Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Exposure of chronic obstructive pulmonary disease patients to particulate air pollution : an assessment… Fisher, Teri Vered 1999

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1999-0505.pdf [ 6.75MB ]
Metadata
JSON: 831-1.0089170.json
JSON-LD: 831-1.0089170-ld.json
RDF/XML (Pretty): 831-1.0089170-rdf.xml
RDF/JSON: 831-1.0089170-rdf.json
Turtle: 831-1.0089170-turtle.txt
N-Triples: 831-1.0089170-rdf-ntriples.txt
Original Record: 831-1.0089170-source.json
Full Text
831-1.0089170-fulltext.txt
Citation
831-1.0089170.ris

Full Text

E X P O S U R E OF C H R O N I C O B S T R U C T I V E P U L M O N A R Y D I S E A S E P A T I E N T S T O P A R T I C U L A T E AIR P O L L U T I O N : A N A S S E S S M E N T OF R E S P I R A T O R Y H E A L T H E F F E C T S by TERI V E R E D F ISHER B . S c , McGi l l University, 1997 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF M A S T E R OF S C I E N C E in T H E F A C U L T Y OF G R A D U A T E STUDIES Department of Experimental Medicine We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y OF BRITISH C O L U M B I A October 1999 © Teri Vered Fisher, 1999 In presenting this thesis in partial fulfillment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Experimental Medicine The University of British Columbia Vancouver, Canada October 12, 1999 A B S T R A C T Epidemiological studies suggest associations between particulate air pollution and health effects, including decreased lung function, and increased respiratory symptoms and medication use. Studies also suggest that persons with cardiopulmonary disease may be at increased risk for particle-related health effects. This study measured personal and ambient fine particulate ( P M 2 . 5 ) and sulfate concentrations for chronic obstructive pulmonary disease (COPD) patients, and evaluated the relationships between exposure and lung function, respiratory symptoms, and medication use. This study was conducted during the summer of 1998 in the Greater Vancouver Regional District, British Columbia. Sixteen (7 male, 9 female) non-smoking COPD patients (mean age: 74, range: 54-86) were recruited from the Vancouver Hospital Respiratory Clinic (13) and B.C. Lung Association (3). Each subject was equipped with a P M 2 . 5 monitor, activity diary, and medication log for seven 24-hour periods (approximately 1.5 weeks apart). Before and after sampling periods, subjects performed forced expiratory maneuvers using portable spirometers. Following sampling periods, subjects completed symptom questionnaire interviews. Ambient P M 2 . 5 measurements were obtained on days corresponding to personal sampling at 5 sites within the study area. Al l samples were analyzed for P M 2 . 5 mass and sulfate concentrations. Mean personal and ambient PM 2 . 5 (and sulfate) levels were 18 and 11 (1.5 and 1.9) pg/m3, respectively. No associations between air pollution and respiratory health outcomes were statistically significant. However, weighted regression analyses suggest that up to a 1% decline in F E V i may be associated with each 10 pg/m3 increase in ambient P M 2 . 5 , and 0.27-1.04% declines in F E V i may be associated with each pg/m3 increase in personal and ambient sulfate. These associations were not observed with respect to personal P M 2 . 5 exposure, suggesting that ambient P M 2 . 5 , personal sulfate, and ambient sulfate were more strongly associated with adverse health effects than was personal P M 2 . 5 . Ozone also appeared to be associated with decreased F E V i . Consequently, it was not possible to differentiate between the effects of particulates and ozone. No consistent associations were observed between any exposure measure and symptom severity or bronchodilator use. Major limitations of this study included the small degree of variability in exposure, and the quality of symptom questionnaire and bronchodilator use data. I l l TABLE OF CONTENTS A B S T R A C T ii T A B L E O F C O N T E N T S iii L IST OF T A B L E S vi L IST OF F IGURES viii A C K N O W L E D G M E N T S ix 1. I N T R O D U C T I O N 1 1.1 Overview 1 1.2 Particulate Matter: Characteristics 1 1.3 Particulate Matter: Health Effects 2 1.4 Epidemiological Study Limitations 6 1.5 Study Population 8 1.6 Study Objectives 9 1.7 Study Design 10 2. M E T H O D S H 2.1 Study Overview 11 2.2 Subjects 12 2.2.1 Recruitment 12 2.2.2 Randomization and Scheduling of Sampling Days 12 2.3 Personal and Ambient Sampling 13 2.3.1 Gravimetric Procedures 13 2.3.2 Gravimetric Quality Control Filters 14 2.3.3 Exposure Monitoring Quality Control Experiments 14 2.3.4 Spirometry Quality Control Experiments 15 2.3.5 Sulfate Analysis 15 2.3.6 Personal Sampling Preparation 16 2.3.7 Personal Sampling 17 2.3.8 Ambient Sampling Preparation 20 2.3.9 Ambient Sampling 22 2.3.10 Dichotomous Sampler Sampling 23 2.4 Exposure Sampling Data Analysis 23 2.4.1 Particulate Sampling 23 2.5 Lung Function Data Analysis 24 2.5.1 Spirometry Maneuver Selection 24 2.5.2 Correlation Between Lung Function Variables 25 2.5.3 Outcome Variables: Post Sampling F E V i and A F E V i 25 2.5.4 Regression Models 25 2.5.5 Individual Ordinary Least Squares Regression Analyses 26 2.5.6 Pooled Ordinary Least Squares Regression Analyses 26 iv 2.5.7 Pooled Weighted Least Squares Regression Analyses 26 2.5.8 Power Calculations 27 2.5.9 Confounder Analyses 27 2.6 Symptom Questionnaire Data Analysis 28 2.7 Bronchodilator Use Data Analysis 29 3. R E S U L T S 30 3.1 Subjects 30 3.1.1 Recruitment Results 30 3.1.2 Subject Characteristics 30 3.1.3 Compliance 31 3.2 Exposure Sampling 31 3.2.1 Balance Room Conditions 31 3.2.2 Exposure Monitoring Quality Control Results 31 3.2.3 Study Data: Clean Up 33 3.2.4 Personal Sampling Results 34 3.2.5 Ambient Sampling Results 36 3.3 Lung Function Testing 39 3.3.1 Spirometry Quality Control 39 3.3.2 Spirometry Database 40 3.3.3 Correlation Between Lung Function Variables 41 3.4 Lung Function Regression Analyses Results 43 3.4.1 Individual Ordinary Least Squares Regression 43 3.4.2 Pooled Ordinary Least Squares Regression 46 3.4.3 Pooled Weighted Least Squares Regression 47 3.4.4 Power Calculations 48 3.4.5 Confounder Results 49 3.5 Symptom Questionnaire Results 50 3.5.1 Symptom Database 50 3.5.2 Pooled Logistic Regression Analyses 51 3.6 Bronchodilator Use Results 53 3.6.1 Bronchodilator Use Database 53 3.6.2 Bronchodilator Use Analyses Results 54 4. D ISCUSSION 56 4.1 Overview of Results and the Implications 56 4.1.1 Personal and Ambient Fine Particulate and Sulfate Exposure 56 4.1.2 Particulate Pollution and Lung Function 56 4.1.3 Particulate Pollution and Respiratory Health Symptoms 59 4.1.4 Particulate Pollution and Bronchodilator Use 61 4.2 Factors Affecting Study Outcome 61 4.2.1 Exposure Assessment 61 4.2.2 Pollution Levels 62 4.2.3 Study Population 63 4.2.4 Sample Size 63 4.3 Conclusions 63 V 4.4 Recommendations 64 B I B L I O G R A P H Y 66 A P P E N D I X A: I N T R O D U C T O R Y L E T T E R 71 A P P E N D I X B: C O N S E N T F O R M 73 A P P E N D I X C: F I L T E R WEIGHING F O R M 77 A P P E N D I X D: P E R S O N A L F L O W L O G 79 A P P E N D I X E: A C T I V I T Y D I A R Y 81 A P P E N D I X F: S Y M P T O M Q U E S T I O N N A I R E 83 A P P E N D I X G: M E D I C A T I O N C H E C K L I S T 85 A P P E N D I X II: A M B I E N T F L O W L O G 87 A P P E N D I X I: K I T S I L A N O A M B I E N T C O N C E N T R A T I O N S 89 A P P E N D I X J : F V C R E S U L T S 91 A P P E N D I X K: INDIV IDUAL L E A S T S Q U A R E S R E G R E S S I O N R E S U L T S 92 A P P E N D I X L: C O R R E L A T I O N COEFFIC IENTS F O R S A M P L I N G SESSIONS A N D I N T E R V A L S 96 A P P E N D I X M: P O O L E D O R D I N A R Y L E A S T S Q U A R E S R E G R E S S I O N S C A T T E R P L O T S 99 A P P E N D I X N: E S T I M A T E D V A R I A N C E S F O R P O O L E D W E I G H T E D L E A S T S Q U A R E S R E G R E S S I O N A N A L Y S E S 103 LIST OF T A B L E S Table 2.1: Study Measurements 11 Table 2.2: Lung Function Models 26 Table 2.3: Symptom Models 28 Table 3.1: Results of Trial Experiments 1.1-1.2 32 Table 3.2: Results of Trial Experiments 2.1-2.4 33 Table 3.3: Personal PM 2 .5 Exposure Summary 35 Table 3.4: Personal Sulfate Exposure Summary 35 Table 3.5: Ambient PM 2 . 5 Concentration Summary 37 Table 3.6: Ambient Sulfate Concentration Summary 37 Table 3.7: Spirometer Quality Control Experiment Results 39 Table 3.8: Number of Maneuvers per Session 40 Table 3.9: Individual F E V , Results 41 Table 3.10: Pearson Correlation Coefficients for Pre and Post-Sampling Outcome Measurements 41 Table 3.11: Lung Function Models 45 Table 3.12: Summary of Individual Regression Models 46 Table 3.13: Pooled Ordinary Least Squares Regression Results 47 Table 3.14: Pooled Weighted Least Squares Regression Results 47 Table 3.15: Absolute and Percentage Changes in FEVi per 10 pg/m3 increase in PM 2 . 5 48 Table 3.16: Absolute and Percentage Changes in FEVi per 1 pg/m3 increase in SO,2" 48 Table 3.17: Lung Function Power Calculations for P M 2 5 Models 48 Table 3.18: Lung Function Power Calculations for S0 4 2 ' Models 49 vii Table 3.19: Correlation Coefficients among Exposure Measures 49 Table 3.20: Ozone and Personal Sulfate Regression Results 50 Table 3.21: Summary of Questionnaire Responses 51 Table 3.22: Summary of Adjusted Questionnaire Responses 51 Table 3.23: Pooled Logistic Regression Results 52 Table 3.24: Pooled Logistic Regression Results (Excluding Subject #9) 52 Table 3.25: Number of Bronchodilator Uses per Sampling Session 53 Table 3.26: Individual P M 2 . 5 vs. Bronchodilator Use Regression Results 55 LIST O F F I G U R E S Figure 2.1: Subject Equipped with Personal P M 2 . 5 Exposure Monitor, Holter Monitor, and Activity Log 18 Figure 2.2: Ambient Fine Particulate Monitoring Sites 21 Figure 3.1: Boxplot of Personal PM 2 . 5 Exposures 36 Figure 3.2: Boxplot of Personal Sulfate Exposures 36 Figure 3.3: Boxplot of Ambient PM 2 . 5 Concentrations 38 Figure 3.4: Boxplot of Ambient Sulfate Concentrations 38 Figure 3.5: Scatterplot of Pre vs. Post-Sampling F E V i 42 Figure 3.6: Scatterplot of Pre vs. Post-Sampling F V C 42 Figure 3.7: Scatterplot of Pre vs. Post-Sampling F E V i / F V C 43 Figure 3.8: Boxplot of Individual Post-FEVi Distributions - Raw Data 44 Figure 3.9: Boxplot of Individual Post-FEVi Distributions - Standardized Data 44 Figure 3.10: Boxplot of Individual A F E V i Distributions 45 Figure 3.11: Number of Bronchodilator Uses 54 Figure 3.12: Scatterplot of Bronchodilator Use vs. Personal PM 2 . 5 by Subject 54 ix A C K N O W L E D G M E N T S I would like to thank and acknowledge the subjects in this study who all generously and conscientiously donated seven days of personal sampling to this project. I want to express my thanks to the members of the staff of the Greater Vancouver Regional District Air Quality Department for their work on the ambient PM2.5 monitoring component of this study. Thank you also to Parveen Bhatti for his work on the ambient sampling; this project would not have been completed on schedule without him. Special thanks to Jochen Bruum for all of his guidance and help in the statistical analyses. I am very thankful to my colleague and friend, Stefanie Ebelt, who encouraged and supported me throughout the entire project. I want to say thank you to two members of my supervisory committee, Dr. Sverre Vedal, and Dr. A . John Petkau for their continued support and advice throughout the study. I am appreciative. Finally, my sincerest thanks to my supervisor, Dr. Michael Brauer, for his encouragement, support, guidance, and friendship which has continued unabated from the commencement of the project two years ago until the present. I am most grateful. 1 1.< I N T R O D U C T I O N 1.1 Overv iew More than 50 years ago, associations between human health and air pollution were observed during the dramatic pollution disasters such as the London Fog of 1952 (Logan 1953). Since then, interest has grown steadily among researchers and policymakers to attempt to answer the questions surrounding the relationships between air pollution and adverse health effects. Some of the more recent studies have indicated that current levels of air pollution common to major cities are associated with adverse acute and chronic health outcomes. These outcomes have included increased respiratory symptoms, decreased lung function, increased hospitalizations and other health care visits for respiratory and cardiovascular disease, and increased cardiopulmonary disease mortality (Pope et al. 1995a). These studies have been carried out in a variety of locations, and have been based on a number of different study designs. Air pollutants are ubiquitous by nature and they are comprised of many different classes. Some of the more common pollutants include gases such as carbon monoxide, nitrogen dioxide, ozone and sulphur dioxide, and particles of various sizes. Numerous researchers have recently investigated the various aspects of health effects resulting from exposure to a number of these pollutants, including particles of different size fractions (Lebowitz 1996). However, research into the health effects of exposure to particles less than or equal to 2.5 urn in diameter (PM2.5) is lacking. This is a crucial deficiency in particulate pollution/health effect studies, as it is believed that this size fraction of particles causes more severe health effects than the other pollutants (as described below). For this reason, PM2.5 was selected as the focus for this study. To increase the sensitivity of the study, a population particularly susceptible to adverse health effects was selected, namely elderly people with moderate chronic obstructive pulmonary disease. Therefore, this study makes a contribution to the available information regarding the potential health effects to patients with chronic obstructive pulmonary disease as a result of exposure to fine particulate matter (PM2.5) in the Greater Vancouver Regional District (GVRD). 1.2 Part iculate Matter: Character ist ics Particulate air pollution is composed of an air-suspended mixture of solid and liquid particles of many shapes and sizes. These pollutants can be classified into primary or secondary aerosols. Primary aerosols are those that are emitted directly into the atmosphere from a natural source. Secondary particles originate as gasses. Once these gasses have been emitted from a source, they are converted into particles in the atmosphere (Wilson and Spengler 1996). Some examples of natural sources of particles include marine aerosols, dust, soil, and volcanoes, while anthropogenic sources include soot, vehicle exhaust, and the conversion of gases from the combustion of fossil fuels. Atmospheric aerosols consist of many differently sized particles ranging from less than 0.1 urn to greater than 100 u.m (Wilson and Suh 1997). The size of the particle is in fact the single most important characteristic in determining the origin, composition, air residence 2 time, and removal of that aerosol. The fine particle fraction ( P M 2 . 5 ) consists of tiny aerosol particles having aerodynamic diameters less than or equal to 2.5 pm. These particles are most often formed by the coagulation of ultrafine particles and by gas-to-particle conversions. The typical components of P M 2 . 5 created by gas-to-particle conversions in the industrialized countries are sulfate (SO4 2 ) from sulfur dioxide ( SO2 ) , nitrate (NCV) from nitrogen oxides (NO x ) and ammonium ( N H / ) from ammonia (NH 3). Acids, metal salts, and elemental and organic carbon are also major components of PM 2 . 5 (Wilson and Spengler 1996). One of the most important characteristics of fine particles is that due their size, they tend to remain airborne for extended periods of time (up to weeks) and can be carried over long distances (hundreds to thousands of kilometres). Therefore, once these particles have entered the air, it is extremely difficult for them to settle out. Furthermore, fine particulates are a major cause of smog as the particles are effective light absorbers and diffusers ( G V R D 1997). Fine particles can be found in the air throughout the Greater Vancouver Regional District, as there are many outdoor and indoor sources of this pollutant throughout the B.C. lower mainland. Typical sources of this pollutant include combustion of coal, oil, gasoline, diesel, and wood (high temperature processes). In addition, P M 2 . 5 exists as a result of atmospheric transformation products of N O x , SO2 and organics (Wilson and Spengler 1996). The major sources of particulate matter in Greater Vancouver are the commercial and industrial sites with discharges regulated by G V R D permits and other stationary sources with significant air emissions (point sources). These are followed by mobile sources such as light-and heavy-duty motor vehicle exhaust, ships, planes, trains, and off-road engines ( G V R D 1994). Some examples of indoor sources of PM 2 . 5 include cooking and environmental tobacco smoke (ETS) (Wallace 1996). The concentration of PM 2 . 5 at any one time in Greater Vancouver depends on many factors, some of which include the season, the meteorology, the geography, the time of day, the number of cars on the road, etc. For Greater Vancouver, the geography plays a major role in the restriction of air movement in the region. Air is trapped, in the Lower Fraser Valley between mountains to the north and south-east, and consequently, under certain conditions, the air quality can deteriorate to dangerous levels ( G V R D 1994). Greater Vancouver is one of the fastest growing regions in North America and the corresponding potential increase in fine particulate emissions could have a severe impact on the air quality in the near future. Also, because PM 2 . 5 is believed to exacerbate illnesses (as described below), the G V R D has stated that "fine particulate pollution is now considered as high priority issue in the region as summer ozone smog" ( G V R D 1994). 1.3 Particulate Matter: Health Effects Many epidemiological studies have successfully demonstrated a significant association between particulate air pollution and health effects. These health effects include (by decreasing severity) mortality, increased hospitalizations, increased emergency room visits, increased respiratory symptoms, and decreased lung function (Vedal 1997). Until recently, most of the epidemiological studies focused on PM1 0 (i.e. particles having aerodynamic diameters less than or equal to 10 pm) as the exposure measurement and acute daily mortality as the outcome variable. Many of the most striking longitudinal time series studies involving the health effects of particulate pollution were based on this association, some of which produced statistically significant results. One example of such a study was 3 performed by Dockery et al (1992) in St. Louis, M O and Kingston, K N in which 1.5% and 1.6% increases in mortality for each 10 pg/m 3 increase in PMio were found. Similarly, in another study carried out in the Utah Valley (Pope et al. 1992), for each 10 pg/m 3 increase in PMio, 1.5%, 1.8%>, and 3.7% changes in daily mortality were observed for total, cardiac, and respiratory deaths, respectively. Schwartz similarly found that increases in total, cardiac, and respiratory deaths were associated with increases in PMio in Birmingham, A L (Schwartz 1993). It is very important to note that when the mortality rates were categorized according to causes of death, almost all studies found results where the respiratory deaths were the highest. Numerous authors have documented these results (Lebowitz 1996). It is therefore hypothesized that exposure to particulate matter is associated most strongly with respiratory illness and consequently, respiratory health is the main focus of the current study. Many of the relevant epidemiological studies discussed above have also been designed to control for potential confounders, including temperature, month, season, day of the week, and other pollutants (e.g. ozone, sulfur dioxide, nitrogen dioxide) (Ostro et al. 1996, Dockery et al. 1992). Many of these studies suggest that these confounders do not significantly alter the PMio-mortality relationship. Additionally, recent longitudinal cohort studies have controlled for risk factors such as smoking, occupational exposures, and socio-economic status while assessing the association between particle concentrations and mortality (Dockery et al. 1993, Pope et al. 1995b). Furthermore, in addition to Canada and the United States, similar studies have been performed in a number of different countries including Chile (Ostro et al. 1996), England (Schwartz and Marcus 1990), and Greece (Katsouyanni et al. 1990), among many others, and the majority of these have found similar results. The above studies provide strong evidence for the hypothesis that high levels of particulate matter can indeed result in an increase in deaths. Still, one of the greatest controversies surrounding the relationship between particulate air pollution and adverse health effects is the issue of a biological mechanism, as the relationship between low concentrations of particles and adverse health effects is not well understood. Mechanisms can however be suggested at the subclinical level by evaluating the relationships between particulate air pollution and less severe health effects (i.e. decreased lung function, increased respiratory symptoms). A number of studies in the past decade have begun to address this issue by utilizing lung function testing to obtain at least one health outcome variable. Specifically, forced expiratory maneuvers, a series of well-standardized tests, have been used extensively in epidemiological studies due to their good reproducibility, ease of measurement, and correlation with disease state, morbidity, mortality, and functional status (Wise et al. 1995). One of the earlier studies to use this strategy, performed by Hoek et al, showed a decline in lung function as a result of increases in particulate pollution (PMio) (Hoek and Brunekreef 1993). In this study, children (aged 7-12) performed spirometry on a regular basis 3 weeks apart, independent of ambient air pollution levels. Additionally, extra testing sessions were inserted during one high air pollution episode (PMio levels reached 174 pg/m 3). During this episode, F V C , F E V i and maximal mid-expiratory flow were below baseline levels. Pope et al found similar results during their study in which adverse changes in respiratory health were 4 associated with daily changes in PMio (Pope et al. 1991). Elevated PMio levels were associated with increases in reported respiratory symptoms and use of medication and an approximately 3 to 6% decline in peak expiratory flow (PEF). This association was observed even at levels below the US ambient air quality standard of 150 u.g/m3. In another study performed on a group of sensitive individuals (patients with mild to moderate obstructive pulmonary disease), Pope et al found that a statistically significant decline of approximately 2% in F E V i was associated with an increase in PMio of lOOpg/m 3 (Pope and Kanner 1993). Another study that analyzed the association between PMio (and acidic components) and F E V i was one that was carried out with three panels of children in the Austrian Alps (Studnicka et al. 1995). Significant results (similar to the above mentioned studies) were found and the conclusion from this study was that "summer acid haze particles may be associated with transient decreases in lung function". More recently, numerous other studies have been documented that describe similar particulate pollution - health effect relationships. A number of these have been carried out in the Utah Valley, where the average PMio concentration is moderately high and episodes of high levels of pollution are common. A review of these studies (Pope 1996) indicates that many different health effects are associated with PMio in this area. These include decreased lung function, increased respiratory symptoms and hospital admissions, increased school absenteeism and increased mortality. Another study reported in 1996 documented the effects of a number of pollutants (PMio, O3, and N O 2 ) on the pulmonary function of 154 primary school children (Scarlett et al. 1996). Results showed a small statistically significant 1% decrease in F V C as a result of PMio exposure. An interesting observation made during this study was that a similar relationship was not found for ozone or nitrogen dioxide. This evidence further supports the association between respiratory health effects and specifically particulate pollution. The current study attempts to further explore this relationship, specifically at the subclinical level, by determining the strength of association between less severe respiratory health effects (i.e. lung function and symptom severity) and exposure to particulate air pollution. It is important to note that although most of the documented studies concerning the relationship between respiratory health effects and PMio have shown positive results, there are some studies that have shown significant negative results. For example, Roemer et al performed a large study in which daily peak expiratory flow (PEF) measurements and respiratory symptoms of 2,010 children throughout Europe were recorded for at least 2 months (Roemer et al. 1998). These outcome variables were analyzed in relation to daily levels of PMio, SO2 , N O 2 , and black smoke. No statistically significant associations between any pollutants and any health effects were reported. This type of study however, in which the results show no significant associations between particles and health, are indeed less numerous than those that do show such an association. The vast majority of the studies mentioned above have indicated their respective health outcomes in relation to PMio. However, the number of the studies addressing the health effects of PM2 . 5 is much smaller, as databases of PM2 . 5 concentrations are much more limited than those of P M i 0 . As mentioned previously, the size of the particles has a great effect on the toxicity of the pollutant. The reason for this is that particles of different sizes 5 tend to deposit in different regions of the lungs. Larger particles tend to be removed from the airflow as they hit the walls of the upper airways, while smaller particles (the fine fraction) are small enough that they are easily inhaled into the gas-exchange regions of the lungs. There, these particles deposit on airway surfaces by sedimentation and diffusion. This feature of P M 2 . 5 makes these particles particularly toxic as they have the ability to circumvent many of the defense mechanisms in the respiratory tract (e.g. cilia) and can deliver harmful substances to the alveoli surfaces. In a study performed by Churg and Brauer, it was found that fine particles remained in the lung parenchyma once they had been inhaled (Churg and Brauer 1997). In this study, particles found in the upper lobe apical segment parenchyma of lung tissue from 10 autopsies were counted and sized. It was found that 96% of the particles had aerodynamic diameters less than 2.5pm. This data supports the belief that the health effects observed in the PMio studies may be due to the fine particle fraction ( P M 2 . 