.g. Ackermann-
Liebrich et al. 1997). This approach assures that results are not confounded by smoking. Some studies
did not investigate smokers in the first place, to avoid the problem of confounding (e.g. Abbey et al.
1998; Tager et al. 2005). None of the reviewed studies suggest effect modification by smoking,
although, the sensitivity of the used methods would probably be insufficient to detect such a
phenomenon.
Several studies on lung function change adjusted for baseline lung function. Adjustment of baseline lung
function can lead to regression to the mean if baseline lung function and air pollution are correlated,
resulting in spurious estimates for the effect on lung function change (Brunekreef 2002). Interpretations
of effects on change without considering lung function levels at baseline are particularly problematic in
studies of children (with rapidly growing lungs) over short follow-up periods. For example,
Jedrychowski et al. (1999) conclude that lung function growth over two years was slower among
children living in a more polluted area of Krakow, compared to their counterparts living in a cleaner part
of the city (after adjusting for baseline lung function, growth and other covariates). However, lung
function at baseline was higher in the more polluted area, suggesting that the observed differences in
lung growth rate were a reflection of the differences in how far the children’s lungs were developed at
baseline. Longer follow-up periods and multiple follow-up examinations are a preferable way to
disentangle long term effects on change in lung function from the influence of different lung function
levels at baseline.

37

Selection Bias
Bias due to selective participation is a concern in all epidemiologic studies. Efforts to assure the highest
possible participation rates are mostly the only way to address this problem. In follow-up studies,
comparisons of baseline characteristics can provide an idea of whether drop-out occurred selectively or
not, as shown for example by Abbey et al. (1998) or Tashkin et al. (1994) in a smaller study with high
drop out rates. Studies with large loss of participants or which provide insufficient information on study
participants must be interpreted with caution. For example, Sekine et al. (2004) observed strong year to
year fluctuations in average lung function towards the end of the eight year follow-up period in their
study. Data on participants indicate that in the two last examinations of the study the participation rate
dropped to less than 50% of the analyzed sample. Such variations in the group composition may provide
an alternative explanation for the observed lung function differences.
Measurement Bias
For measurement bias to occur, error in exposure assessment would need to correlate with lung function,
or error in spirometry would need to correlate with air pollution. There is no indication for such a
scenario to have significantly affected the results of any of the reviewed studies. However, cross-
community studies, especially if conducted over relatively few centers may be susceptible to systematic
misclassification of community estimates both for exposure and outcome. For example the timing of
spirometry, (Brunekreef 2002; Janssen et al. 2003) or need for extrapolation of exposure data for
selected communities (Schikowski et al. 2005) may lead to biased community estimates, and thereby to
biased overall health effect estimates. It cannot be excluded that such misclassification may have caused
one or the other significant finding, but it is unlikely that the overall observed trend of air pollution
effects is due to measurement bias.
Publication Bias
Given the variety in lung function measures combined with the numerous pollutants and possibilities for
stratifications, a large number of analyses can be conducted. Statistical significance of findings must

38

therefore be carefully evaluated in the context of multiple comparisons and selective reporting of results.
In many studies the reported significant findings are singled out from several non-significant analyses of
various outcomes and pollutants. Of further concern are significant findings for exotic measures that
have not been used by others, such as for example sulfur transformation rate, particle exceedance
frequency, or flow lability, among others. Given the generally small effects, one would expect some
studies to observe no significant associations at all. Publication bias is likely to have occurred among the
smaller studies, which almost exclusively report positive findings. Further indications of selective
publication are that among the nine studies which reported no significant associations between air
pollution and lung function there are six which reported significant associations for other outcomes
within the same publication (Forastiere et al. 1994; Hirsch et al. 1999; Nakai et al. 1999; Jang et al.
2003; Janssen et al. 2003; Nicolai et al. 2003), and two publications (Ware et al. 1986; Dockery et al.
1989) are from the large 6-cities study which reported important results for other outcomes (i.e.
mortality) (Dockery et al. 1993). Exclusive null-findings are only provided by Devereux et al. in a small
qualitative comparison across two regions in England (1996).
Nevertheless, given that after a critical review broad conclusions on the evidence for long term effects of
air pollution on lung function seem premature, the influence of publication bias on the available
knowledge can be considered minimal.
Conclusions
Due to the diversity of the reviewed studies comparisons of their findings are difficult and the magnitude
of a presumable effect of air pollution on lung function cannot be generalized.
There is strong support for adverse long term effects of air pollution on lung function growth in children,
resulting in significant deficits at the end of adolescence. No study has, however, followed up
adolescents until they reached the plateau phase. It therefore remains unknown whether air pollution
related growth deficits merely reflect a delayed lung growth that may be compensated for by a prolonged

39

growth phase, or whether these subjects will later as young adults enter the lung function decline phase
starting with a reduced lung function.
In adults, the strongest evidence for adverse long term effects of air pollution on lung function comes
from cross-sectional investigations. In the absence of conclusive studies on lung function decline it is not
clear, however, whether these cross-sectional associations reflect growth deficits experienced during
childhood and adolescence or whether they are the result of a further role of air pollution in accelerating
lung function decline.
Based on the published literature it can be concluded that although the studies that suggest adverse
effects of long term exposure to air pollution on lung function are numerous, important questions
regarding the most relevant age period and exposure windows remain unresolved. Moreover, the role of
specific pollutants or pollution sources needs to be clarified. The relevance of endogenous and
exogenous susceptibility factors modifying adverse effects of air pollution on lung function is poorly
understood and needs further investigation. Future studies should implement state of the art exposure
assessment technologies aiming at individual level exposures to tackle issues of relevant exposure
measures and potential limitations due to confounding simultaneously.

40

Chapter 3: Air Pollution Assessment in the European
Community Respiratory Health Survey
This chapter is subject to a manuscript published in Atmospheric Environment (Götschi et al. 2005)
Introduction
The purpose of the exposure assessment in the European Community Respiratory Health Survey
(ECRHS) was to determine estimates of the population average long term exposures to urban
background pollution in the 21 participating centers (see Figure 4).
Figure 4: Locations of the 21 ECRHS study centers which participated in the air pollution module.
(Note: two centers in Antwerp, City and South).
Grenoble
Antwerp
Ipswich
Norwich
Reykjavik
Gothenburg
Erfurt
Uppsala
Umea
Tartu
Paris
Basel
Pavia
Verona
Turin
Barcelona
Albacete
Huelva
Galdakao
Oviedo
Grenoble
Antwerp
Ipswich
Norwich
Reykjavik
Gothenburg
Erfurt
Uppsala
Umea
Tartu
Paris
Basel
Pavia
Verona
Turin
Barcelona
Albacete
Huelva
Galdakao
Oviedo

41

A major challenge of ECRHS was that at the time the study was conducted Europe had no common,
standardized air pollution monitoring network. While evidence of an important role of fine particles
(PM
2.5
) was emerging from US studies (Dockery et al. 1993; Schwartz et al. 1996), there were no data
for PM
2.5
available for Europe. Therefore, a standardized monitoring scheme was developed and
implemented for the 21 participating centers. The aim was to derive annual means of those pollutants for
which exposure could be sufficiently well characterized by a single, central monitor, in particular PM
2.5

and some of its constituents, and possibly NO
2
.
PM
2.5
has been associated with numerous health outcomes (Pope et al. 2006) and is known for its
relatively homogenous distribution in space (Ito et al. 2004; Brunekreef et al. 2005). It was therefore
chosen as the main pollution indicator for the health analysis within ECRHS.
However, PM
2.5
is an unspecific measure for a mixture of scores of particulate air pollutants,
predominantly, but not exclusively originating from combustion processes. Therefore, a primary goal of
air pollution research is aimed towards identifying culprit agents of air pollution to understand and
prevent adverse health outcomes. Besides physical aspects such as particle number, size or surface, the
chemical composition of particles is likely to play a crucial role (WHO 2003). Airborne particulate
matter is a mixture of thousands of different substances, diverse in such critical characteristics as their
solubility, persistence in the atmosphere and in human tissue, reactivity, toxicity and carcinogenicity, as
well as their chemical structure and elemental composition.
A second incentive to more specifically characterize ambient particulate pollution is to identify sources
of emissions to be targeted by policies. PM
2.5
mass is not source-specific and its composition can vary
significantly in time and space, primarily due to variations in sources, emission strength, meteorological
conditions, physical processes, and chemical reactions. Asides of combustion processes, including traffic
and industrial emissions, natural sources such as wind blown dust or sea spray can contribute to PM
2.5

levels. Further characterization of PM
2.5
samples therefore is needed to attribute health effects to specific
pollution sources.

42

Therefore, the elemental composition and reflectance of the PM
2.5
was analyzed within ECHRS with the
purpose to verify the comparability of PM
2.5
across the participating centers, and possibly identify a
constituent within the particle mix, or a marker for a specific source of particles, which might be more
health relevant than the particle mass concentration itself.
NO
2
is commonly monitored at fixed site monitoring stations. Due to the spatial variability of this
pollutant, however, the location of the monitor can strongly influence the observed concentrations and
limit their use as a marker for “background pollution” in epidemiological research. Monitoring locations
within ECRHS differed considerably with regards to several characteristics, such as proximity to traffic,
sampling heights, and vicinity to emission sources. Earlier published NO
2
measurements from the fixed
site monitors were therefore not considered as exposure markers in the health analysis (Hazenkamp-von
Arx et al. 2004). The monitoring location characteristics are listed in Table 4.
In an additional effort, historic air pollution data was collected to the extent possible. However, the
quantitative use of historic air pollution data for health analysis turned out not to be feasible due to the
limited availability of data, differences in measurement technology, and lack of standardized
qualification of sampling locations. Results from the historic air pollution data collection have been
published in a report (Naef et al. 2000). The available data for NO
2
and SO
2
are shown in Figure 7 and
Figure 8, respectively. An overview of the available historic data for the ECRHS centers is listed in the
annex in Table 27 on page 150.

43

Table 4: Characteristics of the ECRHS air pollution monitoring locations.
Center
Altitude [m a. s. l.]
Sampling height [m above ground]
Official measurements station
Type of Zone
A
Characterisation of Zone
B
Emission source within 500m
C
Distance to nearest street [m]
Type of nearest street [m]
D
Traffic volume on nearest street
E
Wide/Canyon (nearest street)
F
Frequency of heavy traffic
G
Street type within 100m
D
Highest traffic vol. within 100 m
E
Relevant sites within 100m
H
Mean distance monitor subjects’ homes [km]
Albacete 704 12 n u r t;c 30 m m c s m b;t;cs 1.8
Antwerp C 10 2 y u c t;c 10 m h w c m h b;t;c; cs;g 5.4
Antwerp S 30 2 n r r no 40 s l w s s l n 5.8
Barcelona 24 3 y u r t;c 20 s m c s m; s h b;t;c 5.1
Basel 260 4 y u r t 5 m m w s m;s m b;c;cs u.k.
Erfurt 220 2 y u r t 30 m h w f m;s h b;t;c;g 6
Galdakao 60 14 n r r no 50 s l w s s l b;c 12
Gothenburg 30 25 y u c t 30 m h w c m;s h b;t;c;r 6.1
Grenoble 220 6 n s r t;c 50 s l w n s l c 6.9
Huelva 50 4 y u c; r t;c 10 s l w s s m b;t;c;g 1.6
Ipswich 50 8 n s r t;c 100 s l w s s l r 4.3
Norwich 50 10 y u c t 15 s l w s m m b;t;c u.k.
Oviedo 276 2 y u r p;t;c 15 s m w s s m b;c;t 2.2
Paris 75 13 y u r;c t;c 25 s m c s s m b;c 9
Pavia 70 2 y u r t;c 6 m m w f m;s m b;c 6
Reykjavik 53 5 n s r t 35 s l w s s l b u.k.
Tartu 84 17 n u r p;t 50 m m w f m;s m b u.k.
Turin 239 2 y u r;c t;c 2 m m c f m m b;t;c;cs 3.2
Umea 10 15 y u r;c t 15 s l c s m;s m b;c;cs 4.3
Uppsala 8 15 n u r;c t 50 m h w f m h b;t;c;r 15
Verona 60 4 y u r t;c 4 m m wc m;sm;l b;c;g 4.8

A
u = Urban; s = Suburban; r = Rural;
B
Main activities in the area: r = Residential; c = Commercial;
i = Industrial;
C
Major emission sources within 500 meters: p = Public power, co-generation or district
heating; t = Traffic; c = Commercial, institutional or residential combustion; i = Industrial activities;
no = none;
D
m = Main street; s = Side street; h = Highway; no = No street;
E
Estimated traffic volume of
the street with the highest traffic volume within 100 meters: h = High traffic (More than 10'000
vehicles/day); m = Medium traffic (Between 2'000 and 10'000 vehicles/day); l = Low traffic (Less than
2'000 vehicles/day);
F
w = D/H>1.5; c = D/H<1.5 ;
G
c = constantly; f = frequently; s = seldom; n = never;
H
b = Bus stop; t = Traffic light; c = Crossing; r = Railway; cs = construction site; g = Gas station

44
This chapter describes the elemental composition and reflectance of this large set of PM
2.5
samples
across the 21 European centers of ECRHS. The correlations between PM
2.5
and elements and reflectance

within each city and their different patterns across Europe, and the cross-community correlations of the
annual means of all particle metrics are presented. Lastly, the exposure data are discussed with regards to
their use in health analyses.
Methods
A standardized PM
2.5
protocol was implemented and has been described in detail (Hazenkamp-von Arx
et al. 2003). Briefly, in each of the 21 study centers we used identical equipment (Basel-Sampler from
BGI, Inc.; Gelman Teflo filters), procedures, and sampling and storage schemes. In each center, a central
monitoring site was chosen, either at a pre-existing air monitoring station, or in collaboration with local
air monitoring authorities. The main characteristics of the monitoring locations are listed in Table 4.
Monitors in Italy and Antwerp City were located close to roads, which is likely to have affected some of
the metrics.
Between June 2000 and December 2001, particle samples were collected on seven days over a two-week
period during each month, yielding 84 days over a one year period. Weekday samples were exposed
24h, whereas weekends were captured on single filters exposed for 48h (in total 72 filters/center). The
predetermined sampling days were the same for all centers. Overall, more than 1600 samples were
collected. All filters were weighed in the same laboratory. Reweighing of selected filters demonstrated a
high reliability of the weighing process and analyses of blank filters did not suggest the need for
adjustment for filter contamination (Hazenkamp-von Arx et al. 2003). Given the restricted sampling
schedule of 84 days per year, the annual mean estimates are expected to fall within a 10% margin of
error from a hypothetical true mean based on daily measurements (Cyrys et al. 2006).
PM
2.5
filter samples were analyzed for 26 different chemical elements, using energy dispersive X-ray
fluorescence spectrometry (ED-XRF), a non-destructive method previously applied in the EXPOLIS
study (ED-XRF, Geochemical Laboratory, Institute of Mineralogy and Petrography, Basel
University/CH-4056 Basel) (Mathys et al. 2001). ED-XRF is capable of detecting elements with an
atomic number above Z = 11, but is neither able to analyze low atomic number elements like H, C, N or

45

O, nor to perform chemical speciation. Elemental analyses provided accurate results for 14 elements (Al,
As, Br, Ca, Cl, Cu, Fe, K, Mn, Pb, Si, Ti, V, Zn). Measurements of sulfur were highly correlated with
declared sulfur contents of standard materials (r
Pearson
= 0.98); however, concentrations were
systematically overestimated by ED-XRF. A correction factor of 0.42 was derived from parallel analyses
of 12 filters and applied to ED-XRF sulfur measurements. Calibration for Mg, Na, and P showed low
correlations (r
Pearson
< 0.8) between different standards, impeding the interpretation of the measured
values. For 7 elements (Bi, Cd, Co, Cr, Ga, Ni, Se,) concentrations were too low on most filters to be
detected reliably. Iodine could not be analyzed due to methodological problems. More details on quality
assurance of the method are published elsewhere (Mathys et al. 2001; 2002).
Since carbon cannot be detected with ED-XRF, reflectance of the filters was measured and the
absorption coefficient (Abs) calculated. A standard method (Reflectometer EEL model 43; Diffusion
Systems Ltd., London, U.K.) was used to measure reflectance, which has been applied and described
earlier (Götschi et al. 2002). Repeated reflectance measurements of 78 filters showed an average relative
difference of 1.2%. Elemental carbon is the dominant light absorbing substance in airborne particulate
matter; therefore, reflectance can be used as a surrogate measure for elemental carbon (EC). In urban
settings reflectance can be considered a diesel-specific traffic indicator, since several studies estimated
that in urban settings the major fraction of EC originates from diesel combustion (66% - 96%) (Schauer
2003).
The specific elements measured on PM
2.5
will be briefly described. Sulfur (S) is assumed to represent a
background portion of PM
2.5
, mainly consisting of sulfate particles (SO
4
2-
), which are oxidation products
formed from sulfur dioxide (SO
2
) emissions during long-range transportation in the atmosphere. Lead
(Pb) and bromine (Br) may reflect aspects of traffic emissions. However, since Pb was banned from
gasoline, the origin of airborne Pb is less clear, possibly stemming from re-suspension of road dust,
brake abrasion, industrial emissions, and waste incineration (Lee et al. 1994; de Miguel et al. 1997;
Chiaradia et al. 2000; Lammel et al. 2002). Bromine is thought to be mainly emitted by vehicles, though,
other sources, i.e. fossil fuel combustion, incineration, sea spray or crustal material, exist. (Lee et al.

46

1994; Lammel et al. 2002). Other metals, particularly iron (Fe), copper (Cu), and zinc (Zn) are of
toxicological interest, since these transition metals may play a crucial role in the oxidative stress
pathway, hypothesized to be part of the causal explanation of many observed air pollution related health
effects (Gilliland et al. 1999). Aluminum (Al), calcium (Ca), and silicon (Si) are the main components of
geogenic matter, or crustal material (Andrews et al. 1996; Press et al. 1997). Chlorine (Cl) is a
significant contributor to PM
2.5
mass. Sources of chlorine are sea salt particles, salt particles from street
de-icing, industrial emissions of hydrochloric acid (HCl) and emissions from waste incineration (U.S.
Environmental Protection Agency 1990). Potassium (K) is associated with biogenic aerosols from wood
combustion, pollen and spores (Matthias-Maser et al. 1994). Vanadium (V) is a trace element emitted
during the combustion of fossil fuels, such as coal and vanadium-rich fuel oil (WHO 2000). Titanium
(Ti) is used as a pigment in paints (TiO
2
) and in metal alloys. It is abundant in the earth’s crust. The main
sources of Ti contamination in the general environment are the combustion of fossil fuels and the
incineration of titanium-containing wastes (WHO 1984). Major sources of arsenic (As) are nonferrous
metal smelters and power plants burning arsenic-rich coal (WHO 2000). Manganese (Mn) emissions can
be increased near foundries and where ferro- and silico-manganese industries are present (WHO 2000).
All data in this manuscript are presented as measures of air concentrations (µg/m
3
for PM
2.5
, ng/m
3
for
all elements, and absorption coefficient *10
5
/m for reflectance).
For descriptive purposes, annual and seasonal means are presented. For calculation of means, single
filter data were weighed by exposure time with each month assigned equal weight (Hazenkamp-von Arx
et al. 2004). Winter is defined as November to February, and summer as May to August. Coefficients of
variance (Standard deviation/mean) presented for the annual means reflect the variability of the monthly
means.
To compare the composition of PM
2.5
across centers, percent of accounted mass of PM
2.5
(for elements
only) and within city Pearson correlations were calculated. Spearman correlations yielded very similar
results and are therefore not presented. To assess whether indicators will provide independent

47

information for health analyses, Pearson correlation coefficients between annual means were calculated
across centers.
Overall, data completeness was high yielding more than 80% of the scheduled sampling time in all
cities, except for Verona where technical problems occurred (data completeness 37% of scheduled
sampling time). Therefore, data for Verona need to be interpreted with caution.
Results
The range of annual mean concentrations across the 21 study centers is large for all indicators, showing
ratios between the 90
th
and the 10
th
percentile from 2.4 for sulfur to 11.1 for zinc. As can be seen from
Table 5, the majority of indicators show the highest annual means in Turin. Exceptions are indicators of
crustal material and some trace metals (Cu, Ti, V, Zn) which show the highest concentrations in some of
the Spanish centers and in Grenoble. Pavia, located in the same air shed as Turin, the plain of the river
Po (see Figure 4), showed similarly high levels. Barcelona, and to a lesser extent, Antwerp City and
Paris, are in the range of the Italian centers for presumably traffic related indicators (Abs, Pb, Br), but
show significantly lower levels of PM
2.5
and sulfur. Reykjavik in Iceland and the Swedish centers Umea
and Uppsala showed the lowest concentrations for most indicators, particularly for those associated with
anthropogenic activities. In Reykjavik, sulfur concentrations were more than ten times lower than in
Turin, and the difference for reflectance was more than forty-fold (0.1 vs. 4.3 abs. coeff./m).

48

Table 5: Annual means (coefficient of variance) for each air pollution indicator and center.
PM
2.5
in µg/m
3
, absorbance in absorption coefficient/m, and elements in ng/m
3
. Mean values based on less than 50% of filter values above the limit of detection are
printed in italic and should be interpreted with caution.
PM2.5 Abs Al As Br Ca Cl Fe K Mn Pb S Si Ti V Zn
Albacete 13.1 1.4 344 1.6 3.9 277 280 49 350 2.0 11.3 1,009 730 5.6 2.7 12.2
(0.25) (0.23) (0.64) (0.53) (0.45) (0.50) (1.22) (0.54) (0.50) (0.28) (0.54) (0.36) (0.60) (0.58) (0.40) (0.64)
Antwerp City 24.1 2.9 177 7.4 5.1 87 1,113 127 182 6.9 28.6 1,465 363 5.3 6.8 52.4
(0.60) (0.35) (0.49) (1.23) (0.87) (0.34) (1.29) (0.46) (0.79) (0.62) (0.72) (0.34) (0.48) (0.47) (0.59) (0.85)
Antwerp South 20.8 1.7 128 6.3 4.7 46 892 66 182 5.1 25.8 1,453 263 3.6 5.7 45.4
(0.44) (0.44) (0.46) (0.90) (0.71) (0.41) (1.11) (0.46) (0.66) (0.54) (0.71) (0.30) (0.58) (0.61) (0.59) (0.86)
Barcelona 22.2 3.1 389 12.5 12.1 226 831 145 430 10.4 52.7 1,388 686 19.0 9.0 80.5
(0.34) (0.36) (0.51) (0.51) (0.76) (0.19) (0.97) (0.32) (1.41) (0.68) (0.56) (0.34) (0.25) (0.67) (0.29) (0.49)
Basel 17.4 1.7 151 3.6 5.3 60 472 78 255 3.3 13.5 1,039 299 3.1 1.6 32.9
(0.52) (0.27) (0.49) (0.68) (0.51) (0.26) (1.06) (0.24) (0.60) (0.36) (0.34) (0.41) (0.34) (0.34) (0.28) (0.43)
Erfurt 16.3 1.7 148 4.5 2.2 52 329 72 157 3.1 14.8 1,144 313 2.8 0.8 38.5
(0.58) (0.43) (0.41) (0.89) (0.53) (0.30) (1.77) (0.41) (0.70) (0.42) (0.96) (0.47) (0.39) (0.43) (0.33) (1.02)
Galdakao 16.3 1.9 197 8.6 3.8 199 411 166 191 23.0 39.0 1,585 453 4.0 9.6 149.7
(0.36) (0.23) (0.45) (0.66) (0.37) (0.34) (0.75) (0.33) (0.46) (0.42) (0.55) (0.63) (0.42) (0.38) (0.84) (0.61)
Grenoble 19.0 2.6 257 5.8 3.4 139 667 125 326 10.7 23.2 888 1,404 5.5 3.3 185.0
(0.47) (0.36) (0.71) (0.72) (0.59) (1.12) (1.48) (0.63) (0.62) (0.71) (0.53) (0.28) (0.65) (0.62) (0.35) (0.81)
Gothenburg 12.6 1.0 97 2.1 2.2 38 540 52 113 2.7 5.2 903 217 2.4 3.9 15.9
(0.27) (0.29) (0.43) (0.37) (0.34) (0.49) (0.99) (0.43) (0.45) (0.42) (0.71) (0.45) (0.57) (0.50) (0.48) (0.43)
Huelva 17.3 1.4 444 12.2 4.9 168 806 76 297 2.9 26.9 1,558 1,259 17.1 6.7 40.9
(0.27) (0.26) (0.48) (0.77) (0.30) (0.33) (0.84) (0.37) (0.47) (0.34) (0.53) (0.57) (0.43) (1.06) (0.49) (0.63)

49

Table 5 continued: Annual means (coefficient of variance) for each air pollution indicator and center.
PM
2.5
in µg/m
3
, absorbance in absorption coefficient/m, and elements in ng/m
3
. Mean values based on less than 50% of filter values above the limit of detection are
printed in italic and should be interpreted with caution. Values from Verona are based on 37% of the scheduled sampling time only and should also be interpreted
cautiously.
PM2.5 Abs Al As Br Ca Cl Fe K Mn Pb S Si Ti V Zn
Ipswich 16.5 1.3 115 6.4 4.7 37 1,147 41 201 3.3 18.8 999 165 4.5 5.6 22.4
(0.41) (0.39) (0.60) (0.93) (0.72) (0.33) (0.92) (0.39) (0.90) (0.56) (1.06) (0.47) (0.31) (1.00) (0.85) (0.60)
Norwich 16.2 1.6 108 4.3 3.9 92 1,027 42 116 2.6 13.6 977 204 2.6 4.5 15.0
(0.32) (0.25) (0.31) (0.57) (0.56) (1.09) (0.76) (0.32) (0.36) (0.57) (0.55) (0.40) (0.39) (0.34) (0.77) (0.42)
Oviedo 15.9 2.1 467 6.2 7.1 281 562 138 232 6.4 22.9 1,181 781 7.4 5.5 31.1
(0.23) (0.36) (0.25) (0.51) (0.48) (0.37) (0.79) (0.29) (0.28) (0.42) (0.37) (0.56) (0.33) (0.27) (0.32) (0.20)
Pavia 35.3 2.9 228 9.2 11.1 85 963 124 364 9.9 37.4 1,783 539 8.2 4.2 47.0
(0.57) (0.34) (0.34) (0.64) (0.68) (0.39) (1.23) (0.36) (0.68) (0.96) (0.58) (0.23) (0.35) (0.55) (0.31) (0.78)
Paris 17.8 2.4 141 3.7 4.1 79 668 98 180 4.4 15.7 1,081 321 3.9 2.2 40.1
(0.39) (0.21) (0.50) (0.45) (0.70) (0.28) (0.94) (0.34) (0.49) (0.56) (0.49) (0.39) (0.52) (0.42) (0.42) (0.47)
Reykjavik 3.7 0.1 111 0.9 1.2 41 936 23 29 0.5 2.6 155 245 3.0 0.4 2.2
(0.45) (0.61) (0.63) (0.17) (0.61) (0.47) (0.88) (0.62) (0.44) (0.78) (1.12) (0.56) (0.57) (0.73) (0.31) (0.55)
Tartu 14.8 1.6 156 2.5 2.2 85 289 32 386 2.8 8.6 892 367 2.7 1.3 32.6
(0.34) (0.31) (0.88) (0.57) (0.63) (0.71) (0.98) (0.49) (0.55) (0.52) (0.65) (0.28) (0.95) (0.93) (0.44) (0.48)
Turin 44.9 4.3 380 14.4 21.3 116 1,322 262 471 13.3 63.8 1,827 744 8.5 3.6 70.1
(0.50) (0.24) (0.27) (0.48) (0.70) (0.26) (1.08) (0.40) (0.69) (0.59) (0.53) (0.24) (0.28) (0.31) (0.40) (0.60)
Umea 5.6 0.6 70 1.1 1.2 22 160 25 63 1.2 3.0 415 172 1.6 0.9 6.2
(0.21) (0.40) (0.66) (0.50) (0.45) (0.09) (1.26) (0.45) (0.33) (0.45) (0.86) (0.38) (0.75) (0.77) (0.56) (0.51)
Uppsala 10.4 1.0 103 1.8 1.7 34 225 54 116 1.9 5.0 752 247 2.0 1.4 14.9
(0.43) (0.33) (0.44) (0.58) (0.53) (0.46) (1.03) (0.56) (0.79) (0.59) (0.86) (0.50) (0.49) (0.43) (0.55) (0.49)
Verona¹ 41.5 4.2 336 19.4 22.9 257 1,099 302 411 30.7 80.1 2,015 759 8.7 1.9 135.5
(0.42) (0.21) (0.29) (0.42) (0.39) (0.36) (0.87) (0.29) (0.56) (0.67) (0.27) (0.45) (0.52) (0.32) (0.44) (0.53)

50

In general, similar pollution levels have been observed for centers located in the same regions, such as
the three Swedish (Gothenburg, Umea, Uppsala), the two British (Ipswich, Norwich), or the three Italian
centers (Pavia, Turin, Verona), respectively. The two centers in Antwerp, one in the city center
(Antwerp City) and the other in the southern suburbs (Antwerp South), 11.5 km apart, do suggest
pronounced differences for reflectance within the same city (2.9 vs. 1.7 abs. coeff. 10
5
/m). In an earlier
ECRHS publication, a comparable difference in annual mean NO
2
was reported for the same two
locations (58 vs. 26 µg/m
3
) (Hazenkamp-von Arx et al. 2004).
Temporal variability of monthly means within cities (coefficients of variance in Table 5) indicate a fairly
consistent pattern across the centers, showing somewhat lower month-to-month variation for reflectance,
Figure 5: Seasonal air pollution patterns
Silicon (squares), reflectance (triangles), and sulphur (dots) chosen to represent particles from
crustal material, traffic, and long range background pollution, respectively. Centers are grouped
into geographical regions. Levels of all three measures have been rescaled so that means equal the
mean of PM
2.5
. Y-scale reflects PM
2.5
levels in µg/m
3
.

51

and a tendency towards elevated variation for rarer elements, as compared to PM
2.5
, S, Fe, or Ca.
Coefficients of variance are exceptionally high for chlorine. Figure 5 visualizes regional and local
differences in the month-to-month variation for silicon, reflectance, and sulfur, chosen to represent
crustal, traffic, and secondary particles, respectively.
For most elements, there is an expected pattern of higher concentrations during winter months; however,
some deviations occur (Table 6). For example, crustal indicators, such as Al, Ca, Fe, or Si, are often
higher in summer, especially in Spain, although the pattern is not entirely consistent. In Galdakao, for all
indicators except bromine and chlorine, levels are higher in summer than in winter. Chlorine levels are
much lower in summer than in winter in all centers. Similarly to PM
2.5
, sulfur levels are increased during
winter in several centers but in many centers sulfur concentrations also increase steadily over the
summer months to reach peak values around August, leading to higher summer means, when compared
to winter means. This phenomenon seems to be most pronounced in the five Spanish centers.