5 ) contained within the PMio (Wilson and Suh 1997). The effects of particles have also been studied according to a different classification system of particle sizes: fine and coarse. The fine particles are virtually the same as PM 2 . 5 , while the coarse particles are those greater than 2.5 pm but less than 10 pm in aerodynamic diameter. Coarse particles originate from mechanically generated sources such as agricultural processes, mining, and road traffic. One example of a study that was designed to investigate the relationships between mortality and a number of these size fractions of particles (fine, coarse, and PMio), in addition to O 3 , C O , N 0 2 , and SO2 , was accomplished in Toronto, Ontario (Burnett et al. 1999). Daily hospital admissions rates for respiratory, cardiac, cerebral vascular, and peripheral vascular diseases were obtained. Cerebral and peripheral vascular diseases were only weakly associated with the air pollutants. However, PMio, P M 2 . 5 , and coarse particles were associated with 1.9%, 3.3%, and 2.9% respective increases in respiratory and cardiac hospital admissions for each increase of 10 pg/m 3 of the respective particle size fraction. It is important to note that the highest increase in hospital admissions was associated with PM 2 . 5 . This evidence provides additional support for the hypothesis that the fine particle fraction is the most toxic component of particulate matter. This size fraction ( P M 2 . 5 ) is therefore the focus of the current study. There are a number of other studies that support the hypothesis that the fine fraction of particulate matter, as opposed to the coarse fraction, is the main contributor to health effects. One of these studies evaluated the effects of ambient P M 2 . 5 (and O 3 ) on the pulmonary function of hiking adults in New Hampshire (Korrick et al. 1998). Pre- and post-hiking lung function measurements were obtained and the percentage changes in F E V i , F V C , F E V i / F V C , F E F 2 5 & 7 5 % , and P E F R were calculated. Declines in F V C and P E F R were consistently associated with PM 2 . 5 . In another interesting health effect study, statistically significant 1.4% increases in daily mortality were observed for increases in P M 2 . 5 by 10 pg/m 3 (Borja-Aburto et al. 1998). This project, located in Mexico City, also provided data that indicated that the mortality increase was larger for people over 65 years of age. During another project in Mexico City, 40 schoolchildren had their respective peak expiatory flows (PEF) tested both before and after school for 59 days (Gold et al. 1999). Daily changes in P E F were calculated and normalized and then mean PEFs were determined. Additionally, respiratory symptom data was collected. PMio, PM 2 . 5 and ozone were assessed during the same time period. The results indicated a relationship between increased PM 2 . 5 and lagged 6 adverse effects on PEF. Also, higher rates of phlegm were observed with higher levels of pollutants. In a project carried out in Finland, the effects of PMio, P M 2 . 5 , black carbon, and the number of particles with various diameters were analyzed with respect to P E F (obtained at different times) and respiratory symptoms in 49 children with chronic respiratory symptoms (Tiittanen et al. 1999). Data was collected on a daily basis for six weeks. One-day lagged P M 2 . 5 was found to have a significant association with morning PEF. Additionally, evening P E F was significantly related to the 1-day lagged number of particles in the size range 0.1-1.0 urn (corresponding to a portion of P M 2 . 5 ) . Finally, all particulate fractions were significantly associated with an increased risk of cough. These studies together provide a strong argument that P M 2 . 5 is indeed a significant cause of adverse respiratory health effects. 1.4 Epidemiological Study Limitations A number of limitations are evident throughout the particulate/health effect epidemiological studies (Vedal 1995). The first is the issue of implausibility. A critical factor in assessing causality in epidemiology is the finding of similar results in numerous studies. In the case of the relationship between particulate and adverse health effects, multiple studies have indeed obtained similar results, identifying such an association (as discussed above). It can also be argued that there is still no definite biological mechanism by which exposure to low concentrations of particles should result in death. At the present time, this is one of the main criticisms of the study of health effects and particulate air pollution. However, due to the many studies performed in this field (a number of which have been described above), several hypotheses regarding biological mechanisms underlying the association of particulate air pollution and adverse health effects have been proposed. These include: increased susceptibility to infection from impaired host defences; airway inflammation leading to impaired gas exchange and hypoxia; provocation of alveolar inflammation with release of mediators that exacerbate underlying lung disease and increase blood coagulability; and increased lung permeability leading to pulmonary edema (Wilson and Spengler 1996). The second major limitation in the particulate/health effect epidemiological studies is the issue of exposure misclassification. This problem is applicable to many of the studies described above. Exposure misclassification occurs when ambient pollution measurements are used as a surrogate for personal exposures. Consequently, this results in error when determining the exposure of the individuals to the pollutant. The reason for this is that personal exposures are in fact different from ambient measures of particulate pollutants, as many factors (e.g. location, time spent indoors, occupational exposure, exposure to environmental tobacco smoke, etc.) actually affect the true personal exposure. Additionally, much of a person's exposure to particles occurs indoors and therefore, using outdoor ambient measures of P M (particulate matter) as a surrogate for personal exposure results in a finite amount of error. This has been shown in a number of studies. As part of the Particle Total Exposure Assessment Methodology ( P T E A M ) study, Ozkaynak et al compared personal PMio exposures to fixed indoor and outdoor particulate concentrations in California (Ozkaynak et al. 1996). Population weighted daytime personal exposures were found to be 150 ± 9 (SE) pg/m 3 . This was approximately 50% higher than the indoor and outdoor concentrations of 95 ± 6 p.g/m3. This data suggested the existence of 7 the "personal cloud", an air mass close to an individual that has a higher particulate concentration than that of the surrounding air. Specific activities, such as house cleaning and smoking were also found to be associated with higher personal exposure levels (Clayton et al. 1993). A different study, performed by Watt et al compared personal PMio exposure of traffic wardens to ambient measurements over two 4-day periods (Watt et al. 1995). Significant differences were observed as the median concentration for the personal measurements were 123 pg/m 3 and 41 pg/m 3 for the two weeks whereas the ambient concentrations were 10 pg/m 3 and 7.5 u.g/m3 respectively. Another researcher, Janssen, has published numerous studies concerning exposure misclassification. These studies typically have followed subjects over time to determine the extent of correlation between personal exposures and ambient concentrations, in addition to observing the absolute differences between the two. In one of these studies, 45 elementary school children in the Netherlands underwent 4-8 24-hour personal PMio measurements concurrently with outdoor PMio measurements (Janssen et al. 1997). The mean personal 3 3 PMio exposure was 105 pg/m while the outdoor concentration was 38 pg/m . Data was collected on the children's personal activities and sources of particles during the personal sampling as well. It was found that the high personal PMio exposure relative to the ambient concentrations could be attributed to environmental tobacco smoke (ETS) and particles resuspended by personal activities. In another of Janssen's studies, personal, indoor, and outdoor (ambient) PMio measurements were obtained during two periods of 1994, (Janssen et al. 1998). Personal measurements were obtained from 37 non-smoking adults (50 to 70 years old). The mean personal, indoor, and outdoor PMio concentrations were found to be 61.7 Lig/m3, 35.0 pg/m 3 , and 41.5 pg/m 3 , respectively. The Pearson's correlation between personal and outdoor measurements over time however was found to be reasonably high at 0.71 after adjusting for personal exposure to E T S , indicating that E T S is a determinant for personal exposure. In a third study, Janssen investigated the relationships between personal fine particle exposures for 13 children and corresponding ambient concentrations (Janssen et al. 1999). The median Pearson's correlation coefficient was found to be 0.86. However, personal fine particle concentrations exceeded ambient concentrations by approximately 11 pg/m 3 . After excluding children exposed to E T S , this difference was reduced to 5 pg/m 3 . Other studies (Monn et al. 1997) have indeed found similar results indicating the validity of exposure misclassification. With these issues in mind, the current study was designed to obtain personal 24-hour PM 2 . 5 exposures concurrently with ambient PM 2 . 5 samples. Additionally, personal activity diaries were completed to help to elucidate the activities/sources of particles that contribute to personal exposures. The study design also included the collection and analysis of the sulfate components of all P M 2 . 5 samples. Recent studies have suggested that sulfate, a marker of outdoor combustion-source particulate, is a better exposure metric than P M 2 . 5 due to high correlation between personal and ambient sulfate concentrations (Lippmann and Thurston 1996). Furthermore, in comparison to particulate matter, the sulfate component appears to penetrate indoors efficiently, exhibit less spatial variability, and have no major indoor sources (Ozkaynak et al. 1996). For these reasons, in addition to the P M 2 . 5 mass samples, the design of this study included the analysis of the sulfate component of the personal and ambient fine particulate samples. By doing so, the limitation of exposure misclassification 8 found in other studies could be confronted from a variety of perspectives, all the while investigating the associations between respiratory health effects and each of the ambient and personal concentrations of PM 2 . 5 mass and sulfate. 1.5 Study Population The epidemiological studies discussed above have been performed with various populations. Most often it has been the healthy working population that has been selected. However, some studies have investigated the effects of particulate (and other) pollutants on populations that are believed to be at increased risk for, or are more susceptible to, the adverse health effects found to be associated with air pollution. One such study examined the death rates from various causes for the period of 1973-1980 (Schwartz 1994). Specifically, the death rates on the 5% of the days with the highest TSP concentrations (mean 141 pg/m 3) were compared with those from the 5% of days with the lowest TSP concentrations (mean 47 u.g/m3). The relative risk of dying on the high pollution days was found to be 1.08 for all deaths and 1.25 for deaths related to C O P D . In another study, the relationship between mortality and each of PMio, fine and coarse particulate fractions, and sulfate, over 8 years in 6 U.S. cities was analyzed (Schwartz et al. 1996). Increases in PM 2 . 5 by 10 u.g/m3 were associated with a statistically significant 1.5% increase in total daily mortality. However, the same increase in PM2 . 5 was associated with an increase of 3.3% in chronic obstructive pulmonary disease deaths. Studies such as these suggest that people with chronic obstructive pulmonary disease are at increased risk for adverse health effects due to exposure to air pollution. In another type of study, using the Weibel lung morphology model, a computer simulation was performed to compare the particle deposition in healthy subjects with that in C O P D patients (Martonen et al. 1999). C O P D was modeled via airway diameter reduction using airway resistance measurements. Simulations were performed for VT (tidal volumes) of 360, 500, and 1000 ml, respiratory times (T) from 2 to 8 s, and particle sizes of 1, 2, 3, and 5 pm. Results indicated that deposition patterns depended on particular ventilatory parameters, particle sizes and level of disease. For example, when VT = 500 ml and T = 2 s, the deposition of 1 u.m particles was 100% greater for C O P D patients, suggesting that particles in the PM 2 . 5 size range are twice as likely to deposit in C O P D patients compared to healthy subjects. For these reasons, this population (COPD patients) was the chosen group for this study. C O P D is a condition in which there is an obstruction to the air flowing into and out of the lungs. The term chronic obstructive pulmonary disease refers to the conditions of emphysema, chronic bronchitis, or a combination of the two. Emphysema is characterized by the enlargement of the air spaces distal to the terminal bronchiole, with destruction of their walls, while bronchitis refers to a condition in which there is excessive mucous production in the bronchial tree (West 1987). People with this condition have decreased F E V i , V C , F E V i / F V C , FEF25-75%, V m a X 5 o % , and V M A X 7 5 % . Consequently, because the clearance of particles relative to their deposition in the airways varies with breathing rate and pulmonary function level, the deposition of particles is enhanced in patients with C O P D (Bennett et al. 1996). Other susceptible populations are believed to be senior citizens, children, and people who already suffer from cardiopulmonary diseases. 9 A number of particle-related health effect studies have been performed on C O P D patients, in which statistically significant results have been observed. Pope and Kanner assessed the association between lung function and PMio exposure in smoking C O P D patients (Pope and Kanner 1993). Using spirometry data from the Salt Lake City participants in the Lung Health Study (Enright et al. 1991), two lung function testing sessions, 10 to 90 days apart, were related to changes in ambient PMio for the corresponding days. Results showed that on average, an increase of 100 pg/m 3 in PMio resulted in a statistically significant 2.2% decline in F E V i . Another study tested the hypothesis that air pollution may have significant effects on the pulmonary function, symptoms, and medication use of C O P D patients (Harre et al. 1997). Forty subjects completed daily diaries regarding respiratory symptoms, peak expiratory flow (PEF), outdoor activity, visits to medical professionals, and medication use for the same days that ambient PMio, N O 2 , SO2 , and C O were measured. No association was found between lung function and air pollution. However, results indicated that a PMio increase of 35.04 pg/m 3 was associated with a relative risk of 1.38 for night time chest symptoms and a relative risk of 1.15 for increased medication use. These examples are just a few of the studies that support the belief that C O P D patients are at risk for adverse health effects due to exposure to air pollution. Consequently, this population was selected for this study to increase the sensitivity of the outcome. 1.6 Study Objectives This study addresses a number of important knowledge gaps in the published literature concerning the health effects of exposure to particulate air pollution. First, the exposure measures that were chosen (PM2 . 5 and sulfate) are believed to be most strongly associated with adverse health effects, even though the number of studies analyzing this specific relationship has been limited. Furthermore, both personal and ambient concentrations were obtained in this study to observe the relationships between them, and the extent of misclassification in studies reporting ambient concentrations as surrogates for personal exposures. Additionally, this study focuses on respiratory health effects, as these are believed to be most strongly associated with air pollution. Finally, this study was carried out with a population of C O P D patients, a group of subjects who are believed to be at increased risk for health effects. Consequently, the sensitivity of the study was greatly increased from what it would have been had a normal population been utilized. The objectives of this study were: • To measure personal fine particulate (PM2 . 5 ) and sulfate (SO42") concentrations for a population of patients with physician-diagnosed chronic obstructive pulmonary disease (COPD), and ambient levels of these pollutants at 5 Greater Vancouver Regional District air pollution monitoring sites. • To evaluate the relationship between personal and ambient fine particulate matter (PM2 . 5 ) and sulfate, and lung function, respiratory symptoms and medication use. 10 1.7 Study Design This study, carried out during the summer of 1998 in Greater Vancouver, was designed to obtain 24-hour averaged personal and ambient PM2 . 5 and sulfate concentrations. Personal samples were obtained from consenting chronic obstructive pulmonary disease patients, while the concurrent ambient samples were measured at five Greater Vancouver Regional District ah\monitoring sites. Personal exposure samples were accompanied by activity diaries to help elucidate the specific activities/sources that may contribute to higher PM2 . 5 concentrations (Ebelt 1999). The other major component of this study was the collection of health effects data from the C O P D subjects. Spirometry was performed to obtain lung function measurements, symptom questionnaires were completed to evaluate the severity of respiratory symptoms, and medication checklists were filled out to determine the extent of bronchodilator use. This data was collected to test the hypothesis that exposure to fine particulate matter and sulfate in patients with C O P D may lead to adverse health effects, including decreased lung function, increased respiratory symptoms, and increased medication use. 11 2. M E T H O D S 2.1 Study Overv iew This study was conducted from April 21 to September 25, 1998. Both personal and ambient PM 2 . 5 samples were collected. For the personal samples, seven 24-hour averaged PM 2 . 5 exposure measurements were obtained from each participant in the study. Concurrent daily 24-hour averaged ambient PM 2 . 5 concentrations were obtained at five stationary air quality monitoring sites within the Greater Vancouver Regional District (GVRD). Al l samples were analyzed for P M 2 . 5 mass and sulfate concentrations. Personal exposure sampling was also accompanied by a variety of health effect measurements to assess respiratory and cardiac function. Additional personal exposure information was obtained from time-activity logs completed by the subjects during each sampling period and a one-time dwelling characteristics questionnaire. Table 2.1 summarizes the data obtained throughout the study. This thesis focuses on the relationships between the various exposure measurements and the respiratory health effects. Table 2.1: Study Measurements Measurement Frequency Procedure Personal PM 2 . S * 24-hrs; 7 per subject Personal P M 2 . 5 Exposure Monitor Ambient PM 2 . 5 * 24-hrs; 4 per week per site PM 2 . 5 Harvard Impactor Lung function* Pre and post-sampling; 7 per subject Pneumotach Spirometer Symptoms* 7 per subject Post-sampling questionnaire Medication use* 7 per subject Post-sampling medication checklist Oxygen saturation Pre and post-sampling; 7 per subject Pulse Oximeter Electrocardiogram 24-hrs; 7 per subject Cassette Holter Recorder Blood pressure Pre and post-sampling; 7 per subject Blood pressure cuff and stethoscope Time-activity log 24-hrs; 7 per subject Log sheet Dwelling characteristics 1 per subject Interview questionnaire * Focus of this thesis 12 2.2 Subjects 2.2.1 Recruitment Prior to subject recruitment, ethical approval for the study was obtained from the Clinical Research Ethics Board, University of British Columbia. Al l subjects that participated in this study received an honorarium in the amount of $250. The study population consisted of patients with physician-diagnosed chronic obstructive pulmonary disease (COPD). Eligibility criteria of the study limited participation to patients with light to moderate C O P D (defined by F E V i ^ 0.75L). In addition, subjects were to be aged 60 or above, living within the Lower Mainland of British Columbia (excluding the North Shore), and currently not smoking nor living with current smokers. Possible candidates were first made aware of the study by their individual respiratory specialists at the Vancouver Hospital Respiratory Clinic. At this time, interested patients were given introductory letters, briefly summarizing the research project (See Appendix A). Subsequently, the candidates were contacted by phone to schedule introductory meetings. Approximately one month into the data collection portion of the study, a brief presentation regarding the study was made to potential candidates at the B .C. Lung Association "Puffers' Club" to aid in the recruitment of subjects. Introductory letters (identical to those distributed by physicians) were given to interested candidates and names and phone numbers of these people were obtained. Introductory meetings, lasting approximately 45 minutes, were arranged by telephone with interested candidates. During these meetings, the candidates were screened to ensure that they met the selection criteria. At this time, technicians also explained the purposes and objectives of the study in greater detail, demonstrated the use of the required equipment, answered any questions, distributed consent forms (see Appendix B), and obtained informed consent for participation in the study from acceptable candidates. In the case of the people initially contacted by physicians, little screening was required. However, the patients contacted at the Puffers' Club were asked about their respiratory disease history, their age, and the smoking status of themselves and any people who may be living in their homes. Additionally, consenting individuals performed one forced expiratory maneuver using a portable spirometer to verify F E V i acceptability for the study. 2.2.2 Randomizat ion and Schedul ing of Sampl ing Days Each consenting individual was assigned a unique three-digit (ID) number, used for all identification purposes. For each subject, 7 dates, corresponding to each of seven 24-hour sampling periods, were selected. Scheduling was randomised according to the following manner. Each possible sampling day (Monday - Thursday, beginning April 30 and ending September 25) was assigned a unique identification number. Using the Microsoft Excel Version 5.0 random number generator, for each study participant, 7 numbers (corresponding to 7 dates) were selected. Restrictions to this randomisation were as follows. If any two dates for one subject were less than 7 days apart, another number was selected, and sampling days 13 were scheduled with 2 subjects per day on most days, and up to 4 subjects per day when necessary. Furthermore, in some cases, subjects living in certain areas of the G V R D were scheduled on the same days to minimize travel time. Individual schedules were distributed to the subjects at the first sampling session and any conflicts were rescheduled. The day before each sampling session, subjects were telephoned to confirm the appointment for the next day. 2.3 Personal and Ambient Sampling 2.3.1 Gravimetric Procedures All filters used in this study for both personal and ambient particulate monitoring were 2 pm pore-sized Gelman Teflon ("Teflo") filters. Those used for personal sampling measured 37 mm in diameter while the filters used for ambient sampling were 41 mm in diameter. In preparation for all weighing, filters were equilibrated to balance room conditions at least 48 hours prior to weighing. The U B C Occupational Hygiene balance room, containing a number of highly sensitive balances, was continually monitored and maintained at target temperature (20 ± 2 °C) and humidity (50 ± 5% RH) levels. These conditions were recorded on a form each day that filter weighing was performed (see Appendix C). For this study, a Sartorius M3P balance (1 pg resolution, ± 2 pg sensitivity) was used for all weighing. The balance was calibrated monthly with external standard weights. At the beginning of each weighing session, the balance internal calibration sequence was initiated, and subsequently, the balance was zeroed. Using clean forceps, each filter was held by its plastic rim near a radioactive source (Po 2 1 0) for approximately 5 seconds to eliminate any static charge, and then weighed. When the weight was found to have stabilized for 5 seconds, the mass was recorded. These steps were repeated three times for each filter. If the weights were found to differ by 10 pg or more, the entire procedure was repeated. When satisfactory weights were obtained, the filter was stored in a labelled petri plate in the balance room until needed. Similar procedures were followed during post-sampling weighing sessions. The used post-weighed filters were stored (in petri plates) temporarily in the balance room until they were packaged in plastic bags for storage in the laboratory until ion chromatography analysis. Filters were always weighed in a specific order. For 37 mm filters, 3 Q C filters were weighed at the beginning of the weighing session. Then, filters to be used for sampling or as field or lab blanks were weighed. Similarly, for 41mm filters, the same 3 Q C filters were weighed at the beginning of each session. Subsequently, 3 lab blank filters were weighed and finally, 5 filters were weighed per ambient site (i.e. 4 samples and 1 field blank). During post-sampling weighing, all filters that were previously weighed in one session (QC, lab blanks, samples, and field blanks) were all weighed at the same time. 14 2.3.2 Gravimetric Quality Control Filters As a control procedure for verifying the accuracy of the filter weighing performed during this study, three unused filters of each filter size were designated as Quality Control (QC) filters. The Q C filters were weighed prior to each weighing session. Before proceeding with the weighing of study filters, the Q C filter weights were compared with their respective Q C charts for accuracy. Q C charts for each filter were constructed displaying the mean and the 95% and 99.7% confidence limits of all previous weighings; these charts were updated every few weeks with the accumulated weights. If, during a weighing session, Q C weights were found to fall outside these limits and no obvious trends in the weights were found, weighing was continued. Otherwise, further weighing was postponed for another time when the Q C weights were within the limits. Specific filters were also designated as lab blanks. These were to comprise 10% of the total number of used filters. For the personal filters, lab blanks were chosen from the pre-weighed filter set at the same time that field blanks (see below) were prepared. Three filters, identical to those used for ambient sampling, were weighed each week to remain as "ambient" lab blanks. All lab blanks remained in the laboratory for the duration of the study. Field blanks also were to comprise 10% of the total number of filters. Field blanks for personal exposure measurements were prepared by assembling filters in personal exposure monitors and transporting them, together with the other loaded samplers, in sealed plastic containers into the field. One to two personal field blanks were prepared each week. For the ambient filters, all of the spare filters prepared (i.e. 1 per site per week) served as field blanks. 2.3.3 Exposure Monitoring Quality Control Experiments A number of trial experiments to test the operating parameters and limits of the personal and ambient P M 2 . 5 samplers and pumps were completed prior to the commencement of the study. The purpose of the first group of experiments was to determine the relationships between PM 2 . 5 concentrations obtained from personal sampling monitors (PEMs) and those obtained from ambient Harvard Impactors (His). For each of these experiments, various combinations of personal samplers and Harvard Impactors were set-up in a number of different locations. Al l samplers were connected to Universal Sampling Pumps (SKC) and operated at flow rates o f 4 L / m i n ± 10%. During these test trial experiments, field and lab blanks were used. Mean mass differences were calculated for each type of filter, and detection limits were calculated from this data. Two experiments (Experiments 1.1 and 1.2), designed to compare the PM 2 . 5 concentrations obtained from 6 personal PM 2 . 5 exposure monitors (PEMs) with those obtained from 2 ambient Harvard Impactors (His), were carried out in the dining room and basement of a home. During the first experiment, six pumps, powered by battery packs, were connected to PEMs. Two other pumps, operating on A C power, were connected to His and all of these were set-up next to each other in a dining room. The second experiment was identical to the 15 first in terms of its set-up except that the location was in the basement of a home. During this second experiment however, two of the pumps operating on batteries were allowed to continue to run until the batteries were completely drained. This was to determine the length of time that the pumps could operate on power only supplied by the battery packs. During Experiments 2.1 - 2.4, U B C technicians wore two sampling pumps and personal samplers on four different occasions. The purpose of these experiments was to determine the reproducibility in the concentrations reported from the two samplers worn by the same person and subjected to the same environments. 2.3.4 Spirometry Quality Control Experiments At the beginning of this study, it was necessary to determine the length of time that the Tamarac Spirometers were able to remain calibrated. This was crucial as the study was designed in such a way that the spirometers would constantly be moved from one subject's home to the next. Therefore, it was important to determine if the spirometers would need to be recalibrated at each new location. To ascertain if this was the case, test trial calibrations were performed and then a series of simulated expiratory maneuvers followed the calibrations. For these test trials, Spirometers were calibrated using a three-litre syringe in a laboratory in the U B C Occupational Hygiene Department. Each calibration was repeated until the exhale error was < 0.25%. Once the spirometers were successfully calibrated, two different methods were used to test them under various conditions. First, a simulated expiratory maneuver was performed by steadily pushing air from the 3L syringe into the spirometer. Three such individual maneuvers were performed per test. The mean volume and standard deviation were calculated for each test. Second, after the calibration and first series of simulated expiratory maneuvers had been completed, the spirometer was transported to various locations over a number of days. At the various locations, additional simulated expiratory maneuvers were performed without prior calibration of the spirometer. This was to determine if transporting the spirometer, or increasing the length of time between calibration and test, reduced the reproducibility in the results. 2.3.5 Sulfate Analysis Sulfate analysis was conducted on all personal and ambient PM2 . 5 filter samples. As the analysis was a destructive procedure, it was not conducted until all gravimetric measurements were completed and the preliminary PM2 . 5 mass results were studied. Al l equipment was cleaned with soap and distilled/deionized water, and wiped with ethanol before commencing. Ethanol was also used to clean forceps, razor blades, and glass plates between the handling of individual filters. Using cleaned forceps, filters were removed from petri plates and placed onto a glass plate. Then using a razor blade, the plastic rim of each filter was cut in six to eight places, allowing the filters to fold easily. Filters were then placed into new, labelled, plastic screw-top 16 containers. Using a micropipette, 100 pi ethanol were added to each container to completely wet each filter. Five ml of distilled/deionized water were then added and the container was sealed. Containers were sonicated for 15 minutes to force the sulfate on the filters into solution. Ion chromatography (Dionex DX-450) was used to analyze the amount of sulfate in each sample. Syringes (equipped with Gelman Ion Chromatography Acrodisc 0.2u.m filters to prohibit the transfer of filter debris) were used to transfer approximately 0.5 ml of solution from the containers to autosampler vials. Between samples, syringes and filters were rinsed three times with distilled/deionized water. Vials were capped and loaded into trays for analysis by the ion chromatograph (IC). Dionex components of the IC included an Automated Sampler, an Eluent Degas Module, Gradient Pumps, a Self-Regenerating Suppressor (ASRS-I 4mm), an IonPac Column (As4A-SC 4mm), and a Conductivity Detector. Eluent used was a weak anion solution (1.8 m M Na2C03 / 1.7 m M NaHCOa). Samples were run within 24 - 72 hours of preparation in batches of 50 to 70 samples. For each batch of samples, an initial set of five SO4 2 " standards (0.4, 0.8, 2, 4, and 10 u.g/ml, respectively) and blanks were analyzed to generate a calibration curve for the batch. Additionally, after every ninth sample, a standard, at approximately the same concentration as the samples, and a distilled/deionized water blank, were inserted. A second set of calibration standards was included at the end of the batch. Except for a few rare cases, these IC standards and blanks did not drift beyond 15% of the standard concentrations determined at the beginning of the run. Results of the IC runs were obtained through Dionex Chromatography Automation Software (AI-450 Release 3.32). The software computed the sulfate concentrations of samples (in ppm) by comparing the samples' peak areas for sulfate to the calibration curve for that run. Calibration curves were manually checked to ensure high coefficients of determination (i.e. R 2 > 0.999). To overcome negative concentration values for some runs that had detectable sulfate in blank samples, calibration equations for each run were forced through zero. Due to slight shifts in retention time, the software did not identify sulfate peaks for some samples. For these samples, peaks were manually selected to obtain a concentration value. Finally, raw data was exported to Microsoft Excel 7.0 to calculate the air concentrations of sulfate per sample using the air sampling volumes for each respective sample. 2.3.6 Personal Sampling Preparation Pumps used for personal sampling were Universal Sample Pumps by S K C . Additional batteries were attached to extend the duration of time that the pumps could operate to more than 24 hours (up to 36 hours as determined during trial experiments) at a flow rate of 4 L/min. The two batteries were charged on the day prior to use. If time permitted, the batteries were first discharged at this time. Alternatively, batteries were discharged weekly. Personal exposures to PM2 . 5 were measured with personal PM2 . 5 impactors (PEM, M S P Corp.), loaded with 37 mm 2 u.m pore size Teflon filters (Gelman Teflo, R2PJ037) and connected to six-inch long aluminium inlets. The samplers each consisted of a bottom piece 17 (the base) to which the pump tubing was attached, a stainless steel backing plate used to support the filters, a circular impactor plate, and a top piece secured in place by two small metal screws. The impactor plates were made of a porous metal, which were saturated with mineral oil prior to being used. The principle behind this procedure was that the larger particles (> PM2 . 5 ) impacted on the plate and remained stuck to the oil, thus being removed from the airflow (Thomas et al. 1993). After samples were obtained (daily), the collected visible particles were scraped from the impactor plates with razor blades. The rest of the sampler components were wiped with kimwipes soaked with distilled water and then allowed to dry. More thorough cleaning of the samplers was performed on a weekly basis. This involved soaking the top and bottom pieces of the samplers in soapy water for 15 minutes, followed by scrubbing and finally rinsing three times with distilled water. Impactor and backing plates were sonicated for 15 minutes in soapy water and then rinsed three times with distilled water. Al l components were then allowed to dry prior to loading any filters. During all filter loading, necessary precautions were taken to ensure that the risk of contamination was minimal. For example, kimwipes were placed on all surfaces coming into contact with forceps, samplers, or filters. Using cleaned forceps, filters were removed from petri plates and placed inside the bases of the samplers. Impactor plates (saturated with oil) were then placed on top of the filters, followed by the top pieces of the samplers. Two screws were used to seal each of the samplers closed. Self-adhesive labels were used to identify the samplers and filters. Each sampler was leak checked using a vacuum pump, a Matheson 603 rotameter, and an adapter used to seal the inlets of the sampler. If the rotameter showed a value of less than 10 units, equivalent to approximately 0.4 L/min (-10% of the flow rate used during the study), the sampler was considered to be sealed. If the rotameter indicated a leak, the sampler was reassembled or exchanged for another sampler until no leak was found. Airtight samplers were then placed in small clean plastic (Rubbermaid) containers lined with kimwipes for transportation to (and from) subjects' homes. Occasionally, some samplers were found to be leaking during the leak check, even after repeated attempts to seal the samplers. These were disassembled and thoroughly cleaned using acetone and distilled water. Subsequently, the o-rings were replaced and finally the samplers were reassembled. This procedure was repeated until the samplers were no longer leaking. Each day, at the completion of the personal sampling sessions, samplers were returned to the laboratory where they were disassembled. Using clean forceps, filters were transferred back to their original petri plates along with the corresponding labels from the sampler heads. These plates were then returned to the balance room where they were stored until the filters were weighed. 2.3.7 Personal Sampling All personal PM2 . 5 sampling sessions commenced on a Monday, Tuesday, Wednesday, or Thursday morning and ended approximately 24 hours later. Procedures carried out during 18 each personal sampling session were performed in a predetermined sequence. First, Holter monitors (model DM-400 cassette Holter recorder) were set-up to allow for 24-hour heart activity monitoring. Once these monitors were operating, subjects were instructed to lie down for 5-10 minutes with the purpose of recording resting heart activity. At the same time, heart rate and blood oxygen saturation were recorded at 1,3, and 5 minute intervals using Healthdyne Hand Held Pulse Oximeters, attached to the left index finger of each subject. During the resting period, a sampling pump was turned on and allowed to warm up for approximately 2 minutes. After this warm up period, a P E M (loaded with a filter) was connected to the pump. The flow was subsequently set to 4 L/min ± 5% using a calibrated rotameter. The rotameters (Matheson 603) had been previously calibrated at the beginning of the study using a frictionless piston meter (Bios Corp.). Sampling was then initiated and the start time and flow was recorded on a flow log (see Appendix D). The operating pump was placed inside a camera bag equipped with a shoulder strap to make the device as convenient as possible for the subjects to carry. Additionally, the bag was insulated with foam to reduce the amount of noise being emitted from the pump. The sampler, attached to the pump by tubing, remained outside of the bag. It was protected from clothes and other nearby objects by an aluminium cylindrical inlet. This inlet was fitted with Velcro, which secured the sampler to the shoulder strap. The sampler and inlet combination was placed in a downward vertical position, as close to the subject's breathing zone as possible, as illustrated in the following photograph. Figure 2.1: Subject Equipped with Personal PM2.5 Exposure Monitor, Holter Monitor, and Activity Log Following the resting periods, a blood pressure measurement was recorded from the left arm of each subject using a blood pressure cuff/stethoscope combination (Sprague/Rapapport). Subsequently, spirometry was performed using Presto F L A S H portable spirometers from Tamarac Systems Corporation. Spirometers were calibrated once every morning on each 1 9 sampling day using a 3 litre syringe until the exhale errors were < 0.25 %. Following each day of sampling, collected data was downloaded as ASCII text files and hard copies were printed. Daily maintenance involved replacing filters and cleaning the spirometer handheld flow sensor with alcohol swabs. During each personal sampling session, spirometry was performed following the 5-minute resting period and blood pressure measurements. This order was chosen specifically so that the resting heart activity could be recorded before the subjects exerted themselves while performing forced expiratory maneuvers. For each spirometry session, a new disposable mouthpiece was used and the handheld flow sensor was cleaned with alcohol swabs. Each subject was asked to perform a minimum of 3, and up to 6, forced vital capacity maneuvers, while seated and wearing nose plugs, until either 3 acceptable maneuvers were achieved or it was decided that the subject should not continue with the maneuvers. The acceptability of these maneuvers was based on the instrument criteria (i.e. maneuvers with unsatisfactory starts, coughing during the first second, valsalva maneuvers (glottis closure), early terminations of expiration, leaks, and obstructed mouthpieces were considered unacceptable). Subjects were provided with an activity diary (see Appendix E) to complete throughout the 24-hour sampling period. For each 30 minute block, subjects were instructed to indicate their location if indoors (home, restaurant, office and/or other), location if outdoors (near to the home, defined as being within 1 block, or away from the home), if they were in transit (car, bus, walk and/or other), their highest activity level (high, medium, low), if they were exposed to cooking and/or environmental tobacco smoke (ETS), if they had used an inhaler, and if they had been wearing the sampler on their shoulder. Subjects were instructed that if any of these situations were applicable they should be noted, regardless of the duration. Finally, subjects were asked if a bronchodilator had been used that morning and if so, the time of that use was recorded on the activity log. Finally, subjects were instructed as to what their responsibilities were for the 24-hour period. Each subject was encouraged to wear the pump using the provided shoulder strap as much as possible, but was told that at a minimum, the pump should remain in the same room as the subject at all times and taken with the subject wherever he/she may go. For sleeping, the subject was instructed to place the sampling device near the bed and that the pump (but not sampling head) may be covered with a pillow to reduce the noise. Subjects were also told that they were not permitted to shower or bathe, as doing so could negatively affect the Holter monitors and the recording. The subjects were also instructed on how to fill out the activity diary. Subjects were provided with emergency contact phone numbers to call if there were any problems with the equipment or sampling session. Finally, a time was confirmed with the subjects to return the next day to complete the 24-hour sampling. Occasionally, the emergency numbers were used to report that sampling pumps had stopped working. If it was not feasible to visit the subjects in their homes at those times, the subjects were instructed on how to turn their pumps back on over the telephone. In these cases, the subjects were told to remove the pump from the bag, and note the number on the display, indicating the length of time that the pump had been operating. The subjects then turned on 20 the pump and this new start time was recorded as well. It was therefore possible to determine the total amount of time that the pumps were in fact operating for these samples. Procedures for the end of each 24-hour sampling session (day 2) were very similar to those at the beginning of the sampling (day 1). A 5-minute resting heart activity session was recorded at the same time as blood oxygen saturation. The Holter monitor was then removed from the subject. During the resting period, the pump was removed from the insulating camera bag. At this time, the pump was put on hold and this time was recorded on the flow log. Additionally, the value on the pump display, indicating the number of minutes that the pump had been running, was recorded as well. The pump was then taken off hold and an airflow reading was obtained using a previously calibrated rotameter; this was recorded on the flow log. The P E M was then disconnected from the pump and placed in a small plastic container for used samplers. Subsequently, blood pressure and spirometry measurements were obtained. Following the spirometry, the activity log that had been filled out throughout the previous 24 hours was discussed with the purpose of clarifying any ambiguous items. Al l subjects then completed a symptom questionnaire at the end of each sampling period. This symptom questionnaire (see Appendix F) included 8 questions. The first seven asked about coughing, sputum, breathing difficulties, chest pain, heartbeat, fatigue, and dizziness. These were phrased in such a way as to determine if the symptom (if present) for the previous 24 hours was different in severity from the normal level for that subject. Therefore, subjects were asked to categorize each individual symptom as being more, less, or about the same as usual (none). The eighth question asked about the amount of bronchodilator use (i.e. ventolin, combivent, berotec, and/or bricanyl) and was simply a count of the number of times that the medication was used during the sampling session. Finally, a list of other medications was compiled for each subject, indicating type, dose, and frequency of use (see Appendix G). At the conclusion of each sampling period, the appropriate checklist was discussed with each of the subjects to determine which medications had been taken during the 24-hour sampling period. 2.3.8 Ambient Sampling Preparation Ambient fine particulate (PM2 . 5 ) sampling was carried out at 5 different Greater Vancouver Regional District (GVRD) air monitoring sites, with the co-operation of the G V R D staff. These sites included Kitsilano, North Delta, North Burnaby (Kensington), South Burnaby, and South Richmond (see Figure 2.2 below). Sampling at Kitsilano began on April 21, 1998 while sampling at all other sites began on April 27, 1998. All sampling ended on September 25, 1998. 21 Staff of the air monitoring division of the Greater Vancouver Regional District operated the South Burnaby, North Burnaby, and North Delta sites. U B C technicians operated the Kitsilano and South Richmond sites. Harvard Impactors (HI) were used to collect PM2 . 5 samples at all five sites. In addition to this sampler, at the Kitsilano site, a dichotomous sampler was used to obtain fine and coarse particulate samples. A variety of pumps were used at the ambient sites as follows. A Universal Sample Pump by S K C , operating on A C power, was used at the Kitsilano site. At the two Burnaby sites, Harvard black box pumps were used. Finally, at the North Delta and South Richmond sites, Gillian pumps were used. During the week of July 8, both of these pumps stopped working. Repairs were attempted but after the pumps failed again, on July 15 both of these pumps were replaced with Gast pumps. Al l pumps were plugged into electronic timers that automatically turned the pumps on at 7:00 a.m. on Monday morning and off at noon on Friday. Located at each G V R D air quality monitoring site, was a temperature controlled shelter. Sampling pumps were placed in these shelters and rubber tubing was run from the pump to the roof, where the tubing was connected to a Harvard Impactor (HI). Located in the tubing, between the pump and the HI, was a filter (Gelman, Bacterial Air Vent) that remained in 22 place throughout the entire study period to protect the pump from particles in the airflow. Additionally, metal rain caps were placed on the Harvard Impactors to protect them from the weather during sampling. A stand for each of the Harvard Impactors was constructed by using metal wire to secure a tripod to a plywood base. Cinder blocks were placed on top of the wood to prevent the structures from moving. Filter loading procedures for ambient samplers were as follows. Teflon filters (41mm 2 u.m pore size) were placed inside plastic filter cassettes, which were subsequently sealed with small pieces of transparent adhesive tape. The loaded cassettes were then placed in Harvard Impactors on top of disposable cellulose backing pads and secured in place by metal clips. The rest of each HI (i.e. two impactor plates saturated with mineral oil, holders, and a top inlet piece) was then assembled. Self-adhesive labels identified the filter inside each sampler. Harvard Impactors were then leak checked and airtight samplers were sealed with plastic caps for transport. Additional loaded filter cassettes and backing pads for each week were placed in ambient filter storage boxes and transported with the Harvard Impactors to each site. Harvard Impactors were thoroughly cleaned each week. The impactor plates were sonicated in soapy water for 15 minutes and then rinsed three times in distilled water. The rest of the sampler components were soaked in soapy water for 15 minutes, followed by scrubbing and finally rinsing three times with distilled water. Before loading any filters into the samplers, all components were allowed to completely air-dry. 2.3.9 Ambient Sampling Ambient PM2 . 5 sampling was begun on Monday morning of each week, as follows. First, a flow measurement was made using a rotameter (Matheson 603) at the end of the sampling train, without the sampler connected, to verify that the pump was set to the appropriate flow rate of 4 L/min ± 5%. If not, the pump was adjusted until the flow rate was within the appropriate range. This value was recorded under the "flow check" column on the flow log (see Appendix H). The loaded HI was then fastened to the tripod and the rotameter was attached, to the top of the HI using an adapter. Another flow reading, through the whole sampling train was obtained and recorded on the flow log. The rotameter was then removed, and the top inlet piece and rain cap were put in place on the sampler. Finally, the pump was attached to the HI and the start time of the sampling was recorded on the flow log. On each weekday morning thereafter, approximately 24 hours later, this procedure was reversed to record the stop time, the flow through the HI, and after removing the sampler, the flow through the pump alone. The filter then was replaced with the one for the next day as follows. The impactor was disassembled and the used filter (in the cassette) and cellulose backing pad were removed. Using clean forceps, a new backing pad was placed inside the sampler and the next filter cassette was placed on top of the pad. The sampler base was then reassembled. The top of the sampler was then taken apart and the two impactor plates were cleaned by scraping visibly collected particles from the plate with a razor blade. The sampler was then reassembled and ready to begin the next day's sample. The procedure of performing a series of flow checks and recording the start time was then initiated (except on Fridays). 23 Filter tears, scratches, or sampling problems were recorded on the log sheets provided. Each site was provided with filters to cover the four sampling periods of the week, plus one spare filter. When not needed, the spare filters were used as field blanks. 2.3.10 Dichotomous Sampler Sampling In addition to PM2 . 