52

Table 6: Seasonal means for each air pollution indicator and center.
Winter means (upper row) and summer means (lower row). PM
2.5
in µg/m
3
, absorbance in absorption coefficient/m, and elements in ng/m
3
. Mean values based on
less than 50% of filter values above the limit of detection are printed in italic and should be interpreted with caution.
PM2.5 Abs Al As Br Ca Cl Fe K Mn Pb S Si Ti V Zn
Albacete 15.4 1.7 206 2.5 3.8 205 678 33 546 1.6 12.0 714 420 3.5 2.4 17.7
11.5 1.2 499 1.0 4.3 405 38 61 236 2.2 12.1 1,195 1,064 7.5 2.9 7.6
Antwerp City¹ 37.0 3.9 216 15.0 9.6 113 2,725 187 319 11.5 45.0 1,621 393 7.0 9.0 96.2
17.6 2.4 142 3.0 3.5 77 287 98 108 4.8 24.2 1,442 310 4.4 5.9 30.5
Antwerp South 24.4 2.1 133 10.0 7.1 51 1,774 82 260 7.2 37.5 1,205 213 3.9 7.3 72.2
17.3 1.4 101 5.3 2.0 40 288 47 124 3.5 15.5 1,581 284 3.1 4.8 27.4
Barcelona 30.2 4.2 379 18.3 22.5 249 1,704 190 401 17.4 75.0 1,298 728 23.5 9.4 105.9
20.0 2.5 512 9.2 4.5 209 237 124 715 6.5 42.3 1,890 767 21.0 10.5 60.6
Basel 23.7 2.1 197 6.7 7.6 66 1,044 85 397 4.2 14.9 1,074 262 3.4 1.8 43.2
13.7 1.6 147 2.2 3.6 65 52 83 174 2.9 13.7 1,251 354 3.2 1.5 29.7
Erfurt 19.9 2.1 150 6.5 2.7 63 782 80 219 3.7 22.3 1,019 325 3.2 0.8 61.0
11.0 1.2 164 1.7 1.4 47 21 59 98 2.4 4.6 1,221 346 2.8 1.0 12.0
Galdakao¹ 10.8 1.6 119 7.2 4.1 141 507 127 136 16.1 25.2 673 266 2.6 3.0 72.8
20.9 1.9 238 7.6 3.3 231 179 169 253 22.6 38.7 2,571 526 4.5 11.8 178.9
Grenoble 28.0 3.4 322 9.4 5.3 121 1,714 176 512 16.3 33.5 882 914 6.6 3.5 319.2
12.9 1.8 205 3.6 2.4 209 75 75 229 5.0 19.3 1,005 1,365 4.8 2.8 109.3
Gothenburg 12.5 1.2 84 2.2 2.6 40 891 56 113 3.0 5.0 869 183 2.0 3.4 20.2
11.2 0.9 113 1.6 2.1 45 410 49 65 2.5 5.0 826 255 3.3 5.3 9.3
Huelva 17.2 1.7 469 16.5 5.5 186 1,419 85 393 3.7 29.1 882 1,250 13.0 5.0 54.8
16.9 1.1 491 10.5 4.8 170 206 80 202 2.6 29.4 2,191 1,428 24.9 9.2 42.9

53

Table 6 continued: Seasonal means for each air pollution indicator and center.
Winter means (upper row) and summer means (lower row). PM
2.5
in µg/m
3
, absorbance in absorption coefficient/m, and elements in ng/m
3
. Mean values based on
less than 50% of filter values above the limit of detection are printed in italic and should be interpreted with caution. Values from Verona are based on 37% of the
scheduled sampling time only and should also be interpreted cautiously.
PM2.5 Abs Al As Br Ca Cl Fe K Mn Pb S Si Ti V Zn
Ipswich 21.3 1.8 162 11.0 8.4 41 2,418 50 335 3.6 32.2 896 153 7.3 3.7 34.7
15.0 1.0 97 4.2 2.2 36 218 37 119 3.4 9.3 1,350 189 2.8 10.0 14.4
Norwich¹ 17.7 2.0 134 6.3 6.4 209 1,811 51 160 4.2 19.5 736 232 3.0 2.8 17.9
14.6 1.3 97 2.2 2.6 49 247 36 85 1.7 12.0 1,255 207 2.3 7.0 12.7
Oviedo 17.5 3.0 483 9.0 10.5 387 996 117 206 6.3 27.9 766 746 7.5 5.8 35.6
16.7 1.5 485 3.2 4.7 213 142 166 272 6.4 22.0 1,901 843 7.8 6.2 29.2
Pavia 55.3 4.1 226 16.1 19.6 99 2,175 167 655 18.8 61.5 1,907 508 8.7 4.9 83.6
19.9 2.1 280 4.7 3.6 87 77 103 181 5.3 18.6 1,928 655 11.0 3.7 29.2
Paris 21.0 2.5 112 4.1 6.3 76 1,309 97 221 5.0 18.4 885 219 3.1 2.3 42.6
15.9 2.3 169 3.8 2.0 95 158 111 166 4.8 11.8 1,363 398 4.9 2.2 39.8
Reykjavik 4.8 0.1 155 0.8 1.5 56 1,515 30 30 0.8 1.1 91 298 4.2 0.4 1.6
3.3 0.2 79 0.9 1.2 34 618 21 32 0.3 4.7 216 214 3.1 0.4 2.1
Tartu 15.6 1.8 68 3.2 2.3 30 538 19 433 2.6 9.7 865 120 0.9 1.7 37.0
10.2 1.1 131 1.4 1.0 120 30 31 160 1.8 3.1 709 341 2.4 0.7 16.3
Turin 69.2 5.4 449 21.7 38.0 137 3,015 379 876 21.9 100.9 2,095 850 11.3 5.2 121.9
23.0 3.3 303 7.0 7.4 88 115 167 205 5.9 34.6 1,875 643 6.5 2.5 35.3
Umea 5.8 0.8 50 1.4 1.1 22 332 21 71 1.4 1.8 397 117 1.0 0.7 7.5
4.9 0.4 66 0.9 0.9 21 56 20 41 0.7 2.1 334 183 1.4 0.8 3.0
Uppsala 11.5 1.2 78 2.5 2.1 26 346 50 139 2.4 5.6 826 176 1.5 1.3 18.5
7.2 0.7 95 0.9 0.8 30 189 35 47 1.0 1.7 529 234 1.8 1.1 6.6
Verona² 51.0 4.7 381 24.7 27.8 319 1,859 358 554 46.8 94.4 2,167 1,006 10.2 1.3 191.2
16.0 2.6 154 5.1 6.3 92 47 142 148 8.7 43.7 1,402 240 3.3 2.1 92.9

54

Among the presented elements, sulfur, by far, accounts for the highest proportion of PM
2.5
mass, ranging
from 4.6% in Reykjavik to 8.8% in Galdakao (average of single filters). Al, Cl and Si are the only
remaining elements accounting on average for more than 1% of the PM
2.5
mass (range 0.3%-3.0%,
1.5%-22.9%, 1.3%-8.9%, respectively). As, Br, Mn, Ti, and V account for less than 0.1% of the PM
2.5

mass in the majority of the centers.
Table 7 provides Pearson coefficients for correlations of daily values of PM
2.5
and the other indicators
within each city. Barcelona, Turin, Pavia, Grenoble, followed by Erfurt and Paris, are the centers with
the highest correlations between PM
2.5
and reflectance, Pb, Br, and Fe, respectively. PM
2.5
and sulfur are
highly correlated in two thirds of the centers; however Barcelona, Turin and Pavia are now among the
centers with lower correlations (r ≈ 0.55). PM
2.5
and K show correlations above 0.7 in eleven centers.
Pair-wise within city correlations between all other indicators were investigated as well. Correlations
were highest between Al and Si, and Al and Ti which showed Pearson coefficients greater than 0.9 in 11
and 9 centers, respectively. Correlations between Pb and Br were above 0.7 in 16 of the centers. In
Barcelona, Turin, and Pavia, Pb and Br were also correlated with reflectance (r>0.7). Also, Pearson
correlation coefficients were often higher than 0.7 for reflectance and Fe (13 centers), Fe and Mn (13), Si
and Ti (13), and Zn and Pb (10).

55

Table 7: Within city Pearson correlations between PM
2.5
and all other air pollution indicators.
Correlations above 0.7 are bolded.
Abs Al As Br Ca Cl Cu K Mn Pb S Si Ti V Zn
Antwerp City 0.63 0.57 0.75 0.71 0.33 0.67 0.43 0.84 0.74 0.67 0.85 0.32 0.70 0.38 0.75
Antwerp South 0.86 0.49 0.51 0.42 0.41 0.53 0.30 0.64 0.42 0.41 0.74 0.19 0.54 0.60 0.46
Albacete 0.55 0.25 0.30 0.30 0.21 0.36 0.12 0.57 0.27 0.11 0.56 0.21 0.26 0.57 0.19
Barcelona 0.79 0.48 0.69 0.81 0.50 0.79 0.59 0.32 0.59 0.74 0.55 0.49 0.57 0.58 0.67
Basel 0.68 0.24 0.56 0.51 0.19 0.74 0.16 0.87 0.64 0.32 0.83 0.16 0.46 0.71 0.66
Erfurt 0.81 0.49 0.87 0.54 0.50 0.81 0.52 0.92 0.77 0.79 0.80 0.39 0.62 0.21 0.84
Galdakao 0.61 0.56 0.45 0.27 0.55 -0.05 0.40 0.76 0.53 0.42 0.89 0.55 0.57 0.64 0.45
Gothenburg 0.39 0.44 0.42 0.49 0.20 0.10 0.03 0.65 0.38 0.50 0.81 0.24 0.27 0.35 0.48
Grenoble 0.81 0.23 0.78 0.74 0.09 0.86 0.47 0.90 0.70 0.75 0.55 0.06 0.26 0.48 0.78
Huelva 0.64 0.20 0.32 0.65 0.39 -0.02 0.24 0.64 0.47 0.41 0.79 0.40 0.11 0.68 0.31
Ipswich 0.80 0.57 0.58 0.78 0.11 0.67 0.51 0.56 0.39 0.55 0.73 0.50 0.51 0.40 0.78
Norwich 0.73 0.54 0.33 0.65 0.08 0.30 0.36 0.64 0.46 0.26 0.80 0.27 0.48 0.51 0.55
Oviedo 0.48 0.61 0.42 0.53 0.35 0.10 0.35 0.73 0.27 0.43 0.72 0.57 0.64 0.61 0.62
Paris 0.62 0.34 0.69 0.65 0.49 0.45 0.38 0.84 0.48 0.54 0.75 0.20 0.49 0.50 0.54
Pavia 0.78 0.13 0.81 0.81 0.32 0.90 0.40 0.83 0.71 0.83 0.56 0.07 0.11 0.36 0.84
Reykjavik -0.05 0.31 0.03 0.29 0.55 0.70 0.00 0.52 0.26 -0.11 0.42 0.17 0.18 0.01 -0.06
Tartu 0.83 0.19 0.69 0.38 0.02 0.64 0.17 0.85 0.58 0.47 0.76 0.16 0.16 0.50 0.81
Turin 0.73 0.45 0.70 0.82 0.42 0.91 0.52 0.86 0.75 0.76 0.56 0.34 0.51 0.84 0.79
Umea 0.50 0.10 0.43 0.27 0.02 -0.06 0.10 0.74 0.45 0.03 0.87 -0.02 -0.07 0.48 0.64
Uppsala 0.64 0.30 0.66 0.43 0.17 0.08 0.13 0.69 0.62 0.25 0.85 0.15 0.27 0.65 0.57
Verona¹ 0.80 0.43 0.82 0.83 0.61 0.77 0.78 0.90 0.57 0.84 0.65 0.38 0.61 0.39 0.57

1
based on 28 filters only

56

Pair-wise correlations of annual means across the 21 centers are mostly above 0.8 among markers of
predominantly anthropogenic pollution, i.e. S, Abs, Br, Pb, As, Fe and PM
2.5
, as well as among crustal
elements, i.e. Al, Ca, Si and Ti (Table 8). In contrast, correlations are low between anthropogenic and
crustal indicators. In addition, correlations appear somewhat weaker between Cl, K, Mn, V, Zn and
many other indicators. Similar patterns were observed for seasonal means with correlations slightly
lower in summer (data not shown). Additional results from the elemental analysis are available in a
report (Mathys et al. 2002).
Table 8: Pearson correlations of annual means across centers.
Correlations above 0.7 are bolded.
PM2.5 Abs Al As Br Ca Cl Cu Fe K Mn Pb S Si Ti V
Abs 0.93 1
Al 0.47 0.54 1
As 0.84 0.82 0.64 1
Br 0.91 0.87 0.58 0.88 1
Ca 0.30 0.45 0.84 0.51 0.45 1
Cl 0.64 0.54 0.23 0.62 0.60 0.03 1
Cu 0.63 0.69 0.74 0.87 0.66 0.56 0.40 1
Fe 0.85 0.90 0.58 0.86 0.89 0.54 0.47 0.76 1
K 0.73 0.76 0.71 0.67 0.72 0.58 0.28 0.64 0.63 1
Mn 0.68 0.72 0.38 0.78 0.72 0.54 0.32 0.66 0.88 0.49 1
Pb 0.88 0.89 0.60 0.96 0.93 0.54 0.59 0.80 0.94 0.71 0.86 1
S0.870.81 0.56 0.86 0.75 0.46 0.48 0.72 0.78 0.66 0.70 0.85 1
Si 0.34 0.44 0.80 0.51 0.34 0.64 0.14 0.74 0.45 0.62 0.34 0.43 0.38 1
Ti 0.41 0.46 0.80 0.70 0.51 0.57 0.36 0.79 0.42 0.62 0.29 0.59 0.53 0.64 1
V 0.18 0.25 0.38 0.43 0.14 0.39 0.33 0.50 0.25 0.20 0.31 0.39 0.50 0.24 0.55 1
Zn 0.46 0.60 0.33 0.58 0.41 0.43 0.18 0.67 0.68 0.46 0.80 0.63 0.49 0.60 0.27 0.35

Discussion
These first trans-European PM
2.5
speciation data show that PM
2.5
composition and levels of constituents
vary significantly across Europe, with levels of some toxic components, i.e. Zn, being up to 80 times
higher in Northern Italy as compared to Iceland. The overall pattern confirms a north-south pollution
gradient across Europe observed by others (Hamilton et al. 1991; Pacyna et al. 1991; Hoek et al. 1997;
Roemer et al. 2000). In general, pollution levels were higher in larger cities and in centers where climate

57

and topography enhanced accumulation of air pollution, such as in Turin, Barcelona and Antwerp. In
contrast, areas of low population density and exposed to small amounts of pollution from long range
transport, such as Reykjavik and the northern Swedish centers Umea and Uppsala, showed the lowest
pollution levels for all constituents.
Annual means for crustal indicators deviate considerably from the PM
2.5
pattern across centers, most
notably in the Spanish cities. Assuming the likely, although simplified, chemical structures of crustal
compounds in PM
2.5
(i.e. SiO
2
; Al
2
O
3
; 50% CaCO
3
, 50% CaSO
4
), this fraction could make up for 13 to
25% of the PM
2.5
mass in the Spanish centers, and 24% in the alpine center Grenoble, while in the other
centers it typically accounts for less than 10%.
Elevated winter levels of indicators associated with anthropogenic activities were observed in several
centers, most dramatically in Northern Italy and Antwerp, where inversion layers are a common
phenomenon. Persistent thermal inversions, combined with low wind velocities cause an accumulation
of exhaust emissions, while at the same time the amount of wind blown particles from crustal origin is
likely to be reduced. Nevertheless, crustal elements may be increased due to the accumulation of re-
suspended road dust.
On the other hand, the warm and dry summer climate on the Iberian Peninsula, which favors the
formation and lift up of dust particles, is likely to account for the increased concentrations of crustal
particles in the Spanish centers, as compared to centers with cooler, more humid climates. Saharan dust
considerably affects this region as well (Rodriguez et al. 2002; Viana et al. 2003; Alastuey et al. 2004;
Rodriguez et al. 2004). Increased sulfur levels over the summer months, observed most notably in Spain,
are consistent with atmospheric conversion of sulfur dioxide (SO
2
) into sulfate particles (SO
4
2-
) taking
place more rapidly in warm air masses, as described in other studies (Querol et al. 2001; Tanner et al.
2004). As mentioned, calibration procedures and method comparisons revealed that ED-XRF sulfur
concentrations had to be corrected, thus the absolute levels need to be interpreted with care. Relative

58

comparisons between measurements, however, are not affected.
The extreme seasonal differences in chlorine levels are a unique finding. These results may not only
represent differences in actual concentrations, or emission levels but may have methodological reasons.
The warmer summer climate may lead to enhanced evaporation of hydrochloric acid (HCl) from the
filters surface, which can be formed there by various mechanisms, such as reactions of acids or ozone
with salt particles (NaCl) (Yao et al. 2001). Large losses of Cl were also observed on an exposed control
filter that was analyzed 191 times for its elemental composition at a temperature of 40°C and under a
partial vacuum. Cl concentration decreased from approx. 1600 ng/cm
2
to approx. 800 ng/cm
2
over the
first 10 measurements. This, and a similar observation for bromine, indicate that evaporation of
halogenated semi-volatile organic compounds may have taken place (Mathys et al. 2002). Although the
temperature of 40°C during the elemental analysis clearly exceeds outdoor temperatures even during
summer, sampling warmer air may similarly lead to increased Cl losses due to evaporation from the
filters during summer months, as compared to winter. Although evaporation of Cl, in some way, is
probably the most important factor, it remains unclear to what extent increased temperature, ozone
levels, differences in emissions (e.g. street de-icing), or other unknown factors are responsible for the
large differences of Cl observed between summer and winter levels.
Increased city-to-city variation for reflectance and other indicators, as compared to regionally fairly
homogeneous levels for sulfur, probably reflects the more relevant contribution of locally produced
emissions, including emissions close to the sampling locations. Five of the ECRHS monitors are located
within 15 meters of a main street (Antwerp City, Basel, Pavia, Turin, Verona). Several studies could
show strong concentration gradients for traffic related pollutants with increasing distance from roadways
(Janssen et al. 1997; Roorda-Knape et al. 1998; Zhu et al. 2002a; Zhu et al. 2002b). As a consequence,
traffic related PM constituents may be over-sampled by these monitors. The reported levels, therefore,
reflect exposure conditions for people living along comparable roads but not necessarily exposure of the
population at large.

59

Within cities, PM
2.5
and sulfur are mostly highly correlated, since sulfates are a major component of
PM
2.5
. However, these correlations appear lower where primary pollution sources become more
important, such as in large cities with high levels of traffic related pollutants. The correlations of silicon
and aluminum reflect a relatively constant proportion between SiO
2
and Al
2
O
3
found in crustal material
in many areas of Europe (Putaud et al. 2004). The highly correlated lead and bromine used to have a
major common source in gasoline. It is unclear though, to what extend this still is reflected in our data,
or whether residues resulting from road dust or industrial activities are responsible for the observed
correlations. In addition, the correlation between reflectance and Fe remains unclear. While reflectance
is an indicator for traffic related soot emissions, iron could reflect crustal components of road dust re-
suspended by vehicles, or particles generated during combustion in engines. Harrison et al. (2003)
reported moderate correlations between Fe and NO
x
, which they used as an indicator of traffic pollution
at a roadside location.
The diversity of temporal within-city correlation patterns is a result of the variety of center specific
characteristics of pollution sources, topography, and meteorology. The observed differences in the PM
composition within and across centers are likely to lead to variability in PM toxicity not reflected in the
PM mass concentrations alone. Comparisons across centers based on PM mass will therefore be limited
by some inherent misclassification due to the varying composition of PM across Europe.
Despite the observed within-city diversity of daily and seasonal patterns for the various indicators, which
may reflect pollution from different origin, our data show that collinearity among annual means of
pollutants compared across cities is large in most cases. Given these high correlations across cities, it is
impossible to interpret potential associations with health effects independently for the various correlated
constituents in cross-community comparisons.

60

Two main factors are likely to be responsible for the correlations of the long-term estimates across cities
observed for most measured constituents of PM
2.5
. First, all study centers are urban areas and thus may
have similar anthropogenic emission sources. In contrast to the absolute amount of emissions, the
relative contribution of the major anthropogenic sources, i.e. traffic, industry, and domestic emissions, to
a city’s total emissions may not be strongly affected by the size and density of the city. Second, apart
from the total emissions, local meteorology and topography has a major impact on ambient
Figure 6: Spatial variability of PM
2.5
, reflectance (Abs), and elements between Antwerp City and
Antwerp South, 11km apart.
Diamonds represent the coefficient of divergence (COD*) (Pinto et al. 2004), where a value of 0
implies that the two data sets are exactly alike. The whisker plots display the distribution of the
relative difference between pairs of measurements, where again a value of 0 implies that
measurements were equal. Data based on 53 parallel measurements over a period of 12 months.
0 1 2
Cu
Br
Pb
As
Ca
Fe
Ti
Abs
Si
Zn
V
Mn
Cl
OH
Al
K
S
PM2.5
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 1 2
Cu
Br
Pb
As
Ca
Fe
Ti
Abs
Si
Zn
V
Mn
Cl
OH
Al
K
S
PM2.5
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

* xij and xik represent the 24-hr average PM
2.5
concentration
for day i at site j and site k, and p is the number of
observations.
()( ) []
∑
=
+ − =
p
i
ik ij ik ij jk
x x x x
p
COD
1
2
1

61

Figure 7: Historic NO
2
data available for ECRHS centers.
Filled bars represent NO
2
means for 1990-92, empty bars are means from 1999. All data are from
stations labeled “background”.
0 20 40 60 80
AS AC AL BA BS ER GAGOGNHU IP NOOV PS PA RE TA TU UMUP VE

µg/m
3
concentrations. Since their effect on many pollutants is similar, this not only introduces short-term
collinearity between pollutants, but also determines their long-term average levels. Consequently, cities
with adverse weather patterns such as frequent inversion layers will have higher pollution levels for all
pollutants, whereas others with weather more favorable to pollution dispersion will have lower levels for
all pollutants.
The observation of high correlations among various constituents does not necessarily imply that
population mean exposures are equally well characterized for each constituent. Within-city spatial
variability differs considerably among various constituents, while PM mass is relatively homogenously
distributed in space, as has been shown by several reports (Röösli et al. 2001; Zhu et al. 2002a; Ito et al.

62

2004; Brunekreef et al. 2005). This pattern was confirmed in a comparison of ECRHS data from the two
monitoring stations located in Antwerp, only 11km apart from each other, as can be seen from Figure 6.
For indicators which are rather heterogeneously distributed in space, cross-community comparisons
based on single central monitor measurements may fail to reflect true average contrasts in population
average exposures. To assess effects of such heterogeneously distributed constituents, exposure data
collected at various locations within cities would be needed.
A further limitation of the ECRHS data with regard to their use as estimates of long-term exposures is
the uncertainty to which extent levels measured in 2000/01 may represent past exposures. Several
studies observed that annual means of particulate matter varied little over several years (Ackermann-
Liebrich et al. 1997; Schikowski et al. 2005). In the European APHEA study highest and lowest seasonal
Figure 8 Historic SO
2
data available for ECRHS centers.
Filled bars represent SO
2
means for 1990-92, empty bars are means from 1999. All data are from
stations labeled “background”.
0 20 40 60
AS AC AL BA BS ER GAGOGNHU IP NOOV PS PA RE TA TU UMUP VE

µg/m
3

63

means of SO
2
were highly correlated over study periods of 3-12 years (Winter Pearson r=0.96, summer
r=0.86), indicating that single annual means may represent the long term ranking of community
exposure levels fairly well (Katsouyanni et al. 1995).
Only limited historic air pollution data are available for Europe, and for ECRHS cities in particular
(Naef et al. 2000). The available data from ECRHS cities for past SO
2
and NO
2
concentrations, the most
commonly measured pollutants, are shown in Figure 7 and Figure 8. Interpretability of such data,
however, is very limited due to the lack of standardization of measurements across cities, and over time.
It may be fair to assume that for many pollutants the ranking across the 20 cities may not have changed
dramatically over the past one or two decades. For Erfurt on the other hand, located in the former
German Democratic Republic, this assumption may probably not hold due to the radical changes in
emission sources since the German reunification (Heinrich et al. 2000)
Conclusions
A number of conclusions can be drawn with regard to the use of ECRHS air pollution data in health
analyses. Annual means of PM
2.5
and several of its constituents appear to be driven by common factors,
such as meteorology and population density, resulting in estimates which are highly correlated across the
21 ECRHS study centers. Consequently, the ability to assess independent effects of PM
2.5
and its
constituents on lung function is limited, making multiple analyses of various pollution indicators
redundant. PM
2.5
and sulfur have been shown to be relatively homogenously distributed within cities in
several studies. They are therefore far better suited for the use in a health analysis based on comparisons
across centers than most other pollution indicators analyzed. PM
2.5
has been associated with adverse
health outcomes in numerous studies, while evidence for health effects by sulfur and pollution related to
it is less clear.
It can therefore be concluded that PM
2.5
is the most appropriate exposure parameter to use in the ECRHS
health analysis. While sulfur may serve as a suitable alternative indicator to be used in sensitivity

64

analysis, the uncertainties implied with the single monitor based estimates for the remaining indicators
prohibit their use in a cross-community comparison.
Nevertheless, the component analysis in ECRHS revealed that the composition of PM
2.5
can vary
considerably across Europe. This points out to the limitation of PM
2.5
as a general marker of a highly
complex pollution mixture. PM
2.5
is only as good an exposure marker as the particle mass concentration
does relate to the concentration of the biologically relevant, yet unknown component of air pollution.
This relationship may very well vary across the centers of ECRHS. Finally, there is also considerable
uncertainty about the temporal precision of PM
2.5
annual means as surrogates for long term exposures,
due to the lack of past PM
2.5
data.

65

Chapter 4: Air Pollution and Lung Function in the
European Community Respiratory Health Survey
This chapter is subject to a manuscript submitted to a peer reviewed journal.
Introduction
As pointed out in chapter 2, there is some evidence for long term effects of air pollution on lung function
in adults based on cross-sectional differences (e.g. Ackermann-Liebrich et al. 1997; Schikowski et al.
2005). It remains unclear, however, whether such findings reflect lung function deficits obtained during
childhood or adolescence, or whether they actually represent adverse effects of more recent exposures
during adulthood. If the latter were the case, one would expect to observe differences in the rates of lung
function decline to be associated with air pollution as well. However, longitudinal studies in adults are
rare (van der Lende et al. 1981; Jedrychowski et al. 1989; Tashkin et al. 1994) and only one, a study in
Japanese women, was published fairly recently (Sekine et al. 2004). They consistently suggest an
adverse effect of air pollution on lung function decline but the diversity in the methods used and the
populations studied is considerable across these studies, and so are their limitations. For example, older
publications dealt with much higher pollution levels than currently prevailing in most industrialized
countries (van der Lende et al. 1981), comparisons across only two or three communities are prone to
confounding (van der Lende et al. 1981; Jedrychowski et al. 1989), and loss to follow-up may have
influenced some findings (Tashkin et al. 1994; Sekine et al. 2004). Overall, there is insufficient
knowledge to draw conclusions on air pollution effects on lung function decline in adults.
The European Community Respiratory Health Survey (ECRHS) is the most comprehensive multi-center
respiratory cohort study in adults in Europe (Burney et al. 1994). The standardized air pollution protocol
of ECRHS was adopted by 21 centers participating in the baseline survey and a follow-up survey after
nine years (Hazenkamp-von Arx et al. 2003). The aim of this project was to test the hypotheses that

66

long-term urban background air pollution was associated with both lung function level (cross-sectional
analysis) and change in lung function (longitudinal analysis), respectively.
Methods
Details on the study protocol (Burney et al. 1994) and lung function data (Chinn et al. 2005) were
published elsewhere. In brief, participants aged 20-44 were randomly selected from twenty cities in ten
countries. Respiratory health assessment was conducted in 1991-93 and repeated nine years later (2000-
2002), using spirometry and administered questionnaires.
Of the 21 centers, 18 used the same spirometer at both occasions, mostly with updated software at
follow-up (see Table 9). Spirometers used at baseline were Biomedin (Padova, Italy) in twelve centers,
SensorMedics (Yorba Linda, USA) in seven centers, and Jaeger Pneumotach (Würzburg, Germany) in
two centers. The Biomedin is a water seal volume-displacement spirometer. The SensorMedics devices
used in Basel and the Scandinavian centers were heated wire flow-sensing spirometers, while the
SensorMedics spirometers used in two centers in Antwerp were dry rolling seal volume-displacement
devices. The Jaeger Pneumotach spirometer determines flows based on the pressure drop across a fixed
resistance. Equipment was changed in Erfurt (replaced Jaeger Pneumotach by same model) and Antwerp
City and South (switched from Sensormedics Dry Spirometer to Jaeger Masterscope).
The spirometry protocol was consistent with American Thoracic Society (ATS) guidelines (American
Thoracic Society 1995). A shortened version of the protocol is available in the annex. The full ECRHS
lung function protocol is available online (ECRHS 2007). Measurements not fulfilling the ATS criteria
for repeatability, extreme outliers, and subjects missing data for either FEV1 or FVC were excluded
from the analysis. FEV1, FVC, and FEV1/FVC ratio, a well established marker of airway obstruction
(Pauwels et al. 2001), were used as outcomes. Flow measures were not part of the ECRHS protocol.

67

Table 9: Spirometers used in ECRHS and changes in equipment and software at follow-up.
Baseline
Spirometer Same spirometer? Same Software?
Albacete Biomedin Yes Updated
Antwerp City Sensormedics dry spirometer No, Jaeger Masterscope No-Jaeger software in ECRHS II
Antwerp South Sensormedics dry spirometer No, Jaeger Masterscope No-Jaeger software in ECRHS II
Barcelona Biomedin Yes Updated
Basel Sensor Medics hot wire Yes Yes
Erfurt Jaeger Pneumotach No, different Jaeger Pneumotach Updated
Galdakao Biomedin Yes Updated
Gothenburg Sensor Medics hot wire Yes Yes
Grenoble Biomedin Yes Updated
Huelva Biomedin Yes Updated
Ipswich Biomedin Yes Updated
Norwich Biomedin Yes Updated
Oviedo Biomedin Yes Updated
Paris Biomedin Yes Updated
Pavia Biomedin Yes Updated
Reykjavik Sensor Medics hot wire Yes Updated
Tartu Jaeger Flowscreen Yes Yes
Turin Biomedin Yes Updated
Umea Sensor Medics hot wire Yes Yes
Uppsala Sensor Medics hot wire Yes Yes
Verona Biomedin Yes unknown
Follow-up

68

Detailed methods and results of the air pollution assessment are described in Chapter 3 and were
published previously (Hazenkamp-von Arx et al. 2003; 2004; Götschi et al. 2005; Cyrys et al. 2006). For
this analysis, fine particulates smaller than 2.5µm in diameter (PM
2.5
) measured at a central monitor in
each center were used, according to the conclusions from Chapter 3. Sulfur content of PM
2.5
derived
from elemental analysis was used as an alternative exposure measure in sensitivity analyses. Twelve
month means were used as surrogates for past long term air pollution levels, as done in other studies
(Ackermann-Liebrich et al. 1997; Kunzli et al. 2005a; Schikowski et al. 2005).
Questionnaire information included respiratory symptoms, asthma status, medication use, smoking
status, occupational history, household characteristics, residential history, socio-demographic
characteristics, sensitization to grass and cat allergens (Bedada et al. 2007), and bronchial hyper-
responsiveness (Chinn et al. 1997).
Statistical analysis
The data set consisted of 4290 subjects with a lung function measurement at baseline and follow-up. A
three level hierarchical model to relate the lung function measurements taken during each examination to
the average air pollution levels measured in each study center was used, similar to the approach
described by Gauderman et al. (2000; 2002; 2004)
Each subject’s two lung function measurements were regressed against age. Age was centered to the
mean, i.e. 38.4 years. The intercept from this regression represents the adjusted lung function level (of an
average or reference subject at the middle of follow-up), while the regression coefficient for age
represents the slope between the two lung function measurements, in other words the change in lung
function over time.
On the first level adjustments were made for visit specific, or time dependent covariates, such as a
subject’s body mass index (BMI). On the second level adjustments were made for subject specific
covariates which do not change over time, such as height (changes are negligible in the age range of the
analyzed sample), and random effects to account for residual variability between subjects were included.

69

On the third level the air pollution variable was introduced and, in some models, adjustments were made
for center level covariates. Again, random effects, this time to adjust for residual variation between
centers, were included. The three levels were integrated into one model using xtmixed in STATA
version 9.0 (Statacorp LP, College Station, TX, USA), allowing to assess separate effects of air pollution
on lung function level and change in lung function simultaneously. The model allows separate
adjustment of these two coefficients, in other words the covariates adjusting lung function level can
differ from those adjusting the change in lung function. All variances and covariances were estimated
distinctly, specifying the unstructured option for the covariance matrix of the random effects. Due to the
large differences in lung function levels between men and women all models were stratified by sex.
Covariate Adjustment
To investigate the effect that adjustment for various covariates had on the PM
2.5
coefficients a series of
models with increasing number of covariates was estimated. The crude model only included a constant
(intercept for lung function level) and age, which served as the intercept for change in lung function.
Then height was added (adjusting both lung function level and change), since it is the strongest predictor
of lung function. In what was considered a minimal set of covariates adjustments were made for height
and smoking status (never, ex, current) (level and change). In the selection of a parsimonious main
model in addition BMI (level and change) and a six dimensional variable for socioeconomic status based
on professional categories (level only; see Table 12), as used before by Sunyer et al. (2006)) was
included, based on a significant Wald test and the magnitude of their effects on lung function. Further, a
maximally adjusted model including a set of all variables with significant Wald tests, i.e. height squared,
reported respiratory symptoms (short and long term), asthma status, self reported truck frequency at
subject’s residence (ECRHS II), and exercise was estimated. In additional models non-linear decline in
lung function (including age squared), adjustment for center means of age at the end of education,
proportion of subjects exposed to environmental tobacco smoke (ETS), and proportion of subjects in
manual professions; and last, center specific adjustment for height and BMI (including center*height,

70

center*BMI interaction terms) were investigated. In depth sensitivity analyses addressing the
adjustment for height and age are described below.
Subsample Analyses
The robustness of the main model determined in the full sample was then tested across various
subsamples. Searching for susceptible subgroups, subjects were stratified by their smoking, health, and
asthma status. Separate analyses were also conducted for long term residents (same home in both
surveys) and subjects who moved between baseline and follow-up. To investigate whether
methodological differences across centers had an influence on the results, the analysis was limited to
centers that used identical spirometers, and whose air pollution monitors were not located close to major
roads, respectively. Since several potential confounding factors varied along a north south gradient
across Europe separate analyses for three regions, northern, central, and southern Europe, respectively,
were conducted. Finally, the influence of any single center on the findings was investigated by limiting
the sample to 20 centers only.
Sensitivity Analyses for Height Adjustment
Because there was a considerable correlation between height and air pollution the challenge of
disentangling a potential air pollution effect on lung function (FEV1, FVC, and FEV1/FVC) from the
effect of height on lung function was addressed in detail. To investigate the robustness of the results
towards various ways of adjustment for height the following series of additional sensitivity analyses
addressing the adjustment for height were conducted.
First, several variations of a basic multilevel model were calculated. The estimates were adjusted for age
and smoking (pack-years). In particular, the main model was calculated without adjustment for height,
with a linear term for height, and both a linear and a squared term for height, respectively. In yet another
model, center specific associations between height and lung function were allowed for by using
interaction terms between center and height.

71

As an alternative to lung function as measured, percent of predicted values of lung function were used as
the outcome. The prediction equations included adjustment for age (age, age
2
, age
3
) and height and were
applied to the healthy (no symptoms reported) never smokers of the sample. As such % predicted values
are adjusted for height prior to being used in the final model. Further, lung function divided by height
squared was used as the outcome, as suggested by Chinn et al. (2006).
To overcome collinearity of air pollution and height along a north to south gradient, models including
random effects for region (north, central, south) were attempted to be calculated, however, these models
did not converge.
The following additional sensitivity analyses were conducted for models calculating effects on lung
function level and change in lung function separately. These hierarchical models only included two
levels, centers and subjects, but not the time dependent level of surveys. They modeled either lung
function level or the change in lung function as the dependent variable, respectively.
To further explore the possibility of a non-linear relationship between height and lung function level,
logarithmic transformation of both lung function and height was explored (log-log, lin-log, log-lin).
Since the multi-level model only estimated effects on absolute change in lung function [ml/y] separate
series of models for change were calculated using relative change [%/y] in addition to absolute change
as the dependent variable. To investigate whether change in lung function was dependent of height these
models were calculated with and without adjustment for height.
All the above described models were calculated for the following samples, stratified by sex: full sample,
subjects of mid height (25th – 75th percentile), and southern, central, and northern centers.
Sensitivity Analysis for Age Adjustment
Lung function follows a non-linear pattern of growth and decline throughout a lifespan. Within the
ECRHS sample, the average decline in lung function deviated only little from linearity. Therefore, in the
main model a linear decline of lung function was estimated. To investigate the robustness of the results

72

with regards to the assumption of a linear decline, models adjusting for age squared, and age to the third
power were calculated. These models were calculated for the following samples, stratified by sex: full
sample, stratified by age 30 (younger vs. older), full sample excluding subjects with positive change in
lung function over the follow-up period, and the full sample trimmed so subjects with a positive change
in lung function would be excluded as well as an equal number of subjects with the most negative
change in lung function.
Loss to Follow-up Analysis
The potential impact of loss to follow up on the results was assessed by comparing lung function at
baseline between those subjects lost to follow-up and those included in the analyses. The effect of loss to
follow-up was quantified by comparing the crude association between lung function at baseline and air
pollution in all subjects with the association in only those subjects who were later followed. Such
comparisons were calculated for the complete sample (stratified by sex) as well as all other analyzed
subsamples which were based on available baseline characteristics of the population.
Results
Lung function was measured in 8864 subjects at baseline and 5610 at follow-up, which corresponds to a
follow-up rate of 63% (range across centers: 43% - 85%). Limiting the data set to subjects with
complete, valid records resulted in a sample of 4290 subjects (ATS criteria not fulfilled: 733 subjects;
exclusion of extreme values: 279; incomplete: 308). Details on the available lung function data are listed
in Table 10.

73

Table 10: Summary of available lung function data for each ECRHS center.
N with FEV1 and FVC for ECRHS 1
N fulfilling ATS criteria for FEV1 and FVC for ECRHS 1
Sample selection rate at ECRHS1, (MQ/SQ)
Sample selection rate at ECRHS1, (LF/SQ)
Participation rate at ECRHS2, (MQ2/MQ1)
Participation rate at ECRHS2, (LF2/MQ1)
Median follow-up period in years
Shortest follow-up period in years
Longest follow-up period in years
N with FEV1 and FVC for ECRHS 2
N fulfilling ATS criteria for FEV1 and FVC for ECRHS 2
N with FEV1 and FVC for ECRHS 1 and 2
N fulfilling ATS criteria for FEV1 and FVC for ECRHS 1 and 2
Albacete 617 585 0.18 0.18 0.71 0.55 8.6 7.2 11.7 355 329 240 211
Antwerp C 386 357 0.22 0.13 0.53 0.4 9.1 8.3 10.7 246 238 179 159
Antwerp S 423 382 0.21 0.14 0.61 0.5 9.8 9.1 11 319 305 234 200
Barcelona 295 291 0.19 0.11 0.69 0.45 8.9 7.2 10.5 233 218 115 105
Basel 943 817 0.2 0.19 0.54 0.51 10.3 9.4 11.4 533 476 411 332
Erfurt 654 611 0.22 0.2 0.39 0.39 9.2 8.6 10 282 266 258 233
Galdakao 489 457 0.2 0.16 0.81 0.69 8.7 7.4 9.8 551 513 412 362
Gothenburg 762 706 0.3 0.26 0.74 0.48 9.2 7.8 10 409 371 309 258
Grenoble 513 477 0.18 0.18 0.82 0.64 9.1 8.1 10.4 341 329 304 273
Huelva 379 360 0.18 0.17 0.77 0.72 8.4 6.4 10.2 339 320 199 180
Ipswich 506 485 0.16 0.15 0.66 0.55 8.5 7.4 10.3 302 281 233 210
Norwich 461 441 0.18 0.15 0.54 0.47 9 8 10.8 275 267 205 190
Oviedo 394 379 0.18 0.14 0.68 0.58 8.4 6.5 10 300 281 172 159
Paris 616 523 0.21 0.2 0.67 0.53 8 6.4 11 343 308 332 257
Pavia 254 249 0.38 0.31 0.63 0.62 8.6 7.3 10.4 198 195 155 150
Reykjavik 609 586 0.23 0.21 0.82 0.78 8.9 8.1 10.6 497 473 430 396
Tartu 484 267 0.23 0.2 0.6 0.52 7 5.8 8.4 289 122 215 49
Turin 319 258 0.14 0.13 0.5 0.47 8.4 7.1 9.3 167 163 107 93
Umea 611 576 0.22 0.19 0.76 0.59 8.4 7.5 9.5 421 386 305 268
Uppsala 721 603 0.26 0.23 0.83 0.59 8.6 7.2 9.6 496 450 352 264
Verona 337 328 0.13 0.12 0.6 0.52 8.7 7.1 10.1 190 186 171 163

MQ = participants with main questionnaire, SQ = participants with screening questionnaire,
LF = participants with lung function data

74

Loss to follow-up is of concern in a longitudinal study if those subjects who dropped out of the study
differed from those who participated in categories related to the study question. Compared to those lost
to follow-up, participants were on average one year older, slightly more often women (52% vs. 49%),
less likely to ever have smoked (55% vs. 60%), and their lung function was slightly lower (FEV1: 3.78 l
vs. 3.84 l; FVC: 4.66 l vs. 4.57 l). More details on baseline comparisons can be seen in Table 11.
Table 11: Baseline comparison of subjects lost and subjects followed-up.
Min and max refer to lowest and highest center means, respectively.
Mean min max Mean min max
FEV1 [l] 3.84 3.46 4.14 3.78 3.54 3.96
FVC [l] 4.66 4.14 5.09 4.57 4.24 4.84
Female [%] 49 3955 534671
Age [y] 32.9 31.2 36.0 34.0 31.6 36.9
Weight [kg] 69.4 65.7 72.0 68.6 64.6 73.2
Age at end of education [y] 19.3 16.6 21.9 19.8 16.7 22.2
Ever smoker [%] 60 4874 554667
ETS [%] 60 4090 543182
Asthma [%] 7 116 7 215
Rhinitis [%] 23 12 33 26 9 34
Symptomatic [%] 44 2364 402356
# of symptoms 2.4 1.0 4.0 2.1 1.3 3.0
Lost to follow-up Followed-up
ETS = environmental tobacco smoke
The main characteristics of the participants are listed in Table 12. Participants at baseline were between
24 and 43 years old. Average length of follow-up was 8.9 years. Fifty one percent of participants were
women. Mean FEV1 at baseline was 3.25 liters in women and 4.36 l in men. The range across centers
was 3.04 – 3.45 l and 4.05 – 4.82 l in women and men, respectively. Mean FVC was 3.88 l (3.60 – 4.15)
and 5.33 l (4.92 – 5.61), respectively. Over the 8.9 years of follow-up, FEV1 declined on average by
26ml per year and 34ml/y in women and men, respectively. Additional lung function values are listed in
Table 12. As can be further seen from Table 12, participants differed considerably across centers in such
characteristics as height, BMI, SES, and smoking habits, among others.

75

Table 12: Means (Standard deviation) of relevant variables for the complete sample and lowest and
highest center, by sex.
All Lowest Highest All Lowest Highest
Sample size 2250 40 201 2040 16 187
Continuous variables Mean (sd) Lowest Highest Mean (sd) Lowest Highest
PM
2.5
[µg/m
3
] †
16.8 (8.8) 3.7 44.9 17.1 (9.2) 3.7 44.9
Length of follow-up 8.9 (0.78) 7.0 10.4 8.9 (0.76) 7.0 10.4
FEV1[l] BL 3.25 (0.46) 3.04 3.46 4.36 (0.63) 4.05 4.82
FEV1[l] FU 3.03 (0.46) 2.87 3.31 4.07 (0.63) 3.78 4.66
Change in FEV1[ml/y] -25.3 (24.1) -14.7 -40.1 -33.1 (30.3) -18.2 -53.5
FVC[l] BL 3.88 (0.53) 3.6 4.15 5.33 (0.74) 4.92 5.61
FVC[l] FU 3.72 (0.54) 3.46 3.94 5.09 (0.76) 4.76 5.44
Change in FVC[ml/y] -17.8 (29.7) 0.0 -33.4 -27.1 (39.0) -8.3 -46.8
FEV1/FVC BL 0.84 (0.06) 0.81 0.88 0.82 (0.06) 0.78 0.87
FEV1/FVC FU 0.81 (0.06) 0.80 0.85 0.80 (0.06) 0.77 0.86
Age BL 33.9 (7.1) 31.6 36.9 34.0 (7.1) 30.0 37.4
Age FU 42.8 (7.0) 40.6 46.0 42.9 (7.1) 37.8 45.7
Height [cm] 163.6 (6.5) 158.3 167.4 176.5 (7.1) 170.5 180.9
BMI BL [kg/m
2
]
23.1 (3.8) 21.6 24.7 24.5 (3.2) 23.3 26
BMI FU [kg/m
2
]
24.7 (4.6) 22.9 27.3 25.9 (3.6) 24.5 27.5
Age end education 21.0 (7.8) 17.1 28.4 21.0 (7.4) 17.2 26.4
# of reported symptoms‡ BL 2.2 (2.95) 1.2 3.2 2.0 (2.72) 1.3 3.3
# of reported symptoms‡ FU 2.3 (3.00) 1.4 3.3 2.2 (2.87) 1.3 3.2
Women Men

BL = baseline investigation, FU = follow-up investigation;
Annual mean concentrations of PM
2.5
ranged from 3.7µg/m
3
to 44.7µg/m
3
across all centers; more than
two thirds of the centers, however, had concentrations within a much narrower range from 13µg/m
3
to
24µg/m
3
. The distribution of PM
2.5
annual means is displayed in Figure 9. For the exact values for the
PM
2.5
annual means see Table 5.

76

Table 12 continued: Proportions of relevant variables for the complete sample and lowest and highest
center, by sex.
Indicator variables % Lowest Highest % Lowest Highest
Ever smoked BL 52 36 65 60 45 73
Ever smoked FU 52 37 60 61 47 73
Current smoker BL 31 13 51 37 20 62
Current smoker FU 27 13 42 30 15 58
Ex-smoker BL 40 17 63 40 15 59
Ex-smoker FU 50 11 72 52 21 69
ETS BL 53 24 82 59 31 85
ETS FU 38 16 72 43 20 81
SES based on occup. group
Managers/professionals; non-manual 25 11 48 32 10 56
Technicians & associate professions 18 4 29 17 6 27
Other non-manual 36 19 50 16 3 36
Skilled manual 2 0 13 19 3 30
Semi-skilled or unskilled manual 7 0 21 13 3 24
Unclassifiable or unknown 11 1 29 3 0 14
Trucks pass home constantly§ 14 4 27 12 6 27
Ever asthma BL 8 2 17 6 0 14
Ever asthma FU 12 3 21 9 1 19
Ever rhinitis BL 27 9 38 24 9 32
Sensitization IgE [>0.7 kU/L]BL¦ 22 5 34 29 14 41
Women Men

BL = baseline investigation, FU = follow-up investigation; ETS = environmental tobacco smoke;
SES = socio-economic status; ¦ D.pteronyssinus, timothy grass, cat and C. herbarium allergens
Since exposure was measured on the community level, correlations across the 21 centers between PM
2.5
,
lung function, and various covariates are of relevance to assess potential confounding. These are listed in
Table 29 in the annex on page 161. Center mean height, mean age at the end of education, and average
number of reported respiratory symptoms, gas use, and occupational exposures showed the strongest
correlations with PM
2.5
(Pearson r = -0.5). As can be further seen from Table 29 many variables were
correlated with latitude, including PM
2.5
and several lung function measures.
The main results from the regression models of FEV1, FVC, and FEV1/FVC ratio are presented in four
tables, each stratified by sex. Table 13 and Table 14 list regression coefficients for the effects on lung
function level (cross-sectional effects). Table 13 shows results from the parsimonious main model,
listing the effects of age, height, smoking, and PM
2.5
. PM
2.5
regression coefficients from various sub-
analyses are shown in Table 14. Table 16 and Table 17 analogously show the estimates for the effects on

77

Figure 9: Distribution of PM
2.5
annual mean concentrations across ECRHS centers.
Units are µg/m
3
.
0 10 20 30 40 50
RE UMUPGO AL TA OVNOGA ER IP HU BS PS GN AS BA AC PA VE TU

µg/m
3

change in lung function (longitudinal effects). After describing these main results a selection of results
from several sensitivity analyses is presented.
Results for PM
2.5
and Lung Function Level
As expected, age and height were strong predictors of lung function level (Table 13). For example, in
women FEV1 was 25.5ml lower for each additional year of age, whereas it was 31.8ml higher for each
additional centimeter of standing height. The negative effects of current smoking were significant in
men, and for the FEV1/FVC ratio in both sexes. There were no significant effects of PM
2.5
on lung
function level.
These null-findings were robust towards varying the degree of adjustment or varying the sub-sample of
participants analyzed (Table 14). Height was the strongest confounder of the crude negative, though

78

non-significant association between PM
2.5
and lung function level. Including additional covariates had
only a minor influence on the non-significant PM
2.5
coefficients. Neither the subsample analyses among
never smokers, the northern, central, and southern centers, centers that all used the same spirometer
(Biomedin), subjects who lived in the same house at both surveys (long term residents), and movers, nor
among any other sub-samples yielded more conclusive coefficients for PM
2.5
(Table 15).

79

Table 13: Coefficients (p-values) for lung function level from main models* for complete samples, by sex.
Negative coefficients represent adverse effects on lung function. Effects for PM
2.5
are per 10µg/m
3
.
Variable FEV1-level [ml] FVC-level [ml] FEV1/FVC-level† FEV1-level [ml] FVC-level [ml] FEV1/FVC-level
Intercept 3349 4115 0.816 3967 4840 0.822
Age -25.5 (<0.001) -18.0 (<0.001) -0.003 (<0.001) -25.2 (<0.001) -21.6 (<0.001) -0.002 (0.001)
Height [cm] 31.8 (<0.001) 44.6 (<0.001) -0.001 (<0.001) 41.6 (<0.001) 61.3 (<0.001) -0.002 (<0.001)
Ex-smoker 36.5 (0.004) 55.2 (<0.001) -0.003 (0.156) 1.5 (0.934) 13.4 (0.539) -0.002 (0.308)
Current smoker -9.1 (0.478) 23.2 (0.122) -0.009 (<0.001) -62.9 (0.001) -54.0 (0.014) -0.006 (0.008)
PM
2.5
[10µg/m
3
]
9.3 (0.578) -6.4 (0.782) 0.004 (0.194) 38.5 (0.097) 32.0 (0.316) 0.002 (0.577)
Women (N=2243) Men (N=2031)

* Coefficients for lung function level are estimated in the same model as coefficients for change in lung function (Table 16) and therefore adjusted for those variables.
In addition coefficients for lung function level are adjusted for BMI and SES (coefficients not shown).

80

Table 14: PM
2.5
coefficients (p-values) for lung function level from sensitivity analyses using different adjustment variables.
Negative coefficients represent adverse effects on lung function. Coefficients are per 10µg/m
3
PM
2.5
. Units for FEV1 and FVC are ml.
FEV1 FVC FEV1/FVC* FEV1 FVC FEV1/FVC
Sub-models N β (p) β (p) β (p) N β (p) β (p) β (p)
Crude 2250 -24.2 (0.33) -54.9 (0.09) 0.006 (0.06) 2040 -24.0 (0.49) -57.8 (0.21) 0.005 (0.26)
Height only 2250 11.2 (0.50) -5.3 (0.82) 0.004 (0.18) 2040 38.3 (0.09) 29.9 (0.34) 0.002 (0.56)
Minimal 2245 11.2 (0.50) -5.4 (0.81) 0.004 (0.18) 2037 38.8 (0.09) 30.5 (0.33) 0.002 (0.56)
Main 2243 9.3 (0.58) -6.4 (0.78) 0.004 (0.19) 2031 38.5 (0.10) 32.0 (0.32) 0.002 (0.58)
Centre level adj. 2243 19.7 (0.23) 2.6 (0.90) 0.005 (0.14) 2031 39.7 (0.14) 25.1 (0.43) 0.004 (0.37)
Age squared 2245 10.3 (0.53) -7.0 (0.76) 0.004 (0.18) 2037 38.2 (0.09) 29.7 (0.33) 0.003 (0.55)
Maximal 2227 9.0 (0.58) -6.4 (0.78) 0.004 (0.24) 2018 39.2 (0.09) 31.1 (0.33) 0.002 (0.55)
Centre specific adj. for height and BMI 2243 -6.0 (0.79) -24.4 (0.40) 0.003 (0.38) 2031 37.2 (0.13) 24.8 (0.41) 0.004 (0.33)
Women Men

Crude = PM
2.5
; Height only = + height; Minimal = + smoking status (never, ex, current); Main = + BMI, SES Age squared = + age2; Centre level adj. = main + center
means of "Education level (age)", "Proportion of non-manual professions", "ETS"; Maximal = main + long and short term respiratory symptoms, exercise, trucks at
home, height squared Center specific adj. for height, BMI = main + centre*height + centre*BMI; * Models did not converge. Estimates are based on iterated
Estimation-Maximization.

81

Table 15: PM
2.5
coefficients (p-values) for lung function level from sensitivity analyses using different subsamples.
Negative coefficients represent adverse effects on lung function. Coefficients are per 10µg/m
3
PM
2.5
. Units for FEV1 and FVC are ml.
Sub-sample N B p B p B p N B p B p B p
Never smokers 1007 63.3 0.74 -62.3 0.80 0.003 0.33 745 167.0 0.52 -38.6 0.91 0.004 0.44
Ever smokers 1234 135.3 0.48 -19.2 0.94 0.004 0.19 1283 575.3 0.03 588.7 0.11 0.002 0.69
Non-asthmatics 1939 95.6 0.55 -85.7 0.70 0.004 0.16 1833 470.5 0.04 372.8 0.26 0.003 0.44
Asthmatics 300 4.5 0.99 35.9 0.93 0.000 0.94 196 -194.1 0.68 38.6 0.94 -0.004 0.62
Asymptomatics 968 63.1 0.73 -158.3 0.52 0.005 0.06 908 485.1 0.05 567.3 0.10 0.001 0.90
Symptomatics 1275 95.3 0.61 14.6 0.95 0.002 0.64 1123 330.5 0.20 103.4 0.77 0.004 0.34
Biomedin spirometers only 1122 144.5 0.47 -93.5 0.77 0.006 0.08 1057 275.4 0.43 -48.5 0.92 0.006 0.28
Northern centers 871 576.4 0.41 754.4 0.35 0.000 0.97 716 744.6 0.39 1555.4 0.17 -0.007 0.69
Central centers 618 1076.4 0.47 1485.2 0.54 -0.006 0.85 541 869.9 0.44 1974.4 0.46 -0.014 0.74
Southern centers 755 91.2 0.67 -144.6 0.58 0.006 0.05 774 202.5 0.59 -84.6 0.85 0.006 0.26
Mid range of height only 1212 52.3 0.78 -139.7 0.58 0.005 0.28 1119 598.8 0.02 598.6 0.07 0.002 0.66
Longterm residents 1132 236.1 0.24 97.6 0.73 0.005 0.15 956 356.6 0.08 196.0 0.51 0.004 0.44
Movers 1108 -51.8 0.80 -220.9 0.39 0.003 0.32 1074 278.4 0.38 283.6 0.50 0.001 0.86
Excl. 4 monitors close to traffic 1965 174.7 0.68 152.0 0.79 0.001 0.86 1762 648.1 0.21 1257.8 0.06 -0.007 0.46
Excl. Erfurt (past pollution) 2131 105.5 0.54 -55.4 0.82 0.004 0.19 1913 396.6 0.09 321.7 0.33 0.002 0.58
Excl. Tartu and Basel (LF variability) 2026 117.3 0.49 -55.1 0.82 0.005 0.07 1866 404.6 0.07 288.1 0.41 0.003 0.47
Women Men
FEV1 FVC FEV1/FVC FEV1 FVC FEV1/FVC

82

Results for PM
2.5
and Change in Lung Function
Results of the main models for change in lung function are shown in Table 16. Due to the age range of
the ECRHS sample the decline in lung function deviated little from linearity, therefore, the main models
presented fit a linear change in lung function. Non-linear adjustment for age was evaluated as part of the
sensitivity analyses (see below). Note that the intercept for change corresponds to the coefficient for age
in table 2 (lung function level), because the effect of time (age) on lung function level is equivalent to
the constant in a model of linear change in lung function over time. Analyses of the complete samples of
women and men revealed no significant effects of PM
2.5
on change in lung function (Table 16).
Again, these findings were not sensitive to varying the degree of adjustment (Table 17). Two statistically
significant coefficients for the effect of PM
2.5
on change in lung function were observed for FEV1
among men of the nine southern European centers ( β=-3.53ml/10µgm
-3
PM
2.5
; p-value=0.02), and FEV1
among men in twelve centers that used Biomedin spirometers ( β=-3.56ml/a×10µgm
-3
PM
2.5
; p-
value=0.01) (Table 18). These two subsamples are nearly identical since all southern centers used
Biomedin spirometers.

83

Table 16: Coefficients (p-values) for change in lung function from main models* for complete samples, by sex.
Negative coefficients represent adverse effects on lung function. Coefficients are per 10µg/m
3
PM
2.5
. Units: FEV1 and FVC ml/a, FEV1/FVC /a.
Variable FEV1-change
[ml/a]
FVC-change
[ml/a]
FEV1/FVC
change [/a]‡
FEV1-change
[ml/a]
FVC-change
[ml/a]
FEV1/FVC
change [/a]
Age† -25.5 (<0.001) -18.0 (<0.001) -0.003 (<0.001) -25.2 (<0.001) -21.6 (<0.001) -0.002 (0.001)
Height [cm] -0.26 (<0.001) -0.08 (0.373) -0.000 (0.101) -0.40 (<0.001) -0.31 (0.006) -0.000 (0.397)
Ex-smoker -1.37 (0.184) -2.73 (0.028) 0.000 (0.292) -0.68 (0.635) -1.18 (0.516) -0.000 (0.716)
Smoker -2.89 (0.004) -2.76 (0.021) -0.000 (0.253) -2.36 (0.074) -1.12 (0.503) -0.000 (0.031)
PM
2.5
[10µg/m
3
] 0.46 (0.755) 1.31 (0.467) -0.000 (0.547) 0.10 (0.952) 2.03 (0.326) -0.000 (0.200)
Women (N=2243) Men (N=2031)

* Coefficients for change in lung function are estimated in the same model as coefficients for lung function level and therefore adjusted for those variables (see Table
13). In addition coefficients for change in lung function are adjusted for BMI (coefficients not shown); † Coefficients for age from the models for lung function level
(Table 13) serve as intercepts for change in lung function; ‡ Model did not converge: Estimates are based on iterated Estimation-Maximization.

84

Table 17: PM
2.5
coefficients for change in lung function from sensitivity analyses using different adjustment variables.
Negative coefficients represent adverse effects on lung function. Coefficients are per 10µg/m
3
PM
2.5
. Units: FEV1 and FVC ml/a, FEV1/FVC /a.
FEV1 FVC FEV1/FVC* FEV1 FVC FEV1/FVC
Sub-models N β (p) β (p) β (p) N β (p) β (p) β (p)
Crude 2250 0.68 (0.64) 1.44 (0.39) -0.00012 (0.57) 2040 0.78 (0.66) 2.39 (0.25) -0.00023 (0.24)
Height only 2250 0.79 (0.57) 1.95 (0.24) -0.00018 (0.40) 2040 0.51 (0.76) 2.56 (0.21) -0.00030 (0.15)
Minimal 2245 0.80 (0.58) 1.93 (0.25) -0.00017 (0.43) 2037 0.57 (0.73) 2.61 (0.20) -0.00029 (0.16)
Main 2243 0.46 (0.76) 1.31 (0.47) -0.00014 (0.55) 2031 0.10 (0.95) 2.03 (0.33) -0.00028 (0.20)
Center level adj. 2243 -0.78 (0.64) -0.99 (0.55) 0.00005 (0.84) 2031 -1.01 (0.59) -0.35 (0.85) -0.00010 (0.67)
Age squared 2245 0.66 (0.64) 1.72 (0.27) -0.00016 (0.44) 2037 0.69 (0.66) 2.91 (0.11) -0.00030 (0.14)
Maximal 2227 0.14 (0.93) 0.96 (0.59) -0.00014 (0.52) 2018 -0.23 (0.90) 1.72 (0.42) -0.00029 (0.16)
Center specific adj. for
height, BMI
2243 -0.95 (0.60) -0.78 (0.75) -0.00011 (0.66) 2031 -0.12 (0.95) 1.59 (0.46) -0.00026 (0.29)
Women Men

Crude = PM
2.5
; Height only = + height; Minimal = + smoking status (never, ex, current); Main = + BMI; Age squared = + age
2
; Centre level adj. = main + center
means of "Education level (age)", "Proportion of non-manual professions", "ETS"; Maximal = main + long and short term respiratory symptoms, height squared;
Center specific adj. for height, BMI = main + centre*height + center*BMI; * Models did not converge. Estimates are based on iterated Estimation-Maximization.

85

Table 18: PM
2.5
coefficients for change in lung function from sensitivity analyses using different subsamples.
Negative coefficients represent adverse effects on lung function. Coefficients are per 10µg/m
3
PM
2.5
. Units: FEV1 and FVC ml/a, FEV1/FVC /a.
Sub-sample N B p B p B p N B p B p B p
Never smokers 1007 0.94 0.52 2.33 0.19 -0.0003 0.31 745 -0.61 0.74 0.58 0.81 -0.0003 0.33
Ever smokers 1234 0.21 0.90 0.68 0.75 0.0000 0.94 1283 0.73 0.68 2.79 0.18 -0.0002 0.27
Non-asthmatics 1939 1.10 0.45 1.98 0.26 -0.0001 0.66 1833 0.23 0.90 1.97 0.36 -0.0002 0.27
Asthmatics 300 -2.06 0.37 -2.02 0.53 -0.0002 0.45 196 -0.61 0.82 2.52 0.49 -0.0005 0.29
Asymptomatics 968 1.07 0.47 1.95 0.32 -0.0001 0.64 908 1.29 0.55 1.96 0.42 0.0000 0.94
Symptomatics 1275 0.10 0.95 0.95 0.61 -0.0001 0.58 1123 -0.55 0.72 2.24 0.29 -0.0005 0.02
Biomedin spirometers only 1122 -2.49 0.11 -2.34 0.27 -0.0001 0.77 1057 -3.66 0.01 -2.59 0.24 -0.0002 0.39
Northern centers 871 9.25 0.01 14.90 0.00 -0.0006 0.24 716 13.45 0.00 15.50 0.01 -0.0002 0.71
Central centers 618 13.74 0.01 22.95 0.00 -0.0010 0.21 541 13.62 0.12 26.07 0.00 -0.0017 0.09
Southern centers 755 -2.66 0.16 -2.63 0.25 -0.0001 0.74 774 -3.65 0.02 -3.13 0.12 -0.0002 0.39
Mid range of height only 1212 0.58 0.72 1.82 0.34 -0.0002 0.37 1119 1.31 0.48 2.69 0.27 -0.0002 0.37
Longterm residents 1132 0.96 0.53 1.50 0.44 -0.0001 0.84 956 1.29 0.53 4.02 0.07 -0.0003 0.19
Movers 1108 0.06 0.97 1.46 0.50 -0.0003 0.25 1074 -0.68 0.71 0.23 0.93 -0.0002 0.44
Excl. 4 monitors close to traffic 1965 8.33 0.00 8.62 0.03 0.0004 0.43 1762 9.26 0.00 8.87 0.06 0.0002 0.67
Excl. Erfurt (past pollution) 2131 0.47 0.76 1.26 0.49 -0.0001 0.59 1913 -0.01 1.00 1.93 0.35 -0.0003 0.21
Excl. Tartu and Basel (LF variability) 2026 0.50 0.74 1.32 0.47 -0.0001 0.59 1866 0.28 0.86 2.15 0.32 -0.0003 0.20
Women Men
FEV1 FVC FEV1/FVC FEV1 FVC FEV1/FVC

LF = lung function

86

Results from Sensitivity Analyses for Height Adjustment
To assess the robustness of the null findings described above towards various ways of adjustment for
height, a series of models was calculated, as described in the methods section. Overall, these analyses
resulted in 1800 separate estimates for the effects of PM
2.5
, including effects on lung function level and
change in lung function, using FEV1, FVC, and FEV1/FVC ratio as outcomes. None of the estimates for
PM
2.5
effects on lung function level were significantly smaller than zero, which would represent an
adverse effect. Among the 780 estimates for PM
2.5
effects on change in lung function there were 31
significant adverse estimates, all except two of which occurred in the sample of southern European
centers. This number of significant findings is less than the 39 that would be expected by chance alone,
given the large number of analyses. Other than the pattern of significant estimates for change in lung
function in the southern European centers, which was already seen in the main analyses (see above), no
systematic trends related to the modeling approaches could be identified.
Results from Sensitivity Analyses for Age Adjustment
The sensitivity analyses addressing the robustness of the main models towards various ways of
adjustment for age yielded 300 separate estimates for the effects of PM
2.5
. Several models returned
significant, yet small coefficients for age squared. Nonetheless, there were no significant adverse effects
of PM
2.5
observed, independent of how age was adjusted for, nor whether the sample was stratified by
age (younger vs. older than 30 years). Models using percent of predicted lung function values, adjusting
for age by means of prediction equations did not result in any significant adverse effects of PM
2.5
either.
Results from Loss to Follow-up Analysis
The possible effect of loss to follow-up was assessed by comparing the associations between PM
2.5
and
lung function (FEV1, FVC) at baseline for the complete baseline sample and the followed-up sample.
The average difference in center means in lung function between the followed–up sample and the

87

complete baseline sample was 20ml. In the most extreme case it differed by as much as 120ml (FVC in
Tartu). The effects on the associations across centers, however, were small. The differences between the
regression slopes for the baseline sample and the followed-up subjects are shown in Table 19, for the
complete samples and various subsamples of participants. There was a small attenuating effect (negative
sign in Table 19) of loss to follow-up on the effect of PM
2.5
on FEV1 and FVC in women (-6ml/10µgm
-
3
, and -13ml/10µgm
-3
, respectively). The strongest attenuating effects of loss to follow-up on the
regression slopes in subsample analyses did not exceed 30ml per 10µg/m
3
.