5 monitoring at the Kitsilano site, fine (and coarse) particulate monitoring was achieved throughout the study period using the G V R D dichotomous sampler (Anderson Series 241 Sampler). This sampling was normally conducted every sixth day by the G V R D staff as part of the National Air Pollution Monitoring Network (Environment Canada). However, from April 21 until September 25, 1998, the dichotomous sampler was operated on a daily basis by the U B C technicians on the days corresponding to HI sampling. This sampler separated two size fractions of particles being drawn through the device: fine (particles having an aerodynamic diameter <2.5pm) and coarse (particles >2.5pm and <10pm in aerodynamic diameter). These individual size fractions were deposited on two different filters. The inlet and filters were located on the roof of the Kitsilano G V R D trailer. These components were connected to a pump that was situated inside the trailer. The sampler was leak checked every Monday morning by replacing the top metal inlet of the device with a valve that could be closed. When closed and there was no leak, the displays on the pump indicated zero flow and the presence of a vacuum inside the device. Once this had been confirmed, the filters were loaded into the dichotomous sampler. Filters had been previously weighed, loaded in cassettes, and supplied by Environment Canada. On each sampling day two filters (one for each of the fine and coarse fractions) were loaded into the sampler. The sampler timer was then reset. Finally, the sampling was commenced and the flow for each size fraction and the sampling start time was recorded. At the end of each 24-hour sampling period, the flows and stop time were recorded. Subsequently, the used filters were removed and stored in the Kitsilano G V R D trailer. These filters were transported back to Environment Canada where they were weighed and the concentrations of fine and coarse particles were calculated. Additionally, on every sixth day, Environment Canada performed sulfate analysis on the filters as a component of the National Air Pollution Monitoring Network. Mass and sulfate concentrations obtained from the dichotomous sampler and Harvard Impactor at the Kitsilano site are listed in Appendix I. 2.4 Exposure Sampling Data Analysis Microsoft Excel 97 was used to create the initial database. SPSS Version 6.0 was used for all statistical purposes. 2.4.1. Particulate Sampling Personal and ambient concentration samples collected during the study were reviewed to verify correct times and flows. Samples with flow rates outside 4 L/min ± 10%, and collected over less than 20 hours, were deemed invalid and deleted. To determine the amount of time the personal samples overlapped with the ambient samples, average ambient sample start and stop times for each day were compared to the personal samples run each day. 24 Before calculating concentrations, particulate weights were adjusted according to field blank values by subtracting the mean field blank weight changes from all sample weights. Limits of detection for ambient and personal PM2 . 5 were calculated by dividing three times the standard deviation of the field blanks by the mean sampled volumes. The distributions of valid personal and ambient P M 2 5 and sulfate samples were analyzed. Personal exposure data was summarized both by subject and by pooling over all subjects. Ambient concentrations were summarized by site (i.e. by determining the mean of daily measurements for each site) and by averaging over all sites per day. One-way Analyses of Variance ( A N O V A ) were performed to test for differences in concentrations between the ambient sites. These analyses, in addition to others regarding personal and ambient PM2 . 5 monitoring, are discussed in greater detail in Stefanie Ebelt's thesis (Ebelt 1999). Correlation coefficients were calculated to determine the relationships among personal PM2 . 5 , personal SO4 2 ", mean ambient PM2 . 5 , mean ambient SO4 2", mean maximum daily ozone, mean temperature, and mean relative humidity (from 9:00 am to 8:00 am the next day). 2.5 Lung Function Data Analysis 2.5.1 Spirometry Maneuver Selection Subjects were expected to perform forced expiratory maneuvers before and after each of the seven sampling sessions to obtain pre- and post-sampling measurements. This would have resulted in a total of 224 completed lung function testing sessions. When selecting the appropriate maneuvers for the data analysis, the American Thoracic Society (ATS) reproducibility criteria (1994 update) were used (ATS 1995). According to these criteria, maneuvers are considered reproducible if the largest and second largest of each of the F E V i (forced expiratory volume in one second) and F V C (forced vital capacity) are within 0.2L. Furthermore, the largest F E V i (and F V C ) measurement from the reproducible tests is considered to be the "true" value. Therefore, the largest F E V j measurements were used from all of the acceptable curves except for cases in which the differences between the largest and second values were greater than 0.2L. In those cases, the second largest value was used. Two FEV] values selected for the final database were from maneuvers that, according to the error message on the Tamarac Portable spirometer, were classified as having a "Bad stop... keep blowing longer." These were included because the A T S spirometry guidelines state that the "use of data from unacceptable maneuvers due to failure to meet the end-of-test requirements is left to the discretion of the interpreter" (ATS 1995). The fact that these test results were the largest values obtained from the respective reproducible testing sessions indicates that they were indeed the best maneuvers and should be used for analysis. 25 2.5.2 Correlation between Lung Function Variables Pearson correlation coefficients between pre-sampling and post-sampling results for each of 3 outcome variables (FEV i , F V C , and F E V i / F V C ) were determined. These correlation coefficients and plots indicated a very high degree of correlation between measurements obtained 24 hours apart. 2.5.3 Outcome Variables: Post-Sampling FEVi and AFEVi Two outcome variables were selected for the analyses relating lung function to measures of pollutants; both were based on forced expiratory volume in 1 second (FEVi ) data. This was done for a number of reasons. First, F E V i is less effort dependent than F V C and is therefore more easily reproduced. Furthermore, the amount of published literature mentioning F E V i as an outcome variable exceeds that of F V C . The 2 variations of the F E V i variable analyzed in this study were the post-sampling F E V i data and the change in F E V i data (AFEVi ) over the 24-hour sampling period. The post-sampling F E V i variable was selected for analysis as the relationships of interest were between the P M 2 j and sulfate concentrations for the prior 24-hour period and the resulting F E V i . Therefore, the pre-FEVi data was of little use in this case. A standardized post-*FEVi measurement was obtained by calculating the mean post-FEVi for each individual and then subtracting this mean from each individual post-FEVi measurement. These standardized values were used for all post-FEVi analyses. The A F E V i analyses examined the relationship between the change in F E V i (post-sampling F E V i - pre-sampling F E V i ) over the 24-hour sampling periods and the exposure measurements. The crucial issue in this case was to determine whether or not lung function was altered throughout the day as a result of exposure to P M 2 5 (and/or sulfate). One A F E V i distribution, by subject, was produced using raw data. Review of the distribution boxplots for both outcome variables indicated that for one subject (#3), one data point was an extreme outlier. It was found that this subject did not use his inhalers/medications before spirometry on this one occasion, but he did use his medication prior to all other lung function testing. Therefore, this point was removed from the data set and subsequent analyses. Furthermore, this subject's mean post-FEVi was recalculated, excluding this point. 2.5.4 Regression Models Eight different models relating various combinations of the exposure and outcome variables were studied. These models are summarized in Table 2.2. 26 Table 2.2: Lung Function Models Model Exposure Measure Lung Function Measure 1 Personal P M 2 5 Post-samp ling F E V i 2 Personal Sulfate Post-sampling F E V i 3 Ambient P M 2 5 Post-sampling F E V i 4 Ambient Sulfate Post-sampling F E V i 5 Personal P M 2 5 A F E V i 6 Personal Sulfate A F E V , 7 Ambient P M 2 5 A F E V i 8 Ambient Sulfate A F E V , 2.5.5 Individual Ordinary Least Squares Regression Analyses Using the SPSS linear regression function, individual regression analyses for each of the above models were performed. Each subjects' residuals were then plotted against time to determine if there were any trends in lung function over time. For the post-sampling F E V ] data, one subject's lung function was found to increase considerably throughout the study. This subject (#5) also had substantially larger variability in the post-FEVi measurements than those of other subjects. As a result, this subject's data was excluded from all future post-F E V i pooled analyses. For the A F E V i data, no such trends were found. 2.5.6 Pooled Ordinary Least Squares Regression Analyses Pooled ordinary least squares regression analyses were performed for each of the eight models. Residuals obtained and predicted values from these analyses were recorded. Residuals were used to determine if there was any correlation between the outcome variable measurements obtained during different sampling sessions, by creating a series of correlation plots. One plot was constructed for every combination of sampling periods (i.e. sampling sessions 1 vs. 2, 2 vs. 3, 3 vs. 4, 4 vs. 5, 5 vs. 6, 6 vs. 7, 1 vs. 3, 2 vs. 4, etc.). To further characterize the correlation structure, one plot was constructed for every interval of sampling periods (i.e. all data 1 sampling session apart, 2 sampling sessions apart, etc.). No consistent pattern of correlations was found. Therefore, for the regression analyses, the assumption was made that the repeated measurements were independent of each other. In all cases, the pooled ordinary least squares regression analyses indicated that there was substantial variability in the lung function of the subjects and it was therefore necessary to weight the outcome measurements in the final regression models. 2.5.7 Pooled Weighted Least Squares Regression Analyses To verify that weighted least squares regression analyses were necessary, the residual sum of squares (RSS) for each individual's outcome measures were calculated from the pooled (i.e. including all subjects) regression analyses. Variances were then estimated by dividing the RSS by the number of valid residuals for each subject. Since the variances were considerably different for each individual, for each of the eight models, a weighted least squares regression analysis was performed. The structure of these models was based on the following formula: 27 YU = B0 + B}XU+EI 'it where Y is the lung function response (i.e. post-FEVi or A F E V i ) Po is the Y intercept Pi is the slope X i s the exposure (i.e. personal or ambient PM2 . 5 or SO4 2 ) E is the random error / is the individual t is the sampling session Inherent in this model structure are the assumptions that the error terms (Eit) are independent, normally distributed, random variables, with a mean of 0 and different variances for different individuals (cr,2). 2.5.8 Power Calculations The smallest associations between the exposure and lung function measurements that could be detected at a significance level of a = 0.05 with a power of 80% were calculated for each of the 8 models according to the following formula: where A is the true /?; za (1.96) and zp (-0.85) are the quantiles of a standard normal distribution MSE is the mean square error from the ordinary least squares regression models n is the total numbers of samples (individuals x sampling sessions per individual) Vx is the standardized variation of the x-values: Inherent in each power calculation are the assumptions that all variables are independent, and that the estimated variances are assumed to be the "true" variances. 2.5.9 Confounder Analyses The possibility of other atmospheric factors confounding the results of the above models was investigated by examining the correlations between personal PM2 . 5 , personal SO4 2", ambient PM2 . 5 , ambient SO4 2 ", maximum daily ozone, temperature, and relative humidity. Potential confounders with correlation coefficients < 0.5 were considered to be poorly correlated with the exposure measures. One potential confounder (ozone), with correlation coefficients > 0.5, was considered to be moderately correlated with PM2 . 5 and SO4 2 " and therefore, this variable was added to the lung function models to observe if doing so altered the models. Vx = — '^j(Xj - X)2 where Xi is each individual exposure n X is the mean exposure 28 2.6 Symptom Questionnaire Data Analysis For the symptom questionnaire data, the first seven questions were originally phrased such that there were three possible answers: more, less, or about the same. These responses were counted and summarized. Subsequently, it was determined that the use of dichotomous variables allowed for more simple statistical methods than using trichotomous variables. Furthermore, the purpose of the following analyses was to determine if certain PM2 . 5 and/or sulfate levels resulted in more severe symptoms. Therefore, the 3 possible answers to these questions were organized into 2 categories: "more" and "same or less". Adjusted responses from each individual were counted and summarized. The questions concerning chest pain, heartbeat irregularities, and dizziness were eliminated from the analyses as few responses were given indicating more severe levels of these symptoms. Additionally, another category, "Any Respiratory Symptom," was created by collapsing the responses from the coughing, sputum, and breathing difficulties variables. Responses for this category were considered to be "more" if there was an increase in any of coughing, sputum and breathing difficulties. Using the SPSS logistic regression function, 20 pooled models were completed to evaluate the relationships between the symptoms categories and various exposure measures, as indicated in Table 2.3. Table 2.3: Symptom Models Model Exposure Measure Symptom 1 Personal P M 2 5 Coughing 2 Personal Sulfate Coughing 3 Ambient PM2 . 5 Coughing 4 Ambient Sulfate Coughing 5 Personal PM2 . 5 Sputum 6 Personal Sulfate Sputum 7 Ambient P M 2 5 Sputum 8 Ambient Sulfate Sputum 9 Personal PM2 . 5 Breathing Difficulties 10 Personal Sulfate Breathing Difficulties 11 Ambient P M 2 5 Breathing Difficulties 12 Ambient Sulfate Breathing Difficulties 13 Personal P M 2 5 Any Respiratory Symptom 14 Personal Sulfate Any Respiratory Symptom 15 Ambient PM2 . 5 Any Respiratory Symptom 16 Ambient Sulfate Any Respiratory Symptom 17 Personal PM2 . 5 Fatigue 18 Personal Sulfate Fatigue 19 Ambient P M 2 5 Fatigue 20 Ambient Sulfate Fatigue 29 The structure of these models was based on the following formula: logit{Yu)=B0 + B,XH where Y is the symptom response B0 is the Y intercept B/ is the slope Xis the exposure (i.e. personal or ambient PM2 . 5 or SO42") 7 is the individual / is the sampling session Two analyses that produced statistically significant results were examined further. It was found that one individual was highly influential for these two analyses, and therefore the logistic regressions were redone after excluding data from this individual (#9). Symptom questionnaire responses were also analyzed on an individual basis. First, plots of symptom responses vs. exposure measures were created for each individual. Subsequently, this data was stratified by high/low exposures and the number of "more" responses in each exposure category was estimated from the plots. These estimates yielded results in which, on average, the number of "more" responses in each exposure category was approximately equal. Therefore, there were no relationships within the symptom questionnaire data meriting additional analyses. 2.7 Bronchodilator Use Analysis Counts of bronchodilator use for each sampling session by subject were produced from question #8 on the symptom questionnaire. The bronchodilator use, in addition to other medication schedules, were summarized for each subject. To study the relationship between personal exposure to PM2 . 5 and total bronchodilator use, a scatterplot and individual linear regression analyses were completed. Additionally, a similar scatterplot was attempted for P M 2 5 and extra bronchodilator doses. Problems with the data set and preliminary results indicated that there were no relationships within the data meriting additional analyses. 30 3. RESULTS 3.1 Subjects 3.1.1 Recruitment Results Meetings with respiratory specialists at the Vancouver Hospital Respiratory Clinic yielded 35 interested candidates. The presentation to the B.C. Lung Association "Puffers' Club" yielded a further 12 interested parties. Therefore, 47 individuals were considered as potential candidates for the study. O f the 47 interested candidates, 45 were contacted (2 potential subjects could not be reached). O f the 35 names obtained from the physicians, 13 individuals agreed to participate. O f the 12 interested people identified at the "Puffers' Club" meeting, 4 participated in the study. This however, included one woman that was allowed to participate even though it was not clear at the time of data collection whether her respiratory problems were due to COPD. After the study was completed, a physician determined that this person did not in fact have C O P D . Therefore, this person and all results from this subject were excluded from all analyses. Therefore, the total number of acceptable subjects that participated in the study was 16 (36% of potential candidates). Twenty-one subjects did not participate. O f these 21, 6 people indicated that they simply were not interested, 5 indicated that the project would require too much effort, 3 indicated that they were too busy, and 7 did not participate for other reasons. Eight subjects were excluded because they did not meet the study criteria. O f these 8, 4 were excluded because their C O P D conditions were too severe, 3 were excluded because they lived outside the study area, and 1 was excluded because she was living with a smoker. 3.1.2 Subject Characteristics The study population consisted of 7 male and 9 female subjects, aged between 54 and 86 (mean age: 74). All subjects were residents of the Greater Vancouver Regional District (GVRD). Al l subjects were non-smoking, moderate chronic obstructive pulmonary disease (COPD) patients. Thirteen of these patients were identified by physicians as having C O P D while three patients were determined to have moderate conditions by having a forced expiratory volume in one second (FEVi) greater than 0.75 L. Smokers and those patients having severe C O P D were excluded. Al l participants were not living with any smokers at the time of the study, with the exception of one subject. For this subject, the smoker in the home was instructed and agreed to only smoke outdoors or not at all on each sampling day. This subject was not an outlier in any of the analyses. All subjects used various medications throughout the study. All 16 subjects used at least one bronchodilator (salbutamol, berotec, terbutaline, serevent, combivent, atrovent) during at least one sampling session (13 subjects used the medication on all 7 sampling days). Five subjects used a xanthine bronchodilator (theophyline, theodur, uniphyl, choledyl) at least once. Fourteen subjects used at least 1 steroid medication (pulmicort, becloforte, flovent, prednisone, beclomethasone) during the study. Four subjects used cardiac medication (digoxin, sotalol, nitro dur patch), and three 31 subjects used Aspirin, while four used Tylenol during the study. Eleven subjects used miscellaneous medication. 3.1.3 Compliance Adhering to the objective set out at the beginning of the study, each subject underwent seven sampling sessions, yielding a 100% compliance rate. The sampling sessions were randomly spaced, with at least one week between consecutive sessions. In most cases, subjects followed the sampling schedule initially assigned to them. However, when scheduling conflicts arose, sampling days were rescheduled, with consecutive samples at least one week apart. 3.2 Exposure Sampling 3.2.1 Balance Room Conditions Throughout the study, the U B C Occupational Hygiene balance room was continually monitored and maintained at a mean temperature of 22 °C (SD: 0.67) and relative humidity level of 53% (SD: 6). 3.2.2 Exposure Monitoring Quality Control Results A number of trial experiments to test the operating parameters and limits of the personal and ambient P M 2 5 samplers and pumps were completed prior to the commencement of the study. The results of these experiments are described in this section. For all of the trial experiments, field and lab blank filters were used. For the personal (PEM) test trial filters, field blanks and lab blanks each composed 12.5% of all filters. For the ambient (HI) filters, each of the field and lab blanks composed 15% of all filters. Mean mass differences were calculated for these filters. Both personal and ambient field blanks increased by an average of 9 pg (SD: 4) from pre to post weighing. Personal lab blanks increased by 5 pg (SD: 4) on average while ambient lab blanks increased by 2 pg (SD: 0). Personal quality control (QC) filters on average did not change (SD: 0.005), while ambient Q C filters decreased on average by 2 pg (SD: 0.004). Therefore, the weight increases observed for the field blanks appeared to be true increases in mass due to filter handling, etc. Consequently, the mean mass increase on personal and ambient field blanks was subtracted from all personal and ambient samples respectively. Detection limits were calculated as three times the standard deviation in field blanks divided by the mean sampled volume (5.9 m 3 for P E M and 5.1 m 3 for HI samples). Using the mean mass increase on field blanks, the P E M limit of detection was calculated to be 2.1 pg/m 3 . The HI field blanks resulted in a limit of detection of 2.4 Lig/m 3. Two experiments (Experiments 1.1 and 1.2), designed to compare the PM2 . 5 concentrations obtained from 6 personal exposure monitors (PEMs) with those obtained from 2 ambient Harvard Impactors (His), were carried out in the dining room and basement of a home. For 32 the first run, the mean concentration from the six P E M s was 9.9 L i g /m 3 (SD = 1.1) while that from the 2 His was 9.4 L i g /m 3 (SD = 0.12). The difference in these mean concentrations was 4.7%. During Experiment 1.2, two of the pumps operating on batteries (connected to PEMs) were allowed to operate until the batteries were completely drained. These two pumps ran for approximately 35 hours, indicating that this would exceed the anticipated length of time for the personal samples in the study by almost 50%. For this second test trial, the mean concentration from the 4 P E M s connected to A C powered pumps was 9.0 ug/m 3 (SD = 0.34) while that from the 2 His was 6.6 pg/m 3 (SD = 0.54), corresponding to a difference in concentrations from the two types of samplers of 27.3%. From these two experiments, the mean percentage difference between personal and ambient reported concentration was calculated to be 16.0%. Results are summarized in Table 3.1 below. Table 3.1: Results of Trial Experiments 1.1-1.2 PM2.5 Concentrations (u.g/m3) Sampler Experiment 1.1 Experiment 1.2 Individual Mean Individual Mean P E M 1 9.2 9.2 P E M 2 8.9 9.1 9.0 P E M 3 9.3 9.9 8.5 P E M 4 12.0 9.2 P E M 5 9.6 7.5* N / A P E M 6 10.2 7.5* N /A HI 1 9.3 9.4 6.9 6.6 HI 2 9.5 6.2 * Battery-operated; allowed to run until fully drained Experiments 2 . 1 - 2.4 were designed to test the reproducibility in the concentrations reported from two personal samplers worn by the same person and subjected to the same environments. The first of these trial runs produced results where the PM2 . 5 concentrations differed by a factor of greater than 3. The following trials however, produced more stable results as the differences in each pair of reported concentrations ranged from only 3% to 16%. The mean percentage difference from these trials (not including the 71.3% difference from the first trial, as this appears to be an outlier) was 11.0%. Results are summarized in Table 3.2. 33 Table 3.2: Results of Trial Experiments 2.1 — 2.4 Experiment Sampler PM2.5 Concentration (P-g/m3) Percentage of Difference 2.1 P E M 1 31.6 71.3 P E M 2 9.1 2.2 P E M 1 15.8 16.3 P E M 2 18.8 2.3 P E M 1 11.2 3.2 P E M 2 11.6 2.4 P E M 1 25.4 13.4 P E M 2 22.0 The data from experiments 1.1 - 1.2, collectively used with data from experiments 2.1 - 2.4, indicated that the mean difference between personal samplers (11.0%) was less than the mean difference between personal and ambient samplers (16.0%). Therefore, one could expect that when comparing data from two different samplers during the main study, the error was greater than when comparing data obtained from one sampler type, such as in the case of the repeated personal samples. 3.2.3 Study Data: Clean Up All exposure data collected throughout the study was examined for invalid measurements. Samples with pre- or post-sampling flows outside of 4.0 L/min + 10% (i.e. 3.6 - 4.4 L/min) were excluded from the data sets. Samples running less than 20 hours were also excluded. Samples that were indicated on the flow log sheets to have filter damage were flagged, but none were excluded. Ten percent of all ambient filters were lab blanks and 24% were field blanks. For the personal filters, 15% were lab blanks and 10% were field blanks. Mean mass differences were calculated for all lab and field blank filters. Personal lab blanks increased by 2 pg (SD: 7) on average and personal field blanks increased by 16 pg (SD: 7). Both ambient lab blanks and field blanks increased by an average of 3 p,g (SD: 8). The mean mass increase on personal and ambient field blanks was subtracted from all personal and ambient PM2 . 5 samples respectively. The detection limits were defined as three times the standard deviation in field blanks divided by the mean sampled volumes (5.7 m 3 for personal and 5.8 m 3 for ambient samples). Using the mean mass increase on personal field blanks, the personal PM2 . 5 limit of detection was 3.7 u.g/m3. The ambient field blanks resulted in an ambient P M 2 5 limit of detection of 4.2 pg/m 3 . For the sulfate analyses, ambient lab and field blank values deviated from 0 ixg on a run-specific basis. For analytical runs in which sulfate was detectable on blank filters, the amounts of sulfate were similar between runs. Ambient lab and field blanks both increased by an average of 0.3 u.g. All runs of personal filters had detectable sulfate on the blank filters. 34 Personal lab and field blanks both increased by 0.3 u.g. For ambient and personal runs with detectable sulfate on blank samples, the respective mean mass increase of field blanks was subtracted. 3.2.4 Personal Sampling Results For the 16 C O P D subjects, 112 personal exposure measurements were obtained. Six of these measurements were deleted due to invalid flows or times, leaving 106 (95%) valid samples. Two personal measurements (1.9%) were below the detection limit, but were still included in the analyses. The mean personal sample flow rate was 4.0 L/min (SD: 0.1; range: 3.7-4.1). Ninety-eight percent of the samples had flows at 4.0 L/min ± 2% (i.e. 3.9 - 4.1 L/min). Samples were started and stopped between 8:00 am and 11:00 am to aim for 24-hour sample lengths. The mean personal sample duration was 23:47 hours (SD: 50 minutes; range: 20:37 - 25:10 hours). The mean start/stop time for the personal measurements was 9:35 am (SD: 60 minutes; range: 6:49 start - 11:40 stop). The mean percent overlap for the personal samples with ambient samples was 95% (range 85-100%). Therefore, on average for personal samples, 71 minutes of the sampling did not overlap with the ambient samples. Five to seven valid exposure measurements were obtained per subject. Individual distributions for PM2 . 