88

Table 19: Effects of loss to follow-up on the crude association between PM
2.5
and baseline lung
function for various subsamples.
For example in women for FEV1, the association among subjects followed up was -24ml per
10µg/m
3
PM
2.5
. Among all women that participated in the baseline investigation the association was
-31ml per 10µg/m
3
PM
2.5
. In other words, the effect of the loss of some women led to a slightly
attenuated crude association between FEV1 and PM
2.5
at baseline. This difference may be interpreted
as a proxy for the effect of loss of subjects on the association between PM
2.5
and FEV1 in the overall
analysis only including subjects followed-up (under the assumption that there were no differences in
change in lung function between the two groups). An illustration of this example is provided in the
annex in Figure 16 on page 156. Units for slopes are ml/10µgm
-3
.
Sample
N
followed
N
lost
Slope
followed
Slope
all
Difference
in slope
Slope
followed
Slope
all
Difference
in slope
Women (all) 2865 2116 -24 -31 -6 -50 -62 -13
Never smokers 1401 879 -32 -28 4 -58 -52 6
Ever smokers 1495 1268 -13 -33 -20 -38 -71 -33
Non-asthmatics 2625 1965 -16 -23 -7 -46 -57 -11
Asthmatics 278 185 -73 -83 -10 -86 -115 -29
Asymptomatic 1683 1243 -28 -33 -6 -66 -69 -3
Symptomatic 1272 974 -20 -28 -7 -24 -53 -29
Mid height 1573 1243 04 4 -23 -17 6
North 1067 658 -71 -68 4 -123 -121 1
Central 769 731 195 194 -1 216 206 -11
South 886 563 46 41 -5 37 26 -11
Biomedin 1382 967 37 32 -4 23 12 -11
Excl. traffic mon. 2462 1791 -107 -133 -27 -145 -173 -27
Age <30 947 837 -36 -31 5 -76 -68 7
Age >30 1963 1318 -16 -27 -11 -39 -58 -19
Men (all) 2761 2060 -57 -38 19 -104 -80 24
Never smokers 1157 757 -82 -32 50 -143 -73 70
Ever smokers 1677 1350 -38 -41 -2 -76 -87 -11
Non-asthmatics 2594 1925 -57 -34 23 -103 -79 25
Asthmatics 216 167 -57 -67 -10 -118 -107 11
Asymptomatic 1681 1243 -35 -9 27 -66 -40 26
Symptomatic 1151 902 -88 -82 7 -154 -138 16
Mid height 1559 1211 22 43 21 520 15
North 1002 583 -64 -45 20 -76 -74 1
Central 726 747 19 43 24 81 143 62
South 955 592 39 57 19 19 39 20
Biomedin 1315 917 26 44 18 -9 14 23
Excl. traffic mon. 2372 1757 -148 -148 0 -172 -170 2
Age <30 938 837 -2 -3 -1 -34 -28 6
Age >30 1884 1262 -74 -45 29 -131 -97 33
FEV1 FVC

89

Discussion
ECRHS is the first trans-European study to investigate the association between long-term levels of air
pollution and lung function. Neither lung function levels nor change in lung function over nine years
were significantly associated with PM
2.5
across the 21 communities. These null findings stand in contrast
to numerous cross-sectional studies (Chestnut et al. 1991; Ackermann-Liebrich et al. 1997; Kunzli et al.
1997a; Abbey et al. 1998; Schindler et al. 1998; Galizia et al. 1999; Schikowski et al. 2005; Tager et al.
2005), and in particular to the two European multi-center studies, namely SAPALDIA within
Switzerland (Ackermann-Liebrich et al. 1997) and SALIA in the Rhine-Rhur basin in Germany
(Schikowski et al. 2005). They also contradict the few longitudinal studies, although the published
evidence for an adverse effect of air pollution on decline in lung function in adults was weaker in the
first place (van der Lende et al. 1981; Jedrychowski et al. 1989; Tashkin et al. 1994; Sekine et al. 2004).
The observed null-findings are also contrary to expectations from an ever increasing understanding of
the body’s defense mechanisms against air pollutants and the health effects resulting thereof (Gilliland et
al. 1999; Nel 2005; Schlesinger et al. 2006). Increased levels of oxidative stress induced by oxidative
gases and particles in particular affect the redox balance in the lung lining fluid and can cause direct
structural damage to epithelial membranes. This first line of defense is coupled with a persistent
inflammatory response in the lung which leads to tissue damage and affects lung function. These pro-
inflammatory effects are also considered the main link between air pollution and cardio-vascular
outcomes (Kunzli et al. 2005a) for which lung function is a marker (Sin et al. 2005).
There are basically three main approaches on how to interpret the observed null-findings. First, there
could be truly no effect of air pollution on lung function in adults. Second, the null-findings could be due
to some sort of measurement error in both exposure and the outcome. Third, a confounding factor which
was not adjusted for could have biased a true effect towards null. In the following, these issues will be
elaborated in more detail.

90

The observed findings may reflect a true null-effect if either the population studied in ECRHS is less
susceptible to an adverse effect of air pollution than those in other studies, or if other studies’ findings
have been flawed. The first possibility seems unlikely.
Measurement error can occur in several facets. Excessive random misclassification may reduce
statistical power to a level where a true effect cannot be detected anymore. Power calculations (Quanto
version 1.1.1 http://hydra.usc.edu/gxe) indicated sufficient power to detect significant effects on lung
function level of a magnitude comparable to the significant cross-sectional findings in SAPALDIA
(Ackermann-Liebrich et al. 1997). Details on the significantly detectable effect sizes are shown in Table
20 and in Appendix 3: Statistical Power
Table 28 in the annex on page 154. For a comparison to the observed coefficients for lung function level,
see Table 14 on page 81 and Table 15 on page 82. For a comparison to the observed coefficients for
change in lung function, see Table 17 on page 85 and Table 18 on page 86. The estimates for
significantly detectable effect sizes provided in Table 20 are based on center level power calculations for
a sample size of 21 centers, ignoring the clustered nature of the study sample (i.e. the number of
subjects). As such, these estimates can be interpreted as conservative power estimates, with the truly
detectable effects being somewhat smaller.
The size of an effect on FEV1 level significantly detectable with a statistical power of 80%, after
adjustment for height and smoking status, was 39ml (1.2% of mean FEV1) and 58ml (1.4%) per PM
2.5

contrast of 10µg/m
3
, in women and men, respectively. Relative detectable effect sizes were slightly
larger for FVC (approx. 1.5%), and only marginally smaller for subsamples of subjects older than 30
years at baseline. The significant effects reported in SAPALDIA were -1.6% and -3.5% per 10µg/m
3

PM
10
for FEV1 and FVC, respectively (Ackermann-Liebrich et al. 1997). Exclusion of the three most
and the three least polluted centers showed that statistical power was almost entirely driven by these
centers. Significantly detectable effect sizes across the 15 centers in the mid-range of the PM
2.5
spectrum
were approx. 30 times larger, compared to the full study, due to the extremely narrow range of PM
2.5

91

across these centers (Standard deviation = 2.8µg/m
3
) (See Appendix 3: Statistical Power
Table 28 in annex on page 154).
Relative power was considerably smaller to detect an effect on change in lung function. Significantly
detectable effects for change in FEV1 were 4.2ml/y (16.6% of mean change in FEV1) and 4.8ml/y
(14.5%) per 10µg/m
3
of PM
2.5
, in women and men, respectively. There are no longitudinal air pollution
studies in adults available to put these effect sizes in perspective (see Chapter 2). The decline in lung
function observed in the ECRHS sample is very similar to that observed in other studies of adults (Xu et
al. 1995). A comparison to effects of smoking observed in the ECRHS study population may be helpful,
although due to the inherent differences between these exposure measures a quantitative interpretation is
difficult. In a larger sample of ECRHS participants, Chinn et al. (2005) reported accelerated decline in
FEV1 in smokers compared to never smokers, by 5.1ml/y and 4.8ml/y, in women and men, respectively.
The effects of cigarette pack-years in the sample subject to this analysis were 3.4ml/y and 4.2ml/y for an
average female (13 pack years) and male smoker (20 pack years), respectively. Based on this power
calculation a detectable effect of PM
2.5
on change in FEV1 would have to be equivalent to the effect of
approx. 16 pack years and 23 pack years of smoking, in women and men, respectively.
The effects on FEV1 growth observed in the Children’s Health Study (Gauderman et al. 2004), on the
other hand, were only -2.6% and -1.5% per 10µg/m
3
PM
2.5
increase, in girls and boys, respectively.
Effects in adults, in particular in the relatively young age range of ECRHS (20-44y at baseline), are
expected to be smaller than in children, due to the significantly smaller overall changes in lung function,
and the higher susceptibility of young, growing lungs.
Under the expectation that the effects on change in lung function in adults would be smaller than those in
children, and that the effects of cigarette smoking are considerably more detrimental than those of long
term background air pollution, it must be concluded that ECRHS was underpowered to detect significant
effects on the decline of lung function.

92

Lack of power, however, does not explain the fact that no significant effects on lung function level were
observed. The observed coefficients for both lung function level and change in lung function were not
only statistically non-significant, but also did not suggest consistent negative effects of air pollution. The
majority of coefficients showed positive signs, opposite of an expected adverse effect, even though these
coefficients were small and not significant (see Table 13 through Table 18). It remains therefore
worthwhile addressing potential limitations of the study which may have led to misclassification of air
pollution and lung function estimates, and thereby not only reducing statistical power, but also
potentially biasing the association between PM
2.5
and lung function.

93

Table 20: Significantly detectable effects sizes for effects on lung function level and change in lung function.
Detectable effect sizes are shown for 80% power and significance of 95%. Calculations were conducted in Quanto (Version 1.1.1 http://hydra.usc.edu/gxe) and
took into account standard deviation of PM
2.5
, mean of outcome, and standard deviation of center random effects after adjusting for height and smoking status. All
calculations were conducted for a sample of N=21, corresponding to the number of centers. For FEV1/FVC ratio and centers of mid-range PM
2.5
concentrations see
Appendix 3: Statistical Power
Table 28 in annex on page 154. For observed effects see Table 13 through Table 18 on pages 80 to 86.
Sample
Spirometric
Measure
N
Centers
N
Subjects Mean SD Mean SD
Detectable Effect
[ml/10ug/m3] Mean SD
Detectable Effect
[ml/y/10ug/m3]
Men FEV1 21 2040 17.0 9.1 4.22 0.095 58 -0.033 0.0067 4.8
Men <30y FEV1 21 671 17.0 9.1 4.21 0.107 66 -0.032 0.0051 3.6
Men >30y FEV1 21 1369 17.0 9.1 4.13 0.093 57 -0.034 0.0064 4.2
Men FVC 21 2040 17.0 9.1 5.21 0.127 78 -0.027 0.0085 5.7
Men <30y FVC 21 671 17.0 9.1 5.19 0.134 83 -0.024 0.0084 5.7
Men >30y FVC 21 1369 17.0 9.1 5.15 0.133 82 -0.029 0.0075 5.1
Women FEV1 21 2250 17.0 9.1 3.14 0.063 39 -0.025 0.0060 4.2
Women <30y FEV1 21 733 17.0 9.1 3.12 0.068 42 -0.023 0.0041 3.0
Women >30y FEV1 21 1517 17.0 9.1 3.08 0.063 39 -0.027 0.0060 4.2
Women FVC 21 2250 17.0 9.1 3.80 0.092 57 -0.018 0.0072 4.9
Women <30y FVC 21 733 17.0 9.1 3.78 0.089 55 -0.014 0.0059 4.2
Women >30y FVC 21 1517 17.0 9.1 3.77 0.095 58 -0.021 0.0068 4.8
PM2.5 Lung Function Level Change in Lung Function

94

Since air pollution was measured at the center level, factors that may have led to systematic under or
over estimation of center estimates are an important concern in ECRHS. Because the number of centers
is relatively small compared to the number of subjects, any sources of error that will affect center mean
estimates directly (i.e. all subjects from the same center, or air pollution measurements taken at the
center level), will have much larger influence on the final analysis than error on the individuals’ level. Of
particular concern would be if such center level misclassification was associated with either outcome or
exposure, which could lead to biased estimates for the effects of air pollution. In the following the
potential measurement errors for lung function and air pollution measurements in ECRHS are addressed
separately.
Lung Function Assessment
The variability of lung function measurements can depend on a number of factors, such as a subjects’
health status, short term impacts of exposures or medication, the time of the day the test is taken, how
well an expiration maneuver is performed, how the maneuver is administered by the investigator, and
finally, on the device specific variability of the spirometer itself. None of these factors are easily
quantifiable. For the determination of community average lung function estimates, some of these factors
can be assumed to occur at random within the relatively large study samples, while center specific
aspects such as fieldworkers and spirometers could potentially have considerable influence on the
distribution of mean lung function estimates across the 21 centers.
In ECRHS lung volumes were measured following a standardized protocol (see annex on page 157), and
all measurements included in the analysis fulfilled the ATS criteria for reproducibility. Flow measures,
which may have offered to investigate effects on small airways more specifically, were not available. To
prevent that lung function measurements were influenced by short term factors, subjects were advised
not to smoke one hour before the examination and not to use medication affecting the respiratory tract.
Individuals who suffered from a respiratory infection in the three weeks previous to examination were
asked to take the examination at a later point, if possible. Actual adjustment for short term effects of air

95

pollution on lung function was not feasible, however, because air pollution was not measured on a daily
basis (84 out of 365 days only). Effects of daily air pollution levels on lung function measurements are
expected to be small and have been estimated to be less than 1% per changes of 10µg/m
3
in TSP or NO
2

in SAPALDIA (Schindler et al. 2001). The influence of time of measurement is expected to be even
smaller, as reported by Borsboom et al. (1999) who observed only negligible reductions in standard
deviations of measurements taken during daytime hours in a Dutch cohort after adjusting for time of the
day.
Lung function measurements were conducted by well-trained local personnel at each study center using
the local spirometer. Calibration of spirometers was part of the protocol which was followed by all
centers. Nonetheless, differences in equipment and personnel between centers, and changes thereof over
time may have introduced some degree of misclassification in the center mean estimates of lung
function. For example, compliance with ATS criteria was much lower in Tartu (~69%), compared to the
rest of the centers (~92-99%).

96

Figure 10: PM
2.5
levels across ECHRS centers by type of spirometer used.
Wide grey bars in the background reflect mean PM
2.5
levels across centers using the same type of
spirometer. Narrow black bars in the foreground represent center PM
2.5
levels.
0
10
20
30
40
50
AL OVNOGA IP HUPS GN BA PA VE TU
Biomedin
RE UM UPGO BS
SM Hot Wire
TA ER
Jaeger
PM
2.5
Center Means PM
2.5
Means by Spirometer
ug/m
3
AC AS
SM Dry
0
10
20
30
40
50
AL OVNOGA IP HUPS GN BA PA VE TU
Biomedin
RE UM UPGO BS
SM Hot Wire
TA ER TA ER
Jaeger
PM
2.5
Center Means PM
2.5
Means by Spirometer
ug/m
3
AC AS
SM Dry

Biomedin = Biomedin Baires water seal volume displacement spirometer; SM Dry = SensorMedics dry
seal volume displacement spirometer (changed to Jaeger Masterscope at follow-up); SM Hot Wire =
SensorMedics heated wire flow sensing spirometer; Jaeger = Jaeger Pneumotach

As mentioned in the method section of this chapter, not all centers used the same type of spirometer (see
Table 9). The use of different equipment is a potential source of systematic error in cross community
comparisons of lung function which may not be prevented entirely with standardized procedures alone.
Random over or under estimation of true center mean lung function levels due to equipment related
variance may have contributed to the observed null findings. To lead to actual bias towards the null, the
equipment effect would have had to be associated with air pollution levels in a way that centers with
high pollution levels would have over estimated lung function and centers with low pollution would

97

have under estimated it. As can be seen from Figure 10, PM
2.5
values were on average higher in the
centers that used Biomedin or Sensormedics dry spirometers (volume-displacement spirometers),
compared to the centers that used Sensor Medics or Jaeger spirometers based on flow-sensing
technology. Since no direct comparisons between the used devices were conducted, possible systematic
differences between the ECHRS spirometers cannot be quantified. In a comparison of the same type of
SensorMedics and Biomedin spirometers as used in ECRHS across 50 subjects, Künzli et al. (2005b)
measured FEV1 and FVC values within ±1.9% and ±2.3% of a subjects’ mean. The ERS/ATS criteria
for standardization of spirometry require an accuracy of ±3% (Miller et al. 2005). The lung volumes
observed with SensorMedics heated wire flow meters were somewhat increased, as compared to the
Biomedin or SensorMedics volume displacement spirometers. If this same pattern occurred for the
ECRHS devices it would have led to inflated, adverse effects. Differences in equipment are therefore an
unlikely explanation of the observed null-findings. A comparison of lung function values by the type of
spirometer used did not reveal any significant differences (see Appendix 4: Type of Spirometer and
Lung Function
Figure 15 in the annex on page 155.

98

Figure 11 Scatterplot of sub-analysis for change in FEV1 in men across centers which used Biomedin
spirometers.
-5
0
5
Adj. change in FEV1 [ml/a]
15 20 25 30 35 40 45
PM2.5 [ug/m3]
GA
AL
BA
NO
OV
IP
HU
PS
GN
PA
TU
VE
-5
0
5
Adj. change in FEV1 [ml/a]
15 20 25 30 35 40 45
PM2.5 [ug/m3]
GA
AL
BA
NO
OV
IP
HU
PS
GN
PA
TU
VE

Several sensitivity analyses were conducted based on a priori arguments for potential influences on the
results. However, adjustment for various respiratory symptoms prevalent in the two weeks before the
lung function tests, exclusion of Tartu, or inclusion of measurements that did not comply with the ATS
criteria did not affect our findings. The observed significant associations for change in FEV1 among
men across centers using Biomedin spirometers (or across Southern centers) were within the magnitude
of variation across centers with very similar pollution levels (see Figure 11 and Figure 12), and the
statistical significance of these results is strongly driven by the very high PM
2.5
levels in the cluster of
three Italian centers located in the River Po basin, as can be seen in the scatter plots.

99

Figure 12 Scatterplot of sub-analysis for change in FEV1 in men across Southern European centers.
15 20 25 30 35 40 45
PM2.5 [ug/m3]
-5
0
5
Adj. change in FEV1 [ml/a]
GA
AL
BA
OV
HU
GN
PA
TU
VE
15 20 25 30 35 40 45
PM2.5 [ug/m3]
-5
0
5
Adj. change in FEV1 [ml/a]
GA
AL
BA
OV
HU
GN
PA
TU
VE
GA
AL
BA
OV
HU
GN
PA
TU
VE

As in every longitudinal study, loss to follow-up is of concern. For example the drop-out rate was high
in an early study on lung function in California (Tashkin et al. 1994), and findings of a recent study
among Japanese women appear to be affected at least partly by loss to follow-up (Sekine et al. 2004).
Follow-up rates differed considerably across the ECRHS centers (range 43-85%). Systematic differences
between subjects that drop out of the study and those who come back for the follow-up may lead to over
or underestimation of center mean lung function estimates. If in addition, such misclassification is
correlated with air pollution it will distort the air pollution lung function relationship. A comparison
between subjects lost to follow-up and those followed did not reveal any major differences in key
characteristics at baseline (see Table 11). To assess the impact of loss to follow-up on the relation
between PM
2.5
and lung function, the crude regression slopes between PM
2.5
and lung function at

100

baseline were compared between subjects who were lost to follow-up and those included in the final
analysis. This analysis showed that a potential negative association of PM
2.5
with FEV1 or FVC was not
under estimated (attenuated) by more than 30ml per 10µg/m
3
, and for most analyses, by far less (see
Table 19). To put these numbers in perspective it is helpful to compare them to the magnitude of effects
observed by other studies. The smallest significant effects reported by SAPALDIA were 57ml
FEV1/10µgm
-3
PM
10
and 151ml FVC/10µgm
-3
PM
10
, and effect estimates from SALIA were even about
three times larger. Based on these numbers it is unlikely that the null findings observed in ECRHS could
be explained by selection bias due to loss to follow-up. However, the described assessment of loss to
follow-up cannot address a potentially differential response to the baseline investigation, nor can it
estimate potentially differential developments in lung function during the follow-up period.
Characterization of Exposure
ECRHS was the first standardized trans-European PM
2.5
measurement campaign (Hazenkamp-von Arx
et al. 2003). Other studies on lung function, such as SAPALDIA (Ackermann-Liebrich et al. 1997) or
the Southern California Children’s Health Study (Gauderman et al. 2004) have successfully used central
monitors to assess exposure, similar to the method used in ECRHS. Many studies which showed effects
of air pollution on other outcomes also used central monitor based exposure assessment (e.g. Dockery et
al. 1993; Pope et al. 2002). Two other analyses of the ECRHS PM
2.5
data, on the other hand, also
resulted in null findings, namely for allergic sensitization (Bedada et al. 2007) and chronic bronchitis
(Sunyer et al. 2006).
Central Monitor Based Exposure Assessment
The central monitor based exposure assessment approach relies on the assumption that a single monitor
per city is sufficient to capture the contrasts in individuals’ exposure to background pollution across
cities. Thus, contrasts in exposure to background pollution between individuals within the same city are
assumed to be relatively small, compared to the contrasts across cities. Background pollution thereby

101

refers to a pollution mix which at least in theory is not immediately affected by local sources and
therefore shows low spatial within-city variability. Thus, a single monitor location is representative for
the pollution concentration prevailing in the entire city. Presumably small deviations in individuals’
exposures from the city mean concentration are assumed to average out over the whole study sample.
These assumptions are less likely to hold if subjects live farther away from the monitor, if the
topography varies within and/or between cities, or if local sources, such as traffic contribute
considerably to both, individual exposures and overall pollution levels. Although, misclassification of
individuals’ exposures per se does not bias air pollution effect estimates to the null due to a Berkson type
error structure, increased standard errors make it less likely to detect significant associations. Effect
estimates can be biased to the null, however, if a central air pollution estimate itself is biased, in other
words inaccurately reflects population average exposure.
Spatial Variability of Exposure Markers
Besides topographical and meteorological factors pollution concentrations across space are determined
by several pollutant specific factors, such as the spatial distribution of sources and physico-chemical
processes that affect the formation, dispersion, and degradation of pollutants. The validity of the central
monitor based exposure assessment approach directly depends on the spatial variability of the measured
pollutant, which in practice will never be perfectly homogeneous. The lower the spatial heterogeneity of
a pollutant, the more likely are individuals’ true exposures to scatter randomly around and deviate little
from the centrally measured value. Or in other words, the more likely is a central monitor to measure a
value close to the true population average exposure. On the other hand, increased spatial variability
increases the chance that a central monitoring location is not adequately representing the regional levels
of pollution, and the mobility of subjects, for example between their homes and workplaces, has a larger
influence on the cumulative exposure (Briggs 2005). This additional source of variability in exposure
further increases the chance for error in the estimates of exposure of heterogeneously distributed
pollutants.

102

Figure 13 Comparison of PM
2.5
and sulfur annual means across ECRHS study centers.
PM
2.5
concentrations are in µg/m
3
. Note: sulfur* concentrations are standardized so overall sulfur
average equals overall PM
2.5
average.

0 10 20 30 40 50
REUMUPGO AL TA OVNOGA ER IP HU BS PS GN AS BA AC PA VE TU
PM2.5 Sulfur*

µg/m
3

Although PM
2.5
has been shown to have relatively low spatial variability (Hoek et al. 2002b; Zhu et al.
2002a; Zhu et al. 2002b), this is even more the case for its sulfur content (Ebelt et al. 2000; Sarnat et al.
2000; Landis et al. 2001; Brunekreef et al. 2005). Sulfur content of PM
2.5
was available from elemental
analysis (see chapter 3 and Götschi et al. (2005)). The strongest differences in the distribution of sulfur
compared to PM
2.5
were observed in the three Italian centers, where monitors were located very close to
roads (<6m) and therefore PM
2.5
levels may have been somewhat overestimated (see Figure 13). The
exclusion of centers with monitors in close proximity of main roads, (Pavia, Turin, Verona, Antwerp
City), however, did not affect the PM
2.5
coefficients in our models (see Table 15 and Table 18).

103

Considerably increased sulfur levels, on the other hand, were observed in Galdakao and Huelva, Spain.
Overall, however, sulfur content of PM
2.5
and PM
2.5
mass concentration were highly correlated (Pearson
r = 0.87) and using sulfur instead of PM
2.5
in the lung function analysis yielded very similar results (see
Table 30 and following on pages 165-170 in the annex). Sensitivity analyses using considerably more
heterogeneously distributed exposure markers, such as light absorbance, NO
2
or various metals, did not
reveal any significant associations, either (see Table 36 in annex on page 171).
Diversity of Study Areas
Using central monitors, air pollution effects can only be assessed based on cross community contrasts,
which tend to be small across similarly urbanized regions. To compensate for the limited exposure
contrasts and yield sufficient statistical power comparisons across vast geographic areas, or comparisons
between urban and rural communities, as done in SAPALDIA and SALIA, are required. Increased
diversity of the study regions, however, poses a challenge to derive comparable exposure estimates and
increases the potential for confounding by contextual variables (see below). In the following, several
characteristics of the ECRHS study areas that influence the spatial distribution of air pollutants and
thereby affect the quality of centrally derived exposure estimates are discussed.
Most centers of ECRHS had at least 200,000 inhabitants, and some even had over a million inhabitants
(Barcelona, Spain: ~1.5 millions; Paris, France: ~10). Some centers consisted of relatively wide spread
areas leading to distances between subject’s homes and air pollution monitors which exceeded those of
other studies (Galdakao, Spain: 90% within 25km; Paris: 12km) (see Table 21). In contrast, in the
SALIA or the SAPALDIA study the vast majority of subjects lived less than 5km away from a monitor.
The largest study area of SAPALDIA included barely 200’000 inhabitants and six out of the eight
centers were far smaller. On the other hand, the American Cancer Society study showed effects of air
pollution across much larger study areas, albeit analyzing a different outcome in a different study design
(Pope et al. 2002). Differences in population density were extremely large between some ECRHS
centers (e.g. Reykjavik, Iceland: ~400/km
2
; Turin, Italy: ~7’000/km
2
; Barcelona, Spain: ~16’000/km
2
)

104

and contrasts in traffic and building density can be expected to mirror this wide range. All these factors
may have lead to considerable misclassification in the exposure estimates and thereby contributed to the
observed null-findings.
Table 21: Demographic and geographic characteristics of ECRHS centers.
Country
Language
Population [in thousands]
Populationdensity (/km2)
Average distance monitor subejcts homes [km]
Altitude (Air pollution monitor)
Latitude (N)
Longitude
Albacete Spain Spanish 160 140 1.8 704 38°59' 1°50'W
Antwerp City Belgium Dutch 461 2257 5.4 ~10 51°13' 4°24'E
Antwerp South Belgium Dutch 461 2257 5.8 ~30 51°13' 4°24'E
Barcelona Spain Catalan 1,673 15'969 5.1 24 41°23' 2°11'E
Basel Switzerland German 172 7560 u.k. 260 47°34' 7°36'E
Erfurt Germany German 203 753 6 220 50°59' 11°2'E
Galdakao Spain Basque 30 939 12.1 60 43°13' 2°50'W
Gothenburg Sweden Swedish 495 2491 6.1 30 57°42' 11°58'E
Grenoble France French 157 8456 6.9 220 45°11 5°43'E
Huelva Spain Spanish 145 ~1000 1.6 50 37°15' 6°57'W
Ipswich UK English 118 2998 4.3 50 52°02' 1°11'E
Norwich UK English 127 3270 u.k. ~20 52°40' 1°18'E
Oviedo Spain Spanish 212 1137 2.2 276 43°22' 5°50'W
Paris France French ~10,000 3670 9 75 48°52' 2°20'E
Pavia Italy Italian 71 1146 6 70 45°11' 9°09'E
Reykjavik Iceland Icelandic 191 420 u.k. 53 64°05' 21°53'W
Tartu Estonia Estonian 101 2611 u.k. 84 58°23' 26°43'E
Turin Italy Italian 902 6928 3.2 239 45°04' 7°40'E
Umea Sweden Swedish 110 ~50 4.3 10 63°50' 20°15'E
Uppsala Sweden Swedish 183 u.k. 14.6 8 59°51' 17°38'E
Verona Italy Italian 260 1254 4.8 60 45°26' 10°59'E

105

Annual Means as Estimates of Long Term Exposure
A limitation of the ECHRS exposure data is the lack of standardized air pollution measurements for the
years before follow-up, as pointed out in Chapter 3. Although annual means have been widely used as
surrogates for long term air pollution levels (e.g. Ackermann-Liebrich et al. 1997; Schikowski et al.
2005), the available historic air pollution data for ECRHS cities (Naef et al. 2000) indicate that current
annual means may not have represented past exposure equally well in all investigated regions of Europe.
The use of past air pollution data was considered for the ECRHS analysis, however, the lack of
standardized monitoring networks and limited data availability for ECRHS study centers precluded any
quantitative use of this information. In consequence, using measurements from 2000/2001 probably
introduced an unknown amount of misclassification of true long term exposure and may therefore have
contributed to the observed null-findings. In particular the longitudinal analysis of change in lung
function may have been affected by this limitation, however, reducing the analysis to a cross-sectional
comparison at follow-up, in other words only using the most recent lung function data with the
concurrent air pollution data from 2000/2001, did also not reveal any significant associations (data not
shown).
PM
2.5
as a Marker of Exposure
There are only a few commonly measured pollutants which fulfill the requirement of sufficient spatial
homogeneity to be assessed reliably at central monitors. Among the data available in ECRHS, PM
2.5
and
its sulfur content were deemed suitable, as pointed out in the conclusions of Chapter 3. However, PM
2.5
,
or even more so its sulfur content may not necessarily be the most health relevant constituents of urban
air pollution. It must also be kept in mind that PM
2.5
measures the mass concentration of a mixture of
particles of various sizes and composition, including hundreds of different chemical substances. Since
the mass of these particles only serves as a proxy for the actual, unknown causal agents of detrimental
effects, unmeasured variations in the latter across cities could be a source of measurement error, and
hence of bias towards the null. As shown in Chapter 3, PM
2.5
constituents varied substantially across

106

ECRHS centers, and correlations between constituents across cities varied as well (Götschi et al. 2005;
Kunzli et al. 2006). Sensitivity analysis using light absorption, NO
2
, transition metals (Cu, Fe, Mn, Zn),
or potential to form hydroxy-radicals (·OH) (Kunzli et al. 2006) as exposure variables did not yield any
significant associations with lung function, nor did the use of summer or winter means of PM
2.5
and
sulfur (see Table 36 in annex on page 171).
NO
2
at Home Outdoors as an Individual Marker of Exposure
The air pollution protocol of ECRHS was planned in the early nineties at a time when increasing the
number of study centers and thereby the range of exposure to background pollution was considered a
promising design. Since then, however, several studies found associations between markers that reflect
within-community contrasts in exposure to traffic related pollution and various respiratory health
outcomes including mortality, many of which were conducted in European cities (van Vliet et al. 1997;
Ciccone et al. 1998; Venn et al. 2001; Brauer et al. 2002; Hoek et al. 2002a; Nicolai et al. 2003; Sekine
et al. 2004; Schikowski et al. 2005; Gauderman et al. 2007). Most of these studies use proximity of
subjects’ homes to traffic or residential nitrogen dioxide (NO
2
) measurements (Braun et al. 1992;
Schindler et al. 1998; Kramer et al. 2000), all of which capture within-city variability. These exposure
measures may reflect a different, more toxic type of pollution than PM
2.5
, in particular fresh tailpipe
emissions such as for example ultrafine particles (Nel 2005). NO
2
measured outdoors is generally
considered an indicator of traffic related air pollution (Lebret et al. 2000; Reungoat et al. 2003).

107

Figure 14 Comparison of central monitor based PM
2.5
and NO
2
annual means, and NO
2
2-week
measurements at participants’ homes outdoors. (All measures in µg/m
3
)
AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS AS
AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC AC
ER
BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA BA
GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA GA
AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL AL
OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV OV
HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU HU
GN
PS
PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA PA
TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU TU
VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE VE
IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP IP NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO NO
RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE RE
GO
UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM UM
UP
BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS BS
TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA TA
0
50
100
150
RE
UM
UP
GO
AL
TA
OV
NO
GA
ER
IP
HU
BS
PS
GN
AS
BA
AC
PA
VE
TU
.
PM2.5 NO2 central
NO2 at home outdoors

µg/m
3

Within ECRHS, in addition to central monitor based NO
2
measurements, outdoor residential NO
2
was
measured in a sub-sample of participants (N=2011) from 16 study centers. NO
2
was measured at homes
of subjects who agreed to a home visit for the purpose of taking a dust sample of their mattress. Passive
samplers were deployed indoors and outdoors for a period of two weeks. Measurements were repeated
for 813 subjects. Measurements did not follow a strict schedule and deployment of samplers was
handled differently across centers, although repeated measurements were attempted to be conducted
approximately six months after the first measurement, where feasible. No effect of NO
2
on lung function
could be detected (see Table 36 in annex on page 171). Insufficient precision of these measurements as
long term exposure estimates is a probable explanation for null findings. Evaluations of NO
2
time-series

108

data from central monitors showed that one or two two-week periods poorly reflect long term (annual
mean) concentrations of NO
2
. An approach presented by Hoek et al. (2002b) was used to adjust the
residential short term measurements of NO
2
for temporal variability throughout the year. In brief, the
ratio between each residential 2-week measurement and the corresponding 2-week concentration from a
continuous monitor were calculated. The annual mean from the continuous monitor was then multiplied
with the ratio and used as the seasonally adjusted long term NO
2
estimate. The analogous procedure was
applied using the difference between the residential concentration and the continuous monitor
concentration, instead of the ratio. An evaluation of this approach using NO
2
time series from 40 EMEP
(Co-operative Programme for Monitoring and Evaluation of the Long-range Transmission of Air
pollutants in Europe) monitors showed, however, that seasonal adjustment based on continuous monitor
data does often not improve the precision of NO
2
estimates (Götschi et al. 2006). Consequently, the use
of seasonally adjusted NO
2
estimates did also not reveal any significant associations with lung function.
In contrast, Sunyer et al. (2006) observed significant associations between these same NO
2

measurements and self-reported chronic phlegm among female participants of ECRHS, but not among
males. Prevalence of phlegm might be more strongly affected by traffic related air pollution than lung
function. An alternative explanation could be a stronger contrast in phlegm, compared to lung function,
leading to larger statistical power to detect a significant effect and therefore larger tolerance for
measurement error in residential NO
2
. On the other hand, it cannot be excluded that potential
confounders may have affected the self-reported symptoms differently than lung function. Consistent
with the null-findings for lung function, central monitor based estimates of PM
2.5
or sulfur were not
associated with phlegm or other respiratory symptoms.
In either analysis, residential NO
2
measurements clearly underscore the inherent limitation of the
assumption that the relevant exposure contrasts could be estimated by the average pollution level of the
community of residence. As can be seen from Figure 14, within-community contrasts for NO
2
are likely
to exceed between-community contrasts observed for PM
2.5
. Hence, future efforts will be directed
towards estimating individual exposure levels using exposure modeling approaches.

109

Confounding
ECRHS is well suited to adjust for potential confounders on the individual level, such as physiological
characteristics, tobacco smoke and other exposures, socio-economic variables, and various measures of
respiratory health. Adjustment for some variables of short term health status turned out to have a small
impact on measured lung function levels. Height was by far the strongest predictor of lung function.
The studied population is considerably heterogeneous across centers with regard to basic physiological
characteristics, such as height or body mass index (BMI) (See Error! Reference source not found.).
When allowing for center-specific effects of these variables, coefficients across centers varied
considerably (e.g. coefficients for height predicting FEV1 level ranged from 29 ml/cm to 40 ml/cm) but
they did not have a significant impact on the PM
2.5
coefficients (Table 14 and Table 17). The fact that
after adjustment for height, the analyses did not show any major sensitivity towards various degrees of
further adjustment on the individual level indicates that residual confounding on the individual level is
probably small.
Since air pollution was assessed on the center level, the distribution of PM
2.5
, lung function, and
potential confounders across centers is of particular relevance. As can be seen from Table 29 in the
annex on page 161, air pollution concentrations and lung function levels roughly follow a north to south
gradient across Europe, with the highest levels of air pollution in northern Italy (Turin, Pavia, Verona)
and Spain (Barcelona), and bigger lungs in northern Europe. This crude negative association is
confounded by height, as can be seen from Table 13, due to the fact that people in northern Europe are
taller (Mean difference approx. 5cm). The relatively high negative correlation between height and PM
2.5

across centers (Pearson r=-0.46, see Table 29) makes it inherently difficult to disentangle a potential,
small effect of PM
2.5
from the strong effect of height on lung function. The concern that over-adjustment
for height may have diminished a potential air pollution effect was addressed by evaluating numerous
approaches of adjustment for height, as described in the methods section. These approaches consider a
variety of possible relationships between height and lung function, including several non-linear ones,

110

absolute and relative ones, and stratified analyses by height and geographical reason. The null findings
were not sensitive to the way height was adjusted for. It can therefore be concluded that neither residual
confounding by height nor over-adjustment for height did cause the null findings.
Identifying, quantifying, and controlling for other factors that might act as confounders or effect
modifiers on the center level is challenging. As can be seen from Table 29, several other co-factors
showed high correlations with latitude (North to south gradient). Center means of SES (measured as the
proportion of participants having a manual profession), level of education (measured as age at the end of
education), and exposure to environmental tobacco smoke showed correlations with PM2.5 of similar
magnitude as height; however, the influence of these factors on lung function is smaller than that of
height, hence the effects of adjustment on PM2.5 coefficients were small. Table 22 illustrates the
diversity in socio-economic parameters across the ECRHS countries. As can be seen from the last row of
Table 22, several factors, such as smoking habits in men, education in women, or proportion of diesel
powered vehicles, correlate considerably with latitude.

111

Table 22: Country specific socio-economic parameters that illustrate the diversity within ECRHS with regards to potential center level confounders. Countries are
sorted by latitude (North to south). Pearson correlation with latitude is shown in the last row. Data source: (UNECE 2003)
Country
Regular smokers 15+ women [%]
Regular smokers 15+ men [%]
GDP per capita 2001 [US$]
Women with tertiary education [%]
Men with tertiary education [%]
Total health expenditure in % GDP
Life expectancy women [y]
Life expectancy men [y]
Unleaded petrol [%]*
Leaded petrol [%]*
Diesel [%]*
Meters of motorways/km2
Number of passenger cars/1000 pers.
Number of trucks/1000 pers.
New cars registered/y*1000 pers.
Iceland 22.8 24.5 29294 22.8 23.0 9.3 81.4 77.6 75.0 n.a. 25.0 u.k. 561 68 27
Sweden 19.9 17.9 26157 26.9 21.7 7.9 82.1 77.5 67.8 n.a. 32.2 3 455 44 33
Estonia 25.7 50.6 10049 32.5 22.8 6 76.1 65.2 58.4 0.2 41.6 2 299 59 19
UK 25.0 29.0 25894 27.4 29.9 7.3 79.7 75.0 53.7 2.3 44 15 422 50 45
Belgium 26.0 36.0 27746 22.0 24.3 8.7 81.4 75.1 51.8 n.a. 48.2 57 462 51 49
Germany 22.3 34.7 26428 15.1 27.7 10.6 80.8 74.8 52.3 n.a. 47.7 33 534 32 41
France 21.0 33.0 26820 18.4 19.5 9.5 83.0 75.5 28.3 7.8 63.9 15 483 89 38
Switzerland 24.4 36.3 30329 12.9 32.5 10.7 82.8 77.2 75.3 n.a. 24.7 32 506 39 44
Italy 17.3 32.8 26357 8.4 9.4 8.1 82.9 76.7 28.7 27.8 43.5 21 557 56 40
Spain 27.2 44.9 21577 18.7 21.3 7.7 82.9 75.6 22.8 4.9 72.3 19 437 99 38
Pearson r
latitude
-0.08 -0.44 -0.11 0.66 0.14 -0.19 -0.50 -0.17 0.73 -0.46 -0.66 -0.36 -0.13 -0.31 -0.60

* Fuel data for Italy is from 1997, for the remaining countries from 2000 or 2001

112

Diet may have played a role as an uncontrolled confounder or effect-modifier. Several studies found an
effect of diet on lung function, namely for fruits and vegetables, vitamin C, E and A, carotenoids, fatty
acids, and some minerals (Romieu 2005). The effect of a Mediterranean diet, rich in olive oil and
therefore mono-unsaturated fatty acids, could be of relevance in the cross European context of ECRHS.
Mono-unsaturated fatty acids appear to have a protective effect against air pollution effects on lung
development in the Southern California Children’s Health study (Personal communication F. Gilliland
10/05/06). Since the largest contrasts in diet again are expected to occur between northern and southern
Europe, some degree of confounding cannot be excluded. Detailed individual dietary data would be
required to adjust for a protective effect of diet; however, dietary data were not available for this
analysis. It is unknown whether the protective effect of a Mediterranean diet would be sufficient to
entirely compensate for effects of air pollution on lung function of a magnitude as suggested by other
studies. Not in support of such a scenario are findings from the APHEA study on acute effects of air
pollution which showed stronger effects in the southern parts of Europe (Katsouyanni et al. 2001).
However, neither the nature of the outcomes analyzed in short term studies (mortality, cardio-vascular
events, etc.), nor the effect measures determined (relative risks) may necessarily be directly comparable
to long term effects on lung function.
Meteorological factors have been associated with asthma in an Italian multi-center study, but may have
been confounded by differences in pollination, prevalence of mildew, or ozone levels (Zanolin et al.
2004). Meteorological contrasts throughout Europe may have affected respiratory health of participants
as well as the composition of the urban air pollution mixture (see Chapter 3 and (Götschi et al. 2005)),
but similar to height and other factors mentioned above, meteorological factors within ECRHS follow a
strong north to south gradient, which precludes the identification of their potential effects on lung
function (see Table 23). Finally, unknown genetic factors which influence lung function or the body’s
defense mechanisms against air pollution could act as confounders or effect modifiers as well
(McCunney 2005).

113

Table 23: Annual means of meteorological parameters for ECRHS centers.
Centers sorted by latitude (North to south). Pearson correlation with latitude is shown in the last row.
Country
Temperature
[°C]
Rain
[mm/a]
Humidity
[%]
Wind speed
[m/s]
Pressure
[hPa]
Reykjavik Iceland 5.7 n.a. 77.7 4.1 1007
Umea Sweden 5.0 951 82.8 3.3 1011
Uppsala Sweden 7.7 657 82.5 3.3 1011
Tartu Estonia 7.5 787 80.6 2.7 1005
Gothenburg Sweden 8.6 895 86.0 3.7 1012
Antwerp Belgium 12.2 999 77.4 3.5 1015
Erfurt Germany 9.9 543 79.5 n.a. 977
Paris France 13.5 891 71.8 n.a. 1015
Basel Switzerland 12.0 985 75.7 2.3 979
Grenoble France 12.7 1165 72.9 1.5 1017
Verona Italy 14.6 935 78.8 0.7 1009
Pavia Italy 14.9 866 80.7 1.2 1003
Turin Italy 15.2 845 68.5 n.a. 1015
Oviedo Spain 13.7 897 80.3 1.9 977
Galdakao Spain 15.5 885 76.1 2.5 1014
Barcelona Spain 16.5 469 67.9 4.2 968
Albacete Spain 16.6 288 61.8 2.3 940
Huelva Spain 19.0 643 70.5 2.9 1016
-0.97 0.29 0.68 0.46 0.38 Pearson r latitude

Data not available for Norwich and Ipswich.
Center specific differences in unknown or unmeasured factors affecting lung function are particularly
problematic in small studies (in particular across two centers only). Compared to other studies, the
number of centers in ECRHS is relatively large. The distribution of air pollution, however, is such that a
few centers with very low (e.g. Umeå, Uppsala, Reykjavik) and very high PM
2.5
levels (Pavia, Verona,
Turin) are most influential. Uncontrolled for differences in lung function predictors between these few
centers from very different parts of Europe may have confounded the results. In contrast, SAPALDIA
and SALIA investigated relatively homogeneous populations, recruited from within the same country,
and from areas not larger than 300km in diameter. As pointed out earlier, excluding the three ECHRS
centers with the lowest PM
2.5
levels (Reykjavik, Iceland; Umeå, Uppsala, Sweden), and the three
northern Italian centers (highest levels) would also dramatically reduce statistical power.
Stratification by numerous potential confounding factors did not alter the findings (see Table 15 and
Table 18). Since it is difficult to assess the degree of residual confounding after stratification,

114

comparisons across more homogeneous populations or even within communities would be clearly
preferable to reduce the chance of confounding.
Modeling Approach
When estimating effects on lung function level, the applied model took into account lung function
measurements at baseline and follow-up, rather than providing a separate cross-sectional estimate for
each survey. This may limit the comparability of the results to purely cross-sectional studies, such as
SAPALDIA and others. Further, the PM
2.5
concentrations of 2000/01 may in some cities have been poor
proxies for the exposure relevant for the baseline lung function. To test whether the data analysis
approach chosen was the reason for not being able to reproduce the results of other studies, various
cross-sectional models were investigated. Ackermann-Liebrich et al. (1997) reported a 3% difference in
FVC per 10µg/m
3
increase in PM
10
among 18 to 60 year old adults from eight Swiss communities.
When using data at follow-up with the same model specifications as SAPALDIA the ECRHS sample
provided slightly positive estimates (<1%) which were not statistically significant (see Table 37 in annex
on page 173).
Gauderman et al. (2004) observed strong associations between PM
2.5
and the proportion of children with
lung function of less than 80% of the predicted value, a common measure for clinically low lung
function. When applying the same approach to the ECRHS sample, odds ratios for PM
2.5
indicated
decreased probability for low lung function with increasing PM
2.5
although not statistically significant,
except for low FVC in men at follow-up (data not shown). Numerous other variations of the modeling
approach have been pursued, mainly with a focus on the influence of height and age (see methods
section). All of these calculations confirmed that the observed null-findings were robust towards varying
the analytical approach.

115

Conclusions
Urban background air pollution measured as PM
2.5
was not associated with lung function level or change
in lung function in this large longitudinal study across 21 European study centers. Lack of statistical
power was a limitation of the analysis of effects on change in lung function, but not for the analysis of
effects on lung function levels. The lack of exposure characterization on the individual level probably
led to considerable exposure misclassification in this comparison across urban areas, given the large
spatial variability of predominantly traffic-related pollution within European cities. Moreover, the
heterogeneity of the studied populations and their living environments, and the correlation of air
pollution, lung function, and various potential confounders along a common north-south gradient within
Europe quite likely decreased the ability to identify potential effects of air pollution on lung function. It
is therefore concluded that the presented null findings should not be interpreted as a null effect of air
pollution on lung function in adults, and that the particular question of possible long-term effects of air
pollution on decline in lung function in adults remains unresolved, as previously pointed out by
Sunyer.(2001)
Recent developments of spatial analysis technologies (Geographic information systems (GIS), modeling
capacities, etc.) (Jerrett et al. 2005) offer promising tools to (retrospectively) derive individual exposure
estimates for spatially heterogeneous pollutants, such as traffic exhaust. Availability of individual
exposure estimates would also provide a significant increase in statistical power. Future analyses of
ECRHS health data could therefore offer the unique opportunity to investigate long-term effects of
traffic-related pollution, as well as the modifying role of local or regional factors across Europe, as done
previously for acute air pollution effects (Samoli et al. 2006).

116

Chapter 5: Synthesis and Outlook
The study of air pollution effects on lung function has received a great amount of attention over the past
two to three decades. More than fifty studies on long term effects of air pollution on lung function have
been published since the late 1980’ies. Overall, there is a clear predominance of studies reporting
adverse effects of air pollution on lung function, while only nine publications besides ECRHS reported
null findings. Despite this considerable number of publications the scientific basis for several key
questions related to the topic remains weak. This is partly due to the fact that a number of studies are
small and of limited quality, namely those comparing air pollution and lung function across only two or
three communities. In addition, the diversity in study specific aspects ranging from the studied
populations, the methods applied, and the results reported is of such magnitude that many findings can
barely be compared to those of more than one or two other studies.
Significant gaps in knowledge remain on the temporal sequence of exposure to air pollution and the
thereof resulting effects on the lung. The relevance of exposures to air pollution in early life is unknown.
Due to the limited feasibility of spirometry for epidemiologic studies in very young children there are no
studies published investigating children younger than five, and although theoretically possible, no study
on lung function has taken on the challenges of (retrospectively) assessing exposures in early life
(Gilliland et al. 2005). It could therefore be possible that effects on lung growth observed in teenagers at
least partially reflect consequences from exposures that occurred much earlier in life. The study by Avol
et al. (2001) is the only direct evidence for effects of concurrent exposures, showing effects on lung
function growth in teenagers only a few years after the occurrence of significant changes in air pollution
levels. Although not replicated by any other major study, the findings on lung function growth form the
Children’s Health Study which followed teenagers until age 18 are among the strongest evidence for
adverse effects of air pollution on lung function. However, as of today neither this nor any other study
has followed subjects throughout the plateau phase. It therefore remains to be seen whether reduced
growth rates will be compensated for by a prolonged growth phase or whether they will lead to

117

permanent lung function deficits. The latter scenario appears more likely in the light of the findings from
several cross-sectional studies in adults that associated lung function deficits with air pollution, such as
SAPALDIA (Ackermann-Liebrich et al. 1997) and SALIA (Schikowski et al. 2005). Yet it remains
unknown to what extend adverse effects of air pollution on lung function decline during adulthood
contribute to such differences, or whether they entirely reflect the deficits obtained during adolescent
growth.
Considerable efforts have been conducted to identify specific pollutants that could be more harmful for
lung health than others. Large multi-center studies typically measured several pollutants at fixed site
monitors, while more recently several studies used exposure surrogates for traffic related pollution. The
findings are not consistent enough to allow major conclusions on the culprit agents. Numerous studies
including ECRHS demonstrate that a central monitor based exposure assessment approach is not suited
to identify pollutant specific effects because cross-community correlations of all major pollutants are
generally high, while at the same time, their within-community variability can be substantial. In addition,
the selection of pollutants that can be assessed by such monitors, namely those with relatively low
spatial variability, is fairly limited. As such, fine particles are the most persistent measure associated
with adverse effects of lung function. Studies addressing traffic pollution shift the paradigm from
identifying a harmful pollutant to identifying a harmful source of pollution. Based on the available lung
function studies it would be premature to conclude on the effects of traffic on lung health, but there are
an increasing number of findings that highlight the importance of further investigating this source.
The null-findings observed in ECRHS stand in contrast to the majority of previously reported results.
ECRHS was designed to overcome significant limitations of earlier studies and to address some of the
gaps of knowledge mentioned above. Most notably, ECRHS is the largest longitudinal study
investigating air pollution effects on lung function in adults. The extent and standardization of the
assessment of fine particulate matter and its characteristics is unprecedented in Europe. The large
number of study centers spread across wide parts of Europe assured large exposure contrasts and at the
same time limited the influence of single centers to a minimum. Consequently, the observed contrasts in

118

exposure and lung function levels were sufficient to potentially detect significant cross-sectional
associations. However, the combination of limited contrasts in both population mean declines in lung
function and centrally measured urban background pollution levels provided insufficient power to
observe significant effects on change in lung function across the 21 centers.
Independent of the lack of power to detect effects on decline in lung function, however, it retrospectively
becomes clear that what initially was considered this study’s major advantage, namely its size and
geographical spread, at the same time was a significant pitfall and burdened the analysis with several
sources of misclassification.
Ideally, epidemiologic studies are designed to maximize variation in observed exposure and outcome,
while keeping variation in co-factors (related to exposure or outcome) to a minimum. Some amount of
unavoidable variation in co-factors will always remain, and to the extent possible this is measured and
adjusted for statistically. Cross-community comparisons such as ECRHS face the particular challenge
that such co-factors may be center specific. In other words they affect all records from a single center
systematically, and thereby can influence the overall analysis considerably. In studies across a large
enough number of communities such sources of misclassifications are assumed to occur at random. In
ECRHS an excessive amount of such random misclassification may have significantly reduced the
capability to detect a potential association between lung function and air pollution. ECRHS faces the
challenge that due to the diversity in participating centers there are a considerable number of sources of
center specific misclassification.
The air pollution assessment in cross-community comparisons and in the case of ECRHS especially is
most susceptible to center specific misclassification because measurements rely on a single (in the case
of ECRHS) or very few monitors (in some other studies). The location of the monitor, the distribution of
pollution sources and participants’ homes in relation to the monitor, the composition of the air pollution
mix in general and fine particulate matter in particular, the way pollutants distribute throughout the study
areas, and the long term trends of pollution levels are all factors which influence the accuracy of central
monitor based exposure estimates, and which cannot not be standardized or measured easily, if at all.

119

Similarly, lung function measurements are conducted with a local spirometer in each center,
administrated by a local team. Since a systematic comparison of the spirometers used in ECRHS was not
feasible and would have come at a prohibitive cost, potential center specific misclassification in lung
function measurements cannot be quantified. Precision standards for spirometers, standardized
procedures for conducting spirometry, and the clinical setting of these tests, however, justify the
assumption that the validity of these measurements was sufficient.
The diversity across the study centers and populations of ECRHS in socio-economic, cultural,
meteorological, topographical, nutritional, physiological and several other co-factors inherently evokes
the concern of confounding. The statistical analysis assured that the findings were not confounded by the
most relevant predictors of lung function, namely height, age, gender, and smoking, and several other
individual and a few community level confounders. However, some uncertainty remains about whether
or to what degree center specific differences in unmeasured co-factors, in particular those following a
north-south gradient, may have confounded the relationship between air pollution and center mean lung
function towards the null.
Although the possibility of a true null-effect of air pollution on the various lung function measures
cannot be entirely excluded, it seems quite likely that the accumulation of several yet small sources of
random misclassification could be a sufficient explanation for the observed null findings. Although it is
impossible to quantify exactly how much each one of the above mentioned sources of center specific
misclassification contributed, it is apparent that the central monitor based exposure assessment approach
provides ample potential for misclassification of exposure estimates. Further, it is not suitable to capture
major parts of the exposure contrasts between subjects who live in urban environments. And in addition,
the fact that only a single exposure estimate was available for each center made the analysis prone to
center level confounding.
Fortunately enough, the exposure assessment also is a promising target for retrospective improvements
of the analysis. Substituting central monitor based ecological exposure assessment with individually
assigned exposure promises to capture a significantly larger exposure range. Such within-community

120

exposure contrasts are likely to reflect more health relevant constituents of urban air pollution, such as
traffic emitted ultrafine particles which are considered potentially most dangerous (Nel 2005). At the
same time individual exposure assignment would almost entirely eliminate the potential for center level
confounding. Shifting the analysis to an individual level would also considerably increase the statistical
power due to increased exposure contrasts and taking into account the larger number of subjects, as
compared to 21 centers only.
Recently developed spatial analysis techniques based on geographic information systems (GIS) and
smart interpolation models provide the opportunity to retrospectively assess air pollution levels with
relatively high spatial resolution. These models, such as for example the European APMoSPHERE
project (APMoSPHERE 2007) are not only based on monitored air pollution data, but in addition
consider emission inventories, land use information, and transport network data, among others. Using
land use information to determine local pollution concentrations may be particularly interesting for the
assessment of past exposure levels, since such data may be more widely available than actual pollution
measurements. It yet remains to be seen how successful such models will be at predicting relevant
exposure contrasts for the study areas of ECRHS, since exposure contrasts in urban environments can
vary dramatically on very small spatial scales, and relationships between pollution sources and actual
pollution concentrations are complex. Nonetheless, the perspective of having individual level exposure
estimates available to reanalyze the ECRHS health data at some point in the future is promising.
New studies, on the other hand, may choose not to rely on modeled exposure estimates alone but in
addition conduct personal, residential, or other types of within-community measurements. Shifting
exposure assessment to the individual level will also warrant increasing efforts to gather questionnaire
based information related to individual’s exposures, such as their residential history, time-space activity
patterns, in particular work location and commute time, and characteristics of their residential and work
environment.
In summary, both the analysis of ECRHS and the review of the published literature mandate that future
epidemiologic studies aim at comparisons between individuals within communities, thereby capturing

121

large, potentially health relevant exposure contrasts without running the risk of confounding by
community-level co-factors. This methodological necessity is strongly supported by the recently
achieved feasibility of spatial analyses technologies, and offers new opportunities to clarify the health
relevance of traffic and other major sources of urban air pollution.

122

References
Abbey, D. E., R. J. Burchette, S. F. Knutsen, W. F. McDonnell, M. D. Lebowitz and P. L. Enright
(1998). "Long-term particulate and other air pollutants and lung function in nonsmokers."
Ackermann-Liebrich, U., P. Leuenberger, J. Schwartz, C. Schindler, C. Monn, G. Bolognini, J. P.
Bongard, O. Brandli, G. Domenighetti, S. Elsasser, L. Grize, W. Karrer, R. Keller, H. Keller-
Wossidlo, N. Kunzli, B. W. Martin, T. C. Medici, A. P. Perruchoud, M. H. Schoni, J. M.
Tschopp, B. Villiger, B. Wuthrich, J. P. Zellweger and E. Zemp (1997). "Lung function and
long term exposure to air pollutants in Switzerland. Study on Air Pollution and Lung Diseases
in Adults (SAPALDIA) Team." American Journal of Respiratory and Critical Care Medicine
155(1): 122-9.
Alastuey, A., X. Querol, S. Rodriguez, F. Plana, A. Lopez-Soler, C. Ruiz and E. Mantilla (2004).
"Monitoring of atmospheric particulate matter around sources of secondary inorganic aerosol."
Atmospheric Environment 38(30): 4979-4992.
American Thoracic Society (1985). "Guidelines as to what constitutes an adverse respiratory health
effect, with special reference to epidemiologic studies of air pollution." Am. Rev. Respir. Dis.
131: 666-669.
American Thoracic Society (1995). "Standardisation of spirometry." Am J Respir Crit Care Med 152:
1122.
American Thoracic Society (2000). "What constitutes an adverse health effect of air pollution?" Am. J.
Respir. Crit. Care Med. 161: 665-673.
Andrews, J. E., P. Brimblecombe, T. D. Jickells and P. S. Liss (1996). An introduction to environmental
chemistry. Oxford, Blackwell Science.
Anthonisen, N. R., J. E. Connett and R. P. Murray (2002). "Smoking and Lung Function of Lung Health
Study Participants after 11 Years." Am. J. Respir. Crit. Care Med. 166(5): 675-679.
APMoSPHERE. (2007). Retrieved April 27 2007, 2007, from http://www.apmosphere.org/.
Avol, E. L., W. J. Gauderman, S. M. Tan, S. J. London and J. M. Peters (2001). "Respiratory effects of
relocating to areas of differing air pollution levels." American Journal of Respiratory & Critical
Care Medicine 164(11): 2067-72.

123

Bayer-Oglesby, L., C. Schindler, M. E. Hazenkamp-von Arx, C. Braun-Fahrlander, D. Keidel, R. Rapp,
N. Kunzli, O. Braendli, L. Burdet, L. J. Sally Liu, P. Leuenberger and U. Ackermann-Liebrich
(2006). "Living near main streets and respiratory symptoms in adults: the Swiss Cohort Study
on Air Pollution and Lung Diseases in Adults." Am J Epidemiol 164(12): 1190-8.
Bedada, G. B., J. Heinrich, T. Gotschi, S. H. Downs, B. Forsberg, D. Jarvis, C. Luczynska, A. Soon, J.
Sunyer, K. Toren and N. Kunzli (2007). "Urban background particulate matter and allergic
sensitization in adults of ECRHS II." International Journal of Hygiene and Environmental
Health In Press, Corrected Proof.
Borsboom, Gerard J. J. M., W. van Pelt, Hans C. van Houwelingen, Ben G. van Vianen, Jan P. Schouten
and Philip H. Quanjer (1999). Diurnal Variation in Lung Function in Subgroups from Two
Dutch Populations . Consequences for Longitudinal Analysis. 159: 1163-1171.
Brauer, M., G. Hoek, P. Van Vliet, K. Meliefste, P. H. Fischer, A. Wijga, L. P. Koopman, H. J. Neijens,
J. Gerritsen, M. Kerkhof, J. Heinrich, T. Bellander and B. Brunekreef (2002). "Air pollution
from traffic and the development of respiratory infections and asthmatic and allergic symptoms
in children." Am J Respir Crit Care Med 166(8): 1092-8.
Braun, C., U. Ackermann-Liebrich, J. Schwartz, H. Gnehm, M. Rutishauser and W. HU (1992). "Air
pollution and respiratory symptoms in preschool children." Am Rev Respir Dis 145(1): 42-47.
Briggs, D. (2005). "The role of GIS: coping with space (and time) in air pollution exposure assessment."
J Toxicol Environ Health A 68(13-14): 1243-61.
Brunekreef, B. (2002). "Particulate matter and lung function growth in children: a 3-yr follow-up study
in Austrian schoolchildren.[comment]." European Respiratory Journal 20(5): 1354-5.
Brunekreef, B., D. W. Dockery and M. Krzyzanowski (1995). "Epidemiologic studies on short-term
effects of low levels of major ambient air pollution components." Environ Health Perspect 103
Suppl 2: 3-13.
Brunekreef, B. and S. T. Holgate (2002). "Air pollution and health." Lancet. 360(9341): 1233-42.
Brunekreef, B., N. A. Janssen, J. de Hartog, H. Harssema, M. Knape and P. van Vliet (1997). "Air
pollution from truck traffic and lung function in children living near motorways."
Epidemiology 8(3): 298-303.

124

Brunekreef, B., N. A. H. Janssen, J. J. de Hartog, M. Oldenwening, K. Meliefste, G. Hoek, T. Lanki, K.
L. Timonen, M. Vallius, J. Pekkanen and R. Van Grieken (2005). Personal, Indoor, and
Outdoor Exposures to PM2.5 and Its Components for Groups of Cardiovascular Patients in
Amsterdam and Helsinki. Boston MA, Health Effects Institute.
Burney, P., C. Luczynska, S. Chinn and D. Jarvis (1994). "The European Community Respiratory
Health Survey." Eur Respir J 7(5): 954-960.
Chestnut, L. G., J. Schwartz, D. A. Savitz and C. M. Burchfiel (1991). "Pulmonary function and ambient
particulate matter: epidemiological evidence from NHANES I." Archives of Environmental
Health 46(3): 135-44.
Chiaradia, M. and F. Cupelin (2000). "Behaviour of airborne lead and temporal variations of its source
effects in Geneva (Switzerland): comparison of anthropogenic versus natural processes."
Atmospheric Environment 34(6): 959-971.
Chinn, S., P. Burney, D. Jarvis and C. Luczynska (1997). "Variation in bronchial responsiveness in the
European Community Respiratory Health Survey (ECRHS)." Eur Respir J 10(11): 2495-501.
Chinn, S., T. Gislason, T. Aspelund and V. Gudnason (2006). "Optimum expression of adult lung
function based on all-cause mortality: Results from the Reykjavik study." Respir Med.
Chinn, S., D. Jarvis, R. Melotti, C. Luczynska, U. Ackermann-Liebrich, J. M. Anto, I. Cerveri, R. de
Marco, T. Gislason, J. Heinrich, C. Janson, N. Kunzli, B. Leynaert, F. Neukirch, J. Schouten, J.
Sunyer, C. Svanes, P. Vermeire, M. Wjst and P. Burney (2005). "Smoking cessation, lung
function, and weight gain: a follow-up study." Lancet 365(9471): 1629-35; discussion 1600-1.
Churg, A., M. Brauer, M. del Carmen Avila-Casado, T. I. Fortoul and J. L. Wright (2003). "Chronic
exposure to high levels of particulate air pollution and small airway remodeling."
Environmental Health Perspectives 111(5): 714-8.
Ciccone, G., F. Forastiere, N. Agabiti, A. Biggeri, L. Bisanti, E. Chellini, G. Corbo, V. Dell'Orco, P.
Dalmasso, T. F. Volante, C. Galassi, S. Piffer, E. Renzoni, F. Rusconi, P. Sestini and G. Viegi
(1998). "Road traffic and adverse respiratory effects in children. SIDRIA Collaborative
Group." Occup Environ Med 55(11): 771-8.
Cyrys, J., M. Pitz, M. E. Arx, N. Kunzli and J. Heinrich (2006). "Evaluation of A Sampling Strategy for
Estimation of Long-term Pm(2.5) Exposure for Epidemiological Studies." Environ Monit
Assess.

125

de Miguel, E., J. F. Llamas, E. Chacon, T. Berg, S. Larssen, O. Royset and M. Vadset (1997). "Origin
and patterns of distribution of trace elements in street dust: unleaded petrol and urban lead."
Atmos. Environ. 31(17): 2733-2740.
Detels, R., D. P. Tashkin, J. W. Sayre, S. N. Rokaw, F. J. Massey, Jr., A. H. Coulson and D. H. Wegman
(1991). "The UCLA population studies of CORD: X. A cohort study of changes in respiratory
function associated with chronic exposure to SOx, NOx, and hydrocarbons." American Journal
of Public Health 81(3): 350-9.
Devereux, G., T. Ayatollahi, R. Ward, C. Bromly, S. J. Bourke, S. C. Stenton and D. J. Hendrick (1996).
"Asthma, airways responsiveness and air pollution in two contrasting districts of northern
England."
Diez Roux, A. V. (2004). "The study of group-level factors in epidemiology: rethinking variables, study
designs, and analytical approaches." Epidemiol Rev 26: 104-11.
Dockery, D., F. Speizer, B. J. Ferris, J. Ware, T. Louis and A. Spiro III (1988). "Cumulative and
reversible effects of lifetime smoking on simple tests of lung function in adults." Am Rev
Respir Dis 137: 286-292.
Dockery, D. W. (2001). "Epidemiologic evidence of cardiovascular effects of particulate air pollution."
Environmental Health Perspectives 4: 483-6.
Dockery, D. W. and B. Brunekreef (1996). "Longitudinal studies of air pollution effects on lung
function." American Journal of Respiratory & Critical Care Medicine 154(6 Pt 2).
Dockery, D. W., C. A. r. Pope, X. Xu, J. D. Spengler, J. H. Ware, M. E. Fay, B. G. J. Ferris and F. E.
Speizer (1993). "An association between air pollution and mortality in six U.S. cities." N Engl J
Med 329(24): 1753-9.
Dockery, D. W., F. E. Speizer, D. O. Stram, J. H. Ware, J. D. Spengler and B. G. Ferris, Jr. (1989).
"Effects of inhalable particles on respiratory health of children."
Ebelt, S. T., A. J. Petkau, S. Vedal, T. V. Fisher and M. Brauer (2000). "Exposure of chronic obstructive
pulmonary disease patients to particulate matter: relationships between personal and ambient
air concentrations." J Air Waste Manag Assoc 50(7): 1081-94.
ECRHS. (2007). "Lung function protocol." Retrieved April 27 2007, 2007, from
http://www.ecrhs.org/quests.htm.