5 and sulfate were not normal, but they were not improved when the data was ln-transformed. For descriptive purposes and consistency, both transformed and untransformed summary results are given in Tables 3.3 and 3.4. The boxplots that follow are a visual representation of the same data (Figures 3.1 and 3.2). Table 3.3: Personal PM2.5 Exposure Summary Subject N PM2.5 Concentrations (pg/m3) Arithmetic Mean SD Range . Geometric Mean Geometric SD 1 7 11.0 3.3 6.2-15.8 10.6 1.4 2 7 11.9 5.5 5.3 - 19.9 10.8 1.6 3 7 13.7 6.7 5.4-26.2 12.4 1.7 4 7 14.3 3.0 11.2-18.9 14.0 1.2 5 6 12.6 9.2 2 .6-27.6 9.6 2.4 6 7 46.2 22.8 25.9-90.9 42.3 1.6 7 7 17.7 12.7 10.0-46.1 15.4 1.7 8 7 13.9 4.6 7.5 - 19.6 13.1 1.5 9 7 36.2 19.3 15.2-69.6 31.9 1.7 10 7 12.3 4.2 6.8-17.2 11.6 1.5 11 6 11.0 6.0 6.3-21.7 9.9 1.6 12 6 23.9 16.3 9.6-50.2 20.0 1.9 13 7 9.5 4.4 4.7-17.8 8.7 1.6 14 7 6.9 3.2 2.2-1.1.7 6.2 1.8 15 6 29.3 10.0 12.7-39.8 27.5 1.5 16 5 22.9 13.8 8.0-39.7 19.5 1.9 Group 106 18.2 14.6 2 .2-90.9 14.3 2.0 Table 3.4: Personal Sulfate Exposure Summary Subject N Sulfate Concentrations (pg/m3) Arithmetic Mean SD Range Geometric Mean Geometric SD 1 7 1.3 0.5 0.6-2.1 1.2 1.5 2 7 1.2 0.7 0 .6-2 .5 1.0 1.7 3 7 1.4 0.6 0 .8-2 .5 1.3 1.5 4 7 1.7 0.8 0 .7-3 .0 1.6 1.7 5 6 1.3 0.8 0 .6-2.8 1.2 1.7 6 7 1.7 0.6 1.2-2.9 1.6 1.4 7 7 1.7 0.4 1.4-2.4 1.7 1.2 8 7 1.8 1.1 0 .8-3.5 1.5 1.8 9 7 2.2 1.2 0 .6-4 .4 1.9 1.9 10 7 1.2 0.6 0.5-2.3 1.1 1.6 11 6 2.3 1.3 1.3-4.7 2.1 1.7 12 6 1.3 0.7 0 .7 -2 .6 1.1 1.7 13 7 1.7 1.3 0 .6-4 .5 1.4 1.8 14 7 0.8 0.4 0.2-1.3 0.6 2.3 15 6 1.2 0.7 0.5-2.1 1.1 1.9 16 5 0.7 0.4 0.3 - 1.2 0.7 1.7 Group 106 1.5 0.9 0 .2-4 .7 1.3 1.8 36 Subject Lower and upper boundaries of each box indicate the 25* and 75th percentiles, respectively. Horizontal lines represent medians. Smallest and largest observed values that are not outliers are indicated by the tails. Outliers are indicated by 'o'; extreme outliers are indicated by '*'. Figure 3.2: Boxplot of Personal Sulfate Exposures 5_ 44 34 Subject 3.2.5 Ambient Sampling Results For the 5 ambient sites, 413 P M 2 5 samples were collected over the course of the study. O f these, 26 samples were deleted due to invalid flows or times. Thus, 387 (94%) of the ambient 37 samples were successfully collected. Seven ambient measurements (1.8%) were below the detection limit, but were still included in the analyses. The mean ambient sample flow rate was 4.0 L/min (SD: 0.1; range: 3.7-4.4). Ninety-eight percent of the samples had flows at 4.0 L/min ± 6% (i.e. 3.8-4.2 L/min). The mean ambient sample duration was 23:51 hours (SD: 32 minutes; range: 21:43-25:50 hours). The mean start/stop time for the ambient measurements was 8:29 am (SD: 36 minutes; range: 6:37 start - 10:12 stop). The number of measurements obtained at each of the 5 sites ranged from 65 to 84 during 90 days of sampling in total. As with the distributions of personal exposure data, the PM2 . 5 and sulfate distributions of ambient data were not normal. These were also not improved when the data was ln-transformed. For consistency, both arithmetic and geometric summary results are presented in Tables 3.5 and 3.6. Boxplots of this data are shown in Figures 3.3 and 3.4. The one-way Analyses of Variance ( A N O V A) for both PM2 . 5 and sulfate indicated that arithmetic means between sites were not significantly different from each another. The F probability for these tests were 0.34 and 0.45 for PM2 . 5 and sulfate respectively. Relationships between personal PM2 . 5 and sulfate, and ambient P M 2 . 5 and sulfate are discussed elsewhere in greater detail (Ebelt 1999). Table 3.5: Ambient PM2.5 Concentration Summary Site N P M 2 5 Concentrations (pg/m3) Arithmetic Mean SD Range Geometric Mean Geometric SD Kitsilano 84 11.9 4.7 2.3-29.3 11.0 1.5 North Delta 75 11.6 5.2 1.1-32.0 10.5 1.6 North Burnaby 83 11.4 5.0 3.5-30.8 10.4 1.5 South Burnaby 80 11.6 5.0 4 .7-29.7 10.7 1.5 South Richmond 65 10.3 4.4 3.1-24.0 9.4 1.6 Average of all sites for each day 90 11.4 4.1 4 .2-28.7 10.8 1.4 Table 3.6: Ambient Sulfate Concentration Summary Site N Sulfate Concentrations (pg/m ) Arithmetic Mean SD Range Geometric Mean Geometric SD Kitsilano 84 1.9 1.0 0 .4-5.3 1.7 1.7 North Delta 75 2.0 1.2 0 .3-6 .2 1.7 1.8 North Burnaby 83 2.0 0.9 0 .4-4 .6 1.8 1.7 South Burnaby 80 1.7 0.8 0 .4-4 .6 1.6 1.6 South Richmond 65 1.8 1.1 0 .4 -6 .0 1.5 1.8 Average of all sites for each day 90 1.9 0.9 0 .4-5 .4 1.7 1.7 38 Figure 3.3: Boxplot of Ambient PM2.s Concentrations 4 0 , Ambient Site Figure 3.4: Boxplot of Ambient Sulfate Concentrations 90 Mean Ambient Site 39 3.3 Lung Function Testing 3.3.1 Spirometry Quality Control Tests were performed to determine the length of time that the Tamarac Spirometers were able to remain calibrated under a number of different conditions. These results are summarized in Table 3.7 below. Table 3.7: Spirometer Quality Control Experiment Results Calibration Date Trial Date Mean Volume (L) SD Location 26-Mar 26-Mar 3.39 0.005774 Lab 26-Mar 26-Mar 3.24 0.011547 Lab 26-Mar 26-Mar 3.24 0.015275 Lab 27-Mar 27-Mar 3.25 0.005774 Lab 27-Mar 27-Mar 3.24 0.005774 Lab 27-Mar 27-Mar 3.24 0.011547 Lab 31-Mar 31-Mar 3.24 0.005774 Lab 3-Apr 3.25 0.005774 Lab 7-Apr 3.25 0 Lab 15-Apr 3.28 0.020817 Lab 15-Apr 15-Apr 3.25 0 Lab 15-Apr 15-Apr 3.24 0.005774 Lab 15-Apr 3.25 0.005774 (Car) Home 16-Apr 3.27 0.005774 (Car) Home 17-Apr 3.08 0 (Car) Lab 17-Apr 17-Apr 3.26 4.21E-08 Lab 24-Apr 24-Apr 3.24 0.01 Lab 24-Apr 3.23 0.005774 (Car) Home 25-Apr 3.25 0.005774 Home 26-Apr 3.22 0.01 Home 27-Apr 3.22 0.005774 (Car) Lab 27-Apr 27-Apr 3.22 0.011547 Lab The data indicates that all of the simulated expiratory tests (except for the first one) performed using the 3L syringe immediately after the spirometer calibration were within the range of 3.22 to 3.26 L (1.2% of each other). Volumes reported from the spirometers constantly exceeded 3L by approximately 10% because, by convention, the spirometer results were reported as if the gas was at body temperature (BTPS correction). The fact that exhaled air contracts in volume by approximately 10% as it leaves the body, and that the spirometer automatically corrects for this fact, explains the error in these volume measurements. The data indicates that the spirometer baseline calibration varied by up to 1.2%. When the spirometer was tested in the same location over a few days, there was virtually no loss in the reproducibility of the results (see March 31 s t calibration in the above table). When the spirometer was transported after calibration and simulated maneuvers were performed in different locations, there was still virtually no change in the reproducibility of 40 the data for at least two full days (see April 15 and 24 u calibrations in the above table). Therefore, it was decided that during the course of the main study it was sufficient to calibrate the spirometers once every morning, and then use them throughout that day. 3.3.2 Spirometry Database The number of lung function testing sessions expected to be completed was 224. On June 8, 1998, the spirometer batteries failed and no data was collected on that day. As a result, two spirometry sessions scheduled for that day were not completed. Furthermore, one subject did not perform the forced expiratory maneuvers on three other occasions. Therefore, the number of sessions successfully completed and recorded was 219 (97.8%). One additional session was eliminated due to poor acceptability (i.e. <2 successful maneuvers were completed). The mean number of individual maneuvers performed by the subjects per testing session was 3.5. These results are summarized in Table 3.8. Table 3.8: Number of Maneuvers per Session Number of Individual Number of Maneuvers per Session Spirometry Sessions 3 132 4 62 5 21 6 4 Total number of sessions 219 For all F E V i (forced expiratory volume in one second) values used in the data analysis, 211 (96.8%) were within 0.1L or 0.5% of the second largest F E V i , and 6 F E V i values (2.7%) were within 0.2L of the second largest values. Therefore, 99.5% of all F E V i values used in the data analysis were reproducible according to A T S spirometry criteria (ATS 1995). One F E V i value used in the analysis was the second largest value obtained for that testing session. This value was used because the largest F E V i (1.07L) for that session was 0.32L (29.9%) greater than the second largest value (0.75L). F V C (forced vital capacity) was also measured and similarly summarized (see Appendix J). The final F E V i database from all individual lung function testing is summarized in Table 3.9. The data listed is that obtained from using the best F E V i from each lung function testing session, as outlined above. Table 3.9: Individual FEVi Results ID N Mean Min Max SD 1 14 0.88 0.79 1.01 0.065 2 14 0.63 0.48 0.74 0.066 3 14 2.28 1.72 2.55 0.222 4 14 0.75 0.67 0.81 0.048 5 13 1.59 1.10 1.98 0.300 6 14 1.37 1.15 1.46 0.084 7 14 1.84 1.53 1.98 0.124 8 14 1.06 0.75 1.31 0.133 9 14 0.56 0.49 0.64 0.039 10 14 1.41 1.13 1.71 0.182 11 14 0.65 0.59 0.75 0.041 12 14 1.08 0.98 1.29 0.094 13 14 1.40 1.25 1.51 0.074 14 14 1.02 0.92 1.13 0.062 15 14 0.75 0.64 0.89 0.068 16 9 1.37 0.82 1.54 0.233 Group 218 1.16 0.48 2.55 0.488 3.3.3 Correlation between Lung Function Variables To narrow down the number of possible lung function models, Pearson correlation coefficients between pre-sampling and post-sampling results for each of three outcome variables (FEVi, FVC, and FEVi/FVC) were determined. The results are summarized in following table (Table 3.10) and figures (Figures 3.5 - 3.7). Table 3.10: Pearson Correlation Coefficients for Outcome Variables Correlation Coefficient (N=107) Pre and Post F E V i 0.9534 Pre and Post F V C 0.9381 Pre and Post F E V , / F V C 0.9768 42 Figure 3.5: Scatterplot of Pre vs. Post-Sampling FEVi 3.0 T 2.5 2.0 J 1.5" > LU LL O) c Q- 1.0-to CO D_ .5 0.0 .5 LO 1.5 2.0 2.5 Pre-Sampling FEV1 (L) Figure 3.6: Scatterplot of Pre vs. Post-Sampling FVC 5.0-4.5 4.0 3.5 3.0 2 LL ro 2.5 A 2.0 H CO 1-5-| —^< w o 0- 1.0 1.0 1.5 2.0 Pre-Sampling FVC (L) 2.5 3.0 3.5 4.0 4.5 5.0 43 Figure 3.7: Scatterplot of Pre vs. Post-Sampling FEVj/FVC .8-2 > LU CD CO 1 in o CL .3 .3 Pre-Sampl ing FEV1 /FVC The data suggested an extremely high degree of correlation, indicating that using either the pre or post data in the lung function analyses would have been sufficient. The F E V i measurements (i.e. post-FEVi and A F E V i ) were selected for analyses because F E V i is less effort dependent than F V C and therefore more easily reproduced. Additionally, the amount of published literature mentioning F E V i as an outcome variable exceeds that of F V C . The post-sampling F E V i variable was selected for analysis as the relationship of interest was between the PM2.5 (and sulfate) concentration for the prior 24-hour period and the resulting F E V i . The A F E V i analyses examined the relationship between the change in F E V i (post-sampling F E V i - pre-sampling F E V i ) over the 24-hour sampling periods and the corresponding exposure measurements. 3.4 Lung Function Regression Analyses Results 3.4.1 Individual Ordinary Least Sguares Regression For the post-sampling F E V i data, 2 individual distributions were examined. First, a boxplot was produced using the raw data as illustrated below in Figure 3.8. Due to the fact that there was substantial variability in the baseline lung function between subjects, a second standardized boxplot (Figure 3.9) was obtained by calculating the mean post-FEVi for each individual and determining the deviation from the mean. For the A F E V i data, the data was already standardized and therefore, only one distribution boxplot (Figure 3.10) was produced Subsequently, the outlier point from subject #3 was excluded from all of the data (as discussed in Section 2.5.3). 44 Figure 3.8: Boxplot of Individual Post-FEVi Distributions - Raw Data 3 . 0 -Subject Figure 3.9: Boxplot of Individual Post-FEVi Distributions - Standardized Data .6-> UJ w o Q. c CO CD E o > CD Q Subject 45 Figure 3.10: Boxplot of Individual AFEVi Distributions . 8 1 TO O -1.0 J , I N= 7 7 7 7 6 7 7 7 7 7 7 7 7 7 7 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Using the standardized post-FEVi data and the raw (standardized) AFEV] data, individual ordinary least squares regression analyses were performed for each of eight models (see Table 3.11), using the SPSS linear regression function. Detailed results of these analyses are listed in Appendix K and summarized in Table 3.12. Table 3.11: Lung Function Models Model Exposure Measure Lung Function Measure 1 Personal P M 2 . 5 Post-sampling F E V i 2 Personal Sulfate Post-sampling F E V i 3 Ambient P M 2 5 Post-sampling F E V i 4 Ambient Sulfate Post-sampling F E V i 5 Personal PM 2 .s A F E V , 6 Personal Sulfate A F E V i 7 Ambient P M 2 5 A F E V , 8 Ambient Sulfate A F E V i 46 Table 3.12: Summary of Individual Regression Models Number of Number of Number of Number of Model Positive Negative Significant Results Significant Results Slopes Slopes (p<0.10) (p<0.05) 1 9 7 2 1 2 6 10 3 6 10 1 4 7 9 5 11 5 1 6 9 7 1 7 8 8 3 1 8 9 7 2 Residuals for each subject were obtained from individual regression analyses. These were plotted against time to determine if there were any trends in lung function over time. Of all the analyses involving post-sampling F E V i data, one subject's lung function was found to increase considerably throughout the study. This subject (#5) also had substantially larger variability in the post-sampling F E V i measurements than those of other subjects (as illustrated in Figures 3.8 and 3.9). As a result, this subject's data was excluded from all future post - F E V i analyses. The data in the above table indicated that no consistent results were obtained from the individual ordinary least squares regression analyses. Therefore, pooled regression analyses were completed with the goal of obtaining more consistent results. 3.4.2 Pooled Ordinary Least Squares Regression Pooled ordinary least squares regression analyses were performed for each of the eight models to attempt to increase the consistency of the results relative to those from the individual analyses. For these pooled regression models, data from all subjects were included. To determine the correlation structure of the lung function and exposure measures between sampling sessions, a series of correlation plots were created using the residuals obtained from each of the pooled regression analyses. For each model, one plot was constructed for every combination of sampling periods (i.e. sampling sessions 1 vs. 2, 2 vs. 3, 3 vs. 4, 4 vs. 5, 5 vs. 6, 6 vs. 7, 1 vs. 3, 2 vs. 4, etc.). The correlation coefficients are listed in Appendix L (Tables L. 1 - L .8). These coefficients suggested that there were no systematic correlation patterns and therefore, the conclusion was made that for all models there was no correlation between any measurements from different sampling intervals. To further characterize the correlation structures, one plot was constructed for every interval of sampling periods (i.e. all data 1 sampling session apart, 2 sampling sessions apart, etc.) for each model. These correlation coefficients are listed in Appendix L (Table L .9) . No consistent correlation structures for any of the repeated measurements in any one model were found. Consequently, no correlation structure was incorporated into the final models. Results of these final pooled ordinary least 47 squares regression analyses are summarized in Table 3.13. The scatterplots for these regression analyses can be found in Appendix M. Table 3.13: Pooled Ordinary Least Squares Regression Results Model N Slope Standard Error r2 p value 1 97 0.000256 0.000585 0.00201 0.6629 2 97 -0.003307 0.009966 0.00116 0.7407 3 101 -0.001980 0.002044 0.00938 0.3352 4 101 -0.003831 0.008322 0.00214 0.6463 5 103 -0.000021 0.000907 0.00001 0.9818 6 103 -0.019609 0.015178 0.01626 0.1993 7 106 -0.003118 0.003132 0.00944 0.3218 8 106 -0.004023 0.012917 0.00093 0.7561 Results of these models were more consistent (i.e. all slopes, except one, were negative) than those of the individual ordinary least squares regression analyses. However, no pooled models provided statistically significant results. Therefore, pooled weighted least squares regression analyses were attempted to account for differences in variability between subjects. 3.4.3 Pooled Weighted Least Squares Regression Pooled weighted least squares regression analyses were performed for each of the eight models using estimated variances. These variances for each individual's measurements were estimated by first calculating the residual sum of squares (RSS) for each individual from the pooled ordinary least squares regression analyses, and then dividing each RSS by the number of valid residuals for each subject. Finally, the variance estimates (listed in Appendix N) were included in the final pooled weighted least squares regression analyses to account for the differences in the variances from the ordinary regression models. Results of these analyses are summarized in Table 3.14. Table 3.14: Pooled Weighted Least Squares Regression Results Model N Slope Standard Error r2 P value 1 97 0.000070 0.000347 0.00043 0.8399 2 97 -0.004051 0.006058 0.00468 0.5054 3 101 -0.001266 0.001243 0.01036 0.3112 4 101 -0.005149 0.005158 0.00996 0.3206 5 103 0.000057 0.000465 0.00015 0.9029 6 103 -0.012093 0.006830 0.03010 0.0796 7 106 -0.001240 0.001437 0.00711 0 . 3 9 0 1 8 106 -0.003115 0.006228 0.00240 0.6180 Results indicate that no statistically significant associations were found between post-sampling F E V i measurements and any pollution measure. A statistically significant inverse association was found between personal sulfate exposures and A F E V i over the 24-hour 48 sampling periods (p=0.0796). Furthermore, results indicate that for all models with personal S O / " , ambient SO4 2 -, and ambient PM2 . 5 as the exposure measures, all slopes were negative as originally hypothesized. These results were converted into the absolute and percentage (of mean) changes in F E V i associated with 10 pg/m 3 increases in P M 2 5 and 1 pg/m 3 increases in S0 4 2 " , as indicated in Tables 3.15 and 3.16, respectively. Table 3.15: Absolute and Percentage Changes in FEVi per 10jug/m3 increase in PM2.5 P M 2 5 A F E V i / A F E V i / 10pg/m3 Models 10pg/m 3(mi) (%) Post-FEVi Models 1 0.70 0.06 3 -12.66 -1.09 A F E V i Models 5 0.57 0.05 7 -12.40 -1.07 Table 3.16: Absolute and Percentage Changes in FEVi per 1 jug/m3 increase in SO4' SO4 2 A F E V i / l p - g / m 3 A F E V i / l p g / m 3 Models (ml) (%) Post-FEVi Models 2 -4.05 -0.35 4 -5.15 -0.44 A F E V i Models 6 -12.09 -1.04 8 -3.12 -0.27 3.4.4 Power Calculations The lowest associations (slopes) that could be detected at a significance level of a = 0.05 with a power of 80% was calculated for each model, as indicated in Tables 3.17 and 3.18. Table 3.17: Lung Function Power Calculations for PM2.5 Models P M 2 5 A F E V i / A F E V i / 10pg/m3 Models 10pg/m3 (ml) (%) Post-FEVi Models 1 16.45 1.42 3 57.44 4.95 A F E V i Models 5 25.49 2.20 7 88.01 7.59 49 Table 3.18: Lung Function Power Calculations for SO/' Models so42 A F E V i / lp-g/m3 A F E V i / lu.g/m3 Models (ml) (%) Post-FEVi Models 2 28.00 2.41 4 23.39 2.02 A F E V i Models 6 42.65 3.68 8 36.30 3.13 T h e s e r e s u l t s i n d i c a t e that f o r t h e r e g r e s s i o n a n a l y s e s t o h a v e p r o d u c e d s t a t i s t i c a l l y s i g n i f i c a n t r e s u l t s u s i n g t h e s t u d y s a m p l e s i z e a n d v a r i a n c e , t h e c h a n g e s i n F E V i w o u l d h a v e h a d t o b e r e l a t i v e l y l a r g e (at least 3 t i m e s g r e a t e r ) . C h a n g e s o f t h i s m a g n i t u d e w e r e n o t o b s e r v e d n o r e x p e c t e d a n d t h e r e f o r e , s t a t i s t i c a l l y s i g n i f i c a n t r e s u l t s c a n n o t b e o b t a i n e d f r o m t h i s l u n g f u n c t i o n d a t a . H o w e v e r , t h e r e s u l t s s t i l l m a y s u g g e s t a n i n v e r s e a s s o c i a t i o n b e t w e e n e x p o s u r e t o PM2 . 5 ( a m b i e n t ) a n d SO4 2* ( p e r s o n a l a n d a m b i e n t ) , a n d d e c r e a s e s i n F E V i . S i m i l a r p o w e r c a l c u l a t i o n s w e r e a l s o c o m p l e t e d t o d e t e r m i n e t h e l o w e s t d e t e c t a b l e a s s o c i a t i o n s t h a t w o u l d h a v e r e s u l t e d h a d 30 s u b j e c t s ( the o r i g i n a l s t u d y t a r g e t ) p a r t i c i p a t e d i n t h e s t u d y . T h e s e l o w e s t d e t e c t a b l e a s s o c i a t i o n s ( f o r 30 s u b j e c t s ) w o u l d s t i l l h a v e r e m a i n e d r e l a t i v e l y l a r g e at a p p r o x i m a t e l y t w o t h i r d s o f t h e o r i g i n a l l o w e s t d e t e c t a b l e a s s o c i a t i o n s ( f o r 16 s u b j e c t s ) . T h e r e f o r e , t h e p o w e r t o d e t e c t a n a s s o c i a t i o n b e t w e e n p a r t i c u l a t e p o l l u t i o n a n d h e a l t h e f f e c t s w o u l d n o t h a v e b e e n s u b s t a n t i a l l y i m p r o v e d h a d 30 s u b j e c t s b e e n r e c r u i t e d . 3.4.5 Confounder Results C o r r e l a t i o n c o e f f i c i e n t s f o r p e r s o n a l a n d m e a n a m b i e n t P M 2 5 a n d SO4 2", t e m p e r a t u r e , r e l a t i v e h u m i d i t y , a n d d a i l y m a x i m u m o z o n e are s u m m a r i z e d i n T a b l e 3.19. Table 3.19: Correlation Coefficients among Exposure Measures Exposure Measures Personal P M 2 , Personal S042" Ambient P M 2 5 Ambient S042" Max Daily Ozone Temperature Relative Humidity Personal P M 2 5 1.0000 Personal S04 2 0.1942 1.0000 Ambient P M 2 . 5 0.1463 0.6976 1.0000 Ambient S04 2 0.1201 0.8671 0.8141 1.0000 Max Daily Ozone 0.1266 0.6588 0.6680 0.5911 1.0000 Temperature -0.0725 0.3397 0.4404 0.3889 0.4305 1.0000 Relative Humidity -0.0226 -0.2083 -0.2607 -0.1097 -0.4605 -0.4640 1.0000 50 Results of these correlations indicate that personal SO42", ambient SO42", and ambient PM2 . 5 were all highly correlated with each other. Interestingly, personal PM2 . 5 was poorly correlated with all other variables. These findings are discussed elsewhere in greater detail (Ebelt 1999). Temperature and relative humidity were both poorly correlated with the various PM2 . 5 and sulfate concentrations and therefore, were not considered confounders. Daily maximum ozone (O3) was found to be moderately correlated with personal SO42", and ambient P M 2 5 and SO42". Consequently, this variable was added to both the ordinary and weighted variations of model 6 (personal SO42" vs. A F E V i ) in the attempt to improve the models. For the weighted O 3 models, the variance estimates were the same as those used for the pooled weighted least squares regression analysis (#6) above. Results of these analyses are outlined in Table 3.20. Table 3.20: Ozone and Personal. Sulfate Regression Results Model Type Independent Variable(s) Slope Standard Error r2 p value A Ordinary Ozone -1.222604 1.135619 0.01102 0.2842 B Ordinary Personal S 0 4 2 " -0.019609 0.015178 0.01626 0.1993 C Ordinary Personal S0 4 2 " -0.015323 0.020397 0.01724 0.4543 Ozone -0.489767 1.548697 0.01724 0.7525 D Weighted Ozone -1.051747 0.538769 0.03535 0.0536 E Weighted Personal S0 4 2 " -0.012093 0.006830 0.03010 0.0796 F Weighted Personal S0 4 2 ~ -0.004346 0.009533 0.04304 0.6495 Ozone -0.889024 0.764582 0.04304 0.2477 Results of the ordinary regression analyses indicate that ozone alone was not significantly associated with A F E V i (model A ) , and that the personal SO4 2 " model was not significantly improved with the addition of ozone (models B and C, respectively). However, for the weighted analyses, ozone alone appeared to be associated with A F E V i (Model D). Additionally, ozone appeared to alter the weighted personal SO4 2 " model (Models E and F ) more so than for the ordinary model. This indicates that ozone may have been more strongly associated with changes in lung function than personal SO4 2". 3.5 Symptom Questionnaire Results 3.5.1 Symptom Database All 16 subjects answered the symptom questionnaire at the end of each sampling session (N=7), for a total of 112 completed questionnaires (100%). Responses are summarized in Table 3.21. 51 Table 3.21: Summary of Questionnaire Responses Question Responses (N) Res ponses (%) Total (N) More Less Same More Less Same 1. Coughing 11 9 92 9.8 8.0 82.1 112 2. Sputum 18 10 84 16.1 8.9 75.0 112 3. Breathing difficulties 25 10 77 22.3 8.9 68.8 112 4. Chest pain 0 1 111 0.0 0.9 99.1 112 5. Heartbeat irregularities 6 2 104 5.4 1.8 92.9 112 6. Fatigue 34 5 73 30.4 4.5 65.2 112 7. Dizziness 6 4 •102 5.4 3.6 91.1 112 Another symptom category, "Any Respiratory Symptom," was created by collapsing the responses from the coughing, sputum, and breathing difficulties variables. Responses for this category were considered to be "more" if there was an increase in any of coughing, sputum and/or breathing difficulties. Additionally, all answers were converted from trichotomous variables into dichotomous variables. Subsequently, the answers to the questions were organized such that they were grouped into "more" or "same or less" severe symptom levels. The adjusted results are outlined below in Table 3.22. Table 3.22: Summary of Adjusted Questionnaire Responses Responses (N) Responses (%) Total (N) Question More Less or More Less or Same Same 1. Coughing 11 101 9.8 90.2 112 2. Sputum 18 94 16.1 83.9 112 3. Breathing difficulties 25 87 22.3 77.7 112 4. Any Respiratory Symptom 37 75 33.0 67.0 112 5. Chest pain 0 112 0.0 100.0 112 6. Heartbeat irregularities 6 106 5.4 94.6 112 7. Fatigue 34 78 30.4 69.6 112 8. Dizziness 6 106 5.4 94.6 112 3.5.2 Pooled Logistic Regression Analyses The questions concerning chest pain, heartbeat irregularities, and dizziness were found to have few responses (0, 5.4, and 5.4%, respectively) indicating increased symptom levels. Therefore, this data was excluded from the following analyses. Pooled logistic regression analyses were completed for 20 models. Results are summarized in Table 3.23. 52 Table 3.23: Pooled Logistic Regression Results Model Exposure Measure Symptom Slope Standard Error p value 1 Personal PM2 . 5 Coughing -0.0043 0.0231 0.8519 2 Personal Sulfate Coughing -0.6827 0.5248 0.1933 3 Ambient PM2 . 5 Coughing -0.1116 0.0919 0.2247 4 Ambient Sulfate Coughing -0.3666 0.3809 0.3358 5 Personal PM2 . 5 Sputum 0.0083 0.0164 0.6149 6 Personal Sulfate Sputum -0.0145 0.3040 0.9619 7 Ambient PM2 . 5 Sputum 0.0274 0.0616 0.6569 8 Ambient Sulfate Sputum 0.1342 0.2514 0.5935 9 Personal PM2 . 