126

European Respiratory Society (1993). "Standardized lung function testing. Official statement of the
European Respiratory Society." Eur Respir J Suppl 16: 1-100.
Forastiere, F., G. M. Corbo, R. Pistelli, P. Michelozzi, N. Agabiti, G. Brancato, G. Ciappi and C. A.
Perucci (1994). "Bronchial responsiveness in children living in areas with different air pollution
levels." Arch Environ Health 49(2): 111-8.
Frank, R., M. C. Liu, E. W. Spannhake, S. Mlynarek, K. Macri and G. G. Weinmann (2001). "Repetitive
ozone exposure of young adults: evidence of persistent small airway dysfunction." Am J Respir
Crit Care Med 164(7): 1253-60.
Frischer, T., M. Studnicka, C. Gartner, E. Tauber, F. Horak, A. Veiter, J. Spengler, J. Kuhr and R.
Urbanek (1999). "Lung function growth and ambient ozone: a three-year population study in
school children.[see comment]." American Journal of Respiratory & Critical Care Medicine
160(2): 390-6.
Fritz, G. J. and O. Herbarth (2001). "Pulmonary function and urban air pollution in preschool children."
International Journal of Hygiene & Environmental Health 203(3): 235-44.
Frye, C., B. Hoelscher, J. Cyrys, M. Wjst, H. E. Wichmann and J. Heinrich (2003). "Association of lung
function with declining ambient air pollution." Environmental Health Perspectives 111(3): 383-
7.
Galizia, A. and P. L. Kinney (1999). "Long-term residence in areas of high ozone: associations with
respiratory health in a nationwide sample of nonsmoking young adults [dsee comments][see
comment]." Environmental Health Perspectives 107(8): 675-9.
Gauderman, W. J., E. Avol, F. Gilliland, H. Vora, D. Thomas, K. Berhane, R. McConnell, N. Kuenzli, F.
Lurmann, E. Rappaport, H. Margolis, D. Bates and J. Peters (2004). "The effect of air pollution
on lung development from 10 to 18 years of age.[see comment][erratum appears in N Engl J
Med. 2005 Mar 24;352(12):1276]." New England Journal of Medicine 351(11): 1057-67.
Gauderman, W. J., G. F. Gilliland, H. Vora, E. Avol, D. Stram, R. McConnell, D. Thomas, F. Lurmann,
H. G. Margolis, E. B. Rappaport, K. Berhane and J. M. Peters (2002). "Association between air
pollution and lung function growth in southern California children: results from a second
cohort." American Journal of Respiratory and Critical Care Medicine 166(1): 76-84.
Gauderman, W. J., R. McConnell, F. Gilliland, S. London, D. Thomas, E. Avol, H. Vora, K. Berhane, E.
B. Rappaport, F. Lurmann, H. G. Margolis and J. Peters (2000). "Association between air
pollution and lung function growth in southern California children." American Journal of
Respiratory & Critical Care Medicine 162(4 Pt 1): 1383-90.

127

Gauderman, W. J., H. Vora, R. McConnell, K. Berhane, F. Gilliland, D. Thomas, F. Lurmann, E. Avol,
N. Kunzli, M. Jerrett and J. Peters (2007). "Effect of exposure to traffic on lung development
from 10 to 18 years of age: a cohort study." Lancet 369(9561): 571-7.
Gilliland, F., E. Avol, P. Kinney, M. Jerrett, T. Dvonch, F. Lurmann, T. Buckley, P. Breysse, G. Keeler,
T. de Villiers and R. McConnell (2005). "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." Environ Health Perspect 113(10):
1447-54.
Gilliland, F. D., R. McConnell, J. Peters and H. Gong, Jr. (1999). "A theoretical basis for investigating
ambient air pollution and children's respiratory health." Environ Health Perspect 107 Suppl 3:
403-7.
Gold, D., X. Wang, D. Wypij, F. Speizer, J. Ware and D. DW (1996). "Effects of cigarette smoking on
lung function in adolescent boys and girls." N Engl J Med 335: 931-7.
Gong, H., Jr. (1992). "Health effects of air pollution. A review of clinical studies." Clin Chest Med
13(2): 201-14.
Goren, A., S. Hellmann, Y. Gabbay and S. Brenner (1999). "Respiratory problems associated with
exposure to airborne particles in the community." Archives of Environmental Health 54(3):
165-71.
Goss, C. H., S. A. Newsom, J. S. Schildcrout, L. Sheppard and J. D. Kaufman (2004). "Effect of
Ambient Air Pollution on Pulmonary Exacerbations and Lung Function in Cystic Fibrosis."
Am. J. Respir. Crit. Care Med. 169(7): 816-821.
Götschi, T., B. Forsberg, J. Heinrich, B. Jacquemin, R. McConnell, J. Sunyer and N. Künzli (2006).
Evaluation of Improvements in Accuracy of “Residential” NO2 Annual Mean Estimates when
Using Data from Central Reference Monitors. Annual Conference of the ISEE/ISEA, Paris,
France; Poster available on request: gotschi@usc.edu.
Götschi, T., M. E. Hazenkamp-von Arx, J. Heinrich, R. Bono, P. Burney, B. Forsberg, D. Jarvis, J.
Maldonado, D. Norbäck, W. B. Stern, J. Sunyer, K. Torén, G. Verlato, S. Villani and N. Künzli
(2005). "Elemental Composition and Reflectance of Ambient Fine Particles at 21 European
Locations." Atmos env 39: 5947–5958.
Götschi, T., L. Oglesby, P. Mathys, C. Monn, N. Manalis, K. Koistinen, M. Jantunen, O. Hanninen, L.
Polanska and N. Kunzli (2002). "Comparison of black smoke and PM2.5 levels in indoor and
outdoor environments of four European cities." Environ Sci Technol 36(6): 1191-7.

128

Hamilton, R. S. and T. A. Mansfield (1991). "Airborne particulate elemental carbon: Its sources,
transport and contribution to dark smoke and soiling." Atmospheric Environment. Part A.
General Topics 25(3-4): 715-723.
Harrison, R. M., R. Tilling, M. S. C. Romero, S. Harrad and K. Jarvis (2003). "A study of trace metals
and polycyclic aromatic hydrocarbons in the roadside environment." Atmos env 37: 2391-
2402.
Hazenkamp-von Arx, M. E., T. Götschi Fellmann, L. Oglesby, U. Ackermann-Liebrich, T. Gislason, J.
Heinrich, D. Jarvis, C. Luczynska, A. J. Manzanera, L. Modig, D. Norback, A. Pfeifer, A. Poll,
M. Ponzio, A. Soon, P. Vermeire and N. Kunzli (2003). "PM2.5 assessment in 21 European
study centers of ECRHS II: Method and first winter results." Journal of the Air & Waste
Management Association 53(5): 617-28.
Hazenkamp-von Arx, M. E., T. Götschi, U. Ackermann-Liebrich, R. Bono, P. Burney, J. Cyrys, D.
Jarvis, L. Lillienberg, C. Luczynska, J. A. Maldonado, A. Jaen, R. de Marco, Y. Mi, L. Modig,
L. Bayer-Oglesby, F. Payo, A. Soon, J. Sunyer, S. Villani, J. Weyler and N. Künzli (2004).
"PM2.5 and NO2 assessment in 21 European study centres of ECRHS II: annual means and
seasonal differences." Atmospheric Environment 38(13): 1943-1953.
He, Q. C., P. J. Lioy, W. E. Wilson and R. S. Chapman (1993). "Effects of air pollution on children's
pulmonary function in urban and suburban areas of Wuhan, People's Republic of China."
Heinrich, J., B. Hoelscher and H. E. Wichmann (2000). "Decline of ambient air pollution and respiratory
symptoms in children." Am J Respir Crit Care Med 161(6): 1930-6.
Hirsch, T., S. K. Weiland, E. von Mutius, A. F. Safeca, H. Grafe, E. Csaplovics, H. Duhme, U. Keil and
W. Leupold (1999). "Inner city air pollution and respiratory health and atopy in children."
European Respiratory Journal 14(3): 669-77.
Hoek, G., B. Brunekreef, S. Goldbohm, P. Fischer and P. A. van den Brandt (2002a). "Association
between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort
study." Lancet 360(9341): 1203-9.
Hoek, G., B. Forsberg, M. Borowska, S. Hlawiczka, E. Vaskovi, H. Welinder, M. Branis, I. Benes, F.
Kotesovec and L. Otto Hagen (1997). "Wintertime PM10 and black smoke concentrations
across Europe: results from the peace study." Atmospheric Environment 31(21): 3609-3622.
Hoek, G., K. Meliefste, J. Cyrys, M. Lewne, T. Bellander, M. Brauer, P. Fischer, U. Gehring, J.
Heinrich, P. van Vliet and B. Brunekreef (2002b). "Spatial variability of fine particle
concentrations in three European areas." Atmospheric Environment 36(25): 4077-4088.

129

Hogervorst, J. G., T. M. de Kok, J. J. Briede, G. Wesseling, J. C. Kleinjans and C. P. van Schayck
(2006). "Relationship between radical generation by urban ambient particulate matter and
pulmonary function of school children." J Toxicol Environ Health A 69(3-4): 245-62.
Horak, F., Jr., M. Studnicka, C. Gartner, J. D. Spengler, E. Tauber, R. Urbanek, A. Veiter and T.
Frischer (2002). "Particulate matter and lung function growth in children: a 3-yr follow-up
study in Austrian schoolchildren.[see comment]." European Respiratory Journal 19(5): 838-45.
Ihorst, G., T. Frischer, F. Horak, M. Schumacher, M. Kopp, J. Forster, J. Mattes and J. Kuehr (2004).
"Long- and medium-term ozone effects on lung growth including a broad spectrum of
exposure.[see comment]." European Respiratory Journal 23(2): 292-9.
Ito, K., N. Xue and G. Thurston (2004). "Spatial variation of PM2.5 chemical species and source-
apportioned mass concentrations in New York City." Atmospheric Environment 38(31): 5269-
5282.
Jaakkola, M. S. (1994). The role of active and passive smoking in the evolution of ventilatory lung
function and respiratory symptoms during young adulthood. Helsinki, University of
Helsinki/McGill University, Montreal Canada.
Jammes, Y., S. Delpierre, M. J. Delvolgo, C. Humbert-Tena and H. Burnet (1998). "Long-term exposure
of adults to outdoor air pollution is associated with increased airway obstruction and higher
prevalence of bronchial hyperresponsiveness."
Jang, A. S., C. H. Yeum and M. H. Son (2003). "Epidemiologic evidence of a relationship between
airway hyperresponsiveness and exposure to polluted air."
Janssen, N. A., B. Brunekreef, P. van Vliet, F. Aarts, K. Meliefste, H. Harssema and P. Fischer (2003).
"The relationship between air pollution from heavy traffic and allergic sensitization, bronchial
hyperresponsiveness, and respiratory symptoms in Dutch schoolchildren."
Janssen, N. A., D. F. M. Van Mansom, K. Van Der Jagt, H. Harssema and G. Hoek (1997). "Mass
concentration and elemental composition of airborne particulate matter at street and
background locations." Atmospheric Environment 31(8): 1185-1193.
Jedrychowski, W., E. Flak and E. Mroz (1999). "The adverse effect of low levels of ambient air
pollutants on lung function growth in preadolescent children." Environmental Health
Perspectives 107(8): 669-74.

130

Jedrychowski, W. and M. Krzyzanowski (1989). "Ventilatory lung function and chronic chest symptoms
among the inhabitants of urban areas with various levels of acid aerosols: Prospective study in
Cracow." Environ Health Perspect 79: 101-7.
Jerrett, M., A. Arain, P. Kanaroglou, B. Beckerman, D. Potoglou, T. Sahsuvaroglu, J. Morrison and C.
Giovis (2005). "A review and evaluation of intraurban air pollution exposure models." J Expos
Anal Environ Epidemiol 15: 185-204.
Kan, H., G. Heiss, K. M. Rose, E. Whitsel, F. Lurmann and S. J. London (2007). "Traffic exposure and
lung function in adults: the Atherosclerosis Risk in Communities study." Thorax.
Karita, K., E. Yano, W. Jinsart, D. Boudoung and K. Tamura (2001). "Respiratory symptoms and
pulmonary function among traffic police in Bangkok, Thailand." Archives of Environmental
Health 56(5): 467-70.
Katsouyanni, K. (2003). "Ambient air pollution and health." British Medical Bulletin 68: 143-56.
Katsouyanni, K., G. Touloumi, E. Samoli, A. Gryparis, A. Le Tertre, Y. Monopolis, G. Rossi, D.
Zmirou, F. Ballester, A. Boumghar, H. R. Anderson, B. Wojtyniak, A. Paldy, R. Braunstein, J.
Pekkanen, C. Schindler and J. Schwartz (2001). "Confounding and effect modification in the
short-term effects of ambient particles on total mortality: results from 29 European cities within
the APHEA2 project." Epidemiology 12(5): 521-31.
Katsouyanni, K., D. Zmirou, C. Spix, J. Sunyer, J. P. Schouten, A. Ponka, H. R. Anderson, Y. Le
Moullec, B. Wojtyniak, M. A. Vigotti and et al. (1995). "Short-term effects of air pollution on
health: a European approach using epidemiological time-series data. The APHEA project:
background, objectives, design." Eur Respir J 8(6): 1030-8.
Kim, J. J., S. Smorodinsky, M. Lipsett, B. C. Singer, A. T. Hodgson and B. Ostro (2004). "Traffic-
related Air Pollution Near Busy Roads: The East Bay Children's Respiratory Health Study."
Am. J. Respir. Crit. Care Med.: 200403-281OC.
Kopp, M., W. Bohnet, T. Frischer, C. Ulmer, M. Studnicka, G. Ihorst, C. Gardner, J. Forster, R. Urbanek
and J. Kuehr (2000). "Effects of ambient ozone on lung function in children over a two-summer
period." Eur Respir J 16(5): 893-900.
Kramer, U., T. Koch, U. Ranft, J. Ring and H. Behrendt (2000). "Traffic-related air pollution is
associated with atopy in children living in urban areas." Epidemiology 11(1): 64-70.

131

Kunzli, N., M. Jerrett, W. J. Mack, B. Beckerman, L. LaBree, F. Gilliland, D. Thomas, J. Peters and H.
N. Hodis (2005a). "Ambient air pollution and atherosclerosis in Los Angeles." Environ Health
Perspect 113(2): 201-6.
Kunzli, N., B. Kuna-Dibbert, D. Keidel, R. Keller, O. Brandli, C. Schindler, K. M. Schweinzer, P.
Leuenberger and U. Ackermann-Liebrich (2005b). "Longitudinal validity of spirometers--a
challenge in longitudinal studies." Swiss Med Wkly 135(33-34): 503-8.
Kunzli, N., F. Lurmann, M. Segal, L. Ngo, J. Balmes and I. B. Tager (1997a). "Association between
lifetime ambient ozone exposure and pulmonary function in college freshmen--results of a pilot
study." Environmental Research 72(1): 8-23.
Kunzli, N., I. S. Mudway, T. Gotschi, T. Shi, F. J. Kelly, S. Cook, P. Burney, B. Forsberg, J. W.
Gauderman, M. E. Hazenkamp, J. Heinrich, D. Jarvis, D. Norback, F. Payo-Losa, A. Poli, J.
Sunyer and P. J. Borm (2006). "Comparison of oxidative properties, light absorbance, total and
elemental mass concentration of ambient PM2.5 collected at 20 European sites." Environ
Health Perspect 114(5): 684-90.
Kunzli, N. and I. B. Tager (1997b). "The semi-individual study in air pollution epidemiology: a valid
design as compared to ecologic studies." Environ Health Perspect 105(10): 1078-83.
Lammel, G., A. Röhrl and H. Schreiber (2002). "Atmospheric Lead and Bromine in Germany - Post-
abatement Levels, Variabilities and Trends." Environmental Science & Pollution Research
9(6): 397-404.
Landis, M. S., G. A. Norris, R. W. Williams and J. P. Weinstein (2001). "Personal exposures to PM2.5
mass and trace elements in Baltimore, MD, USA." Atmospheric Environment 35(36): 6511-
6524.
Lebret, E., D. Briggs, H. van Reeuwijk, P. Fischer, K. Smallbone, H. Harssema, B. Kriz, P. Gorynski
and P. Elliott (2000). "Small area variations in ambient NO2 concentrations in four European
areas." Atmospheric Environment 34(2): 177-185.
Lee, D. S., J. A. Garland and A. A. Fox (1994). "Atmospheric concentrations of trace elements in urban
areas of the United Kingdom." Atmospheric Environment 28(16): 2691-2713.
Longhini, E., P. Fanari, E. Giacomotti, P. Palmulli, M. Gabrione, D. Gori, S. Barlera, G. Tognoni and A.
Salvadori (2004). "Lung function and methacholine responsiveness among adolescents in an
air-polluted area."

132

Lubinski, W., I. Toczyska, A. Chcialowski and T. Plusa (2005). "Influence of air pollution on pulmonary
function in healthy young men from different regions of Poland." Annals of Agricultural &
Environmental Medicine 12(1): 1-4.
Mathys, P., C. Oetterli, W. B. Stern, L. Oglesby , M. E. Hazenkamp-von Arx and N. Kuenzli (2002).
Elemental Analysis of PM2.5 Samples Collected in the Framework of the ECRHS II Study:
Description of Methods. http://www.ecrhs.org/reports/wp7report.pdf.
Mathys, P., W. B. Stern, L. Oglesby, C. Braun-Fahrländer, U. Ackermann-Liebrich, M. Jantunen and N.
Künzli (2001). "Elemental analysis of airborne particulate matter by ED-XRF within the
European EXPOLIS study." ICP Information Newsletter 27(3): 190-195.
Matthias-Maser, S. and R. Jaenicke (1994). "Examination of atmospheric bioaerosol particles with radii
>0.2 µm." J. Aerosol Sci. 25(8): 1605-1613.
McConnell, R., K. Berhane, L. Yao, M. Jerrett, F. Lurmann, F. Gilliland, N. Kunzli, J. Gauderman, E.
Avol, D. Thomas and J. Peters (2006). "Traffic, susceptibility, and childhood asthma." Environ
Health Perspect 114(5): 766-72.
McCunney, R. J. (2005). "Asthma, genes, and air pollution." J Occup Environ Med 47(12): 1285-91.
Miller, M. R., J. Hankinson, V. Brusasco, F. Burgos, R. Casaburi, A. Coates, R. Crapo, P. Enright, C. P.
van der Grinten, P. Gustafsson, R. Jensen, D. C. Johnson, N. MacIntyre, R. McKay, D.
Navajas, O. F. Pedersen, R. Pellegrino, G. Viegi and J. Wanger (2005). "Standardisation of
spirometry." Eur Respir J 26(2): 319-38.
Naef, R. and D. Xhillari (2000). Finale Report of Work Package 5: Historic Data of Ambient Air
Pollution in 37 European Cities of ECRHS I and II. http://www.ecrhs.org/reports/FINAL
REPORT WP5 ECRHS II.pdf.
Nakai, S., H. Nitta and K. Maeda (1999). "Respiratory health associated with exposure to automobile
exhaust. III. Results of a cross-sectional study in 1987, and repeated pulmonary function tests
from 1987 to 1990." Arch Environ Health 54(1): 26-33.
Nel, A. (2005). "Atmosphere. Air pollution-related illness: effects of particles." Science 308(5723): 804-
6.
Neuberger, M., H. Moshammer and M. Kundi (2002). "Declining ambient air pollution and lung
function improvement in Austrian children." Atmospheric Environment 36(11): 1733-1736.

133

Nicolai, T., D. Carr, S. K. Weiland, H. Duhme, O. von Ehrenstein, C. Wagner and E. von Mutius (2003).
"Urban traffic and pollutant exposure related to respiratory outcomes and atopy in a large
sample of children.[see comment]." European Respiratory Journal 21(6): 956-63.
Pacyna, J. M., S. Larssen and A. Semb (1991). "European survey for NOx emissions with emphasis on
Eastern Europe." Atmospheric Environment. Part A. General Topics 25(2): 425-439.
Pauwels, R. A., A. S. Buist, P. M. A. Calverley, C. R. Jenkins and S. S. Hurd (2001). "Global Strategy
for the Diagnosis, Management, and Prevention of Chronic Obstructive Pulmonary Disease.
NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop
Summary." Am. J. Respir. Crit. Care Med. 163(5): 1256-1276.
Peters, J. M., E. Avol, W. J. Gauderman, W. S. Linn, W. Navidi, S. J. London, H. Margolis, E.
Rappaport, H. Vora, H. Gong, Jr. and D. C. Thomas (1999). "A study of twelve Southern
California communities with differing levels and types of air pollution. II. Effects on
pulmonary function."
Pinto, J. P., A. S. Lefohn and D. S. Shadwick (2004). "Spatial variability of PM2.5 in urban areas in the
United States." J Air Waste Manag Assoc 54(4): 440-9.
Pope, C. A., 3rd, R. T. Burnett, M. J. Thun, E. E. Calle, D. Krewski, K. Ito and G. D. Thurston (2002).
"Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air
pollution." Jama 287(9): 1132-41.
Pope, C. A., 3rd and D. W. Dockery (2006). "Health effects of fine particulate air pollution: lines that
connect." J Air Waste Manag Assoc 56(6): 709-42.
Press, F. and R. Siever (1997). Understanding Earth. New York, W.H. Freeman.
Putaud, J.-P., F. Raes, R. Van Dingenen, E. Bruggemann, M.-C. Facchini, S. Decesari, S. Fuzzi, R.
Gehrig, C. Huglin and P. Laj (2004). "A European aerosol phenomenology--2: chemical
characteristics of particulate matter at kerbside, urban, rural and background sites in Europe."
Atmospheric Environment 38(16): 2579-2595.
Querol, X., A. Alastuey, S. Rodriguez, F. Plana, C. R. Ruiz, N. Cots, G. Massague and O. Puig (2001).
"PM10 and PM2.5 source apportionment in the Barcelona Metropolitan area, Catalonia,
Spain." Atmospheric Environment 35(36): 6407-6419.
Raizenne, M., L. M. Neas, A. I. Damokosh, D. W. Dockery, J. D. Spengler, P. Koutrakis, J. H. Ware and
F. E. Speizer (1996). "Health effects of acid aerosols on North American children: pulmonary
function." Environmental Health Perspectives 104(5)(506-14).

134

Reungoat, P., M. Chiron, S. Gauvin, Y. Le Moullec and I. Momas (2003). "Assessment of exposure to
traffic pollution using the ExTra index: study of validation." Environ Res 93(1): 67-78.
Rodriguez, S., X. Querol, A. Alastuey and E. Mantilla (2002). "Origin of high summer PM10 and TSP
concentrations at rural sites in Eastern Spain." Atmospheric Environment 36(19): 3101-3112.
Rodriguez, S., X. Querol, A. Alastuey, M.-M. Viana, M. Alarcon, E. Mantilla and C. R. Ruiz (2004).
"Comparative PM10-PM2.5 source contribution study at rural, urban and industrial sites during
PM episodes in Eastern Spain." Science of The Total Environment 328(1-3): 95-113.
Roemer, W., G. Hoek, B. Brunekreef, J. Clench-Aas, B. Forsberg, J. Pekkanen and A. Schutz (2000).
"PM10 elemental composition and acute respiratory health effects in European children
(PEACE project). Pollution Effects on Asthmatic Children in Europe." Eur Respir J 15(3): 553-
559.
Rojas-Martinez, R., R. Perez-Padilla, G. Olaiz-Fernandez, L. Mendoza-Alvarado, H. Moreno-Macias, T.
Fortoul, W. McDonnell, D. Loomis and I. Romieu (2007). "Lung Function Growth in Children
with Long-Term Exposure to Air Pollutants in Mexico City." Am J Respir Crit Care Med.
Romieu, I. (2005). "Nutrition and lung health." Int J Tuberc Lung Dis 9(4): 362-74.
Roorda-Knape, M. C., N. A. H. Janssen, J. J. De Hartog, P. H. N. Van Vliet, H. Harssema and B.
Brunekreef (1998). "Air pollution from traffic in city districts near major motorways."
Atmospheric Environment 32(11): 1921-1930.
Röösli, M., G. Theis, N. Künzli, J. Staehelin, P. Mathys, L. Oglesby, M. Camenzind and C. Braun-
Fahrländer (2001). "Temporal and spatial variation of the chemical composition of PM10 at
urban and rural sites in the Basel area, Switzerland." Atmospheric Environment 35(21): 3701-
3713.
Samoli, E., E. Aga, G. Touloumi, K. Nisiotis, B. Forsberg, A. Lefranc, J. Pekkanen, B. Wojtyniak, C.
Schindler, E. Niciu, R. Brunstein, M. Dodic Fikfak, J. Schwartz and K. Katsouyanni (2006).
"Short-term effects of nitrogen dioxide on mortality: an analysis within the APHEA project."
Eur Respir J 27(6): 1129-38.
Sarnat, J. A., P. Koutrakis and H. H. Suh (2000). "Assessing the relationship between personal
particulate and gaseous exposures of senior citizens living in Baltimore, MD." J Air Waste
Manag Assoc 50(7): 1184-98.
Schauer, J. J. (2003). "Evaluation of elemental carbon as a marker for diesel particulate matter." J Expo
Anal Environ Epidemiol 13(6): 443-53.

135

Schikowski, T., D. Sugiri, U. Ranft, U. Gehring, J. Heinrich, H.-E. Wichmann and U. Kramer (2005).
"Long-term air pollution exposure and living close to busy roads are associated with COPD in
women." Respiratory Research 6(1): 152.
Schindler, C., U. Ackermann-Liebrich, P. Leuenberger, C. Monn, R. Rapp, G. Bolognini, J. P. Bongard,
O. Brandli, G. Domenighetti, W. Karrer, R. Keller, T. G. Medici, A. P. Perruchoud, M. H.
Schoni, J. M. Tschopp, B. Villiger and J. P. Zellweger (1998). "Associations between lung
function and estimated average exposure to NO2 in eight areas of Switzerland. The
SAPALDIA Team. Swiss Study of Air Pollution and Lung Diseases in Adults." Epidemiology
9(4): 405-11.
Schindler, C., N. Kunzli, J. P. Bongard, P. Leuenberger, W. Karrer, R. Rapp, C. Monn and U.
Ackermann-Liebrich (2001). "Short-term variation in air pollution and in average lung function
among never-smokers. The Swiss Study on Air Pollution and Lung Diseases in Adults
(SAPALDIA)." Am J Respir Crit Care Med 163(2): 356-61.
Schlesinger, R. B., N. Kunzli, G. M. Hidy, T. Gotschi and M. Jerrett (2006). "The health relevance of
ambient particulate matter characteristics: coherence of toxicological and epidemiological
inferences." Inhalation Toxicology 18(2): 95-125.
Schmitzberger, R., K. Rhomberg, H. Buchele, R. Puchegger, D. Schmitzberger-Natzmer, G. Kemmler
and B. Panosch (1993). "Effects of air pollution on the respiratory tract of children."
Schwartz, J. (1989). "Lung function and chronic exposure to air pollution: a cross-sectional analysis of
NHANES II."
Schwartz, J., D. W. Dockery and L. M. Neas (1996). "Is daily mortality associated specifically with fine
particles?" J Air Waste Manag Assoc 46(10): 927-39.
Sekine, K., M. Shima, Y. Nitta and M. Adachi (2004). "Long term effects of exposure to automobile
exhaust on the pulmonary function of female adults in Tokyo, Japan." Occupational and
Environmental Medicine 61(4): 350-7.
Sherman, C. B. (1992). "The health consequences of cigarette smoking. Pulmonary diseases." Med Clin
North Am 76(2): 355-75.
Sin, D. D., L. Wu and S. F. P. Man (2005). "The Relationship Between Reduced Lung Function and
Cardiovascular Mortality: A Population-Based Study and a Systematic Review of the
Literature." Chest 127(6): 1952-1959.

136

Stern, B. R., M. E. Raizenne, R. T. Burnett, L. Jones, J. Kearney and C. A. Franklin (1994). "Air
pollution and childhood respiratory health: exposure to sulfate and ozone in 10 Canadian rural
communities." Environ Res 66(2): 125-42.
Sugiri, D., U. Ranft, T. Schikowski and U. Kramer (2006). "The influence of large-scale airborne
particle decline and traffic-related exposure on children's lung function." Environ Health
Perspect 114(2): 282-8.
Sunyer, J. (2001). "Urban air pollution and chronic obstructive pulmonary disease: a review." Eur Respir
J 17(5): 1024-33.
Sunyer, J., D. Jarvis, T. Gotschi, R. Garcia-Esteban, B. Jacquemin, I. Aguilera, U. Ackerman, R. de
Marco, B. Forsberg, T. Gislason, J. Heinrich, D. Norback, S. Villani and N. Kunzli (2006).
"Chronic bronchitis and urban air pollution in an international study." Occup Environ Med.
Tager, I. B., J. Balmes, F. Lurmann, L. Ngo, S. Alcorn and N. Kunzli (2005). "Chronic exposure to
ambient ozone and lung function in young adults." Epidemiology 16(6): 751-9.
Tanner, R. L., W. J. Parkhurst, M. L. Valente and W. David Phillips (2004). "Regional composition of
PM2.5 aerosols measured at urban, rural and "background" sites in the Tennessee valley."
Atmospheric Environment 38(20): 3143-3153.
Tashkin, D. P., R. Detels, M. Simmons, H. Liu, A. H. Coulson, J. Sayre and S. Rokaw (1994). "The
UCLA population studies of chronic obstructive respiratory disease: XI. Impact of air pollution
and smoking on annual change in forced expiratory volume in one second." American Journal
of Respiratory and Critical Care Medicine 149(5): 1209-17.
U.S. Environmental Protection Agency (1990). Air emission species manual, volume II: particulate
matter species profiles, 2
nd
edition. Research Triangle Park.
UNECE. (2003). "Trends In Europe And North America: The Statistical Yearbook of the Economic
Commission for Europe 2003." Retrieved 7 Feb 2007, from
http://www.unece.org/stats/trends/ - ch8.
van der Lende, R., T. Kok, R. Peset, P. H. Quanjer, J. P. Schouten and N. G. Orie (1981). "Longterm
exposure to air pollution and decline in VC and FEV1. Recent results from a longitudinal
epidemiologic study in the Netherlands." Chest 80(1 Suppl): 23-6.
van Vliet, P., M. Knape, J. de Hartog, N. Janssen, H. Harssema and B. Brunekreef (1997). "Motor
vehicle exhaust and chronic respiratory symptoms in children living near freeways." Environ
Res 74(2): 122-32.

137

Venn, A., S. Lewis, M. Cooper, R. Hubbard, I. Hill, R. Boddy, M. Bell and J. Britton (2000). "Local
road traffic activity and the prevalence, severity, and persistence of wheeze in school children:
combined cross sectional and longitudinal study." Occup Environ Med 57(3): 152-8.
Venn, A. J., S. A. Lewis, M. Cooper, R. Hubbard and J. Britton (2001). "Living near a main road and the
risk of wheezing illness in children." Am J Respir Crit Care Med 164(12): 2177-80.
Viana, M., X. Querol, A. Alastuey, G. Gangoiti and M. Menendez (2003). "PM levels in the Basque
Country (Northern Spain): analysis of a 5-year data record and interpretation of seasonal
variations." Atmospheric Environment 37(21): 2879-2891.
Wang, B., Z. Peng, X. Zhang, Y. Xu, H. Wang, G. Allen, L. Wang and X. Xu (1999). "Particulate
matter, sulfur dioxide, and pulmonary function in never-smoking adults in Chongqing, China."
International Journal of Occupational & Environmental Health 5(1): 14-9.
Ward, D. J. and J. G. Ayres (2004). "Particulate air pollution and panel studies in children: a systematic
review." Occup Environ Med 61(4): e13.
Ware, J., B. J. Ferris, D. Dockery, J. Spengler and D. Stram (1986). "Effects of ambient sulfur oxides
and suspended particles on respiratory health of preadolescent children." Am Rev Respir Dis
133: 834-42.
WHO (1984). Environmental Health Criteria 24: Titanium. Geneva,, United Nations Environment
Programme, International Labour Organization, World Health Organization
http://www.inchem.org/documents/ehc/ehc/ehc24.htm#SubSectionNumber:1.1.2.
WHO (2000). Air quality guidelines for Europe. Copenhagen, World Health Organization.
WHO (2003). Health Aspects of Air Pollution with Particulate Matter, Ozone and Nitrogen Dioxide.
Bonn, World Health Organization: 98.
Wilhelm, M. and B. Ritz (2003). "Residential proximity to traffic and adverse birth outcomes in los
angeles county, california, 1994-1996." Environ Health Perspect 111(2): 207-16.
Wjst, M., P. Reitmeir, S. Dold, A. Wulff, T. Nicolai, E. F. von Loeffelholz-Colberg and E. von Mutius
(1993). "Road traffic and adverse effects on respiratory health in children.[see comment]." Bmj
307(6904): 596-600.

138

Wongsurakiat, P., K. N. Maranetra, A. Nana, C. Naruman, M. Aksornint and T. Chalermsanyakorn
(1999). "Respiratory symptoms and pulmonary function of traffic policemen in Thonburi."
Journal of the Medical Association of Thailand 82(5): 435-43.
Wyler, C., C. Braun-Fahrlander, N. Kunzli, C. Schindler, U. Ackermann-Liebrich, A. P. Perruchoud, P.
Leuenberger and B. Wuthrich (2000). "Exposure to motor vehicle traffic and allergic
sensitization. The Swiss Study on Air Pollution and Lung Diseases in Adults (SAPALDIA)
Team." Epidemiology 11(4): 450-6.
Xu, X., S. T. Weiss, D. W. Dockery, J. P. Schouten and B. Rijcken (1995). "Comparing FEV1 in adults
in two community-based studies." Chest 108(3): 656-662.
Xu, X. P., D. W. Dockery and L. H. Wang (1991). "Effects of air pollution on adult pulmonary
function." Archives of Environmental Health 46(4): 198-206.
Yao, X., M. Fang and C. K. Chan (2001). "Experimental Study of the Sampling Artifact of Chloride
Depletion from Collected Sea Salt Aerosols." Environ. Sci. Technol. 35(3): 600-605.
Yu, T. S., T. W. Wong, X. R. Wang, H. Song, S. L. Wong and J. L. Tang (2001). "Adverse effects of
low-level air pollution on the respiratory health of schoolchildren in Hong Kong." Journal of
Occupational & Environmental Medicine 43(4): 310-6.
Zanolin, M. E., C. Pattaro, A. Corsico, M. Bugiani, L. Carrozzi, L. Casali, R. Dallari, M. Ferrari, A.
Marinoni, E. Migliore, M. Olivieri, P. Pirina, G. Verlato, S. Villani and R. Marco (2004). "The
role of climate on the geographic variability of asthma, allergic rhinitis and respiratory
symptoms: results from the Italian study of asthma in young adults." Allergy 59(3): 306-14.
Zhu, Y., W. C. Hinds, S. Kim, S. Shen and C. Sioutas (2002a). "Study of ultrafine particles near a major
highway with heavy-duty diesel traffic." Atmospheric Environment 36(27): 4323-4335.
Zhu, Y., W. C. Hinds, S. Kim and C. Sioutas (2002b). "Concentration and size distribution of ultrafine
particles near a major highway." J Air Waste Manag Assoc 52(9): 1032-42.

139

Appendices

140

Appendix 1: Summary Tables of Reviewed Studies of Less Relevance
Table 24: Cross-sectional studies of medium relevance. Main study characteristics.
Publication Study Country N
Age
Health status
Smoking status
FVC, FEV1
FEV1/FVC
PEF, MMEF, FEF()
<0.7/0.8 pred.
Between # of communities
Within community
Between individuals
TSP
PM10, PM2.5
NO2
O3
SO2
Others
Models
In/out adj., resid. history
Traffic (proximity, density)
Abbey 98 7-Day-Adv. USA 1510 25+ ns x x x x x x x x x
Chestnut 91 NHANES USA 6913 25-75 x x x x 49 x
Dockery 89 6-CITIES USA 5422 10-12 x x 6 x x x x
Fritz 01 Germany 235 3-7 x x x x x x x x VOC
Frye 03 Germany 2493 11-14 x x t x x
Galizia 99 USA 520 17-21 ns x x x x
Hirsch 99 Germany 1137 9-11 x x x x x x x x
Hogervorst 06 Netherlands 429 8-13 x x 6 x x ROS
Kuenzli 97 USA 130 16-19 nc ns x x x x x x x
Schwartz 89 NHANES USA 3922 6-24 x x 44 x x x x
Sugiri 06 Germany 2574 5-7 na TL R 13 (w) x x
Tager 05 USA 255 16-19 nc ns x x x x x x x
Ware 86 6-CITIES USA 10106 6-9 x 6 (w) x x x x x
Wjst 93 Germany 4320 9-11 x x x x
Other
exposures
Study
population
Lung function
measures
Exposure
contrast Air pollutants

NHANES = National Health and Nutrition Examination Survey; nc = no cardio-respiratory disease; na = no asthmatics; ns = non-smokers; TL = total lung capacity;
R = airway resistance; (w) = some consideration of within-community contrasts; VOC = volatile organic compounds; ROS = radical oxygen species;

141

Table 24 continued: Cross-sectional studies of medium relevance. Summary of statistical analyses.
Publication
Sex, age, height, race
Weight, BMI, obesity
Smoking status
ETS
Asthma, atopy, allergy
Childhood resp. infection
Co-exposures, occupational
SES, income, education
Short term effects, season
Housing characteristics
Sex
Age
Respiratory health
Smoke exposure
Occupational
Income, education
Time of residence
Exposure assessment
Other
Abbey 98 xxx xxxx xx x ia
Chestnut 91 xx x xxx
Dockery 89 x x x x x x To
Fritz 01 x
Frye 03 xx x x xx x
Galizia 99 xx x xx
Hirsch 99 xxx xxx
Hogervorst 06 x x x
Kuenzli 97 x xxx x x x
Schwartz 89 xxx x x xx
Sugiri 06 xx x x xx
Tager 05 x xxx x x x sa
Ware 86 xx x xx
Wjst 93 xxx xxx
Adjusted covariates Sensitivity/ subgroup analyses

ia = interaction for smoking and parental respiratory health; To=time outdoors; sa = interaction for low FEF25–75/FVC (small airways);

142

Table 24 continued: Cross-sectional studies of medium relevance. Main results and limitations.
Publication Main results Strengths Limitations
Abbey 98 Sign. neg. assoc. for SO4 in males, PM10, O3, in
males with parental history of resp. dis.
Residential history, individual
exposure assign.
Main effects of interaction variable,
unique measures
Chestnut 91 Sign. neg. assoc., for FVC, FEV1, threshold at
60ug/m3 TSP
Large number of cities Outdated air pollution measure?
Dockery 89 Null findings Sample size Outdated air pollution measure? Limited
exposure range?
Fritz 01 Lower lung function in areas with traffic related
pollution profiles
descriptive, no adjustement for covariates
Frye 03 Neg. assoc. with lung function and air pollution
over time, sig. for FVC
Exposure changes over time Changes of lifestyle factors
Galizia 99 Sign. lower lung function in subjects from high O3
counties. Stronger effect in males.
Individuall assigned exposure Exposure assessment, data analysis
Hirsch 99 Null findings GIS based exposure assessment no particle measures
Hogervorst 06 Sig. neg. assoc. with ROS per PM mass, pos. assoc.
for ROS per m3 air, PM.
Innovative measure of PM toxicity Exposure assessment, interpretation of
results
Kuenzli 97 Sign. neg. assoc. for O3 with FEF, no sign.
associations for volumes, PM10
Lifetime effective exposure,
considering pers. activity
sample size, remaining potential for
community level confounding?
Schwartz 89 Sign. neg. assoc. FVC, FEV1, PEF, treshold at
100ug/m3 TSP
Sample size, non-linear analyses High pollution communities all in
California?
Sugiri 06 Better lung function with decreasing TSP. Benefit
diminished for children living near traffic
Background and traffic exposure,
long and short term
Tager 05 Sign. neg. assoc. for O3 on FEF among subjects
with low FEF25–75/FVC
Lifetime effective exposure Sample size, main effects of interaction
variable not considered
Ware 86 Null findings Sample size Outdated air pollution measure? Limited
exposure range?
Wjst 93 Sig. neg. assoc. between PEF and car counts Traffic exposure on school district

ROS = radical oxygen species;

143

Table 25: Cross-sectional studies of low relevance. Main study characteristics.
Publication Country N
Age
Sex
Health status
Smoking
status
FVC, FEV1
FEV1/FVC
PEF, MMEF,
FEF()
<0.7/0.8 pred.
Between # of
communities
Within
community
TSP
PM10, PM2.5
NO2
O3
SO2
Others
Devereux 96* UK 600 20-44 m x 2 uk
Forastiere 94* Italy 1215 c x 2 uk
Goren 99 Israel 976 7-13 x x 2 x x
He 93* China 604 7-13 nc x 2 x x x
Jammes 98* France ?a p x x 2 x x
Jang 03 S. Korea 67010-13 x 3 xxxx
Karita 01 Thailand 206 20-60 m x 3 (x)
Karnat 92* India 349 a x x 4 x x x
Longhini 04 Italy 531 11-13 x x 2 no
Lubinski 05 Poland 1278 18-23 m n 3 x x x
Matkovic 98* Croatia ?a w n x x x 3 x
Schmitzberger 93* Austria 1626 c x x 3 xxxx
Stern 89* Canada 735 7-12 x 2 xxxxx
Stern 94 Canada 3945 7-11 x x 2/10 xxxxx
Wang 99* China 1075 35-60 ns x x 2 x x
Wongsurakiat 99 Thailand 932 a n x x (w)
Xu 91* China 1440 40-69 ns x 3 x x
Yu 01 China 1294 8-12 x x x 2 (x) x x
Study population
Lung function
measures
Exposure
contrast Air pollutants

* = based on abstract only; c = children; a = adults; m = men; w = women; nc = no cardio-respiratory disease; p = patients; 2/10 = 10 communities across 2 regions;
(w)=occupation based within community contrasts; (x)=not used in analysis; uk=unknown; no=no pollution measurements; (also see Table 1)

144

Table 25 continued: Cross-sectional studies of low relevance. Summary of statistical analyses.
Publication
Sex, age, height,
race
Weight, BMI,
obesity
Smoking status
ETS
Asthma, atopy,
allergy
co-exposures,
occupational
SES, income,
education
Short term
effects, season
Housing
characteristics
Respiratory health
Occupational
Other
Devereux 96*
Forastiere 94*
Goren 99 xxx x x
He 93* xx
Jammes 98*
Jang 03
Karita 01 xx x
Karnat 92*
Longhini 04 xxx x x
Lubinski 05 (?)
Matkovic 98*
Schmitzberger 93*xx x
Stern 89* xx x xxxx x
Stern 94 xx x x x
Wang 99* xx xx x
Wongsurakiat 99 xx
Xu 91*xx
Yu 01 xx
Adjusted covariates
Subgroup
analysis

* based on abstract only. ETS = environmental tobacco smoke; SES = socio-economic status.

145

Table 25 continued: Cross-sectional studies of low relevance. Main results and limitations.
Publication Main results Limitations
Devereux 96* Null-findings 2 community comparison
Forastiere 94* Null-findings for lung function, but difference reported for
bronchio-hyperresonsivness
2 community comparison
Goren 99 Sign. lower flows in polluted community, inconsistent for lung
volumes.
2 community comparison
He 93* Lung function lower in urban area than in suburban area. 2 community comparison
Jammes 98* Lower lung function in COPD patients living in downtown, vs
suburbs (crude analysis)
2 community comparison, data analysis
Jang 03 Null-findings for lung function, but sign. difference for
bronchio-hyperresponsivness
No multivariate analysis for lung function (BHR only)
Karita 01 Lung function lower in Bangkok traffic police. compared to
rural policemen
Qualitative exposure categories based on occupation and
region
Karnat 92* Lower lung function in higher air pollution zones Sample sizen, data analysis?
Longhini 04 Sign. lower lung function in urban area compared to mountain
valleys
no air pollution measurements, 2 community comparison
Lubinski 05 Lower lung function among higher air pollution categories data analysis, adjustment for covariates
Matkovic 98* Lower lung function in higher air pollution communities Sample sizen, data analysis?
Schmitzberger 93* Lower lung function in higher air pollution zones Comparison across 3 exposure categories
Stern 89* Lower FEV1, FVC in conmmunity with higher SO2 and
nitrate levels
2 community comparison
Stern 94 Lower FEV1, FVC in high air pollution area Comparison across 2 exposure areas
Wang 99* Sign. lower lung function in urban than in suburban area 2 community comparison
Wongsurakiat 99 Lower lung function in traffic policemen than control group Comparison across 3 exposure categories
Xu 91* Lower lung function with increasing TSP, SO2 across
3 suburban areas
3 community comparison
Yu 01 Lower lung function in high pollution area 2 community comparison

COPD = Chronic obstructive pulmonary disease; BHR = bronchio-hyper responsiveness;

146

Table 26: Longitudinal studies of lower relevance. Main study characteristics.
Publication Country N
Age
Sex
Health status
Smoking status
FVC, FEV1
PEF, MMEF, FEF()
<0.7/0.8 pred.
# of measurements
Years of follow-up
Between # of communities
Within community
Between individuals
Jedrychowski 89* Poland 1414 a x uk 13 3
Jedrychowski 99 Poland 1001 9 x 2 2 2
Nakai 99 Japan 444 30-59 w nr nsx102.522
Tashkin 94 USA 2625 25-59 x 2 5 3
Van der Lende 81 Netherlands 3-4000 15-50 x x 4 9 2
Study population
Lung function
measures
Exposure
contrast

a = adults; w = women; nr = no respiratory disease history; ns = never smokers; uk = unknown;

147

Table 26 continued: Longitudinal studies of lower relevance. Summary of statistical analyses.
Publication
TSP
PM10, PM2.5
EC, Black smoke, soot
NO2
O3
SO2
Others
Sex, age, height, race
Smoking status
ETS
Asthma, atopy, allergy
Occupational
SES, income, education
Housing characteristics
Baseline lung function
Sex
Smoke exposure
Jedrychowski 89* xxx
Jedrychowski 99 x x x xxxx x x x x
Nakai 99 x x x
Tashkin 94 xx(x)xxxx x xxx
Van der Lende 81 x x x x x x x
Air pollutants Adjusted covariates
Subgroup
analyses

(x) = not used in analysis; ETS = environmental tobacco smoke;

148

Table 26 continued: Longitudinal studies of lower relevance. Main results and limitations.
Publication Main results Strengths Limitations
Jedrychowski 89* Lower FEV1, faster decline in men in areas with higher pollution Length of follow-up
Jedrychowski 99 Faster lung function growth in low polluted area, but lower levels
at baseline
Adjustment for change in
height (growth rate)
2 communities only. adj. for
baseline lung function?
Nakai 99 No significant differences in lung function level of change
between three exposure zones, incl. one in proximity to traffic
Up to ten lung function
measurments
3 zone comparison. Small sample.
Short follow-up period.
Tashkin 94 Sign. faster decline of FEV1 in men, non-smoking women, in
more polluted community
Loss to follow-up, 3 community
comparison
Van der Lende 81 Sign. faster decline in FEV1, VC in more polluted community 3 follow-up assessments 2 community comparison

* = based on abstract only; VC = vital capacity;

149

Appendix 2: Historic Air Pollution Data for the ECRHS Cities
Table 27: Availability of historic air pollution data for ECRHS cities.
Characteristics of stations with past air pollution data and start years for time-series of specific pollutants. (See following page for table).
Footnotes:
Station type: T = Traffic (Station used for monitoring traffic induced air pollution), I = Industrial (Station used for monitoring industrial air pollution), B = Background
(Station used for monitoring background air pollution levels. These stations can be located inside (urban/background) as well as outside (rural/background) cities.
U = Unknown (Station type is not known)
Type of zone: U = Urban (Station is located within the city), S = Suburban (Station is located in the outskirts (fringe) of a city, or in small residential areas outside the
main city), r = Rural (Station is located outside the city), U = Unknown (Location of the station is not known)
Characteristic of zone: R = Residential, C = Commercial, I = Industrial,
Emission source (Major emission source in station environment within 500 meters): P: = Public power, co-generation and district heating, T = Traffic, C = Commercial,
institutional and residential combustion, I = Industrial activities, NO = No emission source within 500 meters),
Street within 100 m (Street type within 100 meter radius): M = Main street, S = Side street, H = Highway, U = Unknown, NO = No street
Traffic volume (Estimated traffic volume of the street with the highest traffic volume within 100 meters radius): H = High traffic (More than 10'000 vehicles/day),
M = Medium traffic (Between 2'000 and 10'000 vehicles/day), L = Low traffic (Less than 2'000 vehicles/day), U = Unknown
Wide/Canyon: W = D/H>1.5 (D = Distance between axis street and buildings), C = D/H<1.5
Years = Reported is the year when the station started to measure the specific pollutant

150

Table 27 continued: Availability of historic air pollution data for ECRHS cities. For footnotes see page 150.
City Country Station name
Station type
Type of Zone
Charact. of zone
Emission source
Street within 100m
Traffic volume
Wide/canyon
SO
2
CO
NO
2
NO
TSP
BS
PM
10
O
3
Antwerp Belgium Borgerhout T U C TC M H W 80 80 80 80 95 95 80
Tartu Estonia Riia-Turu T U RC PTC M H W 93
Grenoble France Villeneuve B U R NO NO 85 85 85 92 92 92
Grenoble France Fontaine B U R NO M M W 85 94 94 94
Grenoble France Saint Martin d'Heres B U R NO S M W 99 99 99 99 99
Paris France Aubervilliers B U R TC M U 92 92 92 92
Paris France Neuilly-sur/Seine B U R TC M U 92 92 92 92
Paris France Paris 12eme B U R TC S 92 92 92 92 97
Paris France Paris 13eme B U RC TC M U 92 92 92 92 92
Paris France Foret de Rambouillet B r rural NO NO 94 93 92
Paris France Champs Elysees T U RC TC M H W 92 92 92
Erfurt Germany Kraempferstr. B U R TC M H 91 92 91 91 91
Pavia Italy Folperti TB U R TCI M M W 94 94 94 94 94 94
Turin Italy Consolata T U C TC M M W 91 91 91 91 91
Turin Italy Lingotto B U R TC S M W 91 92 91 91 91 93
Verona Italy Corso Milano T U R PT M M W 96 96 96 96 96 96
Verona Italy Torricelle B NO L W 9696 9696 96

151

Table 27 continued: Availability of historic air pollution data for ECRHS cities. For footnotes see page 150.
City Country Station name
Station type
Type of Zone
Charact. of
zone
Emission
source
Street within
100m
Traffic
volume
Wide/canyon
SO
2
CO
NO
2
NO
TSP
BS
PM
10
O
3
Albacete Spain Albacete B U R C S 99 99 99 99 99 99
Barcelona Spain S. Geruasi T C T MS H W 86 86 86 87 86
Barcelona Spain Poble Nou T S R T S H W 86 86 86 87 86
Barcelona Spain Sagrera B S R C S H W 92 92 93 95 84
Barcelona Spain Sants B U RC T S H W 92 95 95 95
Barcelona Spain Eixample T U RC T M H W 96 97 96
Galdakao Spain C. Lope de Vega I U RCI TCI S M 80 80
Galdakao Spain Barrio San Miguel I US RCI TCI S M 91
Galdakao Spain Ayto de Etxebarri I U RCI TCI S H 80 80
Galdakao Spain Colegio Pena Lemona I U RCI TCI S M 80 83
Galdakao Spain C. Calderon de la Barca I U RCI TI S M 92 92 92 90 88
Galdakao Spain Polideportivo Municipal I U RCI TI S M 95 95 95 95 95 95
Galdakao Spain Durango I U RCI TI S M 97 97 97 97 97
Galdakao Spain Mundaka B R R no S L 98 98 99 98 98
Huelva Spain La Orden TB U C T M M W 95 95 95 95 98
Huelva Spain Manuel Lois TB U R T M L W 95 96 96 95 95
Huelva Spain Los Rosales TI U RI T M H 95 95 95 95 95
Huelva Spain El Estadio T U RC T MS M 95 95 95 95 95
Huelva Spain Marimas del Titan B S R T S L 96 96 96 96 96
Huelva Spain Pozo Dulce TI U RI T M H 9595 9595 95 95
Oviedo Spain Palacio de los Deportes T U R TC M H W 93 93 93 93 94
Oviedo Spain General Elorza T U R TC M H C 93 93 93 93 97
Oviedo Spain Plaza de Toros T U R PTC M H W 93 93 93 93 95

152

Table 27 continued: Availability of historic air pollution data for ECRHS cities. For footnotes see page 150.
City Country Station name
Station type
Type of Zone
Charact. of zone
Emission source
Street within 100m
Traffic volume
Wide/canyon
SO
2
CO
NO
2
NO
TSP
BS
PM
10
O
3
Goteborg Sweden Femman B U C T M H W 80 87 80 80 80 91 80
Goteborg Sweden Jarntorget B U C PTI H H W 90 90 92 91
Umea Sweden Radhusesplanaden 8 B U RC T S M 88 88
Umea (Lycksele) Sweden Lycksele B U RC TC M M 86 86 86
Umea (Vindeln) Sweden Vindeln B R U L 86
Uppsala Sweden Town Library B U R T M H W 86 96 86
Uppsala Sweden Marsta B 98
Uppsala Sweden Kungsgatan T U R T M H W 98 98 98 98 98
Basel Switzerland St.Johannplatz TB U R T S M 80 88 86 86 97 88
Basel Switzerland Basel-Binningen B S R TC NO 88 88 87 87 82 97 88
Ipswich UK Civic Drive 96
Ipswich UK Chevallier St 96
Ipswich UK Kings Av 96
Ipswich UK Wherstead Rd 96
Ipswich UK Landseer Rd 97
Ipswich UK Tavern St 96
Norwich UK Churchill Road B U R TC S L C 80 80
Norwich UK Norwich Roadside T U TC MH M W 97
Norwich UK Norwich Centre B U R TC S L W 97 97 97 97 97

153

Appendix 3: Statistical Power
Table 28: Significantly detectable effects sizes for effects on FEV1/FVC ratio, and for a subsample of 15 centers in the middle range of PM
2.5
.
Detectable effect sizes are shown for 80% power and significance of 95%. Calculations were conducted in Quanto (Version 1.1.1 http://hydra.usc.edu/gxe) and
took into account standard deviation of PM
2.5
, mean of outcome, and standard deviation of center random effects after adjusting for height and smoking status. All
calculations were conducted for a sample of N=21, corresponding to the number of centers.
Sample
Spirometric
Measure
N
Centers
N
Subjects Mean SD Mean SD
Detectable Effect
[/10ug/m3] Mean SD
Detectable Effect
[/y*10ug/m3]
Men FEV1/FVC 21 2040 17.0 9.1 0.81 0.018 0.011 -0.002 0.0008 0.001
Men <30y FEV1/FVC 21 671 17.0 9.1 0.81 0.016 0.01 -0.002 0.0007 0.001
Men >30y FEV1/FVC 21 1369 17.0 9.1 0.80 0.018 0.011 -0.002 0.0009 0.0011
Women FEV1/FVC 21 2250 17.0 9.1 0.83 0.013 0.008 -0.003 0.0008 0.001
Women <30y FEV1/FVC 21 733 17.0 9.1 0.83 0.012 0.007 -0.003 0.0008 0.001
Women >30y FEV1/FVC 21 1517 17.0 9.1 0.82 0.014 0.009 -0.003 0.0009 0.0011
PM2.5 Lung Function Level Change in Lung Function

Sample,
Mid-range Centres
Spirometric
Measure
N
Centers
N
Subjects Mean SD Mean SD
Detectable Effect
[ml/10ug/m3] Mean SD
Detectable Effect
[ml/y/10ug/m3]
Men FEV1 15 1407 17.2 2.8 4.19 0.645 1500 -0.0289 0.0290 65
Women FEV1 15 1584 17.2 2.8 3.12 0.468 1100 -0.0221 0.0237 55
Men FVC 15 1407 17.2 2.8 5.19 0.750 1700 -0.0238 0.0383 90
Women FVC 15 1584 17.2 2.8 3.78 0.532 1200 -0.0150 0.0293 70
Men FEV1/FVC 15 1407 17.2 2.8 0.81 0.064 0.14 -0.0019 0.0045 0.011
Women FEV1/FVC 15 1584 17.2 2.8 0.83 0.062 0.14 -0.0026 0.0048 0.011
PM2.5 Lung Function Level Change in Lung Function

154

Appendix 4: Type of Spirometer and Lung Function
Figure 15: Lung function levels across ECHRS centers by type of spirometer used.

155
means of % of predicted lung function. Prediction equations adjust for sex, height and age.
dics dry
Wide shaded bars in the background reflect mean lung function levels (% of predicted values) across
centers using the same type of spirometer. Narrow solid bars in the foreground represent center
Biomedin = Biomedin Baires water seal volume displacement spirometer; SM Dry = SensorMe
90
100
110
PS PA HUAL TUGNNOBA IP VE OVGA
Biomedin
REGOUP BS UM
SM Hot Wire
ER TA
Jaeger
FEV1 % pred. FVC % pred.
Mean FEV1 % pred. Mean FVC % pred.
%
pred
.
AC AS
SM Dry
90
100
110
PS PA HUAL TUGNNOBA IP VE OVGA PS PA HUAL TUGNNOBA IP VE OVGA
Biomedin
REGOUP BS UM
SM Hot Wire
ER TA ER TA
Jaeger
FEV1 % pred. FVC % pred.
Mean FEV1 % pred. Mean FVC % pred.
%
pred
.
AC AC AS AS
SM Dry
seal volume displacement spirometer (changed to Jaeger Masterscope at follow-up); SM Hot Wire =
SensorMedics heated wire flow sensing spirometer; Jaeger = Jaeger Pneumotach

Appendix 5: Loss-to-follow-up Effect
Figure 16: Illustration of loss-to-follow-up effect on PM
2.5
FEV1 association at baseline.
Solid circles and line represent FEV1 levels and association with PM
2.5
at baseline among subjects
which were included in the analysis, namely those followed up. Hollow circles and dashed line
represent FEV1 levels and association with PM
2.5
among all subjects at baseline, including those lost
to follow-up. The difference between the two regression lines can be interpreted as a proxy for the
effect that loss-to-follow-up had on the association between PM
2.5
and lung function level in the
presented analysis. For the vast majority of sub-samples investigated this effect was small (see Table
19). Note that the presented associations are crude and only serve the purpose to illustrate the
differences between subjects followed-up and those lost to follow-up. Any apparent associations with
PM
2.5
disappeared after adjusting for height and other covariates.
Upper graph shows data for women, and lower graph shows data for men.
3 3.1 3.2 3.3 3.4 3.5
FEV1[l] at baseline
0 10 20 30 40 50
PM2.5 [ug/m3]
Followed-up women Complete baseline women

4 4.2 4.4 4.6 4.8
FEV1[l] at baseline
0 10 20 30 40 50
PM2.5 [ug/m3]
Followed-up men Complete baseline men

156

Appendix 6: ECRHS Lung Function Protocol
Full version including questionnaire is available at www.ecrhs.org/quests.htm
LUNG FUNCTION TESTS
CRITERIA FOR TESTING
Criteria for baseline spirometry
The purpose of baseline spirometry is to record an accurate Forced Expiratory Volume in one second
(FEV
1
) and Forced Vital Capacity (FVC) from every subject who attends the testing centre.
ACCEPTANCE CRITERIA:
Any subject who is able to attend the testing centre.
EXCLUSION CRITERIA:
If the subject smokes: Lung function testing should be carried out at least one hour after the last cigarette
has been smoked.
If the subject has used an inhaler: Lung function testing should be carried out at least one hour after the
use of any inhaler.
If the subject has used an inhaler that is not a beta-2-agonist or an anticholinergic inhaler in the last
one to four hours: Lung function testing is carried out and the data recorded.
If the subject has used an inhaler that is a long acting Beta-2-agonist in the last 8 hours: If the subject is
willing to come back at another time when they have not taken their lay acting Beta-2-agonist,
another appointment should be made. HOWEVER – this may be difficult for them to do, in which
case, testing should proceed and medication used should be recorded.
If the subject has used an inhaler that is a beta-2-agonist or an anticholinergic inhaler in the last one to
four hours: If the subject is willing to come back another time for lung function testing, another
appointment should be made. If the subject is unable or reluctant to return another time, testing
should proceed and the medication used should be recorded.
If the subject has taken an oral beta-2-agonist or an oral theophylline or an oral antimuscarinic within
the last eight hours: If the subject is willing to come back another time for lung function testing,
another appointment should be made. If the subject is unable or reluctant to return another time,
testing should proceed and the medication used recorded.
If the subject has had a respiratory tract infection in the last three weeks: Another appointment should
be made unless the subject is unwilling to come back, in which case testing should continue. The
number of days elapsed since the end of the respiratory infection should be recorded.

157

If, after a total of nine attempts, a subject is unable to produce a technically satisfactory manoeuvre, no
FEV
1
or FVC will be recorded.
Making the appointment for testing
Ideally, lung function testing should be performed:
1) more than four hours after the use of a beta-2-agonist or anticholinergic inhaler,
2) more than eight hours after inhaled long acting beta-2-agonist, oral beta-2-agonist or theophylline or
oral antimuscarinic.
When the appointment for lung function testing is made the fieldworker should determine if the subject is
taking any of the following medications:
1) beta-2-agonist inhaler (short or long acting),
2) anticholinergic inhaler,
3) oral beta-2-agonist,
4) oral theophylline,
5) oral antimuscarinic.
If the subject is taking any of these medications (or any other inhaler) an appointment time should be
agreed that will cause the least disruption to the subject's normal dosing schedule.
One simple way of ensuring compliance with these instructions is to:
1) avoid early morning appointments for those using inhalers,
2) fix a time for an appointment and then ask the subject to take their inhalers four hours before and oral
medication eight hours before testing. Ask them to avoid taking their long acting beta-2-agonist if
possible.
The fieldworker should ensure that the subject has not had a respiratory tract infection in the three weeks
prior to testing and should advise the subject not to smoke for one hour prior to coming to the testing
centre. A letter should be sent to the subject explaining this.
Subjects who have not followed guidelines
Those who have had a cigarette in the last hour should have the lung function test delayed until one hour
has elapsed. (Most subjects will be in the centre for at least one hour.)
Those who have had an inhaler in the last four hours or oral medication (or long acting beta-2-agonist)
in the last eight hours may fall into one or more of the following categories:
1) misunderstood the instructions,
2) forgot the instructions,
3) ignored the instructions,
4) may have symptoms too severe to follow the instructions.

158

Lung function testing may still be carried out unless the subject is excluded for other reasons, and recent
medication should be noted in the Lung Function Questionnaire.
THE FORCED EXPIRATORY MANOEUVRE
General guidelines
All forced expiratory manoeuvres will be performed:
1) sitting, legs uncrossed
2) with noseclip on,
3) using a plastic or cardboard mouthpiece without teethgrips,
4) tight clothing should be loosened.
Two types of forced expiratory manoeuvre will be used in this protocol:
1) During baseline spirometry and bronchodilator challenge FVC will be measured and all subjects must
exhale fully.
2) During methacholine challenge only the FEV
1
needs to be recorded and the technician may interrupt
the exhalation when this has been achieved.
A technically unsatisfactory manoeuvre (FEV
1
or FVC) is defined as:
1) an unsatisfactory start of expiration characterised by excessive hesitation of false start
2) coughing during the first second of the manoeuvre, thereby affecting the measured FEV
1
value, or
any cough that interferes with the accurate measurement of FVC
3) Valsalva Manoeuvre (glottis closure)
4) A leak in the system or around the mouthpiece
5) An obstructed mouthpiece , e.g. the tongue in front of the mouthpiece.
Manoeuvres which have these faults are technically unsatisfactory and are rejected as failed attempts.
Evidence of poor compliance is shown by:
1) greater than 200ml (NB in ERCHS I this was 5%, this has been changed in line with current
ATS criteria) variation in FEV
1
between blows
2) greater than 150 mL or 5% FVC back-extrapolated volume
3) peak expiratory flow that is less than 85% of the best record
4) expiratory time that is less than six seconds
If these features are noted technicians should encourage the subject to produce a better reading but the
blows should not be excluded as failed attempts on these criteria alone.
A manoeuvre may only be rejected as a failed attempt if it is 'technically unsatisfactory'. Manoeuvres
with evidence of 'poor compliance' only should not be rejected.
The above protocol is consistent with current ATS guidelines (Am J Respir Crit Care Med

159

1995;152:1107-1136). These state that ‘The only criterion for unacceptable performance is fewer
than two acceptable curves. No spirogram should be rejected solely on the basis of its poor
reproducibility……elimination of data from subjects who fail to meet ATS reproducibility criteria
may result in population bias by excluding subjects who have abnormal lung function’
Instructions to subjects
Some of the subjects will never have used any form of lung function testing equipment before and others
will be very familiar with it.
Technicians should explain to the subject that the aim of the test is to find out how much air can be
blown out of the lungs and how forcefully it can be blown out.
This can be done by asking the subject to follow these steps:
1) Take in as deep breath as possible when full-
2) Place the mouthpiece in his/her mouth.
3) Close his/her lips tightly around the mouthpiece.
4) Blast or blow through the mouthpiece into the spirometer, blowing air out as hard, fast,
smoothly and completely as possible.
The subject should continue to push out air actively for as long as possible (FVC manoeuvre) or until the
technician tells him/her to stop (FEV
1
manoeuvre). During this time the technician must offer
positive encouragement to push or squeeze out more air.
Baseline spirometry
1) Ensure that it is appropriate to perform lung function testing.
2) Demonstrate the manoeuvre to all subjects at least once (more often if he/she appears uncertain).
3) Ask the subject to carry out five FVC manoeuvres.
4) Record the FEV
1
and FVC and Peak Expiratory Flow (in litres per second) from at least two and
up to five technically satisfactory manoeuvres.
5) If the subject has failed to produce two technically satisfactory manoeuvres after five attempts, the
technician should show them again how to conduct the manoeuvre and allow them four more
attempts.
6) Any subject who is unable to produce two technically satisfactory manoeuvres after nine attempts
should not be tested further and no FEV
1
/ FVC data should be recorded.
7) The number of rejected attempts should be recorded as appropriate on the Lung Function Data
Collection Sheet.