5 Breathing Difficulties 0.0330 0.0150 0.0280 10 Personal Sulfate Breathing Difficulties 0.0848 0.2637 0.7476 11 Ambient PM2 . 5 Breathing Difficulties 0.0350 0.0546 0.5212 12 Ambient Sulfate Breathing Difficulties 4.75xl0"5 0.2307 1.0000 13 Personal PM2 . 5 Any Respiratory Symptom 0.0350 0.0152 0.0212 14 Personal Sulfate Any Respiratory Symptom -0.1130 0.2466 0.6468 15 Ambient P M 2 5 Any Respiratory Symptom -0.0092 0.0500 0.8532 16 Ambient Sulfate Any Respiratory Symptom -0.1330 0.2103 0.5272 17 Personal PM2 . 5 Fatigue 0.0224 0.0141 0.1124 18 Personal Sulfate Fatigue -0.2816 0.2731 0.3024 19 Ambient PM2 . 5 Fatigue -0.0514 0.0536 0.3375 20 Ambient Sulfate Fatigue -0.3734 0.2348 0.1117 Statistically significant positive associations were found between personal PM2 . 5 and increased "Breathing Difficulties" and "Any Respiratory Symptom" (p < 0.05). However, after observing individual symptom response plots by subject, it was determined that one subject was highly influential for these two analyses. As a result, these two logistic regressions were completed after excluding data from this subject (#9). These analyses yielded results in which the associations between the PM2 . 5 and symptoms were weaker and not statistically significant. These results are summarized in Table 3.24 below. Table 3.24: Pooled Logistic Regression Results (Excluding Subject #9) Model Exposure Measure Symptom Slope Standard Error p value 21 Personal PM2 . 5 Breathing Difficulties 0.0151 0.0172 0.3804 22 Personal PM2 . 5 Any Respiratory Symptom 0.0219 0.0159 0.1673 Symptom questionnaire data was also analyzed on an individual basis. Stratifying each subject's exposure levels into high and low categories, and then counting the number of "more" responses in each category yielded results in which on average, the number of "more" responses in each exposure category was approximately equal. Therefore, there were no relationships within the symptom questionnaire data meriting additional analyses. 53 Overall, no consistent results were observed in the symptom analyses. Consequently, no definitive conclusions can be made regarding the associations between symptoms and exposure to P M 2 5 and/or sulfate. 3.6 Bronchodilator Use Results 3.6.1 Bronchodilator Use Database All 16 subjects provided data at the end of each sampling session (N=7) as to the number of times a bronchodilator (i.e. berotec, bricanyl, combivent, and/or ventolin) had been used during the previous 24 hours, for a total of 112 responses. Nine subjects indicated that they used their medication as needed in addition to according to a schedule. Two subjects used the medication only as needed, while 2 other subjects used the medication only according to a schedule. For 3 subjects, the bronchodilator prescriptions varied throughout the study. One of these 3 subjects (#8) only began taking ventolin during her 6 T H sampling session (i.e. she did not have access to any bronchodilators for the first 5 sampling sessions). For the other 2 subjects, the bronchodilator prescriptions were altered in the middle of the study. This is summarized by subject in Table 3.25. A graphical representation of this data is provided in Figure 3.11. Table 3.25: Number of Bronchodilator Uses per Sampling Session Subject Sampling Session Mean Prescription (# of uses per 24 hours) 1 2 3 4 5 6 7 1 6 6 6 5 7 6 6 6 6 ± as needed 2 1 2 1 5 5 5 5 3 Varied 3 3 3 2 3 3 3 2 3 3 ± as needed 4 4 4 4 3 2 3 3 3 4 + as needed 5 2 3 3 2 1 2 1 2 2 ± as needed 6 4 4 4 4 4 4 4 4 ' 4 7 0 0 0 2 0 1 0 0 As needed 8 0 0 0 0 0 2 0 0 Varied 9 5 11 5 4 6 6 7 6 4 ± as needed 10 4 3 4 5 4 4 4 4 4 + as needed 11 6 6 6 5 6 5 6 6 6 ± as needed 12 4 5 4 4 4 4 3 4 4 ± as needed 13 0 0 0 0 0 0 0 0 As needed 14 4 4 4 4 4 4 4 4 4 15 2 5 4 5 5 3 5 4 4 ± as needed 16 1 1 0 3 3 4 3 2 Varied 54 Figure 3.11: Number of Bronchodilator Uses 12 Subject 3.6.2 Bronchodilator Use Analyses Results A scatterplot of bronchodilator use vs. personal PM 2 . 5 exposures (see Figure 3.12) by subject was produced. Figure 3.12: Scatterplot of Bronchodilator Use vs. Personal PM2.5 by Subject 12-10 w CD W o J O "5 o sz o c o m o 44 24 04 o * • C O B B * • M • -40 • O < 4 I T l S t l l N l f •< » > K- • W> < < o a * < •« O O -I- x i < • A4 tott-A -$oo(- *<x + 16 •< 15 T 14 A 13 • 12 • 11 m 10 * 9 X 8 + 7 > 6 < 5 V 4 & 3 0 2 0 1 0 20 Personal PM2.5 (ug/m3) 40 60 80 100 55 Using the SPSS linear regression function, individual regression analyses were conducted between P M 2 5 and bronchodilator use. The results are summarized in Table 3.26. Table 3.26: Individual PM2.5 vs. Bronchodilator Use Regression Results Subject N Slope SE r 2 p value 1 7 0.050945 0.074349 0.08584 0.5237 2 + 7 0.293567 0.096859 0.64755 0.0291 3 7 0.000314 0.032636 0.00005 0.9927 4 + 7 -0.017201 0.113499 0.00457 0.8855 5 6 -0.057201 0.029066 0.49193 0.1205 6 * N / A 7 7 -0.013753 0.027104 0.04897 0.6335 8 + 7 0.016722 0.073112 0.01035 0.8282 9 7 0.087518 0.035952 0.54238 0.0591 10 7 -0.044026 0.058800 0.10082 0.4877 11 6 0.047151 0.036287 0.29682 0.2636 12 6 -0.024676 0.015026 0.40271 0.1759 13 * N / A 14 * N / A 15 6 -0.055879 0.056626 0.19579 0.3796 16 + 5 -0.004488 0.056126 0.00213 0.9413 * bronchodilator variable was a constant + prescription was altered during the study For three of the subjects (# 6, 13, and 14), regression analyses could not be completed because the bronchodilator variable was a constant (i.e. subjects used the same amount of medication on every sampling day). For the remaining 13 subjects, the analyses resulted in 6 positive and 7 negative slopes. One of the positive slopes was significant at p<0.05 (this was due to an increase in the bronchodilator prescription) and another was significant at p<0.10. After visually inspecting the scatterplot above and examining the results of the regressions, it was decided that the relationship between bronchodilator use and personal P M 2 5 did not warrant any further analyses. This was due to the fact that there were a number of problems with the data. First, the various subjects had used their respective medications for different purposes (i.e. scheduled vs. as needed) and therefore the "number of bronchodilator uses" had different meanings for different subjects. Furthermore, these definitions changed in the middle of the study for certain subjects as their medication prescriptions were altered. Finally, the linear regressions indicated a wide range of slopes; more than half of these slopes were negative, the opposite of expectations. 56 4. DISCUSSION 4.1 Overview of Results and the Implications 4.1.1 Personal and Ambient Fine Particulate and Sulfate Exposure In this study, 24-hour averaged personal exposures to fine particulate matter and sulfate for 16 patients with moderate obstructive pulmonary disease were determined. Additionally, concurrent P M 2 5 and sulfate concentrations were obtained from 5 Greater Vancouver Regional District air quality monitoring sites. Basic comparisons were made between the personal and ambient data to determine the relationships between these sources of exposure data and consequently, the extent of exposure misclassification in previous epidemiological studies. The data from the current study indicated that on average, personal exposure to PM2 . 5 was 60% greater than ambient levels of PM2 . 5 . However, the ambient sulfate concentrations were on average only 27% greater than the personal sulfate exposures. Additionally, personal and ambient sulfate measurements were much more highly correlated with each other (Pearson's correlation coefficient = 0.87) than were personal and ambient PM2 . 5 measurements (0.15). These observations suggest that when using ambient measurements as a surrogate for personal particulate exposures, sulfate is the more appropriate exposure measure. These relationships are discussed elsewhere in greater detail (Ebelt 1999). 4.1.2 Particulate Pollution and Lung Function A number of respiratory health effects, including lung function, symptom severity, and medication use, were investigated with respect to personal and ambient concentrations of PM2 . 5 and sulfate to determine if exposure to particulate air pollution was associated with increased respiratory symptoms. The first health outcome that was investigated in this study was the level of lung function for each of the C O P D patients. During the selection of lung function variables for analyses, the relationship between pre-sampling and post-sampling results for each of F E V i , F V C , and F E V i / F V C were determined. High Pearson correlation coefficients (0.95, 0.94, 0.98, respectively) and approximately equal pre-sampling and post-sampling values suggest that there was little short-term (i.e. 24-hour) variability between each of these variables. This implies that it would have been sufficient to have subjects perform only 1 set of forced expiratory maneuvers per sampling session, reducing the number of tasks required by each subject. This protocol however, would also have had a disadvantage in that analyses would not have been possible for changes in lung function measurements throughout the 24-hour sampling sessions. The forced expiratory volume in one second (FEVi) variable was selected to be the focus of the lung function analyses in this study for various reasons. First, spirometry is an effort-dependent maneuver, requiring understanding, coordination, and cooperation by a subject who must be carefully instructed (ATS 1995). Therefore, it is most advantageous to use data from the variables that are the least effort-dependent so that the results are more easily reproduced, and such is the case for F E V i compared to F V C . Additionally, F V C and F E V 57 attempt to measure different lung diseases, as decreases in F V C are generally associated with restrictive diseases while reductions in F E V are generally associated with obstructive diseases (West 1987). It is the obstructive diseases that are more consistent with the types of effects seen in other studies of air pollution effects. Consequently, variations of the F E V i outcome variable (i.e. post-FEVi and A F E V i ) were selected for the analyses. The relationships between lung function and exposure analyses were determined using a sequential statistical approach. First, lung function measurements for each individual (longitudinal data) were analyzed for any trends over time. Except for the data from one subject (which was excluded), no such trends were discovered. Next, the correlation between measurements obtained from different sampling sessions was investigated using the residuals obtained from ordinary (i.e. not weighted) linear regressions. This was accomplished by creating a scatterplot for every combination of sampling periods and sampling intervals, using the residuals obtained from pooled regression analyses. Examining this data indicated that there were no consistent correlations between measurements obtained during different sampling sessions. Therefore, the assumption was made that the effects of time were the same (i.e. random) for each individual. Finally, the variability between exposure measurements for each person was investigated. In this case, residuals were used to obtained variance estimates used in final weighted least squares linear regression models, such that those subjects with larger variances received less weight in the final models. Results of the weighted least squares regression analyses indicated inverse associations between the lung function variables and each of personal SO4 2", ambient SO4 2 ", and ambient PM2 . 5 analyses. Interestingly, positive, but extremely small, coefficients were observed between personal PM2 . 5 and the two lung function variables. These results were not unexpected as the personal SO4 2", ambient S0 4 2 " , and ambient PM2 . 5 were all highly correlated with each other (Pearson's correlation coefficients >0.70) while the correlation coefficients for personal PM2 . 5 exposure and any of the other particulate measures were lower (<0.20). These results also were not unanticipated from a theoretical point of view. Recent studies indicate that personal exposure to sulfate is highly correlated with outdoor sulfate concentrations, which in turn is highly correlated with outdoor PM2 . 5 (Lippmann and Thurston 1996). Similar results were found in this study (Ebelt et al. 1999). One possible explanation for this is that unlike sulfate, personal exposures to PM2 . 5 are influenced by indoor sources such as cooking, for example. Studies such as this one provide improved personal exposure estimates, and consequently the ability to observe relationships with health outcomes is increased. Furthermore, the relationships between the various exposure measures for C O P D patients have not been previously studied, nor has personal monitoring for C O P D patients been extensively achieved. These issues therefore form some of the main advantages of this study. A number of potential atmospheric confounders were analyzed using Pearson's correlation coefficients. Temperature and relative humidity appeared to have low correlations with each of the particulate measurements, (Pearson's coefficients <0.45). Ozone was found to be moderately correlated with personal and ambient SO4 2", and ambient PM2 . 5 (correlation coefficients: 0.66, 0.59, 0.67, respectively). Furthermore, including ozone in the regression anlayses altered the models considerably, suggesting that ozone may be associated with 58 changes in lung function. Therefore, it is difficult to distinguish between a suggested effect of P M and that of ozone. Additionally, numerous other studies have found associations between reduced lung function and ozone (Pope et al. 1995a). One model analyzed by weighted least squares regression indicated a borderline statistically significant inverse association between personal exposures to sulfate and A F E V i (p=0.08). All other combinations of lung function (post-FEVi or A F E V i ) and exposure measurements (personal PM2 . 5 , personal sulfate, ambient P M 2 5 , ambient sulfate) produced results that were not statistically significant, indicating that the results may be due to chance. After converting these statistical results into changes in lung function per change in exposure measure, the data indicated a range of-1.09 to +0.06% and -0.27 to -1.04% changes in F E V i associated with 10 pg/m 3 increases in PM2 . 5 and 1 pg/m 3 increases in SO4 2 ", respectively. These results are difficult to compare directly to results of other studies as different measurements of particulate pollution have been used depending on the study. However, the published literature generally indicates a physiologically small but statistically significant decline in lung function as a result of exposure to particulate air pollution (Pope et al. 1995a). Overall, these studies have shown decreases of up to 0.35% in F E V i associated with each 10 pg/m 3 increase in daily mean PMio (Wilson and Spengler 1996). The large declines in F E V i observed in this study are neither inconceivable nor inexplicable, considering the facts that the observed results were based on a population believed to be at increased risk for health effects due to particulate polluation and that smaller changes in concentrations of PM2 . 5 and SO4 2 - relative to P M i 0 are associated with larger lung function changes. These larger effects observed for the ambient P M 2 5 (and personal exposure to ambient PM2 . 5 , as measured by sulfate) may provide support for the hypotheses that this particulate size fraction is more toxic and more highly associated with adverse health effects than PMio. Furthermore, the large changes in lung function observed in this study may support the hypothesis that people with C O P D are more susceptible to adverse respiratory effects as a result of exposure to particulate air pollution (Schwartz et al. 1996). As mentioned previously, except for one model (personal SO4 2 " vs. A F E V i ) which was borderline statistically significant, all lung function regression models produced results that were not statistically significant. This was not unexpected, as the power to detect a statistically significant association was greatly reduced from what was expected at the outset of this project. Power calculations were examined to determine the lowest association (slope) that could be detected at a significance level of a = 0.05 with a power of 80% for each of the 8 models in the study. The results of these calculations indicated that in order for the regression analyses to produce statistically significant results using the actual study sample size and variance, the changes in lung function would have had to be relatively large (and unlikely) at approximately 1.4-7.6% and 2.0-3.7% changes in F E V i for every 10 pg/m 3 increase in PM2 . 5 and 1 pg/m 3 increase in SO4 2 -, respectively. Changes of this magnitude were not observed (and have not previously been observed in other studies (Dockery and Pope 1994)) and therefore, no definite (i.e. statistically significant) conclusions can be made. Overall, the lung function component of this study did not provide any statistically significant findings. However, the results suggest that increases of 10 pg/m 3 of PM2 . 5 may be 59 associated with up to a 1% decline in F E V i in people with C O P D . Similarly, 1 u.g/m3 increases in sulfate may be associated with declines in F E V i from 0.27-1.04%. 4.1.3 Particulate Pollution and Respiratory Health Symptoms This component of the study was carried out to investigate the hypothesis that exposure to PM2 . 5 and/or sulfate causes an increase in reporting of respiratory symptoms amongst a group of C O P D patients. The pooled logistic regression analyses completed in this portion indicated that approximately 50% of the slopes were positive (i.e. indicating an increase in reported symptoms with increases in pollutants). The other half of the models indicated negative slopes, suggestive of the opposite situation in which a decrease in reported symptoms was associated with increases in PM2 . 5 or sulfate. It is interesting to note that although all slopes for the coughing models were not statistically significant, they all suggested an inverse association between cough and all exposure measures. For the breathing difficulty models, all 4 models suggested a direct relationship between the symptom prevalence and the respective exposure variable. Furthermore, the relationship between breathing difficulties and personal PM2 . 5 was statistically significant (p=0.03). For the collapsed symptom variable (i.e. Any Respiratory Symptom), the personal PM2 . 5 model was found to be statistically significant (p=0.02). Despite these few interesting observations, overall, no consistent results were noticed, as the probability of reporting increases in symptoms for the various models ranged from less than -100% to greater than 35%, and less than -100%) to greater than 100%, for 10 pg/m 3 increases in PM 2 .5 and sulfate, respectively. These results, similar to the lung function results, are difficult to precisely compare to literature results because different exposure measures have been used in different studies. The documented studies have often reported statistically significant associations between particulate pollution and more severe respiratory symptoms. Typically, these studies have found that a 10 u,g/m3 increase in PMio has been associated with a 1-10% increase in respiratory symptoms (Pope et al. 1995a) and in some cases, the percentage increases in reported respiratory symptoms have been as high as 15% (Wilson and Spengler 1996). Another key issue to consider when interpreting the symptom analyses results is the fact that some individuals had strong tendencies to always provide the same questionnaire response regardless of any other factors. For example, one subject (#9) who had a high mean PM2 . 5 exposure (36.2 u.g/m3) throughout his 7 sampling sessions also tended to state that he always had "more" breathing difficulties. This raises a number of interesting issues in terms of the data analysis of these responses. The situation may have been that these responses actually were an indication that the subject was truly experiencing more breathing difficulties and therefore, these may be associated with the high exposures recorded on those sampling days. However, it may have also been the case that the subject had a tendency to reply "more" to the question, even though he may not truly have been experiencing any more symptom severity than usual. In the former case, the logistic regression analyses would accurately describe the case for this individual and the data could easily be used in pooled analyses. In the latter, the responses are effectively false positives, which must be adjusted for in pooled analyses. Unfortunately, it is virtually impossible to identify which case it is. In the event that the latter situation was the one encountered during the study, the pooled logistic regression 60 analyses for the 2 statistically significant results were repeated after excluding the results from this one subject. These new results were not statistically significant. A different type of analysis was also attempted to interpret the results should the case with the false positives be the true one for the study. In this analysis, each subject's 7 exposures were stratified into high and low groups, based on individual medians. This allowed for the adjustment for differences in exposures between subjects. The number of "more" responses in each category was counted for each individual. These numbers were then to be pooled by exposure category. This however, was not completed as the preliminary data indicated that the number of "more" responses in each category for each individual was approximately equal, suggesting no relationship between exposure to particulate matter and symptom severity. Although this type of analysis effectively eliminates the problem mentioned above regarding false positives, another problem is introduced making results difficult to interpret. Now, as an example, in the case of subject #9, his data suggests no relationship. This is accurate if he was providing false positive responses; if his answers were true, the power to detect this in the new analyses is effectively lost. When this individual's data is pooled with the data from the rest of the group, his component shows no association when in fact it may be that he was truly experiencing more breathing difficulties during all of his high exposure sampling sessions. These types of problems appear to be inherent to all symptom questionnaire analyses due to the facts that the data is largely subjective, subjects may consistently provide the same responses regardless of other factors, and exposure levels for each individual can be substantially different. An additional factor, unique to this study in comparison to others, that must be included when interpreting the results of this component (and the lung function component) is the fact that all subjects continued to use their regular medications throughout the entire study period. For the majority of the subjects, these medications included at least one bronchodilator, designed to improve lung functioning and alleviate respiratory symptoms. Consequently, lung function measurements and reported symptom severity during the study may have been confounded by bronchodilator use, as the use of these medications may have limited the true variability of the respiratory health outcomes. However, it is also possible that the amount of medication use was an indication of the magnitude of effect that the pollutants had on the subjects and therefore, this was the reason for attempting the medication use analyses. Unfortunately, due to problems stated below (see Section 4.1.4), these analyses were not successfully completed. Evident from the results of the current study, especially in comparison to previously documented results, is the fact that the symptom analyses do not provide any substantial additional information regarding the relationship between exposure to particulate air pollution and respiratory symptoms. This is most likely due to a small study population, relatively small variability in pollution levels, the nature of the inherently subjective symptom questionnaire, and the use of respiratory medication. 61 4.1.4 Particulate Pollution and Bronchodilator Use This final component of the study was designed to test the hypothesis that exposure to PM2.5 and/or sulfate results in increased bronchodilator use in patients with moderate chronic obstructive pulmonary disease. Various problems were encountered during the data analyses making it very difficult to arrive at any conclusions. One problem with the data was that each subject used his/her medication for a different purpose, as some subjects used the bronchodilators strictly according to a schedule, some used the medication only as needed, while others used it both according to a schedule in addition to as needed. Another substantial problem was that the medication prescriptions for three of the subjects were altered in the middle of the study. Consequently, the results from pooled data were inconclusive. It was possible however to analyze the data for some of the individuals independently using ordinary least squares linear regression. For three of the subjects, this could not be completed because the bronchodilator variable was a constant (i.e. strict medication schedules were followed). For the remaining 13 subjects, the analyses resulted in 6 positive (i.e. direct association) and 7 negative (i.e. inverse association) slopes. Two of the positive slopes were significant (p=0.03 and p=0.06, respectively). For the first one of these slopes, this was the result of the subject having his bronchodilator prescription increased during the study. Furthermore, the results indicated a wide range of slopes; more than half of these were negative, the opposite of expectations. Although no conclusive results were obtained from this study regarding the relationship between bronchodilator use and exposure to PM2.5 and/or sulfate, previous studies have found that an approximate 2.3 to 12% increase in bronchodilator use in asthmatics is associated with each 10 u.g/m3 increase in PMio (Wilson and Spengler 1996). In one study completed with relatively healthy fourth and fifth grade elementary students and asthmatic patients, elevated levels of PMio pollution were associated with statistically significant increases in the use of asthma medication (Pope et al. 1991). For the patient-based sample, the probability of the use of extra asthma medication was 6.2 times as high when PMio levels were at the highest (approximately 18 times higher) for the study period versus the lowest. In another study performed on children with chronic respiratory symptoms, results similarly indicated a consistently positive association between PMio, black smoke, and SO2 and the amount of bronchodilator use (Roemer et al. 1993). Therefore, the inconclusive results in the current study were most likely due to a small study population, relatively small variability in pollution levels and especially the fact that the bronchodilators were used for different purposes among the subjects. 4.2 Factors Affecting Study Outcome 4.2.1 Exposure Assessment One of the main factors affecting the outcome of the health effect analyses in this study was the selection of specific exposure measures used to investigate the relationships between particulate pollution and respiratory health effects. Four different exposure measures were 62 selected: personal and ambient PM2 . 5 , and personal and ambient SO42". These particle classifications were chosen for a number of reasons. Traditionally, epidemiological studies have relied on estimates of exposure in which single ambient air pollution concentrations have been used to represent the exposure of an entire study population (Pope et al. 1995b, Dockery et al. 1992). Consequently, the estimations of the individual exposures to particles from the central outdoor pollution monitors have resulted in considerable error, referred to as exposure misclassification. Another important issue is the fact that many time series study designs have been based on the assumptions that ambient particles efficiently penetrate indoors, and that ambient monitoring data is therefore an adequate surrogate for indoor concentrations, and consequently for personal exposures to ambient particles. For these reasons, the exposure assessment component of this study was applied at the individual and ambient levels, making it possible to evaluate the effects of both levels of exposure classification (based on personal and ambient monitoring) of particles on respiratory health effects. This was one major advantage of this study. In addition to fine particles, sulfate was also included in the analyses in this study. Sulfate, a marker of outdoor combustion-source particulate, has been suggested as a better exposure metric than PM2 . 5 due to high correlation between personal and ambient sulfate concentrations (as was observed in this study) (Lippmann and Thurston 1996). Furthermore, in comparison to particulate matter, the sulfate component penetrates indoors efficiently, exhibits less spatial variability and has no major indoor sources (Ozkaynak et al. 1996). For these reasons, in this study, the associations between respiratory health effects and both ambient concentrations and personal exposures for aerosol sulfate were examined, over the same averaging period. 4.2.2 Pollution Levels Another important factor to consider when discussing the results of this study is the concentration level of each pollutant analyzed. Some previous studies that have investigated the associations between particulate air pollution and adverse health effects were completed at mean ambient concentrations much greater than those reported in this study. For example, in a study performed by Pope and Kanner on C O P D patients, PMio levels as high as 100 pg/m 3 were associated with small 2% declines in F E V i (Pope and Kanner 1993). In other studies, mean PMio concentrations have ranged from 27-76 pg/m 3 (Vedal 1997). Furthermore, in some cases, 24-hour PMio concentrations have been reported to exceed 150 pg/m 3 . At these extremely high levels of P M , up to a 7% reduction in lung function has been observed (Pope et al. 1995a). As a comparison, the mean PMio concentration in this study was 18 pg/m 3 or 1.58 times the mean ambient P M 2 5 The range of particulate levels encountered in Greater Vancouver during the summer of 1998 was therefore substantially below those reported in other studies where the associations between adverse health effects and exposure to particulates have been more conclusive. If there had been a greater range of concentrations of pollutants during the summer of 1998, it is likely that there would have been more variability in the observed health effects, resulting in increased statistical significance. Therefore, the lack of statistical significance in the results of this study can be at 63 least partly explained by the relatively small variability in particulate levels that were experienced in Greater Vancouver during the summer of 1998. 4.2.3 Study Population Another unique aspect of this project was that the study population was comprised of elderly individuals with mild to moderate chronic obstructive pulmonary disease (COPD). By using this approach, the healthy adult and childhood asthma populations that have traditionally been chosen for many of the health effects exposure studies were avoided. The group of C O P D patients selected for this project was done so with the purpose of increasing the sensitivity of the study, as these individuals are representative of those who are at increased risk for air pollution-associated health effects (Schwartz et al. 1996). Consequently, the severity of health effects observed in this study are valid for such a population, as these effects were most likely more severe than those experienced by the general population during the study period. Therefore, it is important to note that the results of this study are valid for a population subgroup that is considered to be "high risk." 4.2.4 Sample Size Another relevant issue to consider when making conclusions about this study is the number of subjects who agreed to participate in the project. Without information on exposure variability of elderly C O P D patients, formal sample size estimates prior to the beginning of the study were not possible. Therefore, the study was originally designed to be completed with a sample size of 25-30 selected participants. This number of subjects would have provided a broad range of possible relationships between personal exposures and pollutant concentrations in Greater Vancouver and was considered to be a practical limit for the level of exposure monitoring that was planned. Unfortunately, only 16 subjects meeting the study criteria were recruited, reducing the statistical power of the analyses. Power calculations were completed to determine the lowest detectable associations (slopes) at a significance level of a = 0.05 with a power of 80% that would have resulted had 30 subjects participated in the study. For all 8 models, these new lowest detectable associations (for 30 subjects) would have remained relatively large at approximately two thirds of the original lowest detectable associations (for 16 subjects), still well above the actual observed associations. Therefore, the statistical power to detect an association between particulate pollution and health effects would not have been substantially improved had 30 subjects been recruited. 4.3 Conclusions Overall, the results of this study suggest that exposure to ambient PM2 . 5 , and personal and ambient sulfate, but not personal PM2 . 5 , may be associated with decreased F E V i in patients with mild to moderate chronic obstructive pulmonary disease in the Greater Vancouver metropolitan area. Specifically, the results suggest that up to a 1% decline in F E V i may be associated with each 10 u.g/m3 increase in PM2 . 5 , and 0.3-1.0% declines in F E V i may be associated with each 1 u,g/m3 increase in sulfate. Interestingly, these associations were not 64 observed with respect to personal PM2 . 5 , suggesting that ambient PM2 . 5 , personal sulfate, and ambient sulfate may be more strongly associated with adverse health effects than is personal PM2 . 5 It is important to note that no lung function analyses were statistically significant, and furthermore, no consistent associations were observed between symptom severity and bronchodilator use, and exposure to PM2 . 5 and/or sulfate. Ozone appeared to be associated with A F E V i and may have been more strongly associated with lung function changes than personal sulfate. Limitations of this study include little variability in the exposure (due to a small sample size and relatively small range in particulate pollution concentrations), and the quality of the symptom questionnaire and bronchodilator use data. 4.4 Recommendations As a whole, this study design is adequate to investigate the relationship between exposure to particulate air pollution and adverse respiratory health effects. However, a number of improvements could be made to the study protocol to increase the significance of the results. First, one of the limitations of the study was a small sample size; ideally, this should be increased as a larger study population could result in more exposure variability. As an alternative or as a further improvement to the protocol, increasing the number of sampling sessions per subject would have a similar effect. Both of these options have advantages and disadvantages. Increasing the sample population would increase the power of the analyses without requiring the subjects to participate in more than 7 sampling sessions. Furthermore, if these additional subjects were monitored on different days, this would also tend to increase the variability of the exposures. This approach, however, would be labour intensive and costly, requiring additional research technicians. Increasing the number of sample sessions per subject also could improve the power of the analyses, without having to deal with the obstacle of recruiting additional subjects. This appears to be the most efficient way to increase the variability of the exposures. However, when using this approach, the likelihood of recruiting each subject would most likely diminish, as each one would be required to commit more of his/her time to the project. Additionally, subjects might withdraw mid-way through the study if they were asked to participate in too many sampling sessions. Therefore, perhaps the best approach would be to increase the number of sampling sessions slightly (up to 10) and also recruit additional subjects. The subject compliance rate in this study was excellent at 100%, as all of the subjects that consented to participate completed all sampling sessions. Some subjects stated that they were actually looking forward to each sampling session because they enjoyed the visits with the research technicians. Al l efforts were made by these technicians to put the subjects at ease during each visit and to explicitly state how thankful the technicians were for each subject's participation. For similar studies, in which there is considerable personal contact with the subjects, it is recommended that the subjects be explicitly told and made to feel that they are a most vital component to the success of the project. From the experience in this study, this attitude towards the subjects makes them feel useful and valuable, and as a result, they are generally happy to participate in such projects. In certain cases, when it was felt that subjects were growing tired of the project, the technicians brought small pastries to the subjects on the next sampling day. These "treats" 65 were very much appreciated. The subjects also appreciated the phone calls that they received the day before each sampling session. These served as reminders of the following day's sampling session for the subjects and also provided the opportunity for the technicians to discover if any unforeseen circumstances had arisen that would make sampling impossible for the subject during the originally scheduled time. Most subjects were motivated to participate in the project either because they felt that they were contributing to science or they simply wanted to spend time talking with visitors. The honorarium itself was the main motivator for participation in the study for only 1 subject. The other individuals perceived it as a pleasant, but unnecessary gift. If just to serve this purpose, it is recommended that honorariums still be provided for all subjects participating in similar studies as a form of expressing thankfulness and appreciation. The majority of the subjects that were unwilling to participate in the project stated that they were either simply not interested, or that participation would require too much effort. For these individuals, perhaps additional motivation is required to help entice them to participate. This motivation could come in the form of greater honorariums, or other gratuities, such as restaurant or other gift certificates. The symptom questionnaire component of the study had two major limitations. First, the questions were phrased in such a way that allowed for at least partially subjective answers, making pooled analyses extremely difficult. Furthermore, specific subjects tended to always provide the same response regardless of other factors. Therefore, phrasing the questions in such a way that the answers would be more objective and variable would provide more power to the symptom/PM analyses. One recommendation would be to have subjects rate each symptom severity "on a scale of 1-10." This would allow for more variation in the responses which could subsequently be converted into dichotomous variables by splitting the responses into two groups: more or less, defined as being, for example, >6 or <5 respectively. This method would also allow for other types of statistical analyses, as the data would be in a form in which there were 10 possible values instead of 2. The last component of this study also had two major limitations. One was the fact that the "bronchodilator use" question had different definitions to different subjects, despite efforts to ensure consistency. The second was that the medication prescriptions varied among the subjects and it was not possible to account for these differences in the analyses. Therefore, should this type of study be repeated, it is important that the medication use be explicitly described with as much detail as possible. This should include information regarding the regular bronchodilator prescription for each subject and whether or not these prescriptions were followed on a regular basis. Additionally, if the prescriptions were not followed for each sampling session, information should be recorded regarding the changes to the prescription for each sampling session. Furthermore, any physician-directed prescription alterations should be recorded. Possibly the most efficient method of recording this data would be to have each subject record the time of individual bronchodilator uses and the purpose (i.e. scheduled dose, extra dose needed for shortness of breath, etc.) of taking the medication. This would allow for a number of different analyses, based on medication used according to prescriptions, as needed, or combination of the two. These recommendations would likely improve the power of the analyses. 66 BIBLIOGRAPHY ATS. 1995. Standardization of Spirometry, 1994 Update. American Thoracic Society. Am J Respir Crit Care Med 152: 1107-36. Bennett, W. D., K. L. Zeman, and C. Kim. 1996. Variability of fine particle deposition in healthy adults: effect of age and gender. Am J Respir Crit Care Med. 153: 1641-7. Borja-Aburto, V . H., M. Castillejos, D. R. Gold, S. Bierzwinski, and D. Loomis. 1998. Mortality and ambient fine particles in southwest Mexico City, 1993- 1995. Environ Health Per sped 106: 849-55. Burnett, R. T., M. Smith-Doiron, D. Stieb, S. Cakmak, and J. R. Brook. 1999. Effects of particulate and gaseous air pollution on cardiorespiratory hospitalizations. Arch Environ Health 54: 130-9. Churg, A. , and M. Brauer. 1997. Human lung parenchyma retains PM2.5. American Journal of Respiratory & Critical Care Medicine 15 5: 2109-11. Clayton, C. A. , R. L. Perritt, E. D. Pellizzari, K. W. Thomas, R. W. Whitmore, L. A . Wallace, H. Ozkaynak, and J. D. Spengler. 1993. Particle Total Exposure Assessment Methodology (PTEAM) study: distributions of aerosol and elemental concentrations in personal, indoor, and outdoor air samples in a southern California community. Journal of Exposure Analysis & Environmental Epidemiology 3: 227-50. Dockery, D. W., A . C. d. Pope, X. Xu , J. D. Spengler, J. H. Ware, M. E. Fay, B. G. Ferris, Jr., and F. E. Speizer. 1993. An association between air pollution and mortality in six U.S. cities. New England Journal of Medicine 329: 1753-9. Dockery, D. W., and C. A. Pope, 3rd. 1994. Acute respiratory effects of particulate air pollution. Annual Review of Public Health 15: 107-32. Dockery, D. W., J. Schwartz, and J. D. Spengler. 1992. Air pollution and daily mortality: associations with particulates and acid aerosols. Environmental Research 59: 362-73. Ebelt, S. 1999. Particulate Air Pollution and Chronic Obstructive Pulmonary Disease Patients: A n Assessment of Exposure and Cardiovascular Health Effects. Occupational and Environmental Hygiene. University of British Columbia, Vancouver. Ebelt, S. T , T. V. Fisher, A. J. Petkau, S. Vedal, and M. Brauer. 1999. Exposure of Chronic Obstructive Pulmonary Disease (COPD) Patients to PM2.5: Relationship Between Personal and Ambient Air Concentrations. Pages 5. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, U S A . 67 Enright, P. L., L. R. Johnson, J. E. Connett, H. Voelker, and A. S. Buist. 1991. Spirometry in the Lung Health Study. 1. Methods and quality control. Am Rev Respir Dis 143: 1215-23. Gold, D. R , A. I. Damokosh, C. A. Pope, 3rd, D. W. Dockery, W. F. McDonnell, P. Serrano, A. Retama, and M. Castillejos. 1999. Particulate and ozone pollutant effects on the respiratory function of children in southwest Mexico City. Epidemiology 10: 8-16. G V R D . 1994. G V R D Air Quality Management Plan. Greater Vancouver Regional District, Burnaby, B C . G V R D . 1997. Lower Fraser Valley Ambient Air Quality Report 1997. Greater Vancouver Regional District, Burnaby, B C . Harre, E. S., P. D. Price, R. B. Ayrey, L. J. Toop, I. R. Martin, and G. I. Town. 1997. Respiratory effects of air pollution in chronic obstructive pulmonary disease: a three month prospective study. Thorax 52: 1040-4. Hoek, G., and B. Brunekreef. 1993. Acute effects of a winter air pollution episode on pulmonary function and respiratory symptoms of children. Arch Environ Health 48: 328-35. Janssen, N. A. , G. Hoek, B. Brunekreef, H. Harssema, I. Mensink, and A. Zuidhof. 1998. Personal sampling of particles in adults: relation among personal, indoor, and outdoor air concentrations. Am J Epidemiol 147: 537-47. Janssen, N. A. , G. Hoek, H. Harssema, and B. Brunekreef. 1997. Childhood exposure to PMIO: relation between personal, classroom, and outdoor concentrations. Occup Environ Med 54: 888-94. Janssen, N. A. , G. Hoek, H. Harssema, and B. Brunekreef. 1999. Personal exposure to fine particles in children correlates closely with ambient fine particles. Arch Environ Health 54: 95-101. Katsouyanni, K., A. Karakatsani, I. Messari, G. Touloumi, A. Hatzakis, A . Kalandidi, and D. Trichopoulos. 1990. Air pollution and cause specific mortality in Athens. J Epidemiol Community Health 44: 321-4. Korrick, S. A. , L. M . Neas, D. W. Dockery, D. R. Gold, G. A. Allen, L. B. Hill, K. D. Kimball, B. A. Rosner, and F. E. Speizer. 1998. Effects of Ozone and Other Pollutants oh the Pulmonary Function of Adult Hikers. Environ Health Perspect 106: 93-9. Lebowitz, M . D. 1996. Epidemiological studies of the respiratory effects of air pollution. Eur Respir J 9: 1029-54. 68 Lippmann, M. , and G. D. Thurston. 1996. Sulfate concentrations as an indicator of ambient particulate matter air pollution for health risk evaluations. J Expo Anal Environ Epidemiol 6: 123-46. Logan, W. P. D. 1953. Mortality in London fog incident. Lancet 264: 336-8. Martonen, T. B., R. A. Segal, and C. S. Kim. 1999. Comparison of Particle Deposition in Healthy Subjects and C O P D Patients Using Computer Simulations and Human Data. Pages 28. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, U S A . Monn, C , A. Fuchs, D. Hogger, M. Junker, D. Kogelschatz, N. Roth, and H. U. Wanner. 1997. Particulate matter less than 10 microns (PMIO) and fine particles less than 2.5 microns (PM2.5): relationships between indoor, outdoor and personal concentrations. Sci Total Environ 208: 15-21. Ostro, B., J. M. Sanchez, C. Aranda, and G. S. Eskeland. 1996. Air pollution and mortality: results from a study of Santiago, Chile. Journal of Exposure Analysis & Environmental Epidemiology 6: 97-114. Ozkaynak, H., J. Xue, J. Spengler, L. Wallace, E . Pellizzari, and P. Jenkins. 1996. Personal exposure to airborne particles and metals: results from the Particle T E A M study in Riverside, California. J Expo Anal Environ Epidemiol 6: 57-78. Pope, C. A. , 3rd. 1996. Adverse health effects of air pollutants in a nonsmoking population. Toxicology 111: 149-55. Pope, C. A. , 3rd, D. V. Bates, and M. E. Raizenne. 1995a. Health effects of particulate air pollution: time for reassessment? Environmental Health Perspectives 103: 472-80. Pope, C. A. , 3rd, M. J. Thun, M. M. Namboodiri, D. W. Dockery, J. S. Evans, F. E . Speizer, and C. W. Heath, Jr. 1995b. Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. American Journal of Respiratory & Critical Care Medicine 151: 669-74. Pope, C. A. d., D. W. Dockery, J. D. Spengler, and M. E. Raizenne. 1991. Respiratory health and PMIO pollution. A daily time series analysis. American Review of Respiratory Disease 144: 668-74. Pope, C. A. d., and R. E . Kanner. 1993. Acute effects of PMIO pollution on pulmonary function of smokers with mild to moderate chronic obstructive pulmonary disease. American Review of Respiratory Disease 147: 1336-40. Pope, C. A. d., J. Schwartz, and M. R. Ransom. 1992. Daily mortality and PM10 pollution in Utah Valley. Archives of Environmental Health 47: 211-7. 69 Roemer, W., G. Hoek, and B. Brunekreef. 1993. Effect of ambient winter air pollution on respiratory health of children with chronic respiratory symptoms. Am Rev Respir Dis 147: 118-24. Roemer, W., G. Hoek, B. Brunekreef, J. Haluszka, A . Kalandidi, and J. Pekkanen. 1998. Daily variations in air pollution and respiratory health in a multicentre study: the P E A C E project. Pollution Effects on Asthmatic Children in Europe. Eur Respir J 12: 1354-61. Scarlett, J. F., K. J. Abbott, J. L. Peacock, D. P. Strachan, and H. R. Anderson. 1996. Acute effects of summer air pollution on respiratory function in primary school children in southern England. Thorax 51: 1109-14. Schwartz, J. 1993. Air pollution and daily mortality in Birmingham, Alabama. Am J Epidemiol 137: 1136-47. Schwartz, J. 1994. What are people dying of on high air pollution days? Environ Res 64: 26-35. Schwartz, J., D. W. Dockery, and L. M. Neas. 1996. Is daily mortality associated specifically with fine particles? Journal of the Air & Waste Management Association 46: 927-39. Schwartz, J., and A. Marcus. 1990. Mortality and air pollution in London: a time series analysis. Am J Epidemiol 131: 185-94. Studnicka, M. J. , T. Frischer, R. Meinert, A. Studnicka-Benke, K. Hajek, J. D. Spengler, and M. G. Neumann. 1995. Acidic particles and lung function in children. A summer camp study in the Austrian Alps. Am J Respir Crit Care Med 151: 423-30. Thomas, K. W., E. D. Pellizzari, C. A. Clayton, D. A. Whitaker, R. C. Shores, J. Spengler, H. Ozkaynak, S. E. Froehlich, and L. A. Wallace. 1993. Particle Total Exposure Assessment Methodology (PTEAM) 1990 study: method performance and data quality for personal, indoor, and outdoor monitoring. Journal of Exposure Analysis & Environmental Epidemiology 3: 203-26. Tiittanen, P., K. L. Timonen, J. Ruuskanen, A. Mirme, and J. Pekkanen. 1999. Fine particulate air pollution, resuspended road dust and respiratory health among symptomatic children. Eur Respir J 13: 266-73. Vedal, S. 1995. Health effects of inhalable particles: implications for British Columbia. B.C. Ministry of Environment, Lands and Parks, Vancouver, B.C. Vedal, S. 1997. Ambient particles and health: lines that divide. Journal of the Air & Waste Management A ssociation 47: 551-81. Wallace, L. 1996. Indoor particles: a review. J Air Waste Manag Assoc 46: 98-126. 70 Watt, M. , D. Godden, J. Cherrie, and A. Seaton. 1995. Individual exposure to particulate air pollution and its relevance to thresholds for health effects: a study of traffic wardens. Occup Environ Med 52: 790-2. West, J. B. 1987. Pulmonary Pathophysiology - the essentials. Williams & Wilkins, Baltimore, M D . Wilson, R., and J. D. Spengler. 1996. Particles in Our Air: Concentrations and Health Effects. Harvard University Press, Boston. Wilson, W. E. , and H. H. Suh. 1997. Fine particles and coarse particles: concentration relationships relevant to epidemiologic studies. J Air Waste Manag Assoc 47: 1238-49. Wise, R. A. , J. Connett, K. Kurnow, J. Grill, L. Johnson, R. Kanner, and P. Enright. 1995. Selection of spirometric measurements in a clinical trial, the Lung Health Study. Am J Respir Crit Care Med 151: 675-81. APPENDIX A - INTRODUCTORY LETTER T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A E n v i r o n m e n t a l a n d O c c u p a t i o n a l L u n g D i s e a s e R e s e a r c h Un i t 72 Air Pollution Health Study With your cooperation, Dr. Michael Brauer, Dr. Sverre Vedal, Dr. A. John Petkau, Mr. Teri Fisher and colleagues at the University of British Golumbia will be conducting a study to evaluate the exposure of individuals with Chronic Obstructive Pulmonary Disease (COPD) to fine particle air pollution Recent studies suggest that COPD patients have increased susceptibility to particulate air pollution. With this study, we hope to identify factors which predict high particle exposure and to evaluate the health impact of particle air pollution on individuals with COPD. Participants in this project will be compensated with an honorarium of $250. A member of our research staff will be contacting you shortly with further details about the procedures involved in collecting data and to ask about your interest in participating in this project. If you have any questions regarding participation in the project, please feel free to contact Mr. Teri Fisher at 822-1274 or Dr. Brauer at 822-9585. We hope that you will agree to participate and we thank you for your cooperation. Michael Brauer, ScD. Sverre Vedal, MD. A. John Petkau, PhD. Teri Fisher, BSc. Sincerely, M. Brauer , ScD Department ol Medicine Occupational Hygiene Programme 3rd Boor. 2206 East Mall Vancouver. B . C . C a n a d a V 6 T 1Z3 M. C h a n - Y e u n g , M B , F R C P Respiratory Division Department ol Medicine 2775 Heather Street Vancouver , B . C . C a n a d a V 5 Z 3J5 H. D i m i c h - W a r d , P h D Respra tory Division Department ol Medicine 2775 Heather Street Vancouver , B . C . C a n a d a V 5 Z 3J5 S.M. K e n n e d y , P h D Occupational Hygiene Programme Faculty ol Graduate Studies 3rd floor. 2206 East Mall Vancouver . B . C . C a n a d a V6T IZ3 S. V e d a l , M D , F R C P C Respiratory Division Department ol M e t f c n e 2775 Heather Street Vancouver . B . C . Canada V5Z 3J5 Phone: (604) 822-9585 Fax: (604) 822-9588 Phone: (604) 675-4122 Fax: (604) 875-4695 Phone: (604) 875-4122 Fax: (604) 875-4695 Phone: (604) 822-9577 Fax' (604) 822-9588 Phone: (604) 875-4122 Fax: (604)875-4695 APPENDIX B - C O N S E N T F O R M T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A 74 Environmental and Occupational Lung Disease Research Unit •S3*: INFORMED CONSENT FORM Personal exposure to fine particles Principal Investigator: Dr. Michael Brauer, Department of Medicine, The University of British Columbia Phone: 822-9585 Co-Investigators: Dr. Sverre Vedal, Department of Medicine, The University of British Columbia Phone: 875-4122 Dr. A. John Petkau, Department of Statistics, The University of British Columbia Phone: 822-4673 Teri Fisher, Experimental Medicine Program , The University of British Columbia Phone: 822-1274; Cellular: Stefanie Ebelt, Occupational Hygiene Program, The University of British Columbia Phone: 822-1274; Cellular: Purpose: I, understand that I am being invited to participate in a (name in full) research study which will determine the personal exposure levels to fine particle air pollutants of patients with COPD (Chronic Obstructive Pulmonary Disease). COPD patients are being invited to participate because research has suggested that these individuals have increased susceptibility to particulate air pollution. We plan to measure levels of exposure and to evaluate factors associated with high or low exposure. We will also evaluate the impact of these exposures on respiratory and cardiovascular health. The personal exposure levels will be analyzed and compared to average levels of air pollution in Vancouver as measured by the Greater Vancouver Regional District. M. Brauer, ScD Department of Medicine Occupational Hygiene Programme 3rd floor, 2206 East Mall Vancouver, B.C. Canada V6T 1Z3 Phone: (604)822-9585 Fax: (604) 822-9588 U. Chan-Yeung, MB, FRCP Respiratory Division Department of Medicine 2775 Heather Street Vancouver, B.C. Canada V5Z 3J5 Phone: (604)875-4122 Fax: (604) 875-4695 H. Dimich-Ward, PhO Respiratory Division Department of Medicine 2775 Heather Street VanJauwfraClCafrfa^SZ 3J5 Phone: (604)875-4122 Fax: (604) 875-4695 S.M. Kennedy, PhO Occupational Hygiene Programme Faculty ol Graduate Studies 3rd floor, 2206 East Mall Vancouver, B.C. Canada V6T 1Z3 Phone: (604) 822-9577 Fax: (604) 822-9588 S. Vedal, MD, FRCPC Respiratory Division Department of Medicine 2775 Heather Street Vancouver, B.C. Canada V5Z 3J5 Phone: (604) 875-4122 Fax: (604) 875-4695 75 Study Procedures: I understand that: • a technician will contact me to arrange an appointment to visit me at my home and equip me with a personal monitor and sampling pump, to measure the level of air pollutant exposure. When the monitor is worn, I may engage in all normal activities. At the completion of the 24-hour monitoring period, the technician will retrieve the sampler and activity log (described below) from me. In addition, lung function and heart monitoring tests (described below) will be performed at this time. • I will be asked to provide 7 days of personal monitoring (each 24 hours in length) over a period of 4 months. The monitoring will take place at intervals of approximately 1.5 weeks apart, and I will be instructed at the beginning of the project as to when my samplings are to take place. • I will be asked to complete a daily activity diary during each of the 24 hour monitoring periods. To complete this diary, I will be asked to provide information on the proximity of my home and/or workplace to traffic, the hours of the day spent outside, indoors, in transit, and undergoing specific activities associated with particle exposure (exposed to environmental tobacco smoke or cooking). The diary will require approximately 15 minutes to complete on each of the 7 sampling days. • I will be asked to complete a short symptom diary at the end of each monitoring period indicating the occurence of respiratory and cardiac symptoms and the use of a bronchodilator (inhaler) and other medication during the measurement period. The symptom diary will require approximately 2 minutes to complete on each of the 7 sampling days. • Lung function tests (spirometry) will be performed at the completion of each monitoring period. I will be asked to wear nose plugs and perform three forced expiratory maneuvers (taking a deep breath and then breathing out as quickly and forcefully as possible). The lung function tests will require approximately 5 minutes to complete on each of the 7 sampling days. • I will be asked to wear a portable electrocardiograph recorder (Holter monitor) during each of the 7 24-hour monitoring periods. For the Holter monitoring, several electrical leads will be attached to my chest under my clothing. Information on the heart rate is measured by the leads and recorded on a tape. This measurement will require approximately 15 minutes to assemble on each of the 7 sampling days. During the 24-hour period I will engage in normal activities, although I will not be allowed to bathe or shower. • I will also be asked to wear a particle measurement device and a Holter monitor for one 6-hour period during the study. During this period I will also be asked to engage in a series of typical every-day activities such as walking outside, riding on a bus and eating in a restaurant. These measurements are designed to measure any immediate responses of the heart to changes in levels of particles in the air. Page 2 of 3 76 Exclusions: I understand that smoking will affect the results of this research, and therefore current smokers and/or subjects living with smokers at their homes must be excluded from this study. Risks and Discomfort: There are no risks involved with either the air pollution sampling, the lung function or electrocardiograph measurements. For Holter electrocardiograph measurements, the electrodes only detect electrical impulses produced by the heart. No electricity passes through the body, so there is no risk of electric shock. During the Holter monitoring I will not be able to bathe or shower. I understand that wearing the sampler and pump may sometimes be awkward, but technicians will work with me to find a comfortable configuration. I will be trained to remove and wear the sampler should the need arise. For example, when I am sleeping, the sampler will be placed at bedside. The sampling equipment weighs approximately 1.5 kg. Confidentiality: Any information resulting from this research study will be kept strictly confidential. Al l documents will be identified only by code number and kept in a locked filing cabinet. I will not be identified by name in any reports of the completed study. Remuneration: I understand that in order to defray the cost of inconvenience, at completion of the monitoring, I will receive an honorarium in the amount of $250. Contact: I understand that if I have any questions or desire further information with respect to this study, I should contact Dr. Brauer at 822-9585, Dr. Vedal at 875-4122, or Mr. Fisher / Ms Ebelt at 822-1274. If I have any concerns about my treatment or rights as a research subject I may contact the Director of Research Services at the University of British Columbia, Dr. Richard Spratley at 822-8598. For any immediate concerns relating to the measurements or equipment, I will receive a card at the time of measurement listing appropriate emergency contact numbers (cellular telephone) which can be called at any time. Patient Consent: I understand that participation in this study is entirely voluntary and that I may refuse to participate or I may withdraw from the study at any time without any consequences to my continuing medical care. I have received a copy of this consent form for my own records. I consent to participate in this study. Patient Signature Date Witness Signature Date Patient Name (please print) Investigator's Signature Date Page 3 of 3 APPENDIX C - FILTER WEIGHING F O R M COPD H E A L T H STUDY FILTER WEIGHING F O R M Form #8 Data Entered: Pre-Wei ;ht (mg) Post-Weight (mg) Psychrometer Electronic Psychrometer Electronic Temperature: Temperature: Humidity: Humidity: Date: Date: ^ame(s): ^ame(s): Lot #: Filter No. 1 2 3 Average 1 2 3 Average C1 C2 C3 ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-ST-APPENDIX D - P E R S O N A L FLOW LOG Person ID ST-ST-ST-ST-ST-ST- Filter No. Pump Sampler INO. Date ON Time ON Flow Check ON Actual Flow ON Date Time Count OFF Actual Flow Flow Check Notes o l-f 3 o> O n " i o m O Z ^ s 3- 9 n> n> a . 3 APPENDIX E - ACTIVITY DIARY 0 3 O ) CO ho CO o o o I o 3 CD TJ CD GO 01 c —% 03 3 zr CD CD 03 > 03 O o CO o X o 3 CD TJ CD w 53" c *n 03 3 CD —i CD 03 > 03 CO o o o o TJ CD GO ST c —5 03 3 Z CD 03 > 03 o o I CD CO O X o 3 CD TJ CD CO ST c —i 03 3 Z CD 03 > 03 CO o I CT3 O O X o 3 CD TJ CD W CD C 03 3 CD 03 > 03 o o CO o X o 3 CD TJ CD GO. fa c 03 CD CD 03 > 03 CO o o o X C3 3 CD TJ CD CO 5T c —» 03 3 o CD 03 > 03 o o CO o X o 3 CD TJ CD ST c —1 03 3 CD —\ ~z~ CD 03 03 >< CO o o o X o 3 CD TJ CD CO 03 C -1 03 3 CD 03 > 03 >< o o I CO CO o X o 3 CD TJ CD 5T c 3 " CD —% ~z~ CD 03 > 03 CO o o o X o 3 CD TJ CD co ST c o o I ro CO o X o 3 CD TJ CD 00 .—* 03 c —I 03 3 CD 03 > 03 CD 03 > 03 >< CO o I ro o o X o 3 CD TJ CD CO 03 C 03 3 Z CD 03 > 03 >< o o co o X o 3 CD TJ CD 1/3 03 c —\ 03 3 to CO o o o X o 3 CD TJ CD CO sr c —1 03 3 ro o o I I ro CO o X o 3 CD TJ CD CO oT c CD 0 0 CO o ro o o X o 3 CD TJ CD CO 5T c -1 03 3 CD 03 > 03 CD 03 > 03 Z CD 03 > 03 CD 03 > 03 O o CO o X o 3 CD TJ CD CO 03 c 03 3 3 " CD CD 03 > 03 CO o o o X o 3 CD TJ CD CO 5T c —i 03 3 zr CD "z" CD 03 > 03 o o CO o X o 3 CD TJ CD ST c —1 03 3 Z CD 03 > 03 >< CO o o o X o 3 CD TJ CD CO 03 c 03 3 o zr CD —\ ~z~ CD 03 > 03 o o I CO CO o X o TJ CD cn 03 c 03 zr CD CD 03 > 03 CO o o o X o 3 CD TJ CD oT c —1 03 3 CD 03 > 03 O D b o I 0 0 CO o X o 3 CD TJ CD CO 2" c — \ 03 3 Z CD 03 > 03 3 CD 3 Q. O o Ifl a: o 3 CD 3 -CD 5 c ? CD 9* O c 5 ^ O CD 3 Tl TJ =: CD CD 00 o w 3 03 a. o 03 o 03 CD c 00 Q3_ o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 o 03 00 c 00 Q3_ CD c 00 03_ 00 C 00 03_ 7? 00 c 00 Q3_ CD c 00 Q3_ 7? 00 C 00 Q3_ CD c 00 03. I T CD c 00 Q3_ 7? CD c 00 03_ CD c 00 Q3_ CD c 00 00 c 00 03_ 7T CD c 00 03_ CD c 00 03_ 7T 00 c 00 03_ 7T 00 c 00 Q3_ 7T 00 c 00 03_ CD c 00 $ 03. CD c 00 Q3_ CD c 00 03_ CD c 00 CD c 00 Q3_ CD £Z 00 CD C CO 0 ) 3? =V 3 " CD CD CD CD 3 T CD 3 " CD 3 " CD CD CD CD 3 " CD ZX CD zx CD zr CD CD 3 -CD Q. X Q. X C L X o C L X X C L X C L X C L X Q. X C L X C L X C L X C L X C L X C L X C L X C L X C L X C L X C L X QL X C L X C L X C L X 5 CD S -< CD 00 < CD 00 Z o -< CD 00 -< CD 00 -< CD 00 -< CD 00 -< CD 00 < CD 00 •< CD 00 •< CD 00 z o •< CD 00 - < CD 00 < CD 00 -< CD 00 Z o -< CD 00 Z o -< CD 00 < CD 00 < CD 00 - < CD 00 < CD 00 < CD 00 < CD 00 •< CD 00 Z o < CD 00 00 5 o O O 5' < CD 00 Z o -< CD 00 Z o CD 00 z o < CD 00 z o •< CD 00 z o -< CD 00 -< CD 00 Z o •< CD 00 -< CD 00 Z o -< CD 00 - < CD 00 < CD 00 Z o •< CD 00 Z o -< CD 00 < CD 00 < CD 00 Z o -< CD 00 Z o -< CD 00 " < CD 00 Z o -< CD 00 Z o -< CD 00 Z o -< CD 00 Z o •< CD 00 Z o •< CD 00 Z o 3 7 O 6 J CD g - < CD 00 CD 00 • < CD 00 -< CD 00 -< CD 00 • < CD 00 Z o < CD 00 < CD 00 - < CD 00 Z o -< CD 00 Z o •< CD 00 Z o •< CD 00 - < CD 00 Z o •< CD 00 < CD 00 < CD 00 -< CD 00 z o -< CD 00 < CD 00 CD 00 -< CD 00 - < CD 00 < CD 00 W Q) CD — CD - < CD 00 Z o -< CD 00 Z o < CD OO •< CD 00 < CD 00 z o -< CD 00 Z o -< CD 00 Z o < CD 00 Z o •< CD 00 < CD 00 z o < CD 00 •< CD 00 < CD 00 -< CD 00 Z o -< CD 00 Z o •< CD 00 < CD 00 Z o CD 00 • < CD 00 < CD 00 z o -< CD 00 -< CD 00 < CD 00 •< CD 00 ^ CO & CD z o CD 00 APPENDIX F - S Y M P T O M QUESTIONNAIRE 84 Data Entered: A I R P O L L U T I O N STUDY S Y M P T O M QUESTIONNAIRE S T A R T D A T E : ID: FDLTER: ST-B L O O D P R E S S U R E : Q I Q2 Q3 Q4 Pre: Systolic Diastolic Post: Systolic Diastolic Q5 Q6 Q7 Q8 1. C O U G H : Did you have M O R E / LESS / A B O U T T H E S A M E (NO) cough today compared to most days? 2. S P U T U M : Did you produce M O R E / L E S S / A B O U T T H E S A M E (NO) sputum today compared to most days? 3. DD7FICULTY B R E A T H I N G Did you experience M O R E / L E S S / A B O U T T H E S A M E (NO) difficulty breathing today compared to most days? 4 C H E S T PAIN: Did you experience M O R E / LESS / A B O U T T H E S A M E (NO) chest pain today compared to most days? 5. H E A R T B E A T : Were you aware of your heart beating rapidly or throbbing (palpitations) M O R E / LESS / A B O U T T H E S A M E (NOT A T A L L ) today compared to most days? 6. F A T I G U E : Did you feel M O R E / LESS / A B O U T T H E S A M E level of fatigue today compared to most days? 7 DIZZINESS: Did you experience M O R E / LESS / A B O U T T H E S A M E (NO) dizziness today compared to most days? 8. B R O N C H O D I L A T O R USE: A bronchodilator (inhaler) was used times today. 9. P U L S E O X I M E T E R R E A D I N G S : SpQ 2 Pulse SpO z Pulse Pre: 1 min Post: 1 min 3 min 3 min 5 min 5 min 10. NOTES: APPENDIX G - MEDICATION CHECKLIST 86 ID: MEDICATION C H E C K L I S T Data Entered: Actual Usage per Sampling Day Medication Dose per pill/puff Procfnnorl 1 2 3 4 5 6 7 Usage ST- ST- ST- ST- ST- ST- ST-APPENDIX H - AMBIENT FLOW LOG 89 APPENDIX I - KITSILANO AMBIENT CONCENTRATIONS During April 21 - September 24, 1998, Kitsilano ambient mass and sulfate concentrations were obtained using both a Dichotomous Sampler (DS) and Harvard Impactor (HI). Results of these measurements are outlined in Table 1.1, correlation coefficients for these exposure measures are summarized in Table 1.2, and plots illustrating the relationships between fine and PM2 . 5 mass, and fine and PM2 . 5 sulfate are shown below (Figures 1.1 and 1.2). Table 1.1: Kitsilano Ambient Concentrations Exposure Measure N Concentrations (p.g/m3) Mean SD Range Dichotomous Sampler -Coarse Mass 90 7.5 4.6 1.1-26.0 Dichotomous Sampler -Coarse Sulfate 23 0.3 0.3 0.1 - 1.7 Dichotomous Sampler -Fine Mass 90 7.8 4.2 2 .1-26.6 Dichotomous Sampler -Fine Sulfate 23 1.8 1.1 0 .5 -4 .7 Harvard Impactor -PM2 . 5 Mass 84 11.9 4.7 2.3-29.3 Harvard Impactor -PM2 . 5 Sulfate 84 1.9 1.0 0 .4-5 .3 Table 1.2: Correlation Coefficients for Kitsilano Mass and Sulfate Concentrations Exposure Measures DS Coarse Mass DS Fine Mass DS Coarse Sulfate DS Fine Sulfate HI PM2 . 5 Mass DS Coarse Mass 1.0000 DS Fine Mass 0.5782 1.0000 DS Coarse Sulfate 0.0058 0.2430 1.0000 DS Fine Sulfate 0.5100 0.8116 -0.1006 1.0000 HI PM2 . 5 Mass 0.5197 0.8131 0.0704 0.7506 1.0000 HI PM2 . 5 Sulfate 0.3466 0.7540 0.0940 0.9238 0.8095 90 91 APPENDIX J - FVC RESULTS O f the F V C values obtained during lung function testing that were selected for the final database, 186 (85.3%) were within 0.1L or 5% of the second largest F V C . Eighteen more F V C values (8.3%) were within 0.2L of the second largest values. Therefore, 93.6% of all F V C values were reproducible according to the A T S criteria. For the remaining 14 tests, the largest F V C results were at least 0.21L (5.5 to 25.1%) greater than the second largest F V C . Therefore, for these 14 cases, the second largest F V C maneuvers were selected for the final database. Three F V C values selected for the final database were from maneuvers that, according to the error message on the Tamarac Portable spirometer, were classified as having a "Bad stop... keep blowing longer." These were included because the A T S spirometry guidelines state that the "use of data from unacceptable maneuvers due to failure to meet the end-of-test requirements is left to the discretion of the interpreter" (ATS 1995). The fact that these test results were the largest values obtained from the respective testing sessions indicates that they were indeed the best maneuvers. Table J.I: Individual FVC Results ID N Mean M i n Max SD 1 14 2.59 2.06 2.90 0.216 2 14 2.62 2.00 2.84 0.199 3 14 4.29 3.21 4.76 0.430 4 14 2.37 2.04 2.60 0.127 5 13 3.35 2.71 3.90 0.385 6 14 2.26 1.77 2.46 0.158 7 14 2.88 2.50 3.15 0.173 8 14 1.86 1.40 2.08 0.199 9 12 2.14 1.70 2.55 0.303 10 14 2.97 2.50 3.46 0.339 11 14 1.96 1.80 2.24 0.104 12 14 1.62 1.46 1.94 0.121 13 14 4.18 3.90 4.46 0.159 14 14 2.21 2.06 2.46 0.110 15 14 1.52 1.41 1.63 0.074 16 9 1.88 1.18 2.12 0.280 Group 216 2.56 1.18 4.76 0.836 APPENDIX K - INDIVIDUAL LEAST S Q U A R E S R E G R E S S I O N R E S U L T S Table K.J: Model J - Personal PM2.s vs. Post-FEVi by Subject Subject N Slope Standard Error r2 p value 1 7 -0.000016 0.006009 0.00000 0.9979 2 7 0.000923 0.004297 0.00914 0.8385 3 7 -0.005296 0.018993 0.01531 0.7915 4 7 0.015286 0.004475 0.70006 0.0189 5 6 0.004315 0.016672 0.01647 0.8085 6 7 0.000968 0.001071 0.14046 0.4075 7 7 0.002351 0.002770 0.12598 0.4347 8 7 0.021489 0.010551 0.45342 0.0973 9 7 -0.000301 0.000906 0.02160 0.7532 10 7 -0.011413 0.013502 0.12503 0.4365 11 6 -0.000142 0.004093 0.00030 0.9739 12 6 0.002698 0.003763 0.11388 0.5130 13 7 -0.014745 0.007203 0.45597 0.0960 14 7 -0.000507 0.006444 0.00124 0.9403 15 6 0.000808 0.001930 0.04194 0.6971 16 3 0.012623 0.004441 0.88984 0.2154 Table K.2: Model 2 - Personal Sulfate vs. Post-FEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.004071 0.038423 0.00224 0.9197 2 7 -0.043476 0.028895 0.31167 0.1928 3 6 -0.164686 0.089340 0.45931 0.1391 4 7 0.037395 0.024284 0.32169 0.1842 5 6 0.014774 0.193368 0.00146 0.9428 6 7 -0.019946 0.043176 0.04094 0.6635 7 7 -0.086174 0.098510 0.13273 0.4217 8 7 0.073973 0.053489 0.27668 0.2252 9 7 -0.013669 0.013343 0.17348 0.3526 10 7 0.137630 0.086314 0.33709 0.1717 11 6 0.013735 0.016911 0.14157 0.4622 12 6 -0.051213 0.088819 0.07674 0.5951 13 7 -0.033730 0.029742 0.20460 0.3082 14 7 -0.008462 0.048050 0.00616 0.8671 15 6 -0.013374 0.029233 0.04973 0.6711 16 3 -0.128571 0.123718 0.51923 0.4878 Table K.3: Model 3 - Ambient PM2.s vs. Post-FEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.001163 0.008157 0.00405 0.8922 2 7 -0.002761 0.005193 0.05187 0.6233 3 6 -0.029386 0.017178 0.42250 0.1623 4 7 0.006001 0.008069 0.09962 0.4905 5 7 -0.028776 0.027878 0.17566 0.3493 6 7 -0.004295 0.006023 0.09233 0.5076 7 7 -0.000275 0.006435 0.00036 0.9676 8 7 0.012310 0.017219 0.09274 0.5066 9 7 -0.001774 0.002302 0.10619 0.4757 10 7 0.016998 0.015454 0.19483 0.3215 11 7 0.004464 0.004439 0.16823 0.3608 12 7 -0.012898 0.014434 0.13770 0.4125 13 7 -0.026493 0.006343 0.77722 0.0087 14 7 0.002767 0.005458 0.04890 0.6337 15 7 -0.004869 0.011390 0.03526 0.6868 16 4 -0.012383 0.008924 0.49053 0.2996 Table K.4: Model 4 - Ambient Sulfate vs. Post-FEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.000216 0.022738 0.00002 0.9928 2 7 -0.035993 0.020680 0.37728 0.1423 3 6 -0.128571 0.064776 0.49620 0.1181 4 7 0.026630 0.024750 0.18800 0.3311 5 7 -0.042254 0.189036 0.00989 0.8320 6 7 -0.021708 0.025547 0.12618 0.4343 7 7 0.012083 0.040529 0.01747 0.7776 8 7 0.053335 0.048307 0.19602 0.3198 9 7 -0.011256 0.009664 0.21342 0.2967 10 7 0.091344 0.054663 0.35835 0.1556 11 7 0.017901 0.014648 0.23000 0.2761 12 7 -0.049680 0.052176 0.15349 0.3847 13 7 -0.032012 0.033136 0.15730 0.3784 14 7 -0.000055 0.022336 0.00000 0.9981 15 7 0.004583 0.040919 0.00250 0.9152 16 4 -0.077434 0.049330 0.55197 0.2571 Table K.5: Model 5 - Personal PM2.s vs. AFEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.015728 0.008969 0.38081 0.1399 2 7 0.001799 0.006376 0.01568 0.7891 3 6 -0.019272 0.015558 0.27727 0.2832 4 7 0.008602 0.005159 0.35732 0.1563 5 6 0.007632 0.007454 0.20764 0.3638 6 7 0.000700 0.002550 0.01485 0.7947 7 7 -0.003735 0.002536 0.30259 0.2008 8 7 0.017340 0.013618 0.24488 0.2589 9 7 0.000514 0.000894 0.06205 0.5901 10 7 0.004262 0.021847 0.00755 0.8530 11 6 0.001307 0.004678 0.01913 0.7938 12 6 0.003359 0.005125 0.09698 0.5480 13 7 -0.011870 0.004118 0.62429 0.0345 14 7 -0.001372 0.011110 0.00304 0.9065 15 6 0.002290 0.001257 0.45338 0.1426 16 3 -0.024542 0.010486 0.84562 0.2571 Table K. 6: Model 6 - Personal Sulfate vs. AFEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.050885 0.069326 0.09727 0.4959 2 7 -0.001461 0.051851 0.00016 0.9786 3 6 -0.250169 0.143414 0.43205 0.1560 4 7 0.015344 0.022186 0.08731 0.5200 5 6 -0.094090 0.084130 0.23821 0.3260 6 7 0.026372 0.097383 0.01446 0.7674 7 7 0.013826 0.108256 0.00325 0.9034 8 7 0.017045 0.068638 0.01218 0.8137 9 7 -0.015942 0.012957 0.23241 0.2733 10 7 0.179556 0.139616 0.24857 0.2548 11 6 0.010106 0.020442 0.05758 0.6470 12 6 -0.025202 0.124078 0.01021 0.8490 13 7 -0.029155 0.018880 0.32292 0.1832 14 7 -0.050769 0.080015 0.07452 0.5537 15 6 0.040584 0.016030 0.61576 0.0645 16 3 0.328571 0.000000 1.00000 N / A Table K. 7: Model 7 - Ambient PM2.s vs. AFEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.027357 0.009523 0.62273 0.0349 2 7 -0.000146 0.007941 0.00007 0.9861 3 6 -0.053619 0.023125 0.57338 0.0813 4 7 0.007722 0.005739 0.26583 0.2363 5 6 -0.006504 0.017505 0.03336 0.7291 6 7 -0.016104 0.012083 0.26215 0.2401 7 7 -0.005720 0.006082 0.15034 0.3901 8 7 0.000556 0.019849 0.00016 0.9787 9 7 -0.002080 0.002270 0.14378 0.4015 10 7 0.014005 0.025405 0.05730 0.6052 11 7 0.002826 0.005365 0.05257 0.6209 12 7 -0.014354 0.020503 0.08928 0.5151 13 7 -0.014867 0.006426 0.51708 0.0686 14 7 0.010082 0.008540 0.21798 0.2909 15 7 0.009089 0.005600 0.34510 0.1655 16 3 0.060526 0.006989 0.98684 0.0732 Table K.8: Model 8 - Ambient Sulfate vs. AFEVi by Subject Subject N Slope Standard Error r2 p value 1 7 0.031602 0.040749 0.10737 0.4731 2 7 -0.009515 0.038786 0.01189 0.8160 3 6 -0.212229 0.095766 0.55112 0.0910 4 7 0.013772 0.020738 0.08105 0.5360 5 6 -0.100828 0.099133 0.20548 0.3666 6 7 0.005453 0.060758 0.00161 0.9320 7 7 -0.009451 0.041699 0.01017 0.8297 8 7 0.000545 0.059158 0.00002 0.9930 9 7 -0.012491 0.009452 0.25885 0.2436 10 7 0.103223 0.092833 0.19825 0.3168 11 7 0.013066 0.017980 0.09552 0.5000 12 7 -0.043813 0.075895 0.06249 0.5888 13 7 -0.025817 0.021987 0.21614 0.2932 14 7 0.002753 0.038525 0.00102 0.9458 15 7 0.021667 0.022444 0.15710 0.3787 16 3 0.302632 0.034945 0.98684 0.0732 96 APPENDIX L - CORRELATION COEFFICIENTS FOR SAMPLING SESSIONS AND INTERVALS E a c h d iagonal ( f rom top-left to bottom-right) in the fo l l ow ing 8 tables indicates the correlat ion for al l sampl ing completed N sessions apart (i.e. all sampl ing 1 session apart, 2 sessions apart, etc.). Table L. 1: Model 1 - Correlation Coefficients for Sampling Sessions: Personal PM2.5 vs. Post-FEVi Number of Sampling Session 1 2 3 4 5 6 2 -0 13K) 3 -0 .5169 -n 7183 4 0 0270 -0.2064 0 4400 5 0.1527 -0 5472 0.0669 -0 537o 6 -0 7139 0.0203 0 1038 -0.2283 -0 015O 7 -0.0671 -0 0038 -0.2070 -0 4228 0.2441 1 -0 2082 Table L. 2: Model 2 - Correlation Coefficients for Sampling Sessions: Personal Sulfate vs. Post-FEVi Number of Sampling Session 1 2 3 4 5 6 2 -0 1 132 3 -0.5081 -0 7417 4 0 0123 -0.2293 0 4358 5 0.1866 -0 5261 0.0880 -0 5371 6 -0 6915 0.0013 0 2284 -0.2247 -OOiOR 7 -0.0924 -0 0221 -0.1988 -0 4979 0.2484 -0 2002 Table L. 3: Model 3 - Correlation Coefficients for Sampling Sessions: Ambient PM2.5 vs. Post-FEVi Number of Sampling 1 2 3 4 5 6 Session 2 -0 0512 3 -0 .5059 -0 7 4 0 0304 -0.2348 0 3 0 o o 5 0.1724 -0 4X4"? 0.0444 -0 551 1 6 -0 oo82 0.0025 0 1600 -0.2465 -0 0410 7 -0.0735 -0 0702 -0.2252 -0 47h4 0.1786 -0 2117 Table L. 4: Model 4 - Correlation Coefficients for Sampling Sessions: Ambient Sulfate vs. Post-FEVj Number of Sampling 1 2 3 4 5 6 Session 2 -o i>>84 3 -0.5191 -0 727« 4 <) 0083 -0.2173 0 3928 5 0.1878 -0 5257 0.0773 -0 5413 6 -0 67X7 0.0043 0 10^6 -0.2198 -0 0SS8 7 -0.0939 -0 0160 -0.2248 -0 4484 0.1555 -0 1926 Table L.5: Model 5 - Correlation Coefficients for Sampling Sessions: Personal PM2.5 vs. AFEVi Number of Sampling 1 2 3 4 5 6 Session 2 -0 0703 3 0.3517 -0 1560 4 0 4 1 15 -0.3968 0 o 6 s : 5 0.2719 -0 52SO 0.5070 0 5103 6 0 M 2 4 0.4700 0 7451 0.4838 0 1287 7 -0.0371 0 33XS -0.5740 -0 5286 -0.7558 -0 008S Table L 6: Model 6 - Correlation Coefficients for Sampling Sessions: Personal Sulfate vs. AFEVi Number of Sampling Session 1 2 3 4 5 6 2 -0 027S 3 0.1791 -0 2W5 4 0 3913 -0.3988 D 0 8 0 0 5 0.2599 -0 5207 0.5079 0 4247 6 0 M)4-l 0.4180 0 0971 0.5152 0 | 079 7 -0.0076 0 288^ -0.6315 -0 5742 -0.7411 -0 0837 Table L. 7: Model 7 - Correlation Coefficients for Sampling Sessions: Ambient PM2.5 vs. AFEVi Number of Sampling 1 2 3 4 5 6 Session 2 -0 02H5 3 0.2810 -o 1075 4 0 3724 -0.4017 o 7347 5 0.2114 -0 ? u | 0 0.4653 0 4220 6 0 5032 0.4623 0 6084 0.4237 0.1 U70 7 0.0845 o -0.4011 -0 4822 -0.6880 0 0702 Table L.8: Model 8 - Correlation Coefficients for Sampling Sessions: Ambient Sulfate vs. AFEVi Number of Sampling 1 2 3 4 5 6 Session 2 -•>05<)1 3 0.3244 -0.1005 4 o 1053 -0.3879 0 6875 5 0.2665 -0 5054 0.5063 0 4314 6 0 504S 0.4700 0 6208 0.4097 0 0846 7 0.0123 0 . 3 4 ^ -0.3834 -0 4^0] -0.7106 O 0 8 I 6 Table L.9: Correlation Coefficients for Sampling Intervals: Model Sampling Intervals (Number of Samplin g Sessions Apart) 1 2 3 4 5 6 1 -0.2396 -0.0917 -0.2687 0.0936 -0.3221 -0.0671 2 -0.2462 -0.0910 -0.2731 0.1005 -0.3266 -0.0924 3 -0.2354 -0.1077 -0.2613 0.1039 -0.3076 -0.0735 4 -0.2332 -0.1047 -0.2702 0.1030 -0.3086 -0.0939 5 0.1491 -0.0518 -0.0709 0.2827 0.3992 -0.0371 6 0.1385 -0.0572 -0.0879 0.2443 0.3704 -0.0076 7 0.1662 -0.0586 -0.0727 0.2782 0.3893 0.0845 8 0.1549 -0.0564 -0.0746 0.2997 0.3902 0.0123 99 APPENDIX M - POOLED ORDINARY LEAST S Q U A R E S R E G R E S S I O N A N A L Y S E S Figure M.l: Model 1 - Pooled Personal PM2.5 vs. Post-FEVi .3-100 Personal PM2.5 (ug/m3) Figure M.2: Model 2 - Pooled Personal Sulfate vs. Post-FEVi .3-> LU v> O CL CO CD E 2 o ro > cu Q Personal Sulfate (ug/m3) 100 Figure M.3: Model 3 - Pooled Ambient PM2.s vs. Post-FEVi . 3 . .2 • . > UJ <=> - 3 j . . 1 0 10 20 30 Mean Ambient PM2.5 (ug/m3) Figure M.4: Model 4 - Pooled Ambient Sulfate vs. Post-FEVi . 3 -.2 4 > UJ '5 CD Q -.3 0 1 2 3 4 5 6 Mean Ambient Sulfate (ug/m3) 101 Figure M. 5: Model 5 - Pooled Personal PM2.s vs. AFEVi . 6 -100 Personal PM2.5 (ug/m3) Figure M. 6: Model 6 - Pooled Personal Sulfate vs. AFEVi . 6 -Personal Sulfate (ug/m3) 102 Figure M. 7: Model 7 - Pooled Ambient PM2j vs. AFEVi .6 • > LU CU co c co O -.4 0.0 4 Mean Ambient Sulfate (ug/m3) Figure M.8: Model 8 - Pooled Ambient Sulfate vs. AFEVi Mean Ambient PM2.5 (ug/m3) 103 APPENDIX N - ESTIMATED VARIANCES FOR P O O L E D WEIGHTED LEAST S Q U A R E S R E G R E S S I O N A N A L Y S E S Table N.l: Estimated Variances by Subject Subject Model 1 2 3 4 5 6 7 8 1 0.00171 0.00172 0.00176 0.00172 0.00615 0.00672 0.00707 0.00632 2 0.00237 0.00226 0.00226 0.00219 0.00531 0.00556 0.00553 0.00528 3 0.02448 0.02386 0.02300 0.02361 0.06188 0.05784 0.05771 0.06065 4 0.00247 0.00268 0.00272 0.00268 0.00418 0.00432 0.00441 0.00433 5 Excluded Excluded Excluded Excluded 0.02040 0.01850 0.01995 0.02006 6 0.00284 0.00293 0.00277 0.00284 0.01492 0.01543 0.01369 0.01500 7 0.00588 0.00598 0.00614 0.00612 0.00762 0.00810 0.00747 0.00782 8 0.01829 0.01895 0.01907 0.01901 0.04139 0.04100 0.04248 0.04135 9 0.00144 0.00124 0.00122 0.00117 0.00168 0.00108 0.00126 0.00140 10 0.01557 0.01575 0.01625 0.01596 0.06144 0.06180 0.06174 0.06243 11 0.00204 0.00221 0.00229 0.00215 0.00328 0.00388 0.00370 0.00351 12 0.01381 0.01396 0.01183 0.01204 0.02648 0.02647 0.02290 0.02341 13 0.00797 0.00753 0.00697 0.00755 0.00434 0.00311 0.00339 0.00405 14 0.00191 0.00185 0.00204 0.00188 0.00642 0.00552 0.00733 0.00650 15 0.00171 0.00164 0.00285 0.00291 0.00230 0.00346 0.00367 0.00302 16 0.00353 0.00346 0.00264 0.00291 0.01333 0.01383 0.01395 0.01352 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0089170/manifest

Comment

Related Items