160

Appendix 7: Tables of Center Level Correlations
Table 29: Center level correlations between latitude, PM
2.5
, and various lung function measures.
Variables are sorted according to their correlation with latitude. Pearson correlation coefficients > 0.5 are bolded.
Variable name Latitude PM2.5
FEV1
BL
FVC
BL
FEV1/FVC
BL
FEV1
FU
FVC
FU
Change
in FEV1
Change
in FVC
FEV1/FVC
FU
Non-electric stove -0.90 0.71 -0.73 -0.72 0.29 -0.60 -0.56 0.57 0.62 0.08
Height 0.90 -0.48 0.78 0.75 -0.22 0.65 0.60 -0.49 -0.49 -0.07
Exercise -0.83 0.36 -0.74 -0.69 0.09 -0.75 -0.67 0.24 0.21 -0.05
Annoyance from traffic -0.80 0.52 -0.69 -0.66 0.20 -0.47 -0.49 0.61 0.52 0.20
Mostly heating with gas -0.78 0.71 -0.58 -0.48 -0.02 -0.46 -0.33 0.53 0.55 -0.20
Years lived in home BL -0.75 0.49 -0.47 -0.57 0.50 -0.34 -0.43 0.54 0.57 0.37
Sulfur -0.75 0.84 -0.34 -0.40 0.30 -0.19 -0.28 0.53 0.45 0.23
Exposed to ETS FU -0.75 0.35 -0.58 -0.62 0.34 -0.55 -0.60 0.33 0.33 0.27
FEV1 BL 0.74 -0.32 1.00 0.95 -0.25 0.90 0.83 -0.35 -0.33 -0.08
2 week NO2 at home outdoors -0.74 0.77 -0.51 -0.53 -0.39 -0.42 0.38 0.41 0.14 -0.02
Age at end of education 0.71 -0.47 0.60 0.57 -0.14 0.56 0.44 -0.50 -0.51 0.05
FVC BL 0.71 -0.38 0.95 1.00 -0.53 0.87 0.90 -0.32 -0.31 -0.35
Mostly cooking with gas -0.70 0.77 -0.66 -0.64 0.25 -0.57 -0.47 0.43 0.55 -0.08
Electric stove 0.69 -0.75 0.65 0.65 -0.30 0.56 0.46 -0.47 -0.59 0.04
Cars pass home constantly -0.66 0.20 -0.59 -0.55 0.12 -0.36 -0.37 0.61 0.55 0.17
Current smoker BL -0.65 0.14 -0.48 -0.53 0.32 -0.46 -0.52 0.27 0.24 0.28
Ex-smoker BL 0.63 -0.18 0.40 0.48 -0.39 0.37 0.44 -0.26 -0.30 -0.30
Current smoker FU -0.61 0.13 -0.48 -0.59 0.51 -0.39 -0.51 0.43 0.41 0.49
Ever smoked FU -0.60 0.10 -0.49 -0.52 0.20 -0.53 -0.60 0.15 0.00 0.24
FEV1 FU 0.60 -0.30 0.90 0.87 -0.25 1.00 0.90 0.02 -0.04 -0.01
FVC FU 0.58 -0.33 0.83 0.90 -0.52 0.90 1.00 0.04 0.07 -0.41

BL = at baseline, FU = at follow-up, ETS = environmental tobacco smoke

161

Table 29 continued: Center level correlations between latitude, PM
2.5
, and various lung function measures.
Variables are sorted according to their correlation with latitude. Pearson correlation coefficients > 0.5 are bolded.
Variable name Latitude PM2.5
FEV1
BL
FVC
BL
FEV1/FVC
BL
FEV1
FU
FVC
FU
Change
in FEV1
Change
in FVC
FEV1/FVC
FU
Ever smoked BL -0.58 0.09 -0.50 -0.54 0.24 -0.51 -0.59 0.20 0.07 0.28
ETS BL -0.58 0.35 -0.53 -0.58 0.37 -0.48 -0.54 0.40 0.33 0.31
PM2.5 -0.57 1.00 -0.32 -0.38 0.31 -0.30 -0.33 0.26 0.31 0.11
High occup. exp. to biol. dusts 0.55 -0.62 0.51 0.50 -0.14 0.53 0.57 -0.14 -0.01 -0.10
Change in FEV1 -0.55 0.26 -0.35 -0.32 0.03 0.02 0.04 1.00 0.87 0.08
Technicians & associate professionals 0.53 -0.32 0.27 0.30 -0.25 0.18 0.13 -0.50 -0.63 -0.09
Unclassifiable occup. status -0.52 0.34 -0.39 -0.32 -0.03 -0.16 -0.07 0.76 0.72 -0.06
Cars pass home seldom 0.50 -0.27 0.55 0.44 0.07 0.26 0.23 -0.68 -0.55 -0.03
Change in FVC -0.49 0.31 -0.33 -0.31 0.15 -0.04 0.07 0.87 1.00 -0.04
Trucks pass home constantly -0.48 0.14 -0.33 -0.36 0.21 -0.08 -0.11 0.65 0.64 0.23
Ex-smoker FU 0.48 -0.13 0.37 0.52 -0.63 0.24 0.39 -0.48 -0.53 -0.58
Trucks pass home seldom 0.46 -0.17 0.56 0.48 -0.02 0.27 0.26 -0.62 -0.54 -0.10
Gas stove -0.46 0.84 -0.46 -0.46 0.23 -0.42 -0.35 0.28 0.39 -0.08
No occup. exp. to biol. dusts -0.45 0.71 -0.37 -0.43 0.28 -0.37 -0.47 0.19 0.04 0.25
Manual profession -0.42 0.34 -0.18 -0.21 0.13 -0.11 -0.09 0.43 0.43 0.09
BMI BL -0.41 -0.09 -0.27 -0.25 0.05 -0.26 -0.18 0.24 0.32 -0.03
Resp. inf. <2y 0.40 -0.30 0.31 0.16 0.41 0.33 0.17 -0.06 0.09 0.46
Other stove -0.40 -0.08 -0.53 -0.50 0.10 -0.32 -0.22 0.53 0.52 -0.02
Open heating -0.40 0.08 -0.21 -0.11 -0.16 -0.04 0.12 0.51 0.62 -0.24
Resp. infect. <5y 0.40 -0.02 0.41 0.20 0.57 0.36 0.25 -0.10 0.20 0.38

BL = at baseline, FU = at follow-up, ETS = environmental tobacco smoke; BMI = body mass index

162

Table 29 continued: Center level correlations between latitude, PM
2.5
, and various lung function measures.
Variables are sorted according to their correlation with latitude. Pearson correlation coefficients > 0.5 are bolded.
Variable name Latitude PM2.5
FEV1
BL
FVC
BL
FEV1/FVC
BL
FEV1
FU
FVC
FU
Change
in FEV1
Change
in FVC
FEV1/FVC
FU
Ever rhinitis BL 0.35 -0.23 -0.08 0.04 -0.30 0.02 0.10 -0.10 -0.16 -0.24
Medium occup. exp. to biol. dusts 0.35 -0.69 0.25 0.35 -0.33 0.23 0.37 -0.21 -0.05 -0.31
Semi-skilled or unskilled manual
profession
-0.32 -0.10 -0.10 -0.05 -0.09 -0.08 0.04 0.20 0.28 -0.19
Ever asthma FU 0.28 -0.22 -0.23 -0.09 -0.34 -0.34 -0.16 -0.41 -0.40 -0.42
High occup. exp. to mineral dusts -0.28 -0.20 0.02 0.00 0.01 0.05 0.05 0.24 0.26 0.07
Medium occup. exp. to gases and fumes 0.25 -0.60 0.33 0.41 -0.34 0.32 0.45 -0.16 -0.06 -0.34
Trucks pass home frequently -0.23 0.12 -0.52 -0.37 -0.17 -0.34 -0.29 0.31 0.20 -0.07
Ever asthma BL 0.19 0.09 -0.26 -0.14 -0.22 -0.24 -0.11 -0.16 -0.17 -0.30
No occup. exp. to gases and fumes -0.19 0.51 -0.27 -0.34 0.30 -0.23 -0.38 0.13 0.01 0.34
FEV1/FVC BL -0.19 0.31 -0.25 -0.53 1.00 -0.25 -0.52 0.03 0.15 0.84
Microwave -0.18 0.00 -0.54 -0.48 0.02 -0.42 -0.39 0.16 0.07 0.02
IgE Sensitization [>0.7 kU/L] FU -0.17 0.32 -0.14 0.01 -0.04 0.17 0.19 0.17 -0.35 -0.44
Managers and professionals; non-manual 0.16 -0.01 -0.12 -0.19 0.22 -0.15 -0.28 -0.16 -0.22 0.27
Medium occup. exp. to mineral dusts 0.14 -0.37 0.21 0.28 -0.30 0.09 0.25 -0.19 -0.12 -0.36
Cars pass home frequently 0.14 0.18 -0.06 0.06 -0.31 0.10 0.17 0.26 0.12 -0.21
BMI FU -0.12 -0.28 -0.07 -0.03 -0.06 -0.15 0.00 0.00 0.14 -0.19
Symptomatic BL -0.12 0.04 -0.49 -0.46 0.13 -0.31 -0.33 0.28 0.21 0.15
Solid fuel stove 0.11 -0.05 0.21 0.00 0.60 0.30 0.16 0.30 0.47 0.53

BL = at baseline, FU = at follow-up,

163

Table 29 continued: Center level correlations between latitude, PM
2.5
, and various lung function measures.
Variables are sorted according to their correlation with latitude. Pearson correlation coefficients > 0.5 are bolded.
Variable name Latitude PM2.5
FEV1
BL
FVC
BL
FEV1/FVC
BL
FEV1
FU
FVC
FU
Change
in FEV1
Change
in FVC
FEV1/FVC
FU
Skilled manual 0.08 -0.27 0.27 0.23 0.06 0.18 0.25 -0.11 0.11 -0.09
FEV1/FVC FU -0.07 0.11 -0.08 -0.35 0.84 -0.01 -0.41 0.08 -0.04 1.00
# of reported symptoms BL -0.06 0.00 -0.46 -0.39 0.04 -0.24 -0.17 0.33 0.32 -0.01
No occup. exp. to mineral dusts 0.05 0.34 -0.14 -0.18 0.18 -0.09 -0.18 0.00 -0.06 0.19
Kerosene stove 0.04 -0.02 0.08 0.04 0.06 0.14 0.07 0.12 -0.12 0.16
IgE sensitization [>0.7 kU/L] BL 0.03 0.20 0.16 0.28 -0.45 0.18 0.26 -0.01 -0.16 -0.31
Length of follow-up -0.02 0.03 0.23 0.40 -0.72 0.15 0.28 -0.04 -0.27 -0.54
# of reported symptoms FU 0.02 -0.40 -0.31 -0.16 -0.25 -0.27 -0.09 -0.12 -0.06 -0.32
Age BL -0.02 0.19 -0.24 -0.26 0.11 -0.33 -0.41 -0.31 -0.43 0.14
Other non-manual -0.02 0.22 0.28 0.29 -0.11 0.21 0.22 -0.17 -0.12 -0.13
Age FU -0.01 0.22 -0.14 -0.10 -0.17 -0.25 -0.27 -0.29 -0.47 -0.10
High occup. exp. to gases and fumes 0.01 -0.17 0.05 0.08 -0.12 -0.03 0.12 -0.02 0.08 -0.24
Symptomatic FU 0.00 -0.44 -0.36 -0.22 -0.19 -0.31 -0.16 -0.09 -0.03 -0.26

BL = at baseline, FU = at follow-up,

164

Appendix 8: Results of Main Analysis using Sulfur instead of PM
2.5
as Exposure Marker
Table 30: Sulfur coefficients (p-values) for lung function level from main models* for complete samples, by sex.
Negative coefficients represent adverse effects on lung function. Effects for sulfur are per 1µg/m
3
.
Variable FEV1-level [ml] FVC-level [ml] FEV1/FVC-level† FEV1-level [ml] FVC-level [ml] FEV1/FVC-level
Intercept 3321 4092 0.82 3892 4762 0.82
Age -30.0 (<0.001) -23.6 (<0.001) -0.003 (<0.001) -31.1 (<0.001) -27.920 (<0.001) -0.002 (0.006)
Height [cm] 31.9 (<0.001) 44.6 (<0.001) -0.001 (<0.001) 41.7 (<0.001) 61.409 (<0.001) -0.002 (<0.001)
Ex-smoker 36.3 (0.004) 55.1 (<0.001) -0.003 (0.153) 0.9 (0.959) 13.292 (0.543) -0.002 (0.306)
Current smoker -9.3 (0.468) 23.2 (0.122) -0.009 (<0.001) -63.8 (<0.001) -54.263 (0.014) -0.006 (0.007)
Sulfur [µg/m
3
]
4.0 (0.267) 0.9 (0.855) 0.001 (0.260) 12.7 (0.005) 11.9 (0.074) 0.001 (0.628)
Women (N=2243) Men (N=2031)

* Coefficients for lung function level are estimated in the same model as coefficients for change in lung function (Table 33) and therefore adjusted for those variables.
In addition coefficients for lung function level are adjusted for BMI and SES (coefficients not shown).

165

Table 31: Sulfur coefficients (p-values) for lung function level from sensitivity analyses using different adjustment variables.
Negative coefficients represent adverse effects on lung function. Coefficients are per 1µg/m
3
sulfur. Units for FEV1 and FVC are ml.
FEV1 FVC FEV1/FVC* FEV1 FVC FEV1/FVC
Sub-models N β (p) β (p) β (p) N β (p) β (p) β (p)
Crude 2250 -8.4 (0.109) -16.2 (0.016) 0.0013 (0.054) 2040 -7.9 (0.294) -16.6 (0.094) 0.0011 (0.231)
Height only 2250 3.8 (0.290) 0.7 (0.892) 0.0007 (0.288) 2040 10.9 (0.019) 9.8 (0.144) 0.0004 (0.635)
Minimal 2245 3.7 (0.296) 0.5 (0.929) 0.0008 (0.266) 2037 11.3 (0.013) 10.2 (0.124) 0.0005 (0.621)
Main 2243 4.0 (0.267) 0.9 (0.855) 0.0008 (0.260) 2031 12.7 (0.005) 11.9 (0.074) 0.0005 (0.628)
Centre level adj. 2243 7.3 (0.052) 0.4 (0.471) 0.0012 (0.157) 2031 14.3 (0.017) 12.1 (0.097) 0.0009 (0.438)
Age squared 2245 3.6 (0.295) 0.3 (0.946) 0.0008 (0.262) 2037 11.3 (0.014) 10.2 (0.120) 0.0005 (0.612)
Maximal 2227 3.9 (0.254) 0.9 (0.855) 0.0007 (0.317) 2018 12.9 (0.003) 12.1 (0.070) 0.0004 (0.641)
Women Men

Crude = Sulfur; Height only = + height; Minimal = + smoking status (never, ex, current); Main = + BMI, SES Age squared = + age2; Centre level adj. = main + center
means of "Education level (age)", "Proportion of non-manual professions", "ETS"; Maximal = main + long and short term respiratory symptoms, exercise, trucks at
home, height squared. * Models did not converge. Estimates are based on iterated Estimation-Maximization.

166

Table 32: Sulfur coefficients (p-values) for lung function level from sensitivity analyses using different subsamples.
Negative coefficients represent adverse effects on lung function. Coefficients are per 1µg/m
3
sulfur. Units for FEV1 and FVC are ml.
Sub-sample N B p B p B p N B p B p B p
Never smokers 2014 4.6 0.24 2.4 0.66 0.0007 0.36 1489 9.1 0.08 6.6 0.38 0.0007 0.51
Ever smokers 2467 3.9 0.34 0.5 0.93 0.0008 0.25 2566 15.1 0.00 15.4 0.05 0.0004 0.70
Non-asthmatics 3877 3.4 0.31 0.2 0.96 0.0008 0.29 3665 12.4 0.01 12.1 0.08 0.0006 0.53
Asthmatics 600 2.1 0.76 1.3 0.88 0.0002 0.87 392 -3.7 0.72 9.0 0.43 -0.0021 0.24
Asymptomatics 1935 3.6 0.36 -0.6 0.91 0.0011 0.08 1815 14.4 0.00 14.6 0.04 0.0004 0.63
Symptomatics 2550 3.0 0.45 1.7 0.77 0.0002 0.80 2246 11.6 0.02 9.8 0.18 0.0005 0.61
Biomedin spirometers only 2243 7.6 0.15 -0.3 0.97 0.0020 0.04 2114 15.6 0.07 5.3 0.68 0.0024 0.12
Northern centers 1741 11.1 0.31 14.5 0.23 -0.0003 0.91 1431 13.1 0.38 26.8 0.13 -0.0011 0.72
Central centers 1235 29.0 0.10 31.1 0.37 0.0003 0.95 1082 26.0 0.06 37.1 0.33 -0.0015 0.82
Southern centers 1509 7.2 0.22 -2.5 0.75 0.0024 0.00 1548 15.2 0.14 3.2 0.81 0.0026 0.09
Mid range of height only 2424 4.9 0.22 1.3 0.82 0.0009 0.31 2238 16.9 0.00 15.6 0.02 0.0006 0.55
Longterm residents 2264 5.1 0.24 2.6 0.67 0.0009 0.24 1911 8.5 0.05 8.3 0.18 0.0004 0.75
Movers 2216 3.8 0.38 0.2 0.98 0.0008 0.27 2148 12.5 0.04 11.9 0.18 0.0004 0.67
Excl. 4 monitors close to traffic 3929 6.4 0.22 5.9 0.41 0.0003 0.76 3524 13.9 0.02 18.7 0.03 -0.0004 0.77
Excl. Erfurt (past pollution) 4261 4.1 0.27 1.0 0.86 0.0008 0.26 3825 12.5 0.01 11.8 0.09 0.0004 0.65
Excl. Tartu and Basel (LF variability) 4051 4.7 0.18 1.0 0.85 0.0009 0.09 3732 12.9 0.00 11.9 0.10 0.0005 0.53
Women Men
FEV1 FVC FEV1/FVC FEV1 FVC FEV1/FVC

LF = lung function

167

Table 33: Coefficients (p-values) for change in lung function from main models* for complete samples, by sex.
Negative coefficients represent adverse effects on lung function. Coefficients are per 10µg/m
3
sulfur over 1 year.
Units: FEV1 and FVC ml/a, FEV1/FVC /a.
Variable FEV1-change
[ml/a]
FVC-change
[ml/a]
FEV1/FVC
change [/a]‡
FEV1-change
[ml/a]
FVC-change
[ml/a]
FEV1/FVC
change [/a]
Age -3.00 (<0.001) -2.36 (<0.001) -0.000 (<0.001) -3.1 (<0.001) -2.79 (<0.001) -0.000 (0.006)
Height [cm] -0.03 (0.001) -0.01 (0.425) -0.000 (0.099) -0.04 (0.000) -0.03 (0.008) -0.000 (0.403)
Ex-smoker -0.14 (0.183) -0.27 (0.027) 0.000 (0.291) -0.07 (0.625) -0.12 (0.521) -0.000 (0.710)
Smoker -0.29 (0.004) -0.28 (0.019) -0.000 (0.262) -0.24 (0.069) -0.12 (0.485) -0.000 (0.032)
Sulfur [µg/m
3
] 0.47 (0.125) 0.70 (0.059) -0.000 (0.652) 0.52 (0.127) 0.87 (0.038) -0.000 (0.445)
Women (N=2243) Men (N=2031)

* Coefficients for change in lung function are estimated in the same model as coefficients for lung function level (Table 30) and therefore adjusted for those variables.
In addition coefficients for change in lung function are adjusted for BMI (coefficients not shown); † Coefficients for age from the models for lung function level
(Table 2) serve as intercepts for change in lung function; ‡ Model did not converge. Estimates are based on iterated Estimation-Maximization.

168

Table 34: Sulfur coefficients for change in lung function from sensitivity analyses using different adjustment variables.
Negative coefficients represent adverse effects on lung function. Coefficients are per 1µg/m
3
sulfur over 10 years.
Units: FEV1 and FVC ml/10 a, FEV1/FVC /10 a.
FEV1 FVC FEV1/FVC* FEV1 FVC FEV1/FVC
Sub-models N β (p) β (p) β (p) N β (p) β (p) β (p)
Crude 2250 5.1 (0.084) 6.6 (0.054) -0.0001 (0.829) 2040 6.8 (0.053) 9.3 (0.024) -0.0002 (0.591)
Height only 2250 4.9 (0.090) 7.6 (0.023) -0.0003 (0.536) 2040 5.6 (0.097) 9.1 (0.027) -0.0004 (0.392)
Minimal 2245 5.0 (0.090) 7.6 (0.025) -0.0003 (0.585) 2037 5.7 (0.086) 9.0 (0.030) -0.0004 (0.453)
Main 2243 4.7 (0.125) 7.0 (0.059) -0.0002 (0.652) 2031 5.2 (0.127) 8.7 (0.038) -0.0004 (0.445)
Center level adj. 2243 2.5 (0.546) 0.9 (0.839) 0.0006 (0.306) 2031 3.5 (0.449) 1.2 (0.786) 0.0005 (0.369)
Age squared 2245 4.0 (0.172) 6.0 (0.063) -0.0002 (0.676) 2037 5.4 (0.094) 8.4 (0.024) -0.0003 (0.488)
Maximal 2227 3.8 (0.237) 6.1 (0.096) -0.0003 (0.591) 2018 4.3 (0.258) 7.8 (0.074) -0.0004 (0.404)
Women Men

Crude = PM
2.5
; Height only = + height; Minimal = + smoking status (never, ex, current); Main = + BMI; Age squared = + age
2
; Centre level adj. = main + center
means of "Education level (age)", "Proportion of non-manual professions", "ETS"; Maximal = main + long and short term respiratory symptoms, height squared; *
Models did not converge. Estimates are based on iterated Estimation-Maximization.

169

Table 35: Sulfur coefficients for change in lung function from sensitivity analyses using different subsamples.
Negative coefficients represent adverse effects on lung function. Coefficients are per 1µg/m
3
sulfur over 10 years.
Units: FEV1 and FVC ml/10 a, FEV1/FVC /10 a.
Sub-sample N B p B p B p N B p B p B p
Never smokers 1007 5.7 0.05 9.3 0.01 -0.001 0.32 745 3.7 0.33 6.9 0.17 -0.001 0.37
Ever smokers 1234 4.2 0.25 5.4 0.24 0.000 0.88 1283 6.9 0.04 9.8 0.02 0.000 0.72
Non-asthmatics 1939 5.5 0.06 7.6 0.03 0.000 0.74 1833 5.9 0.12 8.7 0.04 0.000 0.55
Asthmatics 300 0.8 0.87 1.8 0.80 0.000 0.49 196 -0.5 0.93 8.5 0.28 -0.002 0.16
Asymptomatics 968 6.1 0.04 7.5 0.06 0.000 0.96 908 8.7 0.05 9.3 0.06 0.000 0.55
Symptomatics 1275 3.9 0.24 7.0 0.06 0.000 0.50 1123 3.2 0.32 8.5 0.04 -0.001 0.08
Biomedin spirometers only 1122 -3.4 0.48 -2.8 0.65 0.000 0.83 1057 -4.8 0.30 -2.1 0.75 0.000 0.59
Northern centers 871 13.1 0.03 21.1 0.02 -0.001 0.59 716 19.9 0.01 21.5 0.05 0.000 0.75
Central centers 618 21.0 0.00 34.6 0.00 -0.002 0.16 541 18.0 0.22 33.3 0.01 -0.002 0.26
Southern centers 755 -4.4 0.52 -2.5 0.74 0.000 0.67 774 -5.3 0.37 -4.2 0.54 0.000 0.76
Mid range of height only 1212 5.1 0.13 7.7 0.04 0.000 0.63 1119 8.1 0.02 11.1 0.02 0.000 0.49
Longterm residents 1132 6.0 0.05 8.2 0.03 0.000 0.74 956 9.6 0.01 12.7 0.00 0.000 0.91
Movers 1108 4.4 0.19 6.9 0.12 0.000 0.52 1074 2.8 0.46 6.0 0.25 -0.001 0.38
Excl. 4 monitors close to traffic 1965 11.9 0.00 13.5 0.01 0.000 0.72 1762 13.5 0.00 13.4 0.02 0.000 0.67
Excl. Erfurt (past pollution) 2131 4.7 0.13 7.0 0.06 0.000 0.68 1913 5.0 0.15 8.8 0.03 0.000 0.41
Excl. Tartu and Basel (LF variability) 2026 5.0 0.12 7.0 0.06 0.000 0.79 1866 5.8 0.08 9.2 0.04 0.000 0.52
Women Men
FEV1 FVC FEV1/FVC FEV1 FVC FEV1/FVC

LF = Lung function

170

Appendix 9: Sensitivity Analyses using Alternative Exposure Markers.
Table 36: Sensitivity analyses using alternative exposure markers in the main model (Effects on lung function level).
Effects for PM
2.5
and sulfur are for seasonal mean estimates. Effects for remaining elements and hydroxyl-radical formation are for annual mean estimates. Effects
for NO
2
at home are for one or the average of two measurements over 14 days at home outdoors. Negative coefficients represent adverse effects on lung function.
Coefficients are per unit indicated. Units for FEV1 and FVC are ml.
Effects on lung function level N B p B p N B p B p
PM2.5 summer [ug/m3] 2243 2.13 0.51 0.69 0.88 2031 7.73 0.08 10.21 0.09
PM2.5 winter [ug/m3] 2243 0.17 0.87 -0.68 0.64 2031 1.35 0.38 0.94 0.64
Sulfur summer [ng/m3] 2243 0.02 0.52 0.01 0.80 2031 0.08 0.04 0.09 0.07
Sulfur winter [ng/m3] 2243 0.02 0.45 0.00 1.00 2031 0.07 0.12 0.06 0.32
Absorbance [abs.coeff./m] 2243 6.31 0.70 -5.15 0.82 2031 39.46 0.08 41.22 0.17
Copper [ng/m3] 2243 -0.44 0.84 -1.23 0.68 2031 4.74 0.12 5.66 0.17
Iron [ng/m3] 2243 0.15 0.51 0.06 0.86 2031 0.76 0.01 0.77 0.07
Manganese [ng/m3] 2243 2.74 0.20 1.62 0.60 2031 9.17 0.00 9.83 0.01
Zinc [ng/m3] 2243 0.15 0.65 0.05 0.91 2031 1.00 0.02 1.33 0.02
OH-formation [AU/m3] 2166 39.86 0.46 17.28 0.82 1948 102.72 0.12 122.87 0.21
NO2 central [ug/m3] 2243 -0.06 0.95 -0.06 0.96 2031 1.21 0.38 2.23 0.21
NO2 at home [ug/m3] 636 -0.57 0.46 -0.76 0.40 547 1.46 0.16 1.04 0.41
Women Men
FEV1 FVC FEV1 FVC

AU = arbitrary unit

171

Table 36 continued: Sensitivity analyses using alternative exposure markers in the main model (Effects on change in lung function).
Effects for PM
2.5
and sulfur are for seasonal mean estimates. Effects for remaining elements and hydroxyl-radical formation are for annual mean estimates. Effects
for NO
2
at home are for one or the average of two measurements over 14 days at home outdoors. Negative coefficients represent adverse effects on lung function.
Coefficients are per unit indicated and year. Units for FEV1 and FVC are ml.
Effects on change in lung function N B p B p N B p B p
PM2.5 summer [ug/m3] 2243 0.49 0.06 0.72 0.02 2031 0.55 0.06 0.95 0.01
PM2.5 winter [ug/m3] 2243 -0.02 0.85 0.03 0.76 2031 -0.05 0.66 0.08 0.54
Sulfur summer [ng/m3] 2243 0.00 0.07 0.01 0.03 2031 0.01 0.03 0.01 0.00
Sulfur winter [ng/m3] 2243 0.00 0.66 0.00 0.45 2031 0.00 0.77 0.00 0.38
Absorbance [/m] 2243 1.20 0.39 1.67 0.33 2031 0.77 0.63 2.79 0.15
Copper [ng/m3] 2243 0.14 0.45 0.17 0.47 2031 0.03 0.89 0.26 0.33
Iron [ng/m3] 2243 0.01 0.62 0.01 0.67 2031 0.00 0.94 0.03 0.32
Manganese [ng/m3] 2243 0.16 0.41 0.22 0.37 2031 0.16 0.46 0.42 0.12
Zinc [ng/m3] 2243 0.03 0.40 0.02 0.66 2031 0.03 0.44 0.05 0.27
OH-formation [AU/m3] 2166 3.51 0.46 4.08 0.49 1948 3.54 0.51 6.85 0.31
NO2 central [ug/m3] 2243 0.07 0.38 0.10 0.31 2031 0.03 0.73 0.15 0.19
NO2 at home [ug/m3] 636 0.02 0.75 0.02 0.73 547 0.12 0.04 0.12 0.12
Women Men
FEV1 FVC FEV1 FVC

AU = arbitrary unit

172

Appendix 10: Sensitivity Analysis using a “SAPALDIA-like” Model
Table 37: Results from Sensitivity Analysis using a “SAPALDIA-like” Model.
Log-transformed cross-sectional model at follow-up. Coefficients for PM
2.5
are to be interpreted as percent differences per 10µg/m
3
. For example, an increase in
PM
2.5
by 10µg/m
3
was (non-significantly) associated with 0.7% increase in FEV1 among women. The only negative coefficient for PM
2.5
was -0.012% per10µg/m
3

PM
2.5
for FVC among women (non-significant).
Variable ln FEV1 [ml] ln FVC[ml] ln FEV1/FVC ln FEV1[ml] ln FVC[ml] ln FEV1/FVC
Intercept -5.556 -5.605 0.117 -3.0479 -4.583 1.492
Sensitisation -0.014 (0.059) -0.0046 (0.493) -0.009 (0.028) -0.013 (0.053) -0.001 (<0.000) -.0124 (0.002)
ln Height 1.814 (<0.000) 2.117 (<0.000) -0.291 (<0.000) 1.866 (<0.000) 2.249 (<0.000) -.384 (<0.000)
ln weight -0.042 (0.017) -0.068 (<0.000) 0.025 (0.011) -0.101 (<0.000) -0.129 (<0.000) 0.029 (0.031)
ln age 3.568 (<0.000) 3.548 (<0.000) -0.019 (0.966) 2.379 (0.003) 3.172 (<0.000) -0.771 (0.096)
ln age
2
-0.528 (<0.000) -0.507 (<0.000) -0.016 (0.792) -0.361 (0.001) -0.455 (<0.000) 0.091 (0.147)
Smoking (py) -0.001 (<0.000) -0.000 (0.248) -0.001 (<0.000) -0.001 (<0.000) -0.001 (0.001) -0.001 (<0.000)
PM
2.5
[µg/m
3
]
0.007 (0.299) -0.000 (0.986) 0.006 (0.886) 0.002 (0.237) 0.005 (0.458) 0.002 (0.774)
Women (N=1840) Men (N=1704)

173

Abstract (if available)

Abstract Lung function is an important measure of respiratory health and a predictor of cardio-respiratory morbidity and mortality. More than fifty publications have addressed long term effects of ambient air pollution on lung function. The vast majority of studies reported some significant adverse effects on lung function, including several studies on traffic. There is strong support for air pollution effects on lung function growth in children, while in adults the evidence is limited to comparisons of lung function levels and knowledge on effects on lung function decline remains inconclusive. The diversity in study designs and investigated air pollution and lung function measures is high, limiting the comparability of studies and the ability to draw quantitative conclusions.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Ambient air pollution and lung function in children

PDF

Examining exposure to extreme heat and air pollution and its effects on all-cause, cardiovascular, and respiratory mortality in California: effect modification by the social deprivation index

PDF

Spatial modeling of non-tailpipe emissions and its association with children's lung function

PDF

Airway inflammation and respiratory health in the Southern California children's health study

PDF

Persistence of pollution-induced lung function deficits in early adulthood: evidence from the Children's Health Study

PDF

Prenatal air pollution exposure, newborn DNA methylation, and childhood respiratory health

PDF

A cohort study of air-pollution and childhood obesity incidence

PDF

Association of traffic-related pollution and stress on childhood lung function

PDF

Air pollution, mitochondrial function, and growth in children

PDF

Air pollution and childhood obesity

PDF

Spatial analysis of PM₂.₅ air pollution in association with hospital admissions in California

PDF

Evaluation of new methods for estimating exposure to traffic-related pollution and early health effects for large population epidemiological studies

PDF

Genetic variation in inducible nitric oxide synthase promoter, residential traffic related air pollution and exhaled nitric oxide in children

PDF

Disparities in exposure to traffic-related pollution sources by self-identified and ancestral Hispanic descent in participants of the USC Children’s Health Study

PDF

A pilot study of a global approach to assessing air pollution exposure in port communities: passive air monitoring of nitrogen dioxide concentrations

PDF

Associations between ambient air pollution and hypertensive disorders of pregnancy

PDF

Air pollution and breast cancer survival in California teachers: using address histories and individual-level data

PDF

Chemical and toxicological characteristics of particulate matter in urban environments with a focus on its sources, associated health impacts and mitigation policies

PDF

The effect of renal function on toxicity of E7389 (eribulin) among patients with bladder cancer

PDF

Effect of biomass fuel exposure on infant respiratory health outcomes in Bangladesh

Asset Metadata

Creator Thomas Götschi (author)

Core Title Long term effects of air pollution on lung function in the European community respiratory health survey

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Publication Date 07/30/2007

Defense Date 05/25/2007

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag adults,Air pollution,Europe,FEV1,FVC,long term,longitudinal,lung function,OAI-PMH Harvest,PM2.5

Place Name Europe (continents)

Language English

Advisor Kuenzli, Nino (committee chair), Gauderman, W. James (committee member), Gilliland, Frank D. (committee member), Jerrett, Michael (committee member), Sioutas, Constantinos (committee member)

Creator Email gotschi@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m718

Unique identifier UC1451752

Identifier etd-Gotschi-20070730 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-540202 (legacy record id),usctheses-m718 (legacy record id)

Legacy Identifier etd-Gotschi-20070730.pdf

Dmrecord 540202

Document Type Dissertation

Rights Thomas Götschi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu