Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Particulate air pollution and chronic obstructive pulmonary disease patients: an assessment of exposure… Ebelt, Stefanie 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_2000-0051.pdf [ 10.64MB ]
Metadata
JSON: 831-1.0089356.json
JSON-LD: 831-1.0089356-ld.json
RDF/XML (Pretty): 831-1.0089356-rdf.xml
RDF/JSON: 831-1.0089356-rdf.json
Turtle: 831-1.0089356-turtle.txt
N-Triples: 831-1.0089356-rdf-ntriples.txt
Original Record: 831-1.0089356-source.json
Full Text
831-1.0089356-fulltext.txt
Citation
831-1.0089356.ris

Full Text

PARTICULATE AIR POLLUTION AND CHRONIC OBSTRUCTIVE PULMONARY DISEASE PATIENTS: AN ASSESSMENT OF EXPOSURE AND CARDIOVASCULAR H E A L T H EFFECTS By Stefanie Ebelt B.Sc, The University of British Columbia, 1997 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIRMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Occupational Hygiene Program) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA October 1999 © Stefanie Tania Ebelt, 1999 In presenting this thesis in partial fulfilment 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. 1 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 QccupochonoJ rH c^y ene_ The University of British Columbia Vancouver, Canada Date r % b r u o r ^ 1*4- , d?OCQ DE-6 (2/88) 11 ABSTRACT Epidemiologic studies have repeatedly demonstrated associations between particulate air pollution and adverse health effects. One concern in time-series studies is the assessment of exposure of the study population using fixed site outdoor measurements. To address the issue of exposure misclassification, we evaluate the relationship between ambient and personal particulate concentrations of a population expected to be at risk of particle health effects. Biologically plausible mechanisms of particle health effects are also lacking; thus, we evaluate several cardiovascular outcomes of our population. Sampling was conducted within the Vancouver metropolitan area during April-September 1998. Sixteen subjects (non-smoking, ages 54-86) with physician-diagnosed COPD wore personal PM2.5 monitors for seven, randomly spaced, 24-hour periods. Time-activity logs, dwelling characteristics data, blood pressure (BP) and 24-hour ambulatory ECG recordings were obtained for each subject. Daily 24-hour ambient PM10 and PM2.5 concentrations were measured at five fixed sites spaced throughout the study region. Sulfate, a marker of ambient combustion-source particulate, was measured in all PM2.5 samples. Regression analyses were conducted to assess the relationship between personal and ambient levels. Ambient concentrations were expressed either as an average of the five values obtained for each day of personal sampling, or the concentration obtained at the site closest to each subject's home. The median Pearson's r of individual regressions between personal and average ambient PM2.5 concentrations was 0.48 (range: -0.68 to 0.83). Using sulfate as the exposure metric, the median correlation was 0.96 (range: 0.66 to 1.00). The mean personal to ambient concentration ratio of all samples was 1.75 for PM2.5 and 0.75 for sulfate. Use of the closest ambient site did not improve the median correlation of the group for either exposure variable. Inclusion of time-activity and dwelling characteristics data in a regression model for PM2.5 exposure improved model fit, but was not highly predictive (R2: 0.27). The model for sulfate was predictive (R2: 0.82) as personal exposures were largely explained by ambient levels. BP, supraventricular ectopic beats (SVE), heart rate (HR) and heart rate variability (HRV) indices, were regressed against exposure. Temperature, relative humidity, carbon monoxide, ozone and bronchodilator use were tested for confounding. Decreases in BP and increases in SVEs were observed with increasing exposure. HR and HRV models produced inconsistent results and were unstable upon the addition of secondary variables. These results indicate a relatively low degree of correlation between personal and ambient concentrations for PM2.5 compared with a high correlation when using sulfate as a marker of outdoor combustion-source particulate. These data also suggest that BP and SVE are sensitive cardiovascular indicators, however the implications of our findings remain to be assessed. iii TABLE OF CONTENTS ABSTRACT ii TABLE OF CONTENTS iii LIST OF TABLES viii LIST OF FIGURES xi ACKNOWLEDGEMENTS xv CHAPTER ^INTRODUCTION 1 1.1 Particulate Air Pollution 1 1.1.1 Origin and characteristics of particles 1 1.1.2 Particle measurement and standards 2 1.1.3 Typical ambient concentrations 3 1.2 Health Effects of Particulates 4 1.2.1 Epidemiology 4 1.2.1.1 Mortality 4 1.2.1.2 Morbidity 5 1.2.1.3 Populations at increased risk 5 1.2.2 Inhalation of particles and deposition in respiratory tract 6 1.2.3 Uncertainties in associations between particulates and health 6 1.3 Part I - Exposure Assessment 8 1.3.1 Ambient measurements can cause exposure misclassification 8 1.3.2 Efforts to reduce exposure misclassification 9 1.3.3 Target population 10 1.4 Part II - Assessment of Cardiovascular Health Effects 11 1.4.1 The question of biological plausibility 11 1.4.2 Possible biological mechanisms 12 1.4.2.1 . Inflammation 12 1.4.2.2 Blood coagulation 13 iv 1.4.3 Measures to assess cardiovascular health 13 1.4.3.1 Blood pressure 14 1.4.3.2 Cardiac arrhythmias 15 1.4.3.3 Heart rate and heart rate variability 16 1.5 Study Design 19 1.5.1 Hypotheses 19 1.5.2 Objectives 19 CHAPTER 2-.METHODS 20 2.1 Overview 20 2.2 Pre-Study Sampler Experiments 20 2.3 Ambient Concentration Measurements 21 2.3.1 Ambient site locations and sampling equipment 21 2.3.2 Lab preparation 22 2.3.3 Field sample collection 22 2.4 Personal Exposure and Health Effects Measurements 23 2.4.1 Study population - eligibility and recruitment 23 2.4.2 Subject identification and sampling schedules 24 2.4.3 Personal sampling equipment and forms 25 2.4.3.1 Particulate sampling 25 2.4.3.2 Lab preparation of particulate samplers 26 2.4.3.3 Information about particulate sources 26 2.4.3.4 Electrocardiograms 27 2.4.3.5 Blood Pressure Measurements 27 2.4.3.6 Medication information 27 2.4.3.7 Other 27 2.4.4 Field sample collection 28 2.5 Gravimetric Analysis 29 2.5.1 Filter weighing procedures 29 2.5.2 Quality Control filters 29 2.5.3 Lab blanks and field blanks 29 2.6 Sulfate Analysis 30 2.7 Data Analysis 30 2.7.1 Particulate sampling - data quality and descriptive statistics 30 2.7.2 Relationship between personal exposures and ambient concentrations 31 2.7.3 Predictors of personal exposure 31 2.7.4 Assessment of cardiovascular health effects 33 CHAPTER 3: RESULTS 37 3.1 Study Population 37 3.1.1 Recruitment 37 3.1.2 Characteristics 37 3.1.3 Compliance 38 3.2 Particulate Sampling - Data Quality 38 3.2.1 Data clean-up .38 3.2.2 Quality control, lab blanks, field blanks and limits of detection 39 3.3 PM2.5 and Sulfate Concentration, Time-Activity and Dwelling Data 40 3.3.1 Personal exposures 40 3.3.2 Ambient Concentrations 43 3.3.3 Time-activity information 47 3.3.4 Dwelling characteristics 47 3.4 Part I - Relationship Between Personal Exposures and Ambient Concentrations ..48 3.4.1 Correlation between personal and ambient concentrations 48 3.4.2 Ratios and differences 52 3.4.3 Is the difference between personal and ambient concentrations dependent on level of personal exposure? 55 3.4.4 Are correlations between personal and ambient measures dependent on level of personal exposure or ambient concentration? 58 3.4.5 Use of ambient data from sites closest to each subject 61 3.4.5.1 Comparing sets of ambient data 61 3.4.5.2 Are P vs. A correlations a function of living distance from closest site? 62 3.4.6 Predictors of Personal Exposure 64 3.4.6.1 PM2.5 multiple regression 64 3.4.6.2 Sulfate multiple regression 67 3.5 Part II - Assessment of Cardiovascular Health Effects in Relation to Exposure 68 3.5.1 Descriptive statistics for six cardiovascular health indicators 68 3.5.2 Regressions against pollution for each variable 78 3.5.2.1 Systolic blood pressure 80 3.5.2.2 Diastolic blood pressure 83 3.5.2.3 Supraventricular ectopy 85 3.5.2.4 Heart rate 88 3.5.2.5 SDNN 90 3.5.2.6 R-MSSD 93 3.5.3 Summary of relationships 95 CHAPTER 4:DISCUSSION 96 4.1 Study Population 96 4.1.1 Participation... 96 4.1.2 Characteristics 97 4.2 Data Quality 100 4.2.1 Co-located sampler experiments 100 4.2.2 Study data 100 4.2.3 Particulate Concentrations 102 4.3 Part I - Relationship between Personal and Ambient Concentrations 102 4.3.1 Correlations 102 4.3.2 Ratios and differences 104 vii 4.3.3 Use of closest site data 106 4.3.4 Predictors of Personal Exposure 107 4.3.4.1 PM2.5 exposure 107 4.3.4.2 Sulfate exposure 109 4.3.5 Conclusions and implications of the exposure assessment 110 4.3.6 Limitations and recommendations for future exposure assessments Ill 4.4 Part II - Assessment of Cardiovascular Health Effects in Relation to Exposure ..112 4.4.1 Hypothesis 112 4.4.2 Blood pressure 112 4.4.3 Cardiac arrhythmia 113 4.4.4 Heart rate 115 4.4.5 Heart rate variability 116 4.4.6 Comparing effect estimates of exposure metrics 117 4.4.7 Conclusions and implications of cardiovascular health effects 118 4.4.8 Limitations and recommendations for future health studies 119 REFERENCES 120 APPENDIX 1: PRE-STUDY SAMPLER EXPERIMENTS 134 APPENDIX 2: AMBD2NT FLOW L O G SHEET 136 APPENDIX 3: STUDY RECRUITMENT LETTER 137 APPENDIX 4: OFFICIAL INFORMED CONSENT FORM 138 APPENDIX 5: TIME-ACTIVITY L O G 141 APPENDIX 6: DWELLING QUESTIONNAIRE 143 APPENDIX 7: MEDICATION FORM 145 APPENDIX 8: SYMPTOMS QUESTIONNAffiE 146 APPENDIX 9: PERSONAL FLOW L O G SHEET 147 APPENDIX 10: QUALITY CONTROL CHARTS 148 viii LIST OF TABLES TABLE 2.1. STUDY MEASUREMENTS 20 TABLE 2.2. VARIABLES CONSIDERED FOR INCLUSION IN THE MULTIPLE REGRESSION MODELS 33 TABLE 2.3. POTENTIAL BLOOD PRESSURE VARIABLES FOR ANALYSIS 36 TABLE 2.4. POTENTIAL ECG VARIABLES FOR ANALYSIS 36 TABLE 3.1. WARNING AND CONTROL LIMITS FOR QUALITY CONTROL FILTERS 39 TABLE 3.2. PERSONAL PM 2 5 EXPOSURE SUMMARY 41 TABLE 3.3. PERSONAL SULFATE EXPOSURE SUMMARY 41 TABLE 3.4. KITSILANO SITE AMBIENT CONCENTRATIONS 43 TABLE 3.5. AMBIENT PM 1 0 CONCENTRATION SUMMARY 44 TABLE 3.6. AMBIENT PM 2 5 CONCENTRATION SUMMARY 44 TABLE 3.7. AMBIENT SULFATE CONCENTRATION SUMMARY 44 TABLE 3.8. TIME-ACTIVITY SUMMARY 46 TABLE 3.9. CORRELATION COEFFICIENTS FOR ALL RELATIONSHIPS BETWEEN PERSONAL AND AMBIENT PARAMETERS 49 TABLE 3.10. INDIVIDUAL CORRELATION COEFFICIENTS RELATING PERSONAL AND AMBIENT CONCENTRATIONS 51 TABLE 3.11. INDIVIDUAL MEAN RATIOS RELATING PERSONAL AND AMBIENT CONCENTRATIONS 53 TABLE 3.12. AVERAGE DIFFERENCES BETWEEN PERSONAL AND AMBIENT CONCENTRATIONS 55 TABLE 3.13. CORRELATION COEFFICIENTS BETWEEN AVERAGE AMBIENT CONCENTRATIONS AND CONCENTRATIONS FROM EACH SITE 61 TABLE 3.14. CORRELATION COEFFICIENTS RELATING CONCENTRATIONS BETWEEN INDIVIDUAL AMBIENT SITES 62 TABLE 3.15. COMPARISON OF MEDIAN PEARSON'S R-VALUES FOR PERSONAL VS. AMBIENT RELATIONSHIPS USING AVERAGE, CLOSEST OR INDIVIDUAL SITE AMBIENT DATA FOR PM 2 5 62 TABLE 3.16. DISTANCE OF RESIDENCE FROM CLOSEST AMBIENT SITE PER SUBJECT 63 TABLE 3.17. OLS REGRESSION FOR PM 2 5 65 TABLE 3.18. WLS REGRESSION FOR PM 2 5 66 TABLE 3.19. OLS REGRESSION FOR SULFATE 67 TABLE 3.20. FEASIBILITY TEST RESULTS FOR MANUAL CALCULATION OF HRV VARIABLES 69 TABLE 3.21. BLOOD PRESSURE SUMMARY 69 TABLE 3.22. SUPRAVENTRICULAR ECTOPY SUMMARY 72 TABLE 3.23. HEART RATE SUMMARY 73 TABLE 3.24. SDNN SUMMARY 75 TABLE 3.25. R-MSSD SUMMARY 76 TABLE 3.26. INTERQUARTILE RANGES FOR EACH EXPOSURE METRIC 80 TABLE 3.27. CORRELATION COEFFICIENTS BETWEEN POTENTIAL CONFOUNDERS AND EXPOSURE METRICS 80 TABLE 3.28. SYSTOLIC BLOOD PRESSURE REGRESSIONS 82 TABLE 3.29. DIASTOLIC BLOOD PRESSURE REGRESSIONS 84 TABLE 3.30. LN-TRANSFORMED SUPRAVENTRICULAR ECTOPY REGRESSIONS ...87 TABLE 3.31. HEART RATE REGRESSIONS 89 TABLE 3.32. SDNN REGRESSSIONS 92 TABLE 3.33. R-MSSD REGRESSIONS 94 TABLE 4.1. PERCENTAGES OF TIME SPENT IN VARIOUS LOCATIONS/ACTIVITIES OVER A 24-HOUR PERIOD 99 TABLE A. 1. P M 2 5 CONCENTRATIONS FROM EXPERIMENTS 1.1 AND 1.2 135 TABLE A.2. PM 2 5 CONCENTRATIONS FROM EXPERIMENTS 2.1-2.4 135 LIST OF FIGURES FIGURE 1.1. T H E ELECTROCARDIOGRAM 15 FIGURE 2.1. MAP OF STUDY REGION 21 FIGURE 2.2. STUDY SUBJECT WEARING SAMPLER AND PUMP 25 FIGURE 3.1. PERSONAL P M 2 . 5 EXPOSURES 42 FIGURE 3.2. PERSONAL SULFATE EXPOSURES 42 FIGURE 3.3. AMBIENT PMio CONCENTRATIONS PER SITE AND A V E R A G E D OVER A L L SITES 45 FIGURE 3.4. AMBIENT P M 2 5 CONCENTRATION PER SITE AND A V E R A G E D OVER A L L SITES 45 FIGURE 3.5. AMBIENT SULFATE CONCENTRATIONS PER SITE AND A V E R A G E D OVER A L L SITES 46 FIGURE 3.6. MEDIAN OF INDIVIDUAL REGRESSIONS B E T W E E N AMBIENT PARAMETERS AND PERSONAL EXPOSURES 49 FIGURE 3.7. T H E DIFFERENCE BETWEEN POOLING D A T A AND USE OF INDIVIDUAL RESULTS WHEN REGRESSING PERSONAL AGAINST AMBIENT 50 FIGURE 3.8. CORRELATION BETWEEN PERSONAL AND AMBIENT P M 2 5 OVER TIME 50 FIGURE 3.9. CORRELATION BETWEEN PERSONAL AND AMBIENT SULFATE O V E R T I M E 51 FIGURE 3.10. DIFFERENCE BETWEEN A V E R A G E AND CLOSEST AMBIENT D A T A IN PERSONAL VS. AMBIENT REGRESSIONS 52 xii FIGURE 3.11. RATIOS BETWEEN PERSONAL AND AVERAGE AMBIENT PM2.5 PER SUBJECT 54 FIGURE 3.12. RATIOS BETWEEN PERSONAL AND AVERAGE AMBIENT SULFATE PER SUBJECT 54 FIGURE 3.13. PERSONAL-AMBIENT VS. PERSONALPM25 56 FIGURE 3.14. PERSONAL-AMBIENT VS. PERSONAL SULFATE 56 FIGURE 3.15. PERSONAL-AMBIENT VS. AMBIENTPM25 57 FIGURE 3.16. PERSONAL-AMBIENT VS. AMBIENT SULFATE 57 FIGURE 3.17. DEPENDENCY OF P VS. A CORRELATIONS FORPM 2 5 ON PERSONAL EXPOSURE 59 FIGURE 3.18. DEPENDENCY OF P VS. A CORRELATIONS FOR SULFATE ON PERSONAL EXPOSURE 59 FIGURE 3.19. DEPENDENCY OF P VS. A CORRELATIONS FORPM 2 5 ON AMBIENT CONCENTRATION 60 FIGURE 3.20. DEPENDENCY OF P VS. A CORRELATIONS FOR SULFATE ON AMBIENT CONCENTRATION 60 FIGURE 3.21. INDIVIDUAL PM 2 5 CORRELATIONS AS A FUNCTION OF DISTANCE FROM CLOSEST SITE 63 FIGURE 3.22. RESIDUAL PLOT OF OLS REGRESSION FORPM 2 5 65 FIGURE 3.23. RESIDUAL PLOT OF WLS REGRESSION FOR PM2 5 66 FIGURE 3.24. RESIDUAL PLOT OF OLS REGRESSION FOR SULFATE 67 FIGURE 3.25. SYSTOLIC BLOOD PRESSURE MEASUREMENTS PER SUBJECT 70 FIGURE 3.26. STANDARDIZED SYSTOLIC BLOOD PRESSURE MEASUREMENTS PER SUBJECT 70 FIGURE 3.27. DIASTOLIC BLOOD PRESSURE MEASUREMENTS PER SUBJECT 71 FIGURE 3.28. STANDARDIZED DIASTOLIC BLOOD PRESSURE MEASUREMENTS PER SUBJECT 71 FIGURE 3.29. LN-TRANSFORMED SUPRAVENTRICULAR ECTOPY VALUES PER SUBJECT 72 FIGURE 3.30. STANDARDIZED LN-TRANSFORMED SUPRAVENTRICULAR ECTOPY VALUES PER SUBJECT 73 FIGURE 3.31. HEART RATE VALUES PER SUBJECT 74 FIGURE 3.32. STANDARDIZED HEART RATE VALUES PER SUBJECT 74 FIGURE 3.33. SDNN VALUES PER SUBJECT 75 FIGURE 3.34. STANDARDIZED SDNN VALUES PER SUBJECT 76 FIGURE 3.35. R-MSSD VALUES PER SUBJECT 77 FIGURE 3.36. STANDARDIZED R-MSSD VALUES PER SUBJECT 77 FIGURE 3.37. OLS REGRESSION BETWEEN SYSTOLIC BLOOD PRESSURE AND AMBIENT PMio 81 FIGURE 3.38. SYSTOLIC BP EFFECT ESTIMATES AND SE FOR INTERQUARTILE RANGE INCREASES IN EXPOSURE 81 FIGURE 3.39. OLS REGRESSION BETWEEN DIASTOLIC BLOOD PRESSURE AND AMBIENT PMio 83 FIGURE 3.40. DIASTOLIC BP EFFECT ESTIMATES AND SE FOR INTERQUARTILE RANGE INCREASES IN EXPOSURE 85 xiv FIGURE 3.41. OLS REGRESSION BETWEEN L N SVE AND AMBIENT P M 1 0 86 FIGURE 3.42. L N SVE EFFECT ESTIMATES AND SE FOR INTERQUARTILE RANGE INCREASES IN EXPOSURE 86 FIGURE 3.43. OLS REGRESSION BETWEEN HEART R A T E AND AMBIENT P M 1 0 88 FIGURE 3.44. HEART R A T E EFFECT ESTIMATES AND SE FOR INTERQUARTILE R A N G E INCREASES IN EXPOSURE 90 FIGURE 3.45. OLS REGRESSION BETWEEN SDNN AND AMBIENT PMio 91 FIGURE 3.46. SDNN EFFECT ESTIMATES AND SE FOR INTERQUARTILE R A N G E INCREASES IN EXPOSURE 91 FIGURE 3.47. OLS REGRESSION BETWEEN R-MSSD AND AMBIENT PMio 93 FIGURE 3.48. R-MSSD EFFECT ESTIMATES AND SE FOR INTERQUARTILE R A N G E INCREASES IN EXPOSURE 95 FIGURE 4.1. TIME-ACTIVITY CHARACTERISTICS OF COPD STUDY POPULATION 98 FIGURE 4.2. TIME-ACTIVITY CHARACTERISTICS OF RETIRED 65+ REFERENCE POPULATION 98 FIGURE A. 1. AMBIENT FILTER QUALITY CONTROL CHART 1 148 FIGURE A.2. AMBIENT FILTER QUALITY CONTROL CHART 2 148 FIGURE A 3 . AMBIENT FILTER QUALITY CONTROL CHART 3 149 FIGURE A.4. PERSONAL FILTER QUALITY CONTROL CHART 1 149 FIGURE A.5. PERSONAL FILTER QUALITY CONTROL CHART 2 150 FIGURE A.6. PERSONAL FILTER QUALITY CONTROL CHART 3 150 XV ACKNOWLEDGEMENTS This study would not have been possible without the contribution of many dedicated individuals, whom I would like to recognize and gratefully thank. First, many thanks go to our study subjects who graciously donated their time and energy over our five-month sampling period. Only with their consistent devotion and cooperation were we able to collect enough data to reach our goal of assessing personal exposures and health outcomes. The large help we received from the Greater Vancouver Regional District Air Quality Department and staff with the ambient sampling is greatly appreciated. Not only did they grant us access to their ambient sampling shelters, but also helped in the data collection. Thanks also to the BC Lung Association for granting the funds for this research. An enormous amount of effort was required in lab preparation, collecting data and data analysis. No small amount of thanks go to Parveen Bhatti, who helped in the lab and field work, and Jochen Brumm, for his guidance in the statistical analysis. Much appreciation also goes to my colleague Teri Fisher, who was an inseparable partner in all aspects of the study and irreplaceable at every step of the way. I would like to thank my committee members, Dr. Sverre Vedal and Dr. John Petkau, for their continuing help and expert advice throughout the project. The utmost appreciation goes to my supervisor, Dr. Michael Brauer, whose tireless willingness to teach, everlasting patience to coach and incredible ability to inspire made everything in the last two years possible. He provided not only enough fuel for me to complete the work presented here, but motivated me to continue my exploration in the field of occupational and environmental hygiene. Finally, I would like to thank my parents, family and friends, whether beside me in person or in spirit, for their support and their everlasting faith. 1 CHAPTER 1: INTRODUCTION Several well-referenced episodes of extreme air pollution occurred earlier this century, which brought attention to the potential health risks of air pollution. In December, 1930, air pollutants accumulated in the Meuse Valley of Belgium, notably increasing mortality and morbidity among humans and animals (Firket, 1931). Donora, Pennsylvania experienced an episode in October, 1948, during which half the population became sick, with 10% becoming severely ill (Shrenk et al., 1949). The most severe episode of the century was the London Fog of 1952 during which air pollution caused approximately 4000 deaths (Logan, 1953). During the 1930s, 40s, and 50s, coal was the major fuel for industrial and domestic use, leading to high concentrations of particulates, sulfur dioxide and other pollutants in the ambient air. The episodes above were attributed to temperature inversions, which trapped air pollutants near to the ground, rather than letting them disperse. In the years following the major episodes, efforts were made to reduce air pollution levels. For example, many of the particle emitting stacks in London were subsequently built taller, to send the emissions above the inversion layer. In 1956, the UK Clean Air Act led to reduction in urban pollution and drastic decrease in the frequency of winter smogs in this region (Wilson, 1996). The causes for excess morbidity and deaths during the heavy winter smogs were widely accepted as being due to air pollution. Over the last three decades, epidemiologic investigations have repeatedly quantified associations between adverse health effects and particulate air pollution. Moreover, recent studies implicate particulates as contributing to morbidity and mortality at concentrations common to urban areas of the developed world, much lower than the extreme smog episodes from earlier times (Dockery and Pope, 1994; Pope et al., 1995a). 1.1 Particulate Air Pollution Particulate air pollution refers to an air-suspended mixture of solid and liquid particles that vary in size, composition, origin and effects (Dockery and Pope, 1994). These particles, to be kept distinct from fibrous particles such as asbestos, are usually described in terms of particulate matter (PM). 1.1.1 Origin and characteristics of particles Size is the most common approach to classifying particles and it is expressed in terms of the particle's aerodynamic diameter. This is defined as the diameter of a unit-density sphere that has the same settling velocity as the particle in question (Dockery and Pope, 1994). In measurements of size distribution, there is a division of ambient particles into a small and large size fraction, termed fine and coarse particles. Fine particles vary in size between <0.01 to 3 u.m aerodynamic diameter whereas coarse particles are in the range between 1 and >50 urn (Wilson and Suh, 1997). To define these categories more precisely, the fine fraction contains those particles with 2 aerodynamic diameter less than or equal to 2.5 um ( P M 2 . 5 ) and coarse particles are those with aerodynamic diameter between 2.5 and 10 um (PMio-2.5)-These two size fractions differ in their sources, their mechanisms of formation and their composition. Coarse particles are largely formed by mechanical means. Major sources include windblown dust from soil, unpaved roads, evaporation of sea spray, demolition of buildings, pollen, mold spores, as well as parts of plants and insects. They can be suspended and dispersed by wind and anthropogenic activity. Coarse particles are mainly composed of aluminosilicates and other oxides of crustal elements in soil and street dust (Wilson and Suh, 1997). Fine particles are mainly formed from combustion sources, such as motor vehicles, power plants and wood burning. These particles are either primary particulate matter (when they arise through the coagulation or condensation of volatilized materials) or secondary particulate matter (when particles are formed by precursor gases undergoing chemical reactions in the atmosphere). The major components of fine particles often include sulfates, nitrates, elemental carbon and organic compounds (Wilson and Suh, 1997). Size also affects the particles' atmospheric lifetimes, spatial distribution and indoor-outdoor ratios. The atmospheric half-life of coarse particles varies between minutes to hours, whereas the half-life for fine particles is as long as days to weeks. Comparing travel distances, coarse particles travel between <1 to 10s of kilometers as opposed to fine particles, which travel up to 1000s of kilometers. Thus, due to their larger size, coarse particles do not travel as far, settle out faster and are thus not suspended indoors to the same extent as fine particles (Wilson and Suh, 1997). 1.1.2 Particle measurement and standards Sampling instruments are designed to collect a specific portion of the particle size distribution. The standard reference technique for measuring air concentrations of particles is based on gravimetric analysis of filters, which collect the particles while a measured volume of air is being pumped through. Canada has National Ambient Air Quality Objectives (NAAQO) for particulate matter, as opposed to the United States (U.S.), which has National Ambient Air Quality Standards (NAAQS) which are set by the U.S. Environmental Protection Agency (EPA). Canada's objectives currently only consider total suspended particulates (TSP), which includes all particles with an upper size limit between 25 um and 45 um. The objective defines the maximum acceptable 24-hour average as 120 ug/m3 and an annual average as 70 ug/m3 (Johnson, 1999). Before 1987, the U.S. NAAQS for particles also considered TSP as the only exposure indicator. In 1987, however, the NAAQS for particles was redefined based on particulate matter smaller than 10 um aerodynamic diameter (PMio). This size cutoff focused monitoring and regulatory efforts on particles of a size that would be deposited in, and damaging to, the lower airways and the gas-exchanging portions of the lung (see Section 1.2.2). The current U.S. standards for PMio are 150 ug/m3 24-hour average and 40-50 ug/m3 as an annual standard. Some Canadian provinces have established their own objectives for PMio, including B.C., Ontario and 3 Newfoundland. In 1995, for example, B.C. defined an "Interim Air Quality Objective criteria for fine particulate (PMio)" as 50 |!g/m3 24-hour average (Johnson, 1999). Over the last few decades, therefore, most epidemiologic studies considered PMio and TSP measurements as the basis of exposure estimation due to the availability of data from national monitoring networks (Dockery and Pope, 1994). However, recent years have provided a growing amount of evidence that fine particles are a better surrogate for those particle components linked to mortality and morbidity at or below the current standards. In response to these findings, the U.S. EPA has recently set new PM2 5 standards in addition to the PMio standards. The new PM 2 5 standards are 65 ug/m3 24-hour average and 15 ug/m3 annual average. Canada is now also considering Canada-wide Standards (CWS) for both P M i 0 and P M 2 5 . The CWS Development Committee will present two alternatives to the Canadian Council of Ministers of the Environment, in the Fall of 1999. Either Canada could implement only 24-hour metrics for each of the size classes or it could implement 24-hour metrics as well as annual average standards. Two 24-hour target values are being considered, 60/30 ug/m3 or 50/25 ug/m3 for PMi 0/PM 2 5, both of which are lower than the current standards in the U.S. (CCME, 1999). The World Health Organization (WHO) also has Air Quality Guidelines to guide governments in standard setting for pollutants. Actual standards resulting from these guidelines may vary between nations depending on which proportion of the general population and which susceptible groups are to be protected. For many pollutants, the WHO guidelines use thresholds, which describe the pollutant levels at which health effects should not occur. However, as it is believed that thresholds for PMio and PM 2. 5 do not exist, the guidelines are in the form of risk-concentration relationships. WHO provides figures for estimating the risk of certain health endpoints, such as percentage increase in daily mortality, percent change in hospital admissions or percent change in symptoms, at the relative pollution level for PMio, P M 2 5 or sulfates (WHO, 1999). 1.1.3 Typical ambient concentrations Ambient particulate concentrations vary greatly between geographical locations. Differences in local sources and the impact of windblown crustal particles can cause regions to differ in their relative fractions of fine and coarse particles. Particulate standards and objectives set to protect public health have been criticized due to such regional variations in composition and uncertainties in how best to assess the exposure of the public. Ambient concentrations in Vancouver, the location of this study, are slightly lower than found in urban areas of eastern Canada, which have both lower or higher population estimates. Brook et al. describe the relationships between the TSP, PMio and PM 2 . 5 concentrations obtained in various Canadian locations (Brook et al., 1997). From 14 urban sites, operating between 1986 and 1994, the mean TSP, PMio and PM 2 . 5 levels were 55.2, 27.6 and 13.9 ug/m3, respectively. In general, the concentrations of these particulate measures were lower in western Canada when compared to eastern parts of the country. The mean fine mass in Vancouver was 15.5 ug/m3 whereas the 4 larger urban centers of Toronto and Montreal reported concentrations between 15.9-20.9 ug/m3; Windsor, a smaller center than Vancouver also reported 16.8-18.1 ug/m3. Most locations reported higher PM2.5 concentrations during the winter season compared to the summer; the summer-winter difference was greatest for Vancouver and Victoria. 1.2 Health Effects of Particulates 1.2.1 Epidemiology Numerous epidemiologic studies have been conducted to assess the relationship between particulate air pollution and adverse health effects, which are defined by mortality or various indicators of morbidity. These studies have repeatedly found significant associations between particulates and health at concentrations found currently in urban areas and similar to those reported above in Section 1.1.3. 1.2.1.1 Mortality Associations between low concentrations of TSP and daily mortality have been found in Detroit, Michigan (Schwartz, 1991), Steubenville, Ohio (Schwartz and Dockery, 1992b) and Philadelphia, Pennsylvania (Schwartz and Dockery, 1992a). In these studies, increases of 100 ug/m3 TSP led to 4-7% increases in total mortality. A meta-analysis of multiple studies suggested a similar result of 6% increase in total mortality for each 100 ug/m3 increase in TSP (Schwartz, 1994a). Current levels of PM 1 0 have also been significantly associated with increased daily mortality in a number of different settings, which all show similar estimates of mortality. Studies conducted in Utah Valley, Utah (Pope et al., 1992), St. Louis, Missouri (Dockery et al., 1992), Kingston, Tennessee (Dockery et al., 1992) and Birmingham, Alabama (Schwartz, 1993) demonstrated 11-17% increased mortality for 100 ug/m3 increases in PMio. In a review by Dockery and Pope, the effect estimates for various U.S. studies using different exposure metrics were compared (Dockery and Pope, 1994). Effect estimates ranged between 0.7% and 1.6% increase in total deaths per day associated with each 10 ug/m3 increase in the PMio level. Similar results have been found in Santiago, Chile where the effect estimate was 1% (Ostro et al., 1996). Referring to the WHO guidelines, effect estimates for daily mortality range between 0.58% and 0.82% for every 10 ug/m3 increase in PMio (WHO, 1999). Many of these epidemiologic studies of the time-series design have been criticized for lack of control of individual risk factors. The "daily time-series" studies evaluate a regression relationship between air pollution and daily deaths or daily hospital admission counts over time. Two prospective cohort studies, assessing the long-term effects of particulates, have been conducted, the Six-Cities study (Dockery et al., 1993) and a study involving 151 U.S. metropolitan areas (Pope et al., 1995b). Both of these studies accounted for individual risk factors, such as smoking status, gender, age, weight, height and education level, which could each 5 affect mortality rates of the study population independently of the particulate measure in question. However, even after adjusting for these health risk factors, significant associations between particulates and mortality remained when comparing the most to least polluted areas. Although quantitative comparisons between studies of long-term exposure and the time-series studies are difficult to make, these results suggest that the positive associations between air pollution and mortality are not due to the influence of other risk factors. 1.2.1.2 Morbidity Mortality studies are supported by studies that have reported associations between particulate air pollution and decreased lung function (Hoek and Brunekreef, 1993; Hoek and Brunekreef, 1994; Pope and Dockery, 1992; Pope and Kanner, 1993), exacerbation of symptoms (Ostro et al., 1993; Schwartz et al., 1994), and increased hospitalization or emergency room visits (Choudhury et al., 1997; Ostro et al., 1999; Poloniecki et al., 1997; Schwartz, 1994b; Schwartz, 1997; Schwartz, 1999; Schwartz and Morris, 1995; Sunyer et al., 1993). These studies indicate associations for various exposure metrics, including Black Smoke, TSP, PM 1 0, PM 2 5 and sulfates. In addition, the associations have been found in various populations including healthy and health-compromised children, adults and the elderly, and in various locations, including various parts of the U.S. and Europe. Dockery and Pope calculated combined effect estimates for morbidity across various studies (Dockery and Pope, 1994). For each 10 ug/m3 increase in PMio, 1.0% increase in emergency department visits, 0.15% decrease in daily lung function (using the Forced Expired Volume in one second (FEVi)), and 3.0% change in daily lower respiratory symptom reporting were found. t 1.2.1.3 Populations at increased risk Studies have compared particulate associated mortality and morbidity between the general population and sub-populations potentially at increased risk of being adversely affected by particulates. Higher relative risk estimates have been observed for senior citizens and for those with pre-existing respiratory or cardiovascular diseases (Delfino et al., 1994; Pope et al., 1992; Schwartz, 1994c; Schwartz and Dockery, 1992a). It is believed that among susceptible individuals such as asthma patients, observations of increased symptoms, lower lung function, increased medication use and ultimately higher use of hospital services on days of elevated pollution are plausible (Bates, 1992). In summary, despite differences in study design, geographical location and target populations, the epidemiology of particulate air pollution consistently demonstrates significant associations between various sizes of particulates and mortality and morbidity. These studies suggest that the relationships are neither statistical anomalies nor due to confounding by weather or other possible factors. However, the specific biological mechanisms that lead to increased morbidity or mortality are not well understood. 6 1.2.2 Inhalation of particles and deposition in respiratory tract The biological effects of particles are assessed by studying their physical and chemical nature, their distribution in the respiratory tract and the physiological events that occur in the lungs in response to the deposition of particles (Dockery and Pope, 1994). Particle dosimetry addresses sites of deposition of particles in the respiratory tract, independent of specific particle composition, which is considered a determining factor in the biological mechanisms of particle health effects. Aerodynamic diameter is a major determinant which influences how and where particles deposit in the respiratory system. By convention, there are three regions of the respiratory system where particles may deposit: the nasopharyngeal, tracheobronchial or pulmonary (gas-exchange) regions. Most particles greater than 10 um in diameter and approximately 60-80% of particles in the 5-10 um size range are deposited in the first and upper region. Fine particles, PM 2 5 and smaller, penetrate deep into the lungs to the gas exchange region (Spengler and Wilson, 1996). Depending on various factors such as particle diameter, density, inspiration time and flow rate, deposition can occur in the airways and alveoli by impaction, sedimentation or diffusion (Heyder, 1982b). Once particles are deposited, they either can be cleared or be retained by the body. Clearance can be achieved by action of the mucocilliary ladder when particles are deposited in the trachea and bronchioles or by lung macrophages when the particles are deposited in the lower respiratory system. Particles may also migrate through the alveolar tissue directly into lymphatic circulation (Lippmann et al., 1980). In tune with epidemiologic studies, particles smaller than 2.5 um in diameter are expected to present a greater risk than a comparable mass concentration of larger particles. By penetrating to the gas-exchange region, these particles can evade many of the respiratory system's defense mechanisms, such as cilia, and are able to deliver high concentrations of potentially harmful substances to cells of the lung (Spengler and Wilson, 1996). 1.2.3 Uncertainties in associations between particulates and health Between the many epidemiologic and toxicologic studies, a number of uncertainties remain. In general, controversy exists whether or not the associations between PM and health indices are a causal or statistical association due to confounding by meteorological or other factors. Some reviews claim the associations have merely been statistical (Gamble and Lewis, 1996) whereas others believe the associations are true (Thurston, 1996) and not due to confounding (Samet et al., 1998; Schwartz, 1994a). Assuming the associations are true, association does not indicate causation. It is unclear which characteristics of particulates may be causing the increased morbidity and mortality, or whether particles act in concert with other variables in the air pollution mix. Since the measurement of particles has largely been by size, (i.e. TSP, PMio and PM25), these are the variables that have been used in epidemiologic studies to link health effects with particulate pollution. At the same 7 time, it is important to keep in mind that particulate matter within a size fraction is not one contaminant, but rather a mixture of different compounds. Studies using multiple particle indicators have demonstrated fine particles such as PM 2 5 and sulfates were associated with health effects stronger than for larger size fractions (Burnett et al., 1999; Schwartz et al., 1996; Thurston et al., 1989). This includes the Six-Cities study in which daily mortality was associated with PM 2 5 but not coarse particles. The hypothesis that coarse particles do not contribute to mortality was tested by Schwartz (Schwartz et al., 1999). Little association between mortality and dust storm episodes (composed mainly of coarse particles) in Spokane, Washington was found. The stronger relationships between mortality and PM 2 5 perhaps indicate that the composition or dimensions of fine particles increase their ability to exert a toxic effect compared to larger size fractions. The hypothesis that excess mortality and morbidity due to particulates is mainly due to the fine fraction is supported by studies demonstrating that these particles are able to penetrate (Heyder, 1982a) and be retained by the alveolar spaces of the lung (Churg and Brauer, 1997), where they can cause damage. Particles deposited in the periphery of the lung, such as the bronchioles and alveoli, also have slower clearance relative to particles deposited on airways (Falk et al., 1997; Lay et al., 1994; Pinkerton et al., 1995). The evidence pointing to a higher risk associated with fine particles compared to particles of larger size fractions is not conclusive. Both the coarse fraction and PMio have shown greater associations with health effects than PM 2 5 and smaller particles such as sulfate and acid aerosols (Dockery et al., 1992; Schwartz et al., 1994; Van Den Eeden et al., 1998). Nevertheless, the new PM2.5 standards in the U.S. and those being proposed in Canada were developed in response to the studies showing greater association with health for fine particles than larger size fractions. Thus, PM2.5 is one of the exposure metrics used in the present study. In conjunction with the above uncertainties, there is limited scientific information about the contributions of particles of outdoor origin to actual human exposures and the toxicological mechanisms by which these outdoor-source particles might cause adverse health effects. These uncertainties have implications for risk assessment and risk management. In the U.S. EPA's Fiscal 1998 appropriations, the U.S. Congress provided $49.6 million for particulate-matter research. Congress also called for an independent study by the National Research Council (NRC) to identify the most important research priorities relevant to setting NAAQS for PM. The first of four planned reports from this committee was published in 1998 (NRC, 1998). The present study focuses on two critical research needs that were identified by the NRC committee: 1. Investigate quantitative relationships between particulate-matter concentrations measured at stationary outdoor monitoring sites and the actual breathing-zone exposures of individuals to particulate matter [and gaseous copollutants], taking ambient outdoor and indoor pollutant sources and human time-activity patterns into account, especially for potentially susceptible subpopulations. 8 2. Investigate the toxicological mechanisms by which particulate matter produces mortality and acute or chronic morbidity, using [laboratory-animal models], human clinical studies, [and in vitro test systems]. 1.3 Part I - Exposure Assessment This study, in part, focuses on the assessment of exposure to particulates for a population expected to be at increased risk of particle health effects. 1.3.1 Ambient measurements can cause exposure misclassification Most time-series epidemiologic studies assess exposure of the study population using fixed site outdoor measurements of particulates. There may be considerable error, referred to as misclassification of exposure, from using central ambient pollution concentrations as surrogates for individual exposures. Use of single ambient measurements as an indicator for personal exposures is based on the assumption that the spatial variability of the exposure indicator is low throughout the area of interest (Schwartz and Morris, 1995). Thus, a measurement from one location will be representative of the concentrations in surrounding locations. Some studies have shown little spatial variability over study regions especially for fine particles (Wilson et al., 1980), however this may vary depending on particulate sources and their location compared to the area of study. In regions where particulate sources are local in origin, spatial variation may be evident (Cyrys et al., 1998; Ozkaynak et al., 1996). In addition, it is likely that those in city traffic are commonly exposed to short bursts of particle concentrations higher than would be recorded by an averaging ambient monitor. For example, roadside measurements with rapid response nephelometers indicate that individuals can receive short bursts of PM 2 5 , up to 45 ug/m3, from diesel buses (Balogh et al., 1994). Time-series studies also rely on the assumption that ambient particles effectively penetrate indoors such that ambient monitoring data are good surrogates for indoor levels. Indoor levels are subsequently assumed to adequately represent personal exposures to ambient particles, where people spend up to 90% of their time. Lastly, these studies assume that within-individual correlations between personal and ambient concentrations are constant. However, indoor-source particles are ubiquitous and can contribute to variable personal exposures, which cannot be recorded by ambient monitors. In terms of absolute amounts, studies of personal exposure to particulate matter have demonstrated increased personal exposures compared to both indoor and outdoor concentrations (Wallace, 1996). Excess personal exposure, labeled the "personal cloud", has been attributed to proximity to particle-generating sources, such as cooking or environmental tobacco smoke as well as indoor activities, such as cleaning or walking on carpet, which can re-entrain house dust (Ozkaynak et al., 1996). These sources are not constant and could lead to variability in individuals' correlation between personal and ambient measures. 9 Thus factors such as spatial variability over the study region and the exposure of people to both indoor- and outdoor-source particulate may lower the correlation between personal and ambient measures. Several studies have found low correlation between personal exposures and centrally located ambient levels (Spengler et al., 1985), or levels averaged over multiple ambient monitors (Sexton et al., 1984). These low correlations suggest that, for most individuals, the variation in outdoor levels of PM is not tightly linked to variation in personal exposures. Thus, use of outdoor concentrations as a surrogate for personal exposures would tend to misclassify personal exposures. Consequently, exposure-response relationships would be attenuated, implying that the effects of particulate pollution could be greater than estimated from the epidemiologic studies (Utell and Samet, 1993). 1.3.2 Efforts to reduce exposure misclassification Spatial variability and indoor-source particulates may contribute to exposure misclassification when using ambient monitors. In terms of public health protection, the aspect of exposure most relevant to regulatory policy is the portion of total exposure that is attributable to outdoor air. Indoor sources, while contributing to personal exposures, may produce particles of differing composition than those generated outdoors and they may have a different health outcome as well. For the purposes of validating epidemiologic studies that use ambient measurements, a personal exposure metric is needed that is tightly linked to the variation of outdoor concentrations. Since in previous exposure assessment studies, ambient concentrations of particles have correlated poorly with personal exposures, it could be argued that the population is not in fact exposed to ambient particulate matter. If a component of outdoor air pollution can be found for which personal exposure and outdoor concentrations are highly correlated and if these outdoor concentrations are in turn highly correlated with ambient PM2.5 concentrations, this would verify that the population is exposed to ambient particles. Sulfate (SO42'), a component of the PM 2 5 mass, is a marker for outdoor combustion-source particulate. It has been suggested as a better exposure metric than either PMio or PM2.5 (Lippmann and Thurston, 1996). Firstly, indoor and outdoor levels of sulfate have been shown to be comparable (Spengler et al., 1981; Suh et al., 1992). Secondly, high correlations between personal and ambient concentrations have been found for various populations including children (Suh et al., 1992), adults (Brauer et al., 1989) and a population of older adults with cardiorespiratory conditions, which spent little time outdoors (Stieb et al., 1998). Sulfate is one of the predominant inorganic species of fine particulate matter, others being ammonium, nitrate, elemental carbon and crustal material. The main precursor of sulfates is S02, emitted to the atmosphere from the combustion of coal, gasoline, diesel and wood (Wilson, 1996). Sulfates are formed through the oxidation of S02, leading to sulfuric acid aerosols (H2SO4). These can react rapidly with atmospheric ammonia (NH3) leading to various molecular forms of SO42" (e.g. ammonium bisulfate (NH4HSO4) or ammonium sulfate (Na2S04)) (Lioy and Waldman, 1989). The average sulfate aerodynamic diameter is 0.48 um and concentrations range from a low of about 0.04 ug/m3 over the remote oceans to over 80 ug/m3 under polluted conditions over the continents (Whitby, 1978). 10 High correlation between outdoor and personal sulfate concentrations, in comparison to PM mass, has been attributed these particles exhibiting low spatial variability and having no major indoor sources. Submicron particles, such as sulfate, also have a low deposition rate and they are not subject to resuspension due to indoor activities, as compared to larger size particles. This explains findings of little difference between indoor and outdoor particle concentrations when there are no indoor combustion sources (Thatcher and Layton, 1995). In addition, particulate sulfate is found to be chemically and physically stable in the air and on sampling filters. The PM mass, in comparison, is noted for losing semi-volatile materials, such as nitrates and organics, over time leading to underestimation of concentrations (Lippmann and Thurston, 1996). In an effort to reduce exposure misclassification, this study compares the use of ambient particulate mass and sulfate measures for assessing personal exposures. Use of sulfate as the exposure metric is hypothesized to be a better predictor of personal exposures than PM mass. A study design with repeated exposure measurements of individuals in the study population also allows for assessment of individual variability in the relationship between personal and ambient concentrations. 1.3.3 Target population Another aspect of this study's exposure assessment is that the target population includes individuals who are at increased risk for particle health effects. Patients with chronic obstructive pulmonary disease (COPD) have been chosen to form the study population. COPD is a chronic condition is which there is insufficient flow of air into or out of the lungs. The result is either constant oxygen deficiency in the blood stream or constant shortness of breath due to increased breathing rate aimed at maintaining adequate oxygen levels. Bronchitis and emphysema are the most common forms of COPD and cigarette smoking is the most common cause. Other factors, such as air pollution, occupational exposures and infections may contribute to disease progression (Cherniack, 1991). COPD patients are thought to be affected to a higher degree from particulates since they suffer almost universally from ventilation-perfusion inhomogeneity. Hence, a disproportionate amount of their respiration occurs in a small part of their lungs, and it is in this portion where the particulate exposure will be concentrated. The combination of greater effective dose in the portion of the lung that remains functional and decreased reserve capacity may enhance their response to particulates (Bates, 1992). While epidemiologic studies have demonstrated certain health-compromised groups, such as COPD patients, as more susceptible to being affected adversely by particulate air pollution, most exposure assessment studies have focused on the personal exposures and activity patterns of healthy children or adults. Exposures of susceptible individuals may differ considerably from the healthy population, for example, due to reduced mobility. Thus, information regarding exposure and activity patterns of these individuals are needed and these are addressed by this study. 11 1.4 Part II - Assessment of Cardiovascular Health Effects While Part I of this study deals with some of the uncertainties surrounding exposure assessment of humans to outdoor-source particles, Part II deals with the uncertainty regarding the biological mechanisms by which these outdoor-source particles cause adverse health effects. 1.4.1 The question of biological plausibility Epidemiologic studies, as described under Section 1.2.1, have demonstrated increased mortality and morbidity of various populations in numerous geographical locations due to particulate air pollution. The increases appear to be due in part to elevated risk among the elderly and for cardiovascular and respiratory disease mortality, such as for COPD patients discussed above. Both time-series studies (Borja-Aburto et al., 1998; Pope et al., 1992; Schwartz, 1994c; Schwartz and Dockery, 1992a) and prospective, long-term exposure studies (Dockery et al., 1993; Pope et al., 1995b) of sulfates, PM 2 5 , PMio and TSP have demonstrated associations with deaths due to respiratory and cardiovascular causes. In a summary of time-series studies, the effect estimates for cardiovascular and respiratory deaths were 1.4% and 3.4% increases, respectively, for each 10 ug/m3 increase in PMio (Dockery and Pope, 1994). Even though the effect estimate for cardiovascular causes is below respiratory causes, the actual number of excess deaths associated with particulate air pollution is higher for cardiovascular disease since its baseline mortality rate is higher (Hessel et al., 1997). For example, Kinney and Ozkaynak found positive associations for cardiovascular deaths in relation to air pollution, whereas none were found for respiratory deaths (Kinney and Ozkaynak, 1991). The authors suggest this was likely due to much lower daily death counts for respiratory causes (8/day) than for cardiovascular causes (87/day). Hospital admissions for cardiovascular causes, as opposed to total admissions, are also increased in association with particulate air pollution. In the U.S., 25 and 23 ug/m3 interquartile range increases in PMio exposures have been associated with 2.48% (Schwartz, 1999) and .2.75% (Schwartz, 1997) increases in admissions for cardiovascular disease in the elderly. Investigating the cause of cardiovascular deaths, Schwartz and Morris found relative risks of hospitalization, the highest for congestive heart failure (RR=1.032), followed by cardiac arrhythmias (RR=1.019) and ischemic heart disease (RR=1.018) for 32 ug/m3 increases in PM i 0 (Schwartz and Morris, 1995). Burnett et al. reported similar results for cardiovascular hospitalization and sulfates in Ontario, also ranking risks due to heart failure as the highest (Burnett et al., 1995). Therefore, a wealth of epidemiologic studies has indicated firstly that morbidity and mortality are increased during, or just after, periods of elevated particulate levels. These studies have also demonstrated that increased numbers of death are due largely to respiratory and cardiovascular causes. However, some scientists call into question epidemiologic studies describing the acute health effects of particulate air pollution, in particular those dealing with cardiovascular mortality, due to lack of a biologically plausible mechanism (Utell and Samet, 1993; Waller and Swan, 1992). Many authors have suggested that air pollution episodes, like any environmental factors, 12 are an additional environmental stress that may cause death in otherwise-compromised patients (Dockery and Pope, 1994). With respect to observed increases in cardiovascular mortality due to air pollution, it has been argued that, since the relation between heart and lung disease is extremely complex, the cardiovascular deaths observed during air pollution episodes may be attributable to a compromised respiratory system (Bates, 1992). Thus, death due to respiratory causes may often be recorded on death certificates as cardiovascular disease, which may result in biased estimates of pollution-related mortality (Pope et al., 1995b). Indeed, with the design of time-series studies, it is not possible to identify mechanisms by which air pollution might cause death, since there is usually no information about individuals' medical status (Borja-Aburto et al., 1998). Studies assessing the clinical status of subjects, therefore, can be undertaken to help support hypotheses concerning the biological mechanisms in the cause of particle health effects. 1.4.2 Possible biological mechanisms The association of ambient particles with cardiovascular function is supported by recent hypotheses, epidemiologic and human/animal toxicological studies. Hypotheses generally suggest that air pollutants diminish the ability of the lungs to oxygenate the blood or the ability of the blood to deliver oxygen (Morris et al., 1995; Schwartz and Morris, 1995). For example, the effects may be caused by acute respiratory responses (such as bronchospasms) and acute inflammatory responses leading to decreased oxygenation of the blood and increased demand on the myocardium (Schwartz and Morris, 1995). Alternatively, Seaton et al. suggest that inflammation caused by particulates in the alveoli may release mediators into the blood leading to increases in blood coagulability (Seaton et al., 1995). If true, these hypotheses can explain the observed increases in cardiovascular deaths associated with urban air pollution episodes. 1.4.2.1 Inflammation Inflammatory responses due to particulates have been observed in a multitude of studies using various subjects (different cell types, animals, or humans) and assessing various particulate mixtures. Residual oil fly ash (ROFA), a byproduct of power plant and industry fuel-oil combustion, diesel exhaust particles (DEP) and concentrated ambient particles (CAPs) are some examples. Inflammation is usually measured by increases in neutrophils (PMNs) in bronchoalveolar lavage fluid (BALF), or the synthesis and expression of the inflammatory cytokines TL-8, JX-6, and TNFalpha. Elevations in total protein and lactose dehydrogenase (LDH) activity in BALF are other indicators of pulmonary injury. Using such indicators of inflammation, inhalation studies of have demonstrated short-term exposures to high ambient concentrations can produce well-defined pulmonary inflammatory responses in both rats (Gordon et al., 1998) and in healthy human volunteers (Blomberg et al., 1999; Salvi et al., 1999). These inflammatory responses may be due in part to bio-available transition metals adsorbed to urban air pollution particles, which result in oxidant generation in the lung (Carter et al., 1997; Ghio et al., 1999; Goldsmith et al., 1998; Lay et al., 1999). 13 1.4.2.2 Blood coagulation An aspect to the hypothesis presented by Seaton et al. was for alveolar inflammation to cause acute changes in blood coagulability (Seaton et al., 1995). Blood coagulation is the process of clot formation and has effects on blood viscosity, the property of blood that makes it resist flow. Viscosity increases when the number of cells increases above normal, or when plasma concentration of large molecules (e.g. protein) is higher than normal. This causes the heart to work harder to maintain the normal blood flow (Rhoades and Pflanzer, 1992). Hematological changes have been observed in both rats and humans following air exposures to particulates. Rats exposed for 3 hours to CAPs consistently showed hematological changes at 3 hours post-exposure, assessed by increased circulating blood neutrophils and a decrease in lymphocytes (Gordon et al., 1998). By sampling the volunteers' blood after diesel exhaust exposures, Salvi et al. demonstrated the inflammatory response in humans is not restricted to the lung, but can extend to the circulatory system in a systemic response (Salvi et al., 1999). Significant increases in neutrophils and platelets were observed in peripheral blood following exposure. It may also be possible that particulates induce secondary polycythemia (an increase in circulating red blood cells above normal), which is triggered in situations of arterial hypoxia and stimulates erythropoiesis (Rhoades and Pflanzer, 1992). Thus, hematologic changes in the systemic circulation (due to particulates) may lead to increased blood coagulation as seen by increases in plasma viscosity, platelets, fibrinogen/fibrin or other factors involved in clot formation. Plasma viscosity has been directly tested as a possible contributor to pollution-induced cardiovascular effects. During the MONICA survey in Augsberg, Germany, there was a period of increased air pollution, at which time increased risk of extreme values of plasma viscosity were observed in both men and women (Peters et al., 1997). Plasma viscosity has been associated with congestive heart failure (Timmis and Nathan, 1993), ischemic heart disease (Yarnell et al., 1991) and myocardial infarctions (Koenig et al., 1992). "An increase in platelet numbers during a PM pollution episode, especially in elderly people compromised with cardiovascular functions, may increase their risk of developing strokes and coronary vessel thrombosis, thereby increasing cardiovascular mortality and morbidity" (Salvi et al., 1999). 1.4.3 Measures to assess cardiovascular health Many recent studies have examined the relationship between particulates and respiratory hospital admissions; and many have used lung function testing and symptom questionnaires as a clinical measure for testing respiratory health effects. Comparatively few studies have focussed on cardiovascular morbidity indicators. Our goal is to assess the cardiovascular health of COPD patients in relation to particulate air pollution. In COPD, persistent airflow limitation can lead to inadequate blood oxygenation. Therefore, though it is a respiratory disease, COPD can have severe consequences on other organs of the body, including the heart, kidneys and brain because these require certain levels of oxygen and carbon dioxide to operate normally. Taking the heart as an example, pulmonary 14 hypertension and cor pulmonale are major factors contributing in the morbidity and mortality associated with COPD (Pietra, 1991). As discussed previously, COPD patients appear to be at increased risk of death in association with particulates. The cause of adverse particle effects in individuals affected by COPD may involve a cardiac mechanism that is equally or more important than other mechanisms, which have greater involvement of the lungs. Therefore, our goal was to assess whether exposure to particulates can affect the existing cardiovascular state of COPD patients. A variety of non-invasive, subclinical indicators are used to assess the cardiovascular health of this population: blood pressure, arrhythmia, heart rate and heart rate variability indices obtained from 24-hour electrocardiogram (ECG) recordings. Observations of association between particulate exposure and these variables might help understand the biological pathways thought to be involved in particle health effects. For example, changes in blood pressure would be suggestive of changes in blood viscosity or vascular tone. Increases in arrhythmia may suggest an increase in myocardial stress due to particulate exposure. Additionally, heart rate and heart rate variability can be used to assess cardiac autonomic function. Before describing these measurements in more detail, a brief review of the nervous system and how it pertains to the cardiovascular system may be helpful for the later discussion of heart rate and heart rate variability. The cardiovascular system is under control of the autonomic nervous system (ANS). The ANS is one branch of the peripheral nervous system, the other being the somatic nervous system. While the somatic nervous system controls the voluntary movement of skeletal muscles, the ANS involves the involuntary control of internal organs, glands and muscles to coordinate bodily functions needed for survival. The ANS can be divided into two components: the sympathetic nervous system and the parasympathetic nervous system. While the sympathetic nerves prepare the body for stressful situations, the parasympathetic system coordinates the non-stress related bodily functions such as digestion. Most organs, including the heart, are innervated by nerve fibers from both the sympathetic and parasympathetic nervous systems and relay different messages to that organ at different times. For example, sympathetic nerves act to increase heart rate whereas parasympathetic nerves cause heart rate to decrease (Rhoades and Pflanzer, 1992). 1.4.3.1 Blood pressure Blood pressure (BP) is a measure of the driving force that causes blood to flow through the vascular system. It is usually expressed in units of millimeters of mercury (mm Hg). During the cardiac cycle, diastole refers to when the ventricles are filling with blood and systole is when the ventricles are actively pumping blood out of the heart. Arterial pressure is lowest at the end of ventricular diastole (known as diastolic blood pressure) and is highest during ventricular systole (systolic blood pressure) (Rhoades and Pflanzer, 1992). Blood pressure is measured by placing a cuff around the subject's arm and inflating it with air such that the artery under the cuff is completely collapsed. A stethoscope is then used to listen to blood flow through the artery. With the artery fully collapsed, no blood will flow through. As pressure is released from the cuff, the pressure at which blood is initially able to spurt through the 15 artery creates pulse-like sounds; this is the systolic blood pressure. Releasing the pressure further allows the blood to flow more continuously through the artery. The pressure at which no more pulse sounds can be heard is the diastolic blood pressure. Hypotheses concerning biological mechanisms have suggested that chemical mediators are released into the systemic circulation upon particulate exposure, which cause increases in blood viscosity. It is feasible that changes in blood viscosity could cause alterations in peripheral resistance to blood flow, which in turn would positively affect blood pressure (Rhoades and Pflanzer, 1992). However, blood pressure as an endpoint has only been used in limited studies dealing with particulate pollution, which have demonstrated both positive changes or no changes in blood pressure upon exposure (Gold et al., 1998; Gong et al., 1999; Linn et al., 1998). More studies are needed to define the blood pressure effects due to particulates more precisely. 1.4.3.2 Cardiac arrhythmias The proper function of the heart as a pump depends on the correct sequence of contraction of different parts of the muscle, which arise from electrical impulses travelling through it. A correct sequence of these impulses depends on the normal function of the conducting system through the heart. This electrical activity of the heart can be detected at the body surface by an instrument called an electrocardiograph. The electrocardiograph produces a record of the electrical impulses that immediately precede contraction of the heart muscle, known as the electrocardiogram (ECG, Figure 1.1). This study uses Holter monitoring to record ECGs of study participants. A Holter monitor is a portable electrocardiograph that can be used in ambulatory (i.e. non-hospital) settings for long-term recordings (i.e. 24-hours). Electrodes placed in specific locations on the subject's chest relay electric impulses from the heart to the monitor, which records the impulses onto cassette tape. Tapes are later scanned and analyzed by a trained technician using computer software to obtain the ECG information. Figure 1.1. The electrocardiogram. A normal impulse is initiated by the S-A node. This impulse spreads through the atria, thereby activating the atria (called atrial depolarization) and producing the P wave on the E C G . Once the atria are depolarized, the impulse reaches the A-V node. From the A-V node, the impulse travels through other heart fibers to the ventricles. Activation of the ventricles forms the QRS complex, which is followed by ventricular repolarization (T wave) (Chung, 1983). 16 ECGs can be used to diagnose disorders of the heart, many of which produce deviations from normal electrical patterns. Any variations from the normal rhythm of the heart are called arrhythmias. In general, cardiac arrhythmias are divided into two major categories: abnormal impulse formation and abnormal conduction (Chung, 1983). Any cardiac impulse originating from a site other than the sinuatrial (S-A) node is termed "ectopic". Such beats can be formed in the atria, the atrioventricular (A-V) junction or the ventricles. In this study, we assess supraventricular ectopic (SVE) and ventricular ectopic (VE) heartbeats. Supraventricular is defined as 'above the ventricles', thus in our case pertains to impulses formed in an atrium or the A-V node. Cardiac arrhythmias can be caused by various factors including sympathetic and parasympthetic effects, chemical mediators, drug toxicity, electrolyte imbalances, heart rate as well as cardiac (e.g. coronary heart disease) and non-cardiac diseases. Atrial and ventricular arrhythmias can be asymptomatic and benign, however may also produce congestive heart failure, or aggravate preexisting CHF and can lead to myocardial ischemia and even death (Timmis and Nathan, 1993). Particulates may increase stress on the myocardium due to hypoxemia or an increased blood viscosity. An increased occurrence of cardiac arrhythmic events may demonstrate such stress. Only one study was found in the literature, which cited results of an analysis of SVEs in relation to particulates; a positive relationship was found (Linn et al., 1998). Some studies have quantified adverse ECG events by analyzing the shape of the ECG waveforms. For example, Godleski et al. exposed dogs to CAPs (-200 ug/m3 PM 2 5 ) (Godleski et al., 1997). While the ECG had no changes during the clean air exposures, concentrated air exposures produced substantial P-Q and S-T segment changes in the dogs. ECG changes in normal and health-compromised rats exposed to ROFA and CAPs have also been demonstrated (Sato et al., 1999; Watkinson et al., 1998; Watkinson et al., 1999). In one study, increases in the incidence and duration of serious arrhythmic events appeared to be associated with impaired atrioventricular conduction and myocardial hypoxia (Watkinson et al., 1998). However, for healthy human volunteers exposed to CAPs did not result in symptoms, decrements in lung function or EKG abnormalities (Ghio et al., 1998). These studies have demonstrated that animals exposed to particles can cause abnormal cardiac rhythms, indicative of life-threatening cardiac arrhythmias. The effects on healthy and susceptible human populations are questionable as studies assessing these effects are lacking. 1.4.3.3 Heart rate and heart rate variability In addition to arrhythmias, other variables that can be obtained from ECG recordings include heart rate and heart rate variability (HRV) indices. The importance of heart rate in disease is not completely understood (Dockery et al., 1999). However, high heart rate has been associated in recent studies with hypertension (Gillum, 1988), coronary heart disease (Dyer et al., 1980) and mortality (Goldberg et al., 1996). The effect of particulates on heart rate is also debatable. For example in normal and health-compromised rats, 17 both significant increases (Gordon et al., 1998) and dose-dependent decreases (Watkinson et al., 1999) in heart rate have been reported in response to exposure. Two studies using human subjects have witnessed increased heart rates in association to air pollution (Peters et al., 1998; Pope et al., 1999a). With respect to the possibility particulates increasing blood coagulation, persons with elevated plasma viscosity measurements have shown greater increases in heart rate (mean increase: 5.1 bpm) than those with normal plasma viscosity measurements (mean increase: 1.4 bpm) during an air pollution episode (Peters et al., 1999). Heart rate relates to measures of heart rate variability (HRV), which indicate the beat-by-beat change in length of the interval between beats, measured by the R-R intervals between QRS complexes. In the normal resting heart, time intervals between beats vary around their mean. This variability results from the combined action of both the sympathetic and parasympathetic nervous systems. The two systems act in balance to maintain and modify cardiovascular parameters through neural, humoral and other physiological mechanisms. Thus, HRV is used to quantify the state of the autonomic nervous system (Cerutti et al., 1995). Time-domain analysis, which will be used in this project to study HRV, is based on different ways of measuring the standard deviation (SD) of sinus (i.e. non-arrhythmic) R-R intervals over the sampling period. Data are reported in terms of interval length in milliseconds (ms). Time-domain variables can be derived either from direct measurements of the beat-to-beat intervals or from differences between adjacent intervals. Variables from the first category that are considered in this study include SDNN and SDANN. SDNN, the standard deviation of the normal R-R intervals, is an estimate of overall HRV in the entire ECG recording. SDANN is calculated by averaging the normal R-R intervals within a 5-minute block, then determining the standard deviation of the 5-minute means for the duration of the ECG recording. These variables are primarily influenced by diurnal and secular trends and short-term sympathetic influences, and less by parasympathetic tone (Kleiger et al., 1995). The second category of time-domain variables, derived from differences between adjacent intervals, includes the r-MSSD and the pNN50. The r-MSSD, standing for root mean square successive difference, is the square root of the mean squared differences of successive normal R-R intervals over the duration of the ECG recording. The pNN50 is the percent of the absolute differences between successive normal R-R intervals that exceed 50 ms. These variables are estimates of the short-term components of HRV and are independent of diurnal or other long term trends. They are affected by alterations in autonomic tone that are mostly parasympathetically mediated (Kleiger et al., 1995). Time-domain variables are used to assess autonomic function both in cardiac patients as well as those with non-cardiac conditions. The autonomic effects of drugs, exercise and other stresses have been assessed by use of these variables, where decreased HRV marks decreased parasympathetic and increased sympathetic tone (Kleiger et al., 1995). HRV has been shown to be decreased in patients with heart failure (Casolo, 1995) and in post-myocardial infarction patients (Bosner and Kleiger, 1995). Decreased HRV after acute myocardial infarction is also a risk factor for subsequent morbidity and mortality, including all cause mortality, ventricular arrhythmias (Vybiral and Glaeser, 1995) and sudden cardiac death (Singer and Ori, 1995). 18 Studies assessing time-domain H R V indices in relation to particulate exposure have reported mixed results (Gold et al., 1998; Liao et al., 1999; Pope et al., 1999b). The findings of altered heart rate or H R V in relation to particulate air pollution would help link together the hypothesis that air pollution causes alterations in blood parameters, which could directly affect the action of the heart. It is possible that a particulate-induced inflammatory response in the lung induces the release of chemical mediators which alter the autonomic nervous system control of cardiac rhythm (Godleski et al., 1996). The studies outlined above give an indication that research using subclinical signs as a measure of effect in particulate pollution studies can aid in determining mechanisms of action in the body. However, these studies are limited and not conclusive. While some studies have obtained repeated health measurements on individuals, most have not conducted personal exposure monitoring. In addition, most studies, while perhaps focussing on the elderly population, do not address other susceptible groups such as C O P D . patients. Detailed exposure and health assessments for a susceptible group, unique aspects of the present study, may yield important information regarding the cardiovascular health effects due to particulates and their measurement. 19 1.5 Study Design 1.5.1 Hypotheses a) Exposure misclassification, when using ambient measures as surrogates for personal exposures, is greater for P M mass than for sulfate, due to spatial variation across the study region and indoor sources of the particle species of concern. b) Exposure to fine particle air pollution is associated with alterations in blood pressure, arrhythmic events, heart rate and heart rate variability in individuals with physician-diagnosed chronic obstructive pulmonary disease. 1.5.2 Objectives a) To take repeated measures of personal PM 2 . s , sulfate, electrocardiograms and blood pressure in a population of C O P D patients and to measure ambient levels of PMio, PM2.5 and sulfate at five locations within the Greater Vancouver Regional District. b) To determine the correlation between personal and ambient measures of PMio, P M 2 . 5 and sulfate over time for individuals susceptible to particle health effects. c) To evaluate factors influencing the correlation between personal and ambient measures, such as subjects' distance of residence from ambient monitors, and the influence of individual activities and housing characteristics on personal exposures. d) To assess the relationship between particulate exposure and blood pressure, arrhythmia, heart rate and heart rate variability of the study population. 20 CHAPTER 2: METHODS 2.1 Overview The study was conducted from Apr i l 21 to September 25, 1998. Ambient P M i o and P M 2 . 5 concentrations were measured at five sites within the Greater Vancouver Regional District ( G V R D ) . Hourly P M i o and 24-hour averaged P M 2 . 5 concentrations were measured at each site for the duration of the study. The study population consisted o f patients with physician-diagnosed chronic obstructive pulmonary disease (COPD). Seven sampling sessions were planned for each subject. Sampling sessions consisted of 24-hour measurements of personal P M 2 . 5 exposure as well as tests to assess respiratory and cardiac function. Time-activity logs were completed by the subjects during each sampling period and a one-time dwelling characteristics questionnaire was also administered. Particulate sulfate concentrations were measured by ion chromatography. Table 2.1 outlines the measurements taken throughout the study and Figure 2.1 displays the locations of ambient monitoring sites and subject homes within the study region. 2.2 Pre-Studv Sampler Experiments Two sets of experiments were undertaken before the data collection phase o f the study to determine the P M 2 . 5 concentration relationships between different sampler types and within samplers o f one type. The first experiments compared the P M 2 . 5 concentrations obtained from personal exposure monitors ( P E M , M S P Corp.), used for personal sampling, with those obtained from Harvard Impactors, used for ambient sampling, during equal sampling conditions. The second experiments were conducted to determine the reproducibility in the concentrations reported by two P E M s worn simultaneously by one person. Appendix 1 presents a more complete description of the methods and results of these experiments. Measurement Frequency Method Ambient P M i 0 * Hourly, 24-hrs, 4 days/wk at 5 sites Tapered Elemental Oscillating Microbalance (TEOM) Ambient P M 2 . 5 and sulfate* 24-hrs, 4 days/wk at 5 sites PM 2 5 Harvard Impactor, 4 L/min sample flow; sulfate by ion chromatography Personal PM2.s and sulfate* 24-hrs, 7x/subject Personal PM25 Exposure Monitor, 4 L/min; sulfate by ion chromatography Time-activity log* 24-hrs, 7x/subject Log sheet completed by subject Dwelling characteristics* lx/subject Interview questionnaire Electrocardiogram* 24-hrs, 7x/subject Cassette Holter recorder Blood pressure* Pre and post sampling, 7x/subject Blood pressure cuff/stethoscope (used in seated position) Lung function 3 blows pre/post sampling, 7x/subject Portable pneumotach spirometer Pulse oximetry 3 recordings pre/post sampling, 7x/subject Hand-held pulse oximeter using red + infrared light (through finger clip sensor) Symptoms 7x/subject Post sample interview questionnaire Medication use 7x/subject Post sample medications checklist Table 2.1. Study Measurements; *focus of this thesis. 2 1 Figure 2.1. Map of study region; • = ambient monitoring site (KT = Kitsilano, ND = North Delta, NB = North Burnaby, SB = South Burnaby, SR = South Richmond); • = subjects' homes. 2.3 Ambient Concentration Measurements 2.3.1 Ambient site locations and sampling equipment With the cooperation of the GVRD Air Quality department, GVRD air monitoring sites within the study region were used for collecting ambient particulate samples. These sites, shown in Figure 2.1 were Kitsilano, North Delta, North Burnaby, South Burnaby, South Richmond and Port Moody. Each site consisted of a temperature-controlled shelter in which sampling equipment was set up. Due to the location of participants' homes within the study region, data from the Port Moody site was not used in the analysis. PMio data was collected on a continuous basis at each ambient monitoring site for the duration of the study with Tapered Elemental Oscillating Microbalances (TEOM) as part of the regular monitoring conducted by the GVRD Air Quality department. This data was obtained from the GVRD at the completion of the study. The following procedures refer to sampling for P M 2 . 5 . P M 2 . 5 samples were collected with Harvard Impactors (HI) (Marple et al., 1987) loaded with 41 mm 2 um pore size Teflon filters (Teflo, R2PJ041, Gelman Sciences). Samplers were set-up on tripods, secured with metal wire to plywood platforms, on the roof top of the shelters. The plywood platforms were weighted with cinder blocks to prevent the structure from moving. Metal rain caps were placed on the tops of the samplers to protect them from the weather. Air 22 was sampled at 4 L/min ± 5% using flow-controlled pumps located inside the shelters. The types of pumps varied between sites (Universal Sample Pump by SKC, Harvard black box pumps, Gillian pumps, Gast pumps). All 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 of each week. Flows were measured before and after each sampling period using precision rotameters (Matheson, model FM-1050) calibrated with a frictionless piston meter (Bios Corp.). Rotameters were calibrated at UBC and placed at each site at the beginning of the study. Dichotomous samplers (Anderson Series 241) were also operated at the Kitsilano and Port Moody sites however this data was not used in the analysis. These samplers collected both course and fine mass fractions at flow rates of 16.7 L/min and 1.67 L/min, respectively. Flows were recorded using the instruments' built-in rotameter. Leak checks were conducted every Monday morning by replacing the sampler inlet with a valve that could be sealed. With the valve closed, absence of leaks was indicated by flows showing zero units on the rotameter and the presence of a vacuum inside the system. 2.3.2 Lab preparation His from each site were cleaned thoroughly once per week. Impactor plates were sonicated in soapy water for 15 minutes and then rinsed three times with distilled/deionized water. Other sampler components, including the filter holder, plastic filter slides and inlet pieces, were soaked in soapy water for 15 minutes, scrubbed and then rinsed three times with distilled/deionized water. All components were allowed to dry completely. Before loading His, clean impactor plates were saturated with mineral oil. Filter slides, holding pre-weighed filters (see Section 2.5.1), were placed in the His on top of disposable cellulose-backing pads. Before sealing the samplers with plastic caps for transport, the His were leak checked. An adapter was fitted to seal the inlet of the sampler. The sampler was connected to a rotameter, which in turn was connected to a vacuum pump used to pull air through the sampler. If the rotameter showed less than 10 units (approximately 0.4 L/min or 10% of the flow rate used during the study), then the sampler was considered sealed. If the rotameter showed greater than 10 units, samplers were taken apart, o-rings and filter slides were inspected for possible damage and replaced if necessary. At the completion of every week, His, flow logs and used filters were returned to the laboratory. Using clean forceps, filters were transferred from their filter slides back to their original petri plates along with the corresponding labels from the slide. 2.3.3 Field sample collection The GVRD Air Quality department staff helped UBC technicians in operation of the five sites. The GVRD completely operated the North and South Burnaby and North Delta sites, whereas UBC was responsible for the Kitsilano and South Richmond sites. 23 Four samples were collected every week at each location. These were collected on weekdays only, starting Mondays, Tuesdays, Wednesdays and Thursdays, with the last sample ending on Friday. At each site, samples were to be started between 7:00 am and 10:00 am and to be ended the following morning approximately 24-hours later. Flows and times were recorded before and after each sampling period on the ambient flow log sheet (Appendix 2). Before starting a new sample, a 'flow check' was conducted to verify that the pump was set to the appropriate flow rate of 4 L/min ± 10%. This was achieved by connecting the rotameter to the end of the flow chain without the sampler connected. This value was recorded and, if needed, the pump was adjusted to be within the 4 L/min range. The HI, loaded with a new filter, was then connected to the sampling train and the 'actual flow' through the sampler was measured and recorded. Leak checks were not performed on the His in the field, however a large difference between the flow check and the actual flow was an indication of a possible leak. Once flows were determined, the top inlet piece and rain cap was placed on the HI, the pump was attached to the sampler and the start time of the sample was recorded. At the end of each sampling period, the HI was disconnected from the pump and the stop time was recorded. The flows through the sampler ('actual flow') and without the sampler ('flow check') were recorded as above. The HI base was disassembled and the used filter was placed into the ambient filter storage box. The used backing pad was removed and a new backing pad was placed in the filter holder using clean forceps on which a new filter was placed. The HI base was reassembled. Finally, the HI inlet was disassembled to remove visible particle build up, using razor blades, from the impactor plates. 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. 2.4 Personal Exposure and Health Effects Measurements 2.4.1 Study population - eligibility and recruitment The target sample size was 25-30 participants. Before subjects were recruited, ethical approval for the study was obtained from the Clinical Research Ethics Board, University of British Columbia. All participants in the study received an honorarium of $250. The study population consisted of patients with physician-diagnosed COPD. Eligibility criteria of the study limited participation to patients with a form of light to moderate COPD (forced expiratory volume in 1 second, (FEVi) ^ 0.75L). In addition, subjects were to be 60 years or older, residents of the metropolitan Vancouver region (excluding the geographically isolated municipalities of North and West Vancouver) and currently not smoking nor living with current smokers. 24 Participants for the study were recruited through the Respiratory Clinic at Vancouver Hospital and Health Sciences Center (VHHSC) and the B.C. Lung Association Puffers' Club. Respiratory physicians at the VHHSC clinic were asked to outline the study to potential candidates and provide interested individuals with a study recruitment letter (Appendix 3). Recruitment through the B.C. Lung Association was initiated by presenting the study at a meeting of the Vancouver "Puffers' Club" approximately one month into the data collection phase of the study. Study recruitment letters were distributed at this meeting and the names and telephone numbers of interested individuals were recorded on a sign-up sheet. Upon collecting names of interested individuals from either route, candidates were initially contacted by UBC technicians via telephone. The study was briefly described and, if the candidate remained interested in participating, a 30-60 minute introductory meeting was arranged. During the introductory meeting, the official informed consent form (Appendix 4) was provided as a source of information about the study. Sampling equipment and procedures were demonstrated and any remaining questions from the candidate were answered. For individuals recruited through the Puffers' Club, the study eligibility criteria were explicitly explained. Each potential candidate was also asked to perform a forced expiratory maneuver using a portable spirometer to screen for severity of disease (to assess that FEVi > 0.75L). Candidates could sign up for the study at the end of the introductory meeting if interested, or they could review the consent form on their own before making a decision. 2.4.2 Subject identification and sampling schedules Each consenting individual was assigned a unique three-digit identification (ID) number, used for all identification purposes. Seven 24-hour sampling sessions were planned for each subject. Initially, sampling schedules were randomized using the random number generator in Excel 5.0, by assigning numbers to each possible sampling day over the course of the study. Constraints included: sampling on weekdays only; spacing sampling dates per subject at least one week apart; scheduling the seventh sampling session for each subject no later than the end of September 1998; scheduling a maximum of three subjects per day. As more subjects were enrolled in the study, travel time was also considered in scheduling sampling sessions. Thus, in some cases, subjects living in certain areas of the GVRD were scheduled on the same day. Sampling schedules were brought to each subject's home on their first sampling day. Subjects were informed that the schedule was tentative and sessions could be rescheduled at any time should conflicts arise. If patients needed to reschedule a sampling date, the constraints outlined above were maintained. 2 5 2.4.3 Personal sampling equipment and forms 2.4.3.1 Particulate sampling Personal exposures to P M 2 . 5 were measured with personal P M 2 . 5 impactors (PEM, MSP Corp.), loaded with 37 mm 2 um pore size Gelman Teflon filters (Teflo, R2PJ037, Gelman Sciences) and connected to six-inch long aluminum inlets. Air was sampled at 4 L/min using a flow-controlled battery operated pump (Aircheck Sampler 224-PCXR4, SKC Inc.). To provide sufficient power for 24-hour samples, secondary battery packs were connected to the pumps in series. Flows were measured at the beginning and end of each 24-hour sampling period with precision rotameters (Matheson, 603) calibrated with a frictionless piston meter (Bios Corp.). To aid in carrying the pump, as well as to reduce noise levels, pumps were placed inside small bags lined with foam as the insulation material. Participants were asked to wear the monitor on their shoulders using the shoulder strap on the bag (Figure 2.2). The sampler, attached to the pump by tubing, was kept outside of the bag. It was protected from clothes and other nearby objects by the aluminum inlet. The inlet was fitted with Velcro, which secured the sampler to the shoulder strap. The sampler and inlet combination was placed in a downward facing position, as close to the subject's breathing zone as possible. Subjects were asked to wear or carry the exposure samplers/pumps, whenever possible during the sampling period. They were allowed to place the monitor beside them while sitting in one room, but were asked to take it with them if they walked to other rooms in the house. At night, the monitor was placed beside the subject's bed. If pump noise levels were a problem, subjects could cover the pump with pillows or blankets to help muffle the sound, however, they were reminded not to cover the inlet to the sampler. Figure 2.2. Study subject wearing sampler and pump. 26 2.4.3.2 Lab preparation of particulate samplers P E M samplers were cleaned before every use. Daily cleanings consisted o f removing visible particles from the impactor plate with a razor blade and wiping down other sampler components with distilled/deionized water. More thorough cleaning of the samplers was performed on a weekly basis. Impactor and backing plates were sonicated for 15 minutes in soapy water and then rinsed three times with distilled/deionized water. Top and bottom pieces of the samplers were soaked in soapy water for 15 minutes, scrubbed and rinsed three times with distilled/deionized water. A l l components were allowed to dry completely. Samplers were loaded with pre-weighed fdters (see Section 2.5.1) in the lab before being taken to the field. Before, loading, impactor plates were saturated with mineral oil . Once all sampler components were assembled, adhesive labels were used to identify the filters in each sampler. Samplers were leak checked using the same criteria as for the Harvard Impactors and any leaks were investigated by examining the filter and o-rings and replacing them i f necessary. Airtight samplers were placed in small tupperware containers lined with kimwipes for transportation to and from the field. A t 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. 2.4.3.3 Information about particulate sources In order to gain an understanding about the sources of PM2.5, subjects were asked to fill out time-activity logs (Appendix 5) for each personal sampling session. The rows of the time-activity log represented 30-minute periods over the course of the sampling session and the columns represented various locations/activities with which the subjects may have been involved. For each half-hour block, subjects were asked to highlight all the columns that apply, regardless of actual duration in minutes. Locations included various types of indoor, outdoor, or transit locations. For the outdoor category, "near" was defined as being within one block from home. Subjects were also asked to assess their highest level of activity for each half-hour. Exposure to cooking and tobacco smoke was defined as being in the same room as the activity. The "locations" and "exposure" categories were not mutually exclusive since exposures to cooking or tobacco smoke would have occurred simultaneously with being present in a certain location. Subjects were also asked about inhaler use and i f they had worn the sampler over their shoulder. From this data, the percent of time subjects spent in each location throughout each sampling period was calculated by counting the number o f half-hour blocks highlighted for each location and dividing by the total number of blocks used for the whole sampling.). A dwelling questionnaire (Appendix 6) was used to obtain information about the participants' place o f residence and location within the city. This questionnaire was filled out once for each participant. These categories were also not mutually exclusive since, for example, a subject could be l iving in a home with carpeting and a gas heating system. 27 2.4.3.4 Electrocardiograms Cardiac function was measured throughout the 24-hour sampling period using Holter monitors (DM-400 cassette Holter recorder). These were battery-operated ambulatory electrocardiogram ( E C G ) recording devices, which recorded three independent channels of E C G . The recorder was connected through a seven wire bonded lead wire set attached to disposable silver/silver chloride self adhesive electrodes which were affixed to the subject. The Holter monitors were fitted with normal bias audiocassette tapes (Maxell UR60) to record signals. Tapes were labeled in a standardized fashion showing tape number, patient ID and date. To aid in carrying the instrument, as well as protecting it, the monitor was placed inside a pouch provided by the manufacturer. Subjects could then wear the monitor on their own belt, placed inside their pockets, or on belts/shoulder straps provided. At night, subjects placed the Holter monitors beside them in their beds. Subjects were asked not to bathe or shower during the sampling periods since doing so could negatively affect the monitors and the recording. 2.4.3.5 Blood Pressure Measurements Blood pressure measurements using a blood pressure cuff/stethoscope combination (Sprague/Rapapport) were conducted at the start and end of each sampling period. Measurements were taken from each subject's left arm, while seated with their elbow resting at mid chest level. With the cuff placed around the upper arm and the stethoscope placed on their brachial artery, the cuff was pumped full o f air until a pressure was created where no sounds from the artery could be heard. Ai r was slowly released from the cuff while checking the pressure gauge on the cuff. The systolic blood pressure was determined from the pressure gauge when the first sounds from the artery were heard through the stethoscope. As air was further released from the cuff, the sounds from the artery were monitored. The sounds usually became louder and then muffled as the pressure in the cuff was released. The diastolic blood pressure was determined from the pressure gauge when no more sounds could be heard through the stethoscope. For some subjects, the muffled sound remained even when pressure in the cuff was reduced to zero mm H g . In these cases, the diastolic pressure was taken to be zero mm H g . 2.4.3.6 Medication information In order to separate the effects of medications from any possible effects of particulate air pollution on the subjects' cardiac and respiratory health, each participant was asked to display their current medications during their first sampling session. The medications were recorded on a medication form including type, dose and prescribed frequency of use (Appendix 7). A t the conclusion of every sampling period, the medication form was discussed with the subject to determine which medications had been taken during the 24-hour sampling period and whether their medication list was still accurate. 2.4.3.7 Other Other health assessment measurements that were conducted during this study, including lung function (Tamarac Systems Corporation, portable spirometer) and the symptoms questionnaire (Appendix 8), are discussed in detail in Fisher, T. V . , M . Sc. Thesis, Experimental Medicine Program, University of British Columbia, 1999 (Fisher, 1999). 28 2.4.4 Field sample collection Sampling equipment and forms were distributed and collected at the homes of participants. The personal sampling periods were planned to start on Mondays, Tuesdays, Wednesdays and Thursdays between 7:00 am and 10:00 am. Stop times were scheduled 24-hours later. The sampling session was initiated by setting up the Holter monitors. Before hook-up, the L E D display was checked for correct clock and date settings and new electrodes were applied to each of the seven lead wires. To prepare the patient, the electrode placement sites on the chest area were located. I f hair was present, this was shaved off with a razor. Sites were then wiped with an alcohol prep to remove body oils and promote better electrode adhesion. These spots were allowed to dry before placing the electrodes on the prepared skin surface sites. Once the subject was prepared, a cassette tape was inserted into the monitor and the headbar latch was pushed to the "start" position. The start time was read off the monitor's L E D display and was recorded on the cassette label. A t this time, the monitor performed a battery test, after which it entered an eight-minute internal calibration mode before starting to record heart function. The wires were secured with medical tape to the subject's skin and he or she was asked to lie down on their back for five to ten minutes to obtain clean, resting heart rate data. During this time, the pulse oximeter was attached to the left index finger and readings were taken at 1 minute, 3 minutes and 5 minutes into the resting period. In addition, during the resting period, the pump was warmed up for two minutes before connecting a pre-loaded sampler. A flow check was conducted through the sampler with a rotameter, and the flow was adjusted i f necessary before starting the exposure sample. F low rate and start time was recorded on a personal flow log sheet (Appendix 9). Following the resting period, the patient's blood pressure was recorded and the lung function tests were performed. Finally, subjects were asked i f a bronchodilator had been used that morning and i f so, the time o f that use was recorded on the activity log. Holter monitors and the personal exposure samplers/pumps were worn by the subjects for the duration of the sampling. Time-activity logs were also completed over the session. Subjects were provided with the telephone numbers of the technicians in case they experienced any problems with the equipment or sampling session. The numbers were used to report pumps that had stopped working. I f it was not feasible to visit the subject in their home at that time, the subject was instructed on how to turn the pump back on. The subject was instructed to remove the pump from the bag, and note the number on the display, indicating the length o f time that the pump had been operating. The pump was turned on by the subject and this new start time was recorded over the telephone. The sampling session was concluded after 24-hours. The subject was asked lie down for another 5-minute resting period and the three pulse oximeter readings were taken. A t this time, the exposure sample was stopped, recording stop time and flows. The sampler was disconnected from the pump and placed back into a tupperware dish for transport back to the lab. Blood pressure was measured directly after the resting period. Subsequently, the Holter monitor was stopped by taking out the cassette tape and recording the stop time on the tape. The seven lead wires were disconnected from the gel pads and the gel pads were removed from the subject. 29 Sticky residue left on their skin was cleaned off with alcohol wipes. The lung function tests were performed once the Holter monitor had been taken down. As a rest in between blows, subjects were asked questions from the symptom questionnaire. Before leaving, the time-activity log and medication form were discussed with the subjects. 2.5 Gravimetric Analysis 2.5.1 Filter weighing procedures Before weighing new fdters, filter packages were opened and allowed to equilibrate for at least 48 hours in a temperature/humidity controlled weighing room. Throughout the study, weighing room conditions were monitored and maintained at a constant temperature of 22 °C (SD: 0.67) and a relative humidity level of 53% (SD: 6). Filters were pre-weighed in triplicate using a micro-balance (Sartorius M3P; lug resolution, + 2ug sensitivity). Filters were weighed until three consecutive weighings were within 10 ug of each other. Pre-weighed filters were placed individually into labeled petri plates. These were stored in the weighing room until needed for sample collection, at which time they were taken to the lab for loading into samplers or ambient filter storage boxes. After sample collection, the filters were unloaded from their samplers or ambient filter storage boxes and placed back into their original petri plates. The plates were returned to the weighing room for equilibration of at least 48 hours. The filters were then post-weighed in triplicate using the micro-balance (until three consecutive weighings were within 10 ug of each other). 2.5.2 Quality Control filters As a control procedure for checking the accuracy of the filter weighing, three unused filters for each filter size were kept throughout the study as Quality Control (QC) filters. The QC filters were weighed prior to each weighing session. Before proceeding with weighing of study filters, the QC filter weights were checked against their respective QC charts for accuracy. QC charts for each filter were constructed displaying the mean, warning (mean + 2 SD) and control (mean ± 3 SD) limits of all previous weighings. These charts were updated every few weeks with the accumulated weights (Appendix 10). 2.5.3 Lab blanks and field blanks For both personal and ambient filters, 10% of each were planned to be field blanks. For the ambient filters, field blanks were the spare filter that was sent out with the batches of filters for each site every week. Field blanks of personal measurements were prepared by assembling filters in a personal impactor and carrying it, together with the other loaded samplers, in the sealed plastic containers to the field. One to two personal field blanks were prepared each week. 30 Lab blanks were also planned to make up 10% of the total filters used. For the ambient filters, three lab blanks were weighed each week together with the sample filters needed that week. Since the ambient lab blanks were given a separate labeling system from the sampler filters, they were designated as such from the pre-weighing stage. For the personal filters, lab blanks were chosen from the pre-weighed filter set at the same time that field blanks were prepared. 2.6 Sulfate Analysis Sulfate analysis was conducted on all personal and ambient PM2.5 filter samples. The analysis, being a destructive procedure, was not conducted until the preliminary PM2.5 results were studied. All 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 different filters. Filters were taken from the petri plates and placed onto a glass plate using forceps. A razor blade was used to make six to eight cuts into the plastic rim of each filter, enough to allow the filters to fold easily. Filters were then placed into new, labeled, plastic screw-top containers. In the containers, 100 ul ethanol was added to completely wet each filter using a micropipette. Five ml distilled/deionized water was then added and the lid to the container was closed. Containers were sonicated for 15 minutes to bring the sulfate into solution. Ion chromatography (IC) was used to analyze the amount of sulfate in each sample. Syringes (attached with filters to prohibit the transfer of large particles) were used to transfer approximately 0.5 ml 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 (Dionex, DX-300). Samples were run within 24-72 hours of being put into solution, in batches of 50 to 70 samples. A set of five IC standards and blanks were run before and after each batch to generate a calibration curve for each run. Dispersed throughout the batch, after every ninth sample, one IC standard and one blank were placed. 2.7 Data Analysis Excel 7.0 was used to create the initial database for data. SPSS Versions 6.0 and 9.0 were used for all statistical analyses. 2.7.1 Particulate sampling - data quality and descriptive statistics Personal and ambient concentration samples taken during the study were reviewed to verify correct times and flows. Some samples were deemed invalid and excluded from further analysis 31 based on pre- or post-sampling flows outside of 4.0 L/min ± 10% (i.e. 3.6-4.4 L/min) or sampling times of less than 20 hours. The hourly ambient PMio data obtained from the GVRD Air Quality department was averaged between 9 am (pre-sample) and 9 am (post-sample) resulting in 24-hour averaged data. For PM2.5 measurements, mean field blank weight changes were subtracted from all sample weights prior to calculating concentrations. For sulfate, mean field blank weight changes were only relevant to specific batches of samples, which were subtracted accordingly. The resulting personal exposure data was summarized both by subject and by pooling over all subjects. Ambient concentrations were summarized for each site and summarized by averaging over all sites on each day. 2.7.2 Relationship between personal exposures and ambient concentrations In analyses relating personal exposures (P) and ambient concentrations (A), the ambient concentrations were expressed either as an average of the five values obtained for each day of personal sampling, or the concentration obtained at the site closest to each subject's home. For eight personal samples, the corresponding closest site had missing data thus the second closest site for these samples were chosen. The correlation between P and A was assessed by means of individual subject regression analyses. The distribution of the individual correlation coefficients was not normal, thus the medians are presented. Regression analyses on pooled data were also conducted. These relationships were assessed visually by means of scatter plots between personal and ambient concentrations per subject as well as for the group. Outlying points were compared with time-activity information and field notes to determine whether their removal would be justified. Autocorrelation between personal exposures due to repeated measurements was tested by visual examination of individual subject autocorrelation function plots. The concentration difference between P and A was determined by P:A ratios and P-A differences. Mean ratios and differences for each person were tabulated, as well as the mean ratio and differences of all samples after they were pooled. Analyses were also conducted to test whether the differences and correlations between P and A were a function of personal or ambient concentration. Finally to determine the usefulness of using ambient data from sites closest to subjects' homes, the spatial variability of particulates over the study region was assessed by Analysis of Variance (ANOVA) and various correlations between the types of ambient data. Also, it was tested whether the P vs. A correlations were a function of distance between the subjects' place of residence and the site. 2.7.3 Predictors of personal exposure The data collected with the time-activity logs and dwelling questionnaire was used to examine which activities or living characteristics, in addition to ambient concentrations, were predictive of personal exposure. Pooled personal PM2.5 and sulfate data were used as the dependent variables in multiple regression analyses. 32 In order to reduce the number of potential independent variables, some categories in the time-activity logs and dwelling questionnaire were collapsed. Of the categories shown on the time-activity log, 'near home' and 'away from home' were combined into one variable 'outdoors'. In addition, 'car' and 'other transit' were combined under one common 'transit' heading. Thus, from nine original time-activity categories, seven variables remained to be analyzed for inclusion in the multiple regression models. For the dwelling questionnaire, some categories did not have enough data and variability to be useful in a regression analysis, such as 'ventilation system' or 'independent air filter'. Others, such as 'heating system' or 'fireplace', were not considered important at predicting exposures since sampling was conducted in the summer, when these would not be in operation. These four variables were left out of the multiple regression. To make use of the 'carpeting' and 'open windows' categories, two variables were created by summing over all rooms in the living space, the extent of carpeting and open windows to give each subject a carpet and a window score. The extent of carpeting in each room was originally coded with 0=no carpet/rugs, l=rugs, 2=carpet. Since not all subjects had the same number and types of rooms in their home, only the types of rooms all subjects had were used in the scoring system. Thus, 'hallways' and 'other' were left out, and the categories for living room/dining room and den were combined into one column, labeled 'living'. Therefore, the 'carpet score' was determined by the extent of carpeting in the bedrooms, bathrooms, kitchen and living room. The 'window score' was determined by the use of windows in the bedrooms, kitchen, living room and by the patio or entrance door. One 'range hood' variable, to describe the use of ventilation during cooking, was created where extent of use was assigned by codes (0=never used, l=sometimes, 2=always). Under 'building type', the house and townhouse categories were combined and coded with a 0, whereas the apartment category was coded with a 1. The remaining variables, 'distance from major road', 'volume of home' and 'number of rooms', were continuous, thus did not require coding. Seven dwelling variables were considered for inclusion in the multiple regression models. Table 2.2 summarizes all variables, other than ambient particulate concentration*, that were considered for inclusion in the multiple regression models. Since time-activity logs were completed for each personal sampling session, each time-activity variable had seven samples per subject. The dwelling variables did not change over time thus only one sample per subject was obtained for each. All time-activity variables were continuous, representing the mean percentage of time subjects spent in the given environments during the sampling period; dwelling variables were either categorical or continuous, depending on the type of variable. In the regressions, the values of these variables were repeated for each of the samples per subject. As described in Section 2.4.3.3, the time-activity and dwelling variables were not mutually exclusive. No time-activity or dwelling variables were normally distributed. Natural log (In) transformations did not improve any variables' fit of normality; thus, all were left untransformed. *For all PM2.5 and sulfate multiple regressions, the ambient concentrations were taken as the average concentration obtained over the five sites for each sampling day. 33 Time-Activity Variables (7 samples/subject) Dwelling Variables (1 sample/subject) Home Distance from major road (continuous) Restaurant Building type (categorical) Other indoors Volume of home (continuous) Outdoors Number of rooms (continuous) Transit Carpet score (continuous) Cooking Window score (continuous) Tobacco smoke = ETS Range hood use (categorical) Table 2.2. Variables considered for inclusion in the multiple regression models. Before developing a multiple regression model, each continuous independent variable was regressed with the dependent variable (personal exposure), to determine which variables may be important in predicting exposure. The two potential categorical variables, 'building type' and 'range hood use', were assessed visually using boxplots of personal exposure vs. each variable. Secondly, correlations between all independent variables were conducted in order to eliminate two highly correlated variables from being included in the multiple regression models. Finally, the distributions of all potential variables were studied to confirm sufficient variability for use in a regression analysis. From these comparisons, core variables were chosen for further analysis (five for PM2.5; one for sulfate). Some variables were chosen as 'additional variables' which would be added in to see their effect on the regressions. Core variables were forced into Ordinary Least Squares (OLS) regression models. The 'additional variables' were added to the core models individually to test for improvement in model fit. Regression diagnostics were analyzed to verify results of the models. These included studying the magnitude of coefficients, standard errors of coefficients and the p-values. Unstandardized residuals were plotted against predicted values to check for equal variances. Variance per subject was estimated by calculating the residual sum of squares (RSS) per subject divided by their number of samples (n). The ranges of RSS/n values over the study population were analyzed. Unequal variances in final models were dealt with by running a Weighted Least Squares (WLS) regression, using the inverse of the estimated variance (1/a2) as the weight factor. For PM2.5, additional regressions were attempted to analyze the effects of interactions between variables (i.e. between 'Home' and 'Ambient PM2.5') and excluding variables from the core model (i.e. 'ETS'). 2.7.4 Assessment of cardiovascular health effects Analysis of blood pressure, arrhythmias, heart rate (HR) and heart rate variability (HRV) was used to assess cardiovascular function of subjects over the study period. The HR and HRV data was obtained by Holter recording and analysis (24 hour ambulatory ECG). Data recorded on the ambulatory ECG cassette tapes was analyzed on a Biomedical Systems Century Advanced Holter System using superimposition and page mode. A certified Holter scanning technologist sorted the beats into various templates (classifications) (i.e. normal, 34 supraventricular, ventricular, paced, artifact, etc.). Only normal R wave to R wave (R-R) intervals were used to determine the HR and HRV values. Some problems were experienced when scanning the cassette tapes. First, the Holter System did not recognize the end of the recording in some cases, causing elongation of the last recorded beat. This affected the 24-hour summary variables for that recording, such as creating very high SDNN values. This problem was overcome by rescanning the affected tapes and ending the play back approximately 5 minutes before the end of the true recording. Secondly, during the middle of the study, the second sides of previously used tapes were used for a number of recordings (N=37). In some cases, tape stretch from previous use, led to high levels of background noise, which made it difficult for the Holter System to distinguish true beats. Full summary reports were obtained for 21 of these recordings. In an attempt to recover data from the other 16 recordings, random strips (~ N=80) over each 24-hour sampling period were printed for use in manual calculations of HRV variables. A feasibility test for the use of manually calculated data was conducted by examining 10 strips over one hour of normal data to the computer summary for that hour. Hand-calculated HRV variables in the feasibility test did not match results of the computer. Therefore, only full report data was used in the data analysis and the 16 recordings with excessive background noise were excluded. All potential BP and ECG variables were entered into the database. These are listed in Table 2.3 and Table 2.4. Histograms of results grouped over all subjects and boxplots for each subject were plotted to check the variability within each measure and the feasibility of using the variables in models against exposure. At this point a number of variables were excluded due to lack of sufficient variability within and between subjects including VE, VT, SVT, atrial fibrillation, bradycardia and pNN50. Remaining variables were assessed for normality. Most variables were not normal but not better represented when ln-transformed. A ln-transformation did represent the SVE variable better; thus the ln-transformed SVE variable (denoted as In SVE for the remainder of the thesis) was used in all analyses. Correlations between the BP variables and between the remaining ECG variables were conducted in order to reduce the number of potential variables by eliminating two highly correlated variables from being used. Correlations between the four BP variables indicated pre-and post-sample results were highly correlated for both systolic (r=0.84) and diastolic BP (r=0.92). Post-sample metrics were chosen for further analysis. Correlations between ECG variables demonstrated all nighttime variables (data recorded from 11 p.m. to 6 a.m.) were highly correlated with their corresponding 24-hour summary values. We decided to focus on the 24Thour variables since for time-domain methods of analysis, 24-hour recordings are thought to be more appropriate than recordings of shorter duration (Task Force, 1996). In addition, the 24-hour values were thought to provide a more accurate health summary for each sample in association with the 24-hour exposure metrics. On this basis, the 24-hour average HR variable was also chosen over the single 24-hour minimum and maximum HR values. 35 Correlations between variables within the categories of Table 2.4 were more highly correlated than across categories. Average HR and mean R-R variables were highly correlated (r>0.93) and SDNN, SDANN and the triangle index were correlated (r>0.82). SDNN5 was related to SDNN (r=0.75) and to r-MSSD (r=0.84) whereas the ln-transformed arrhythmia variable, In SVE, was not correlated with any of the HR or HRV variables. We decided to analyze one variable from each category. For comparability to previous studies, average HR was chosen as the basic rhythm variable and SDNN, derived from direct measurements between normal R-R intervals, was chosen as the variable for assessing overall HRV. R-MSSD, derived from differences between normal R-R intervals, was chosen as a measure of short-term HRV components. Finally, In SVE was the chosen arrhythmia variable. These variables were summarized both by subject and by pooling over all subjects. Subjects' results were then standardized by taking the individual means for each subject over all sampling sessions, and calculating deviations from subjects' mean values for each sampling session. The standardized values were summarized and used in analyses that follow. Relationships between the various ambient and personal exposure indices and each health outcome were initially assessed by OLS regression models. Plots of exposure vs. outcome for all samples pooled aided in the identification of outliers. These were rare cases with extremely large deviations from subjects' mean values, which were greater than 2-4 standard deviations above or below the pooled mean of the variable in question. Information from field notes and ECG reports justified the removal of some of these cases and OLS regression analyses were subsequently repeated excluding outliers. Correlation structures of the pooled OLS regressions were assessed to determine the presence of autocorrelation between sampling sessions due to repeated measurements of subjects. This was conducted by organizing unstandardized residuals for each relationship by subject and sampling session. Correlations between individual sampling sessions one apart to six apart were assessed. Pooled correlations of all one apart sampling sessions to all five apart sampling sessions were also conducted. Consistent trends in these correlations would suggest the presence of autocorrelation in the regression. Regression diagnostics were analyzed to verify results of the OLS models. These included studying the magnitude of coefficients, standard errors of coefficients and the p-values. Unstandardized residuals were assessed by plotting boxplots for each subject. The variance for each subject was estimated by calculating RSS/n. The ranges of RSS/n values over the study population were analyzed. Unequal variances in final models were dealt with by running a WLS regression, using the inverse of the estimated variance (1/a2) as the weight factor. The resulting WLS regressions were further analyzed for potential confounding by several meteorological and co-pollutant variables as well as bronchodilator use. Hourly ambient temperature (T), relative humidity (RH), carbon monoxide (CO) and ozone (O3) data, collected over the study period by the GVRD Air Quality Department, were obtained. From this data, daily 24-hour averages (9 am to 9 am) were calculated for T, RH and CO. For O3, the average of the maximum hourly recordings over each 24-hour period was evaluated. Bronchodilator use was obtained from subjects' time-activity logs in the form of count data. Evaluating deviations 36 from subjects' mean use over the study standardized these values. After assessing the correlation between particulate exposure metrics and the potential confounders, two-variable models were used to assess the effect of these secondary variables on the relationships between particulates and health effects. Blood Pressure Pre-sample systolic BP diastolic BP Post-sample systolic BP diastolic BP Table 2.3. Potential blood pressure variables for analysis. Category ECG Variable Measurement types Arrhythmias Ventricular Ectopy (VE) Total beats Average/hour Average/1000 Ventricular Tachycardia (VT.) Runs Beats Supraventricular Ectopy (SVE) Total beats Average/hour Average/1000 Supraventricular Tachycardia (SVT) Runs Beats Atrial fibrillation Beats % Bradycardia Events Beats Basic rhythm Heart Rate (HR) Average Minimum Maximum Mean R-R 24-hour night only HRV (derived from direct measurements of normal R-R intervals) SDNN 24-hour night only SDANN 24-hour SDNN5 24-hour TRIA 24-hour night only HRV (derived from the differences between intervals) r-MSSD 24-hour night only pNN50 24-hour night only Table 2.4. Potential E C G variables for analysis. 37 CHAPTER 3: RESULTS 3.1 Study Population 3.1.1 Recruitment Through the respiratory physicians at Vancouver Hospital and Health Sciences Centre (VHHSC), names of 35 interested individuals were obtained. The B.C. Lung Association Puffers' Clubs yielded a further 12 interested individuals, resulting in 47 potential candidates. Forty-five individuals were contacted by UBC technicians via telephone. The remaining two individuals could not be reached. After telephone calls and introductory meetings, 21 individuals had decided not to participate in the study. Six individuals were not interested, five indicated the project would require too much effort, three individuals were too busy to take part, one had family problems and six did not participate for other reasons. A further seven individuals were excluded due to study eligibility criteria (four individuals had severe COPD, three were living outside the study area and one person was living with a smoker). Thus, 17 subjects participated in the study, yielding a 36% participation rate. Thirteen of these individuals were recruited through the VHHSC clinic and four individuals were from the Puffers' Club. Participation through the Puffers' Club would have been increased, had more individuals met the study eligibility criteria. One woman participated in the study although it was not clear at the time of data collection whether she in fact had been diagnosed with COPD. At the completion of the study, respiratory physicians were asked to review her forced expiratory maneuver data. It was determined that this participant did not meet the criteria for COPD and this individual was completely excluded from the data analysis. Therefore, the total number of subjects that participated and yielded eligible data was 16, lower than our target population of 25-30. 3.1.2 Characteristics The resulting study population consisted of seven male and nine female subjects, with ages between 54 and 86 years (mean age: 74). All subjects were residents of the Greater Vancouver Regional District (GVRD). All subjects were current non-smokers, and did not live with smokers with the exception of one subject. For this subject, the smoker agreed to smoke outdoors or not at all on each of the subject's sampling days. In addition, all subjects had moderate COPD as defined by FEVi > 0.75L. The COPD condition for 13 subjects was identified by physicians at the VHHSC clinic. Three subjects that were not referred from VHHSC had indicated that they had been diagnosed with COPD and their spirometry data was reviewed by a VHHSC respiratory physician to confirm moderate COPD. Subjects used various medications throughout the study. All 16 subjects used at least one bronchodilator (Salbutamol™, Berotec™, Terbutaline™, Serevent™, Combivent™, Atrovent™) 38 during at least one sampling session (13 subjects used the medication on all seven sampling days). Five subjects used a xanthine bronchodilator (Theophyline™, Theodur™, Uniphyl™, Choledyl™) at least once. Fourteen subjects used at least one steroid medication (Pulmicort™, Becloforte™, Flovent™, prednisone, beclomethasone) during the study. Four subjects used cardiac medication (digoxin, sotalol, nitro dur patch), three subjects used Aspirin and four used Tylenol™ on a regular basis. 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. However, as discussed below, not all personal exposure samples were valid, resulting in less than seven exposure samples for some individuals. The sampling sessions were randomly spaced, with at least 1.5 weeks between consecutive sessions. In most cases, subjects followed the random sampling plan initially assigned to them. When scheduling conflicts arose, sampling days were changed, however consecutive samples for subjects were kept at least 1.5 weeks apart. 3.2 Particulate Sampling - Data Quality 3.2.1 Data clean-up All 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. For the five ambient sites, 413 samples were collected over the course of the study. Of these, 26 samples were deleted due to invalid flows or times. Thus, 387 (94%) of the ambient samples were successfully collected. 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 min.; range 21:43-25:50 hours). For the 16 COPD subjects, 112 personal measurements were taken. Six of these measurements were deleted due to invalid flows or times, leaving 106 (95%) valid samples. 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). The mean personal sample duration was 23:47 hours (SD: 50 min.; range 20:37-25:10 hours). Samples were started and stopped between 8:00 am and 11:00 am to aim for 24-hour sample durations. The mean start/stop time for the ambient measurements was 8:29 am (SD: 36 min.; range 6:37 start-10:12 stop). The mean start/stop time for the personal measurements was 9:35 am (SD: 60 min.; range 6:49 start-11:40 stop). To determine the amount of time the personal 39 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. The mean percent overlap for the personal samples with ambient samples was 95% (range 85-100%). Therefore, on average, 71 minutes of the personal sampling did not overlap with the ambient samples. 3.2.2 Quality control, lab blanks, field blanks and limits of detection Throughout the study, ten percent of all ambient filters were lab blanks (N=63) and 24% were field blanks (N=147). For the personal filters, 11% were lab blanks (N=15) and 10% were field blanks (N=14). Mean mass differences were calculated for all lab and field blank filters. For PM 2 5 , both ambient lab blanks and field blanks increased by an average of 3 ug (SD: 8). Personal lab blanks increased by 2 ug (sd: 7) on average and personal field blanks increased by 16 ug (SD: 7). The mean mass increase on personal and ambient field blanks was subtracted from all personal and ambient samples respectively. The detection limit was defined as three times the standard deviation of the field blanks divided by the mean sample volume (5.8 m3 for ambient and 5.7 m3 for personal samples). Using the standard deviation of the mean mass increase on ambient field blanks, the ambient limit of detection was 4.2 ug/m3. The personal field blanks resulted in a personal sample detection limit of 3.7 ug/m3. Seven ambient measurements (1.8%) and two personal measurements (1.9%) were below their respective detection limits. These were flagged and kept in the data analysis. Table 3.1 summarizes the warning and control limits of the six quality control filters that were weighed throughout the pre-study experiments and throughout the study. Control charts for each filter, displaying the mean, warning (mean ± 2 SD) and control limits (mean ± 3 SD) of all weighings over time, are attached in Appendix 10. The quality control filters did not show a systematic increase over time, thus the systematic increases observed for the lab and field blanks were associated with sample handling as opposed to laboratory conditions and weighing procedures. Quality Control Filter Warning Limit (ug) Control Limit (ug) Ambient 1 ±9 ± 13 Ambient 2 ± 12 ± 18 Ambient 3 ±22 ±33 Personal 1 ±8 ± 11 Personal 2 ±8 ±13 Personal 3 ±6 ± 9 Table 3.1. Warning and control limits for quality control fdters. 40 For sulfate, ambient lab and field blank values differed from non-detectable on an analytical run-specific basis. For analytical runs in which sulfate was detectable on blank filters, the amounts of sulfate were similar between runs. On average for these runs, ambient lab and field blanks both increased by an average of 0.3 ug. A l l runs of personal filters had sulfate detectable on blank filters. Personal lab and field blanks both increased by 0.3 ug, similar to the ambient blanks. For ambient and personal runs with detectable sulfate on blank samples, the respective mean mass increase o f field blanks was subtracted. The limit o f detection for sulfate was determined using Vi the area under the IC chromatogram for the lowest standard samples from each analytical run. In terms of air concentration, using a 5.8 m 3 air volume, the detection limit was 0.2 ug/m 3. N o ambient samples fell under this limit while two personal samples (1.9%) did. 3.3 PM?s and Sulfate Concentration, Time-Activity and Dwelling Data 3.3.1 Personal exposures Five to seven valid exposure measurements were obtained for each subject. Not all subjects had seven valid samples since some samples were excluded due to incorrect times or flows as discussed above. Individual distributions for PM 2 .s and sulfate were not normally distributed but were not represented better when the data was ln-transformed. For descriptive purposes and consistency, both transformed and non-transformed summary results are given in Table 3.2 and Table 3.3. The boxplots that follow are visual representations of the same data (Figure 3.1 and Figure 3.2). For each boxplot in this chapter, the lower boundary of the box represents the 25 t h percentile and the upper boundary is the 75 t h percentile, thus the box-length corresponds to the interquartile range. The horizontal line inside each box represents the median. 41 PM 2 . 5 Concentrations (ug/m3) Subject N A M SD Range G M GSD 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 2.2-11.7 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 3.2. Personal PM2.s exposure summary; A M : Arithmetic Mean, SD: Standard Deviation, G M : Geometric Mean, GSD:Gcomctric Standard Deviation. Sulfate Concentrations (ug/m3) Subject N A M SD Range G M GSD 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 Table 3.3. Personal sulfate exposure summary; A M : Arithmetic Mean, SD: Standard Deviation, G M : Geometric Mean, GSD: Geometric Standard Deviation. 42 Figure 3.1. Personal P M 2 | 5 exposures; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 4 D> 3« Z3 s 2' (/) 1-" r o c o 0' Q. -1 N 7 7 7 7 6 7 7 7 7 7 6 6 7 7 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.2. Personal sulfate exposures; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 43 3.3.2 Ambient Concentrations The number of PM2.5 samples taken at each of the five sites ranged from 65 to 84, with 90 days of sampling in total (including Mondays-Thursdays only). For 32 of these days, dispersed throughout the study period, there were no personal samples taken. PMio data was available from the five ambient sites for each of the 90 days, with the exception of North Delta where only 69 measurements were obtained. The distributions of P M 2 5 and sulfate ambient data were not normally distributed but were not represented better when the data was ln-transformed. Ambient PMio data was approximated by a ln-normal distribution. For consistency, both arithmetic and geometric summary results are presented in Table 3.5, Table 3.6 and Table 3.7 on the following pages. Boxplots of this data are shown in Figure 3.3, Figure 3.4, and Figure 3.5. Two-way Analysis of Variance ( A N O V A ) was performed to test for differences in concentrations between the ambient sites while controlling for overall temporal variation. Site and date were fixed factors and no interaction term was used. The A N O V A s for all exposure metrics indicated significant concentration differences between sites (p<0.001). Including multiple comparison procedures in the analysis, the Least Squares Difference (LSD) test found significant differences (p<0.01) between the South Burnaby and South Richmond sites and all other sites for PMio and sulfate. Lastly, for PM2.5, the L S D test indicated only the South Richmond site to be significantly different from the others. Therefore, overall, the concentrations from the Kitsilano, North Delta and North Burnaby sites were not significantly different from each other. A s a validation of our Harvard Impactor (HI) ambient sampling results for P M 2 5 and sulfate, the Kitsilano site H I and Dichotomous Sampler (Dichot) data were compared. The Dichot collected daily 24-hour fine mass ( P M 2 5 ) samples, which were also analyzed for sulfate every 6 t h day by the National A i r Quality Monitoring Network. The mean, standard deviation and range of measurements for each sampler are presented below in Table 3.4. In general, slightly lower levels were measured by the Dichot although correlations were high between the two samplers for both PM2.5 (r=0.81) and sulfate (r=0.92). Exposure Measure N Concentrations (ug/m3) Mean SD Range Dichot PM2.5 90 7.8 4.2 2 . 1 - 2 6 . 6 H I P M 2 . 5 84 11.9 4.7 2 .3 -29 .3 Dichot Sulfate 23 1.8 1.1 0 . 5 - 4 . 7 H I Sulfate 84 1.9 1.0 0 .4 -5 .3 Table 3.4. Kitsilano site ambient concentrations. 44 PMio Concentrations (uj ?/m3) Site N A M SD Range G M GSD Kitsilano 90 18 8 7-45 16 1 North Delta 69 18 8 5-56 17 1 North Burnaby 90 18 8 6-53 17 2 South Burnaby 90 20 10 6 - 56 18 2 South Richmond 90 15 6 5-45 14 1 Daily Averages* 90 18 7 6-51 16 1 Table 3.5. Ambient P M 1 0 concentration summary; * the average of all five sites over each sampling day, A M : Arithmetic Mean, SD: Standard Deviation, G M : Geometric Mean, GSD: Geometric Standard Deviation. PM2.s Concentrations ( n e/m3) Site N A M SD Range G M GSD 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 Daily Averages* 90 11.4 4.1 4.2-28.7 10.8 1.4 Table 3.6. Ambient P M 2 5 concentration summary; * the average of all five sites over each sampling day, A M : Arithmetic Mean, SD: Standard Deviation, G M : Geometric Mean, GSD: Geometric Standard Deviation. Sulfate Concentrations (ug/m3) Site N A M SD Range G M GSD 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 Daily Averages* 90 1.9 0.9 0.4-5.4 1.7 1.7 Table 3.7. Ambient sulfate concentration summary; * the average of all five sites over each sampling day, A M : Arithmetic Mean, SD: Standard Deviation, G M : Geometric Mean, GSD: Geometric Standard Deviation. 45 60 50 N= 90 69 90 90 90 KT ND NB SB SR AVERAGE Ambient Site Figure 3.3. Ambient P M 1 0 concentrations per site and averaged over all sites; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). CD in Csi 90 AVERAGE Ambient Site Figure 3.4. Ambient P M 2 . 5 concentration per site and averaged over all sites; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box). 46 Figure 3.5. Ambient sulfate concentrations per site and averaged over all sites; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box). Percent time spent per day (minutes) AM SD Range Location Indoors Home 92.4 (1299) 7.3 (111) 73.5- 100.0 (1020- 1560) Restaurant 0.4 (5) 1.3 (17) 0.0-6.5 (0-90) Other 1.9(33) 4.3 (67) 0.0-26.1 (0-360) Outdoors Near* 6.0 (80) 7.7 (104) 0.0-47.7(0-630) Away** 3.0 (46) 4.4 (65) 0.0-21.7 (0-300) Transit Car 2.2 (35) 3.8(58) 0.0-14.9(0-210) Bus 0.0 (0) 0.0 (0) N/A Other o.KD 0.4 (6) 0.0-2.2 (0-30) Exposure to5... Cooking 5.8 (85) 4.4 (60) 0.0- 17.0 (0-240) ETS 0.6 (9) 2.0 (28) 0.0-11.6 (0-150) Table 3.8. Time-Activity Summary; * within one block of home, ** further than one block from home, § in the same room as...; NOTE - minutes are not standardized to 24 hours, thus some values are greater than 1440 minutes; from the time activity form, it is not possible to distinguish between lengths of time < 30 minutes. 47 3.3.3 Time-activity information One hundred and twelve time-activity logs were collected throughout the study. A summary of the variables considered for analysis in conjunction with exposure (see Section 3.4.6) is presented in Table 3.8 (i.e. 'activity level' and 'inhaler use' have not been included). The values represent the mean percentage of time subjects spent in the given environments during the sampling periods. The arithmetic means add to greater than 100% of time since the categories were not mutually exclusive. For example, exposure to cooking or to ETS would have occurred at the same time as being in a certain location, such as indoors at home. Secondly, a subject may have been both at home and in a restaurant within the same half-hour block. The 'walk' category for Transit was not included as this corresponded to time spent in an outdoor location. Subjects spent the majority of their time indoors at home (92.4% of time). Information regarding percentage of time subjects wore the PEM/pump was extracted from the 'wearing sampler' column on the time-activity logs. On average, subjects wore the PEM/pump 42% of the time. If subjects did not wear the pump, it was placed next to them. Due to sleeping (approximately 8 hours/day) the maximum they could have worn the pump was 67% of the time. 3.3.4 Dwelling characteristics One dwelling characteristics questionnaire was collected from each participant, the results of which are summarized below. Only two people lived on a major road (defined as a road with at least 4 lanes). The average distance to the nearest major road from subjects' homes was 175 meters as estimated from a map of the GVRD region. Most subjects lived in a house (7) or an apartment (7); two subjects lived in a townhouse. The average volume of living space was 389 m3 with an average of eight rooms (range: 4-15). Single houses had between two and three floors (including basement) and townhouses had between one and two floors. Apartment buildings were 11-24 stories high and subjects' apartments were located between the second and 23rd floors of the building. Four of the seven suites were corner units. Carpeting was found mostly in the bedrooms, living and dining areas and hallways of subjects' homes (14-15 subjects). Less carpeting was found in bathrooms and kitchens (2-3 subjects). Twelve subjects had kitchens with range hoods, however these were not always used when cooking. Seven subjects always used their range hood, four sometimes used it and one subject never used it. The most common type of heating system in subjects' homes was hot water (8), with other heating systems being gas (4), furnace (2) and electric (2). Only three subjects had an air filter in their home, which in all cases was connected to the heating system, thus would only be in use at the time the furnace was operating. Ten homes were equipped with fireplaces. There were five wood fireplaces and five gas fireplaces in total. These were used by all subjects during the winter only and therefore were not operated during the study period. The majority of the dwellings relied on natural ventilation for air exchange; only one house had a fan and one house had air conditioning. For summer, all but one of the subjects had at least one window open always. Bedroom and balcony windows/doors were the most likely windows to be opened. 48 3.4 Part I - Relationship Between Personal Exposures and Ambient Concentrations 3.4.1 Correlation between personal and ambient concentrations To test our hypothesis of exposure misclassification when using ambient measurements to predict personal exposures and to compare the usefulness of PM and sulfate as exposure metrics, regression analyses between personal and ambient measures were conducted. A summary of regression analyses between ambient PMio, PM 2 5 and sulfate and personal PM 2 5 and sulfate is presented in Table 3.9. For each of the six relationships, results from pooling the data and use of individual regressions are given. In addition, regressions were performed using either average ambient data and using the closest monitor to each subject. Overall, personal sulfate was more highly correlated with all ambient measures than was PM2.s, which is shown visually in Figure 3.6. For all relationships, use of pooled data in the regression analyses resulted in lower correlations compared to the median of the individual regressions (shown in Figure 3.7), suggesting that cross-sectional studies (where a single exposure estimate is used to characterize population exposure) result in greater exposure misclassification than time-series studies (where it is assumed that ambient concentrations and exposures are highly correlated). In these pooled analyses, samples were assumed to be independent. Autocorrelation due to repeated measures was thought to be minimal due to spacing of measurements 1.5 weeks apart. Additionally, visual examination of individual subject autocorrelation function plots demonstrated no consistent pattern in correlations between sampling periods. The level of exposure of the different individuals, which may also contribute to non-independence, was considered by dealing with outliers. For example, one individual was exposed to ETS much more than other study subjects. This individual also had higher exposures, however removing these values from the analysis did not largely affect the pooled correlation results. Results from individual regression analyses are presented in Table 3.10 and in the preceding histograms, Figure 3.8 and Figure 3.9. The median Pearson's r between personal and average ambient PM2.5 concentrations over time was 0.48 (range: -0.68 to 0.83). Using sulfate as the exposure metric, the median Pearson's r between personal and average ambient concentrations over time was 0.96 (range: 0.66 to 1.00). Since Pearson's correlations are susceptible to being influenced by outlying data points, Spearman rank correlations were also conducted. Similar results were obtained: the median Spearman correlation for PM2 5 was 0.59 (range: -0.90-0.89) and the median Spearman correlation for sulfate was 0.89 (range: 0.57-1.00). The histograms demonstrate a large variability in results between subjects for PM2.5, whereas the correlation results for sulfate are much more consistent across subjects. The moderate, variable correlation for PM2.5 compared to the high, stable correlation for sulfate suggests that ambient sulfate concentrations represent personal exposures better than ambient PM 2 5 . Ambient concentrations from sites closest to each subject's residence were tested as a predictor of personal exposure in comparison to data averaged over the five ambient sites. Use of the closest ambient site did not improve the pooled or median correlations for either PM2.5 or sulfate (shown in Figure 3.10). Nine out of the 16 individual correlations decreased when using the closest site 49 for PM2.5 and the resulting median correlation decreased to 0.30 (range: -0.51 to 0.88). Using the closest site for sulfate resulted in the same median value of 0.96 (range: 0.62 to 0.99). These data indicate that use of ambient data from sites closest to each subject's home does not provide a better estimation of personal exposure than averaging ambient data from multiple monitors dispersed throughout the study region. This relationship was analyzed further in Section 3.4.5. Examination of individual scatter plots of personal vs. ambient concentrations revealed some outlying data points. However, when time-activity information and field notes were examined, no explanations could be found for these outlying data points and they were, therefore, included in the analyses. Ambient: PMio vs. P M 2 5 vs. Sulfate vs. Personal: P M 2 5 Suli Pate P M 2 5 Sul Fate P M 2 S Sull Fate Type of Ambient: Avg. Clo. Avg. Clo. Avg. Clo. Avg. Clo. Avg. Clo. Avg. Clo. Pooled (N=106) 0.05 0.05 0.43 0.44 0.15 0.10 0.70 0.72 0.12 0.12 0.87 0.87 Median (N=16) 0.36 0.35 0.62 0.75 0.48 0.30 0.77 0.72 0.39 0.43 0.96 0.96 Table 3.9. Correlation coefficients for all relationships between personal and ambient parameters. c/> "tz o If) 1.0 .6 .4 .2 CO CD CL 0.0 PM10 PM2.5 Ambient Parameters Su l • Personal PM2.5 • Personal Sulfate ate Figure 3.6. Median of individual regressions between ambient parameters and personal exposures. 50 1.0 .8 .2 c/> o (/> m 03 CL 0.0 n P o o l e d regression • M e d i a n of individual PM2.5 Sulfate Figure 3.7. Tbe difference between pooling data and use of individual results when regressing personal against ambient. Figure 3.8. Correlation between personal and ambient PM 2 . S over time. 51 16' 14' PIP 12' § | | § 10' 8' 111 6' Hill 4. | | § § 2, oL I - I -.70 -.50 -.30 -.10 .10 .30 .50 .70 .90 Individual Pearson's r-values for P vs. A Sulfate Figure 3.9. Correlation between personal and ambient sulfate over time. P M 2 5 Sulfate Subject N P vs. Avg Amb* P vs. Clo Amb** P vs. Avg Amb* P vs. Clo Amb** Pearson Spearman Pearson Pearson Spearman Pearson 1 7 0.72 0.64 -0.04 0.98 0.96 0.91 2 7 0.81 0.86 0.53 0.97 0.75 0.94 3 7 0.66 0.75 0.34 0.98 0.85 0.98 4 7 0.74 0.89 0.28 0.95 0.93 0.98 5 6 0.51 0.77 0.27 0.92 0.94 0.98 6 7 -0.28 -0.14 0.01 0.89 0.82 0.96 7 7 0.20 0.25 0.30 0.79 0.86 0.83 8 7 0.44 0.43 0.55 0.97 0.89 0.98 9 7 -0.19 0.00 -0.35 0.99 0.89 0.96 10 7 0.08 -0.07 0.30 0.95 0.96 0.93 11 6 0.68 0.77 0.58 1.00 0.77 0.99 12 6 0.02 0.14 -0.19 0.99 0.94 0.99 13 7 0.83 0.75 0.90 0.97 0.57 0.98 14 7 0.40 0.54 0.58 0.66 0.79 0.62 15 6 0.74 0.89 0.88 0.96 1.00 0.93 16 5 -0.68 -0.90 -0.51 0.92 1.00 0.92 Group # 106 0.15 0.28 0.10 0.87 0.80 0.87 Median## 16 0.48 0.59 0.30 0.96 0.89 0.96 Table 3.10. Individual correlation coefficients relating personal and ambient concentrations; * average ambient; ** closest ambient; * pooled over all subjects, ** of individual regression analyses. 52 1.0 .8 .2 (/) c O CO CD Q- 0.0 PM2.5 Sul ate H A v e r a g e ambient • C l o s e s t ambient Figure 3.10. Difference between average and closest ambient data in personal vs. ambient regressions. 3 .4.2 Ratios and differences The analyses presented in the above section suggest that the correlation between personal and ambient measures was generally lower and more variable between subjects for PM than for sulfate, where high correlations were stable across the group. To aid in characterizing this finding, ratios of personal (P): ambient (A) concentrations as well as the differences between these concentrations were calculated to determine the quantitative differences between the two measures. Table 3.11 presents mean PA ratios for each subject and for all subjects pooled. Individual subject results are also summarized in boxplots, Figure 3.11 and Figure 3.12. The mean pooled personal:average ambient concentration ratio for PM2.5 was 1.75 and the individual means of this ratio ranged from 0.64 to 4.26. Twelve of 16 subjects had ratios above one when comparing their personal exposures to average ambient levels and 13 were above one when using the closest site values. The mean pooled personal:average ambient concentration ratio for sulfate was 0.75 with individual means ranging from 0.39 to 1.05. Using either average ambient or closest ambient levels, 15 out of 16 subjects' ratios were below one. Table 3.12 lists the mean differences (and the corresponding standard deviation) between personal and ambient concentrations for each subject and for the group as a whole. Only average ambient levels were considered for this analysis, When the data was pooled across all individuals, personal PM2.5 concentrations were greater than the ambient by 6.9 ug/m3 (61%) (SD: 14.6 ug/m3); individual mean differences ranged from -4.5 to 33.6 ug/m3 (SD: 2.1 to 24.3 ug/m3). Using sulfate as the exposure metric the mean difference between personal and ambient concentrations was -0.5 ug/m3 (26%) (SD: 0.5 ug/m3) for pooled data. The differences ranged between -1.2 and 53 0.0 |ag/m3 (SD: 0.1 to 0.7 ug/m3) for individuals. For both PM 2 5 and sulfate, these differences were statistically significant (paired sample t-test, p=0.000). These personal to ambient ratios and differences demonstrate that, on average, personal PM 2 5 exposures were significantly higher than the ambient levels whereas personal sulfate exposures were significantly lower. The percent differences also suggest that for PM2.5 the differences were greater than for sulfate. In addition, use of the closest site to each subject's home produced similar results as when the average of all ambient sites was used. P M 2 5 Sul 'ate Subject N P:A (avg.)* P:A (clo.)** P:A (avg.)* P:A (do.)** 1 7 0.99 1.01 0.64 0.67 •2 7 1.29 1.39 0.78 0.99 3 7 1.32 1.43 0.83 0.87 4 7 1.24 1.23 1.05 1.09 5 6 1.05 1.30 0.70 0.77 6 7 4.26 4.53 0.73 0.78 7 7 1.41 1.56 0.72 0.76 8 7 1.47 1.45 0.95 0.95 9 7 3.24 3.48 0.94 0.99 10 7 1.30 1.33 0.54 0.57 11 6 0.95 0.86 0.96 0.86 12 6 2.27 2.16 0.64 0.63 13 7 0.72 0.74 0.71 0.73 14 7 0.64 0.64 0.39 0.37 15 6 3.19 3.27 0.82 0.80 16 5 3.22 3.57 0.62 0.73 Group* 106 1.75 1.81 0.75 0.79 Table 3.11. Individual mean ratios relating personal and ambient concentrations; * average ambient concentration, ** closest ambient concentration; * pooled over all samples. 54 CNJ 0-to g < CL 12 10 8 6 4 2 0 = 1:1 ratio — = mean P:A ratio rrj X ^^ ^^ a. ~T~ I I I I N= 7 7 7 7 6 7 7 7 7 7 6 6 7 7 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.11. Ratios between personal and average ambient P M 2 S per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box); NOTE - scales are different this Figure and Figure 3.12. CO Ui o < CL .2 0.0 = 1:1 ratio — = mean P:A ratio 7 7 7 7 6 7 7 7 7 7 6 6 7 7 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.12. Ratios between personal and average ambient sulfate per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); • = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box); NOTE - scales are different between this Figure and Figure 3.11. 55 PM 2 5 (ug/m3) Sulfate ( ug/m3) Subject N Mean P-A SD Mean P-A SD 1 7 -0.1 2.3 -0.8 0.4 2 7 2.2 3.2 -0.4 0.3 3 7 3.3 5.1 -0.3 0.3 4 7 2.5 2.1 0.0 0.3 5 6 1.0 7.9 -0.5 0.3 6 7 33.6 24.3 -0.7 0.5 7 7 4.1 12.9 -0.9 0.7 8 7 3.8 4.4 -0.1 0.3 9 7 21.4 21.8 -0.4 0.4 10 7 1.8 5.2 -1.0 0.4 11 6 -0.6 4.4 -0.1 0.1 12 6 12.9 16.6 -0.7 0.3 13 7 -3.3 2.5 -0.5 0.3 14 7 -4.5 3.8 -1.2 0.7 15 6 20.1 8.4 -0.2 0.2 16 5 15.1 15.0 -0.5 0.2 Group* 106 6.9 14.6 -0.5 0.5 Table 3.12. Average differences between personal and ambient concentrations; * pooled over all samples. 3.4.3 Is the difference between personal and ambient concentrations dependent on level of personal exposure? To further study the difference between personal and ambient concentrations found in the previous section, we wanted to examine what variables affected the differences. We questioned whether the differences were dependent primarily upon personal exposures or ambient levels. Thus, the differences between personal and ambient concentrations (P-A) were plotted against personal exposures (P) and against ambient concentrations (A) for both PM 2 5 and sulfate. These plots were studied for each subject separately and as a group. Systematic patterns were observed in individual plots, summarized by the group plots shown in the figures below. The difference between personal and ambient PM2.5 concentrations increased as personal exposures increased (Figure 3.13). This demonstrates that increases in personal exposures are not matched by increases in the ambient level and it suggests that the level of personal exposure as well as the excess exposure (i.e. exposure over ambient concentration) is independent of the ambient level. This is confirmed by Figure 3.15, which indicates that the differences between personal and ambient PM 2 5 concentrations are random with respect to the ambient level. In contrast to the result found for PM2.5, the difference between personal and ambient sulfate remained constant throughout the range of personal sulfate exposures (Figure 3.14). This result demonstrates that personal sulfate exposures are matched by increases in the ambient level, 56 suggesting that the ambient level is a determining factor in the level of personal exposures to sulfate. Observing the difference with respect to ambient sulfate concentrations, Figure 3.16 indicates that the difference between personal and ambient levels increases significantly as ambient concentrations increase, suggesting that ambient levels increase to a greater extent than personal exposures do. 100 • I I C L -20 ] . . . . , 0 20 40 60 80 100 Personal PM2.5 (ug/m3) Figure 3.13. Personal-ambient vs. personal P M 2 5 . .5 0.0 -.5 § -1.0 3 - 1 . 5 & I -2.0 CO < -2.5 Q. -3.0 5 ^ it^T ^ 5" 0 1 2 Personal Sulfate (ug/m3) r = -0.02 p = 0.85 Figure 3.14. Personal-ambient vs. personal sulfate. O ) m Q_ < 100 80 60 40 20 0 0- -20 0 10 Ambient PM2.5 (ug/m3) 20 r= -0.13 p = 0.17 30 Figure 3.15. Personal-ambient vs. ambient P M 2 S . Figure 3.16. Personal-ambient vs. ambient sulfate. 58 3.4.4 Are correlations between personal and ambient measures dependent on level of personal exposure or ambient concentration? Results presented in Sections 3.4.2 and 3.4.3 indicated that the difference between personal exposures and ambient concentrations were dependent upon increased personal PM 2 5 exposures. The aim of the analysis presented in this section was to determine whether these differences affected the correlation between personal and ambient measures as well. Personal and ambient concentrations may be different, but if they are highly correlated (such that increases in one measure relate to increases in the other and vice versa), then ambient concentrations could be used to predict personal exposures. However, if the correlation between the measures increase or decrease due to the level of either variable (i.e. personal or ambient concentration), then use of ambient concentrations as a predictor of personal exposure may not be valid. To determine whether the correlations between personal and ambient measures (P vs. A correlation) were associated with personal or ambient concentration, regressions between the individual r-values (obtained in Section 3.4.1, Table 3.10) and mean personal and mean ambient concentrations for each subject were carried out. In all analyses presented, ambient concentrations were taken as the average of the five ambient sites. Using the closest ambient site data resulted in similar values. The analysis of the relationship between mean personal PM2.5 exposures and individual P vs. A correlation coefficients for PM2.5 resulted in a significant negative relationship (p=0.01; Figure 3.17), such that correlation coefficients decreased as personal exposures increased. In contrast, the analysis relating mean personal sulfate concentrations to individual P vs. A r-values for sulfate showed a positive relationship, however not to a significant level (p=0.13; Figure 3.18). Thus for sulfate, increasing personal concentrations resulted in higher correlations between personal and ambient concentrations. Analyses between individual r-values and ambient concentrations did not result in observable trends (Figure 3.19, Figure 3.20). These results demonstrate that the relationship between personal and ambient concentrations is dependent on the level of personal exposure. Thus, ambient concentrations may not be a good predictor of personal exposures, especially for PM2.5. Figure 3.18. Dependency of P vs. A correlations for sulfate on personal exposure. 1.0 9 10 11 12 13 14 15 Individual Mean Ambient PM2.5 (ug/m3) Figure 3.19. Dependency of P vs. A correlations for P M 2 . S on ambient concentration. r = -0.13 p = 0.64 CD CO > Ui > Q_ ro •g '> 1.1 1.0 .9 .8 .7 r = -0.19 p = 0.48 1.4 1.6 1.8 2.0 2.2 Individual Mean Ambient Sulfate (ug/m3) 2.4 2.6 Figure 3.20. Dependency of P vs. A correlations for sulfate on ambient concentration. 61 3 . 4 . 5 Use of ambient data from sites closest to each subject 3.4.5.1 Comparing sets of ambient data In analyses relating personal and ambient measures, ambient concentrations could be taken as the average of all five sites, from one site only or from the closest site to each subject. In comparison to average ambient data, we found that use of the closest monitor resulted in the similar P:A ratios and did not improve the correlation between personal and ambient levels. If the spatial variability of particulate concentrations was large, data from the closest monitor would be expected to improve the correlation. However, if the spatial variability was low, then the closest would not be expected to provide higher correlations with personal exposures than would the average of multiple sites or one centrally-located monitor. In Section 3.3.2, significant differences in concentrations between ambient sites were found to be the result of two sites, South Burnaby and South Richmond, which were found to be significantly different from all others. In this section, regressions between average-closest, average-individual site, and site to site concentration data were completed. As another test for spatial variability, regressions were conducted between personal exposures and each site to determine whether the relationships between each of the five sites and personal exposures were similar. The correlation between average ambient concentrations from each day of personal sampling and closest site concentrations from these days was high: 0.92 for PMio, 0.82 for PM2.5 and 0.95 for sulfate. Also, shown in Table 3.13, high correlations were observed between the average ambient concentrations and the concentrations from each site. Finally, when comparing the concentrations between each site (Table 3.14), correlations were also high, with sulfate showing the lowest and PM2.5 showing slightly higher correlations. Lastly, Table 3.15 displays the median Pearson's r of individual regression analyses between personal PM2.5 exposures and the PM 2 5 data obtained from each site. With the exception of North Burnaby, the sites indicated similar degrees of correlation with personal concentrations. Overall, while some sites were different with respect to concentrations, these high correlations support our results that indicated use of monitors closest to each subject's home did not provide a better estimate of personal exposures than the average of multiple monitors did. With respect to the low correlation between personal exposures and North Burnaby site data, it was noted that North Burnaby was not considered the closest site for any of the subjects. This suggests that the actual distance from a residence to an ambient monitoring site may be a factor in the P vs. A relationship, as assessed in the following section. KT ND NB SB SR Average PMio 0.88 0.92 0.97 0.97 0.93 Average PM2.5 0.85 0.84 0.88 0.83 0.87 Average Sulfate 0.96 0.97 0.93 0.93 0.95 Table 3.13. Correlation coefficients between average ambient concentrations and concentrations from each site. 62 K T ND NB SB SR P M 1 0 K T ~ ND 0.64 --NB 0.85 0.85 — SB 0.79 0.90 0.95 — SR 0.77 0.87 0.86 0.87 --P M 2 5 K T --ND 0.56 « NB 0.63 0.71 --SB 0.54 0.62 0.71 --SR 0.78 0.64 0.73 0.59 --Sulfate K T — ND 0.90 --NB 0.85 0.90 — SB 0.80 0.88 0.87 — SR 0.95 0.94 0.82 0.82 ~ Table 3.14. Correlation coefficients relating concentrations between individual ambient sites. Source of Ambient Data Median Pearson's r Average 0.48 Closest 0.30 Kitsilano 0.42 North Delta 0.41 North Burnaby 0.15 South Burnaby 0.35 South Richmond 0.39 Table 3.15. Comparison of median Pearson's r-values for personal vs. ambient relationships using average, closest or individual site ambient data for P M 1 5 . 3.4.5.2 Are P vs. A correlations a function of living distance from closest site? Since use of the closest site to subjects' homes did not improve the correlation between personal and ambient concentrations, the actual distance to a monitoring site was evaluated as a factor in the relationship. It was hypothesized that for subjects living very close to a monitor, this would be more representative of exposure than would the closest monitor for subjects living greater distance away from a monitor. From the five ambient sites, the site closest to each subject was determined by measuring distances between homes and sites on a map of the GVRD region (Table 3.16). All subjects were within 7 kilometers of an ambient monitoring site and the nearest distance was 400 meters. None of the residences were closest to the North Burnaby site. Individual P vs. A correlation coefficients (using ambient concentrations from the closest site to each subject; Table 3.10) were then regressed against these distances. Although the relationship is not significant, Figure 3.21 indicates that individual P vs. A correlations are a function of living 63 distance from the closest monitoring site. This result indicates that the density of the monitoring network may be important in evaluating personal exposures. Subject Closest ambient site Distance to site (km) 1 South Richmond 7 2 Kitsilano 4 3 South Burnaby 4 4 Kitsilano 3 5 North Delta 1 6 South Richmond 4 7 South Richmond 4 8 Kitsilano 4 9 South Burnaby 7 10 Kitsilano 3 11 Kitsilano 2 12 Kitsilano 3 13 Kitsilano 0.4 14 Kitsilano 5 15 Kitsilano 3 16 North Delta 2 Table 3.16. Distance of residence from closest ambient site per subject. CD CO > .1 > CO Z3 •> TD 1.0 .8 .6 .4 .2 .0 -.2 -.4 -.6 r = -0.35 p = 0.19 0 1 2 3 4 Distance from closest site (km) Figure 3.21. Individual P M J S correlations as a function of distance from closest site. 64 3.4.6 Predictors of Personal Exposure The data collected with the time-activity logs and dwelling questionnaire were used to examine which activities or living characteristics, in addition to ambient concentrations, were predictive of personal exposure. Pooled (over all subjects) personal PM 2 5 and sulfate data were used as the dependent variables in multiple regression analyses. In these analyses, ambient concentrations were taken as the average concentration obtained over the five sites for each sampling day. In order to reduce the number of potential independent variables, some categories in the time-activity logs and dwelling questionnaire were collapsed as described in Chapter 2, Section 2.7.3. Personal exposure data was regressed against each independent variable separately to determine which may be important in predicting exposure. In addition, correlations were conducted between all independent variables in order to eliminate highly correlated variables from being included together in the models. Core variables were then chosen for further analysis. Some variables were chosen as 'additional variables' which would be added to models to determine their importance as predictors. 3.4.6.1 PM2.5 multiple regression Five core variables ([house] volume, [time spent near] ETS, [time spent at] home, [time spent near] cooking, ambient PM2 5) were chosen to be included in the PM 2 5 multiple regression. 'Volume' and 'ETS' were significantly associated with personal PM 2 5 exposure in univariate regressions. 'Home', 'cook' and 'ambient PM 2 5 ' were chosen out of specific interest. 'Volume' was chosen over '# of rooms', to which it was highly correlated, as it was a more accurate measure of home size. 'Transit', 'road distance', 'building type', 'window score' and 'carpet score' were selected from the remaining variables to be added to core models. 'ETS', 'home' and 'transit' were flagged due to their skewed distributions. 'ETS', for example, was largely due to one subject who was exposed to ETS over all personal samples. The core variables were initially forced into an Ordinary Least Squares (OLS) regression including all 106 samples. When 'window', 'road distance' and 'building type' were individually added to this model, no improvement in model fit was observed. 'Transit' and 'carpet score' each slightly improved the fit of the model, however, their coefficients were not predictive in the expected direction. 'Transit' was not further analyzed in this model due to its highly skewed distribution. 'Carpet score' was also left out of the model; when cases with ETS exposure were excluded, 'carpet score' was no longer significant. One subject who has substantial ETS exposure, lived in a house with no carpets, producing the illusion that less carpets lead to higher exposures. Thus, only the five core variables were used for further analysis. Regression diagnostics (residual plot, range of RSS/n values per subject) of this model pointed to unequal variances. Exposure data also indicated one extreme value of personal exposure (90.1 ug/m3) which was much higher than other personal exposures recorded in the study and explained some of the observed variance. Thus, another OLS regression was performed after removing this one case. 65 The results of this model are shown in Table 3.17. The Multiple R and R Square were 0.52 and 0.27 respectively. 'Volume' and 'ETS' were significant predictors of personal P M 2 5 exposure. The residual plot (Figure 3.22) from this regression also indicated unequal variances between residuals and predicted exposures. The RSS/n values per subject ranged from 26.28 to 361.98. Subjects with high variance tended to have higher mean personal PM2.5 exposures and low P vs. A correlations as found in Section 3.4.4. Excluding cases with ETS exposure resulted in decreased Multiple R and R Square values for the model, 0.48 and 0.23 respectively. This did not increase the significance of any other variables. Variable Coefficient SE p-value Intercept -10.018 15.217 0.512 Volume of home (100 m3) -1.687 0.480 0.001* Time spent near ETS (% of sampling)** 1.512 0.571 0.009* Time spent at home (% of sampling) 0.294 0.155 0.061 Ambient P M 2 S (ug/m3) 0.412 0.275 0.137 Time spent cooking (% of sampling) 0.241 0.262 0.360 Table 3.17. OLS regression for PM 2 . S ; * p<0.01; ** due to the ETS exposure of one subject. 40- • 30-• • • 20-to 10-0--10-• • • • - ••• ' • . ••••• • • • • Resid UJ -20--30 • • • 0 0 10 2 0 3 0 4 0 Predicted Value Figure 3.22. Residual plot of OLS regression for PM 2 . S . 66 A Weighted Least Squares (WLS) regression was performed to down-weight individuals with high variances (Table 3.18). The Multiple R and R Square values remained similar to the unweighted version, at 0.52 and 0.27 respectively. The weighted residual plot (Figure 3.23) shows the improvement in the range of variances. Weighting caused the coefficients and p-values to change for most variables, resulting in 'volume' and 'ambient PM 2 5 ' as predictive of personal exposure. Variable Coefficient SE p-value Intercept 8.108 8.614 0.349 Volume of home (100 m3) -1.080 0.244 0.000* Ambient PM2.5 (ug/m3) 0.544 0.190 0.005* Time spent near ETS (% of sampling) 1.070 0.638 0.097 Time spent at home (% of sampling) 0.041 0.096 0.669 Time spent cooking (% of sampling) 0.068 0.161 0.673 Table 3.18. WLS regression for P M 2 S ; * p<0.01. 3' • 2< — 1. CO "D CD 0i or "O CD £ -1' O) • • • • • • • • • • • * • \ • • • . • • . • « • • * *~ 1' • . x • • • • • • • <* -2 5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Weighted Predicted Value Figure 3.23. Residual plot of WLS regression for P M 2 S . 6 7 3.4.6.2 Sulfate multiple regression 'Ambient sulfate' was chosen as the initial core variable for the sulfate multiple regression as it had a much higher correlation with personal sulfate exposures than other variables. Other variables of interest included 'volume' and 'ETS' (these were significantly correlated with personal sulfate when regressed individually) as well as 'outdoors', 'home', 'transit' and 'road distance' out of interest. The final model for sulfate included 'ambient sulfate', 'volume' and 'outdoors'. The results of this OLS regression are presented in Table 3.19. The fit of this model was higher than for PM2.5, with a Multiple R of 0.90 and an R square value of 0.82. The additional variables of 'ETS', 'home', 'transit' and 'road distance' had no added effect on the fit of the model. All three variables were significant, with the ambient concentration measure accounting for the major predictor of personal sulfate exposure. A model including only 'ambient sulfate' had an R Square value of 0.75. The residual plot from this regression (Figure 3.24) and the RSS/n values per subject indicated variances to be less variable than the PM2.5 multiple regression. Thus, a weighted regression was not pursued for sulfate. Variable Coefficient SE p-value Intercept 0.253 0.116 0.031 Ambient Sulfate (ug/m3) 0.738 0.036 0.000* Volume of home (100 m3) -0.086 0.015 0.000* Time spent outdoors (% of sampling) 0.009 0.005 0.045 Table 3.19. OLS regression for sulfate; * p<0.01. 1.5' 1.0' • • .5' 0.0' -.5' • Residual -1.0' • • • Residual -1.5 0 1 2 3 4 5 Predicted Value Figure 3.24. Residual plot of OLS regression for sulfate. 68 3.5 Part II - Assessment of Cardiovascular Health Effects in Relation to Exposure 3.5.1 Descriptive statistics for six cardiovascular health indicators Variables chosen for analysis included post-sample systolic blood pressure (SS BP), post-sample diastolic blood pressure (DS BP), supraventricular ectopy (SVE), average heart rate (HR), and the heart rate variability (HRV) variables, SDNN and r-MSSD. Chapter 2, Section 2.7.4 discusses how these variables were chosen for analysis. Four to seven valid post-sample BP measurements were obtained per subject, with 98 samples in total. This BP sample number was lower than the number of personal samples (N=112) largely because BP measurements were not conducted during the first few weeks of the study. Additionally, on a few occasions, the technicians had difficulty in obtaining BP readings. Distributions for these variables were not normal and they were not represented better when data was ln-transformed. For seven cases (all six cases for subject 14 and one case for subject 15), diastolic BP measurements were not possible to distinguish from 0 mm Hg. These cases were flagged and initially left in the analyses. From the 112 personal sampling days, 8 days at the beginning of the study were also conducted without Holter monitoring. From the 104 Holter samples, two were excluded due to technical problems leading to <6 hours of recording. Full summary reports could not be obtained from a further 16 samples due to high background noise when using the second sides of previously used tapes. To test whether HRV values calculated manually from strip data from these samples would yield accurate results, a feasibility test was conducted by examining 10 strips over one hour of normal data to the computer report for that hour. It was not possible to calculate the SVE data from these strips since the strips did not cover the complete period. The results of the feasibility test are presented in Table 3.20. Mean R-R intervals were mostly underestimated by 1-2% whereas the HRV variables were underestimated by 20-50%. While it may have been justified to use manually calculated mean R-R interval data and convert it to HR, a variable chosen for analysis against particulates, SDNN and r-MSSD were not estimated well by manual calculations. Thus it was decided not to use any manually calculated data. The underestimation of HRV variables can be expected since, due to greater number of R-R intervals being included, the total variance of HRV increases with the length of analyzed recording. It has, therefore, been suggested that it is not appropriate to compare SDNN measures obtained from recordings of different length (Task Force, 1996). Thus only full report data (N=86) was used for the SVE, HR, SDNN and r-MSSD analyses. The 8 days with no reports and the 16 person-days without valid data did not visually have different particulate levels than the days with valid ECG data when plotted in boxplots, confirming that our range of exposures was not limited by losing data from these days. In total, 3 to 7 valid ECG recordings were obtained per subject. SVE was extremely skewed to the right and it was represented better when data were ln-transformed. Average HR was normally distributed. 69 SDNN and r-MSSD were skewed to the right, but were not represented better by ln-transformations thus these variables were left untransformed. Individual and pooled descriptive data for the six cardiovascular health indicators are summarized in the following tables and figures: BP (Table 3.21/Figure 3.25/Figure 3.27), In SVE (Table 3.22/Figure 3.29), HR (Table 3.23/Figure 3.31), SDNN (Table 3.24/Figure 3.33) and r-MSSD (Table 3.25/Figure 3.35). Deviations from each subject's mean values were calculated for each variable to standardize the health outcome results across subjects for subsequent analyses (Figure 3.26, Figure 3.28, Figure 3.30, Figure 3.32, Figure 3.34, Figure 3.36). Sample Time of Day Mean R-R (ms) SDNN (ms) r-MSSD (ms) report manual A report Manual A report manual A 1 13:00-14:00 676 672 -4 38 29 -9 15 11 -4 4:00-5:00 849 851 2 83 68 -15 22 30 8 2 11:00-12:00 759 752 -7 76 60 -16 30 17 -13 21:00-22:00 766 757 -9 23 14 -9 30 14 -16 Table 3.20. Feasibility test results for manual calculation of HRV variables. Blood Pressure (mm Hg) Systolic Diastolic Subject N A M SD Range A M SD Range 1 7 124 6 116-132 74 3 70-78 2 7 150 11 138-168 90 5 82-96 3 7 129 6 120-134 81 3 76-86 4 7 125 5 120-134 66 5 60-75 5 5 107 9 96 - 120 61 5 56-68 6 7 129 9 118-146 72 4 68-78 7 7 129 . . .5.. . . 120-136 77 6 70-86 8 7 129 9 116-138 69 6 62-78 9 5 160 13 142 - 172 82 5 76-88 10 6 142 2 140 - 146 75 5 68-82 11 7 117 4 110-120 57 3 52-60 12 5 151 5 144-156 75 3 70-78 13 4 131 7 124-138 85 6 78-90 14 6 171 4 168-178 0 0 0 15 5 145 6 138-152 44 27 0-66 16 6 115 15 96-136 66 7 56-76 Group 98 134 18 96 - 178 67 21 0-96 Table 3.21. Blood pressure summary; A M : Arithmetic Mean, SD: Standard Deviation. 70 200 180 g 160 E E 140 Q_ 00 120 o B 100 >* CO 80 T o o N= 7 7 7 7 5 7 7 7 5 6 7 5 4 6 5 6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.25. Systolic blood pressure measurements per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). Figure 3.26. Standardized systolic blood pressure measurements per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 71 Figure 3.27. Diastolic blood pressure measurements per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box). O) 40 E E , D_ QQ CO Q "O 0) N T J (0 " D C ro 20 -20 -40 -60 N= 7 7 7 7 5 7 7 7 5 6 7 5 4 6 5 6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.28. Standardized diastolic blood pressure measurements per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box). 72 Supraventricular Ectopy (beats/hr) Subject N A M SD Range G M GSD Range 1 3 1.9 1.4 0.7-3.5 2.7 1.6 1.7-4.5 2 7 1.2 0.5 0.8-1.9 2.2 1.2 1.8-2.9 3 6 6.8 9.1 1.1-24.5 4.9 2.7 2.1-25.5 4 6 1.4 0.9 0.5-2.8 2.2 1.4 1.5-3.8 5 5 54.4 61.7 0.0-160.7 24.5 6.7 1.0-162.4 6 4 5.4 2.2 2.9-7.6 6.1 1.4 3.9-8.9 7 7 7.7 4.8 1.8-15.1 7.4 1.9 2.8-16.1 8 7 32.0 32.5 0.1-93.6 17.9 4.3 1.1-94.6 9 5 21.9 20.7 1.9-55.4 15.1 3.1 2.9-56.3 10 5 4.5 1.1 3.2-5.9 5.4 1.2 4.2-6.9 11 4 2.7 1.4 1.1-4.3 3.5 1.5 2.1-5.3 12 5 3.1 1.9 0.5-5.6 3.7 1.8 1.5-6.6 13 5 29.6 37.7 3.5-94.5 16.7 3.5 4.5-95.6 14 6 9.1 2.5 6.6-13.6 9.8 1.3 7.6-14.6 15 6 361.2 59.7 263.4-443.1 357.8 1.2 265.1 -445.9 16 5 2.5 1.2 0.7-3.8 3.2 1.5 1.7-4.8 Group 86 36.9 93.5 0.0-443.1 8.4 4.5 1.0-445.9 Table 3.22. Supraventricular ectopy summary; A M : Arithmetic Mean, SD: Standard Deviation, G M : Geometric Mean, GSD: Geometric Standard Deviation. Figure 3.29. Ln-transformed supraventricular ectopy (In SVE) values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 73 to JD 2-CO " D 0) N TD ro TD £Z CO 1-0--1--2--3-CO -4, N • o | 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.30. Standardized ln-transfonned supraventricular ectopy On SVE) values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). Heart Rate (bpm) Subject N A M SD Range 1 3 67 1 67-68 2 7 86 5 78-91 3 6 63 1 61-64 4 6 90 2 87-93 5 5 73 13 60-94 6 4 73 3 70-76 7 7 80 5 71-86 8 7 94 3 88-97 9 5 106 2 103 - 108 10 5 82 4 79-88 11 4 88 10 80 - 103 12 5 67 4 64-74 13 5 84 2 82-88 14 6 70 3 67-76 15 6 79 5 70-84 16 5 90 10 82- 105 Group 86 81 12 60-108 Table 3.23. Heart rate summary; A M ; Arithmetic Mean, SD: Standard Deviation. 74 120 110 100 E Q. CO a: •c CD CD 90 80 70 60 50 I 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.31. Heart rate values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 30 £ 20 Q. or 10 ro CO -20 N= 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.32. Standardized heart rate values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 75 SDNN (ms) Subject N A M SD Range 1 3 117 9 112-127 2 7 112 10 97-127 3 6 104 14 90 - 127 4 6 91 14 75-112 5 5 162 82 97 - 278 6 4 84 4 82-90 7 7 107 33 90-180 8 7 53 5 45-60 9 5 46 6 37-52 10 5 75 8 67-82 11 4 92 7 82-97 12 5 103 24 67-135 13 5 82 12 67-97 14 6 121 20 90-150 15 6 146 22 120-180 16 5 61 16 37-75 Group 86 97 38 37-278 Table 3.24. SDNN summary; A M : Arithmetic Mean, SD: Standard Deviation. 300 200 Q CO 0 N = 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.33. SDNN values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 76 E 200 100 Q CO T J CD 1 ° CO "D C ro CO -100 3»l'iT"' N= 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.34. Standardized SDNN values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). r-MSSD (ms) Subject N A M SD Range 1 3 65 16 52-82 2 7 20 3 15-22 3 6 32 10 22-52 4 6 20 6 15-30 5 5 157 144 45 -353 6 4 26 5 22-30 7 7 33 8 22-45 8 7 33 7 22-45 9 5 40 23 15-75 10 5 21 3 15-22 11 4 28 4 22-30 12 5 30 5 22-37 13 5 31 18 15-60 14 6 54 7 45-60 15 6 107 12 90 - 120 16 5 18 4 15-22 Group 86 44 49 15-353 Table 3.25. R-MSSD summary; A M : Arithmetic Mean, SD: Standard Deviation. 77 400-300' 200^ i l -100 | N= 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.35. R-MSSD values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). (/> E, Q CO CO 300 200 100 CD N E CO c ro CO -200 -100 o N= 3 7 6 6 5 4 7 7 5 5 4 5 5 6 6 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Subject Figure 3.36. Standardized r-MSSD values per subject; N = number of samples per subject; o = outliers (cases with values between 1.5 and 3 times the interquartile range from upper or lower edge of the box); * = extreme values (cases with values more than 3 times the interquartile range from upper or lower edge of the box). 78 3.5.2 Regressions against pollution for each variable The six cardiovascular health indicators (SS BP, DS BP, Ln SVE, HR, SDNN, r-MSSD) were used as dependent variables in regressions against various exposure metrics (ambient PMio, ambient P M 2 5 , ambient sulfate, personal P M 2 5 and personal sulfate). Ordinary Least Squares (OLS) regressions of pooled data were conducted initially and scatterplots of each variable vs. particulates were also plotted to identify outliers. In total, 30 relationships between health outcomes and exposure were analyzed. No outliers were observed for systolic BP. One outlier was observed for diastolic BP (subject 15, Figure 3.28). The large deviation from subject 15's mean resulted from one measurement recorded as 0 mm Hg. This case was excluded from further analyses and subject 15's mean DS BP and standardized DS BP values were recalculated before continuing. All other DS BP values which were 0 mm Hg belonged to Subject 14, for whom all DS BP measurements were 0 mm Hg. Analyses excluding this subject did not cause results to change, thus was kept in all analyses. For the ECG data, no outliers were observed for the SVE variable. Considering the HRV variables, for subject 5, two out of five reports presented extremely elevated SDNN and r-MSSD values. Severe atrial fibrillation of this individual, as noted by the ECG technician, may have caused an increase of SDNN and r-MSSD above normal. Rather than only excluding the two reports, subject 5 was completely removed from the SDNN and r-MSSD analyses. Although the values of HR for subject 5 were not visually identified as outliers, this individual was further excluded from the HR analyses due to the correspondence between HR and HRV variables. A further outlying case was removed for SDNN only (subject 7, Figure 3.34). When reviewing the ECG report for this case, the start of the sample presented an extremely elevated SDNN level, thus elevating the SDNN 24-hour summary value above normal. All OLS regressions were repeated excluding the above-identified outliers. Assessment of the correlation structures of these regressions did not identify the presence of autocorrelation due to repeated measures of individuals; no consistent trends were found in the correlations between sampling sessions. For all relationships, regression diagnostics (boxplots of residuals/subject, range of estimated variance (RSS/n) and estimated standard deviation values per subject) pointed to unequal variances. Weighted least squares (WLS) regressions were conducted to account for unequal variance across subjects. Sections 3.5.2.1 to 3.5.2.6 review each relationship in detail. As a general observation, for all relationships, weighting tended to decrease the standard errors of coefficients. The decrease in standard errors was expected with weighted analyses, which are more efficient in slope estimation. In many cases, the magnitudes of coefficients were also decreased with weighting. The reduction in magnitude of coefficients was likely due to individuals whose results were down-weighted due to high variance who also had high exposure and high response values. Plots of WLS regression coefficients (± 2*standard errors) are also presented in the following sections to visually compare the exposure metrics. The standard errors were consistently the largest for 79 ambient and personal sulfate. The large standard errors for sulfate can be explained by sulfate having smaller concentration range over which the health outcome values could be placed, compared with the larger ranges of the P M metrics. Lastly, interquartile ranges were calculated for each exposure metric (Table 3.26). The cardiovascular responses for each interquartile range increase in exposure were calculated, plots of which are also presented in the sections that follow. In most cases, ambient PMio had the largest effects on the health indicators analyzed. The 30 relationships were further tested for confounding by temperature (T), relative humidity (RH), carbon monoxide (CO), ozone (O3) and bronchodilator use (BD) using two-variable models. Since the effects of subjects' regular B D use had already adjusted for by standardizing the health outcome variables, we were interested in testing for potential confounding effects from B D use above or below average. Thus, the B D use variable was standardized by evaluating the deviation from each subjects' mean B D use (similar to the standardization of the health variables). Prior to running two-variable models, the correlation between all main exposure variables and potentially confounding secondary variables was assessed (Table 3.27). Correlations between T, R H , C O and O 3 were variable and not high enough to justify excluding any from the two-variable analyses. B D use had low correlation with the exposure variables. Most exposure metrics were moderately correlated with the other potential confounders, with the exception o f personal P M 2 5 . This can be expected since personal P M 2 5 levels, compared to the other exposure metrics, are more likely affected by non-ambient sources. The two-variable models were weighted using the weights from the simple regression between health indicator and exposure metric of each specific two-variable model In addition to results for the simple O L S and W L S regressions between outcome and exposure, Sections 3.5.2.1 to 3.5.2.6 present the two-variable regression results for each health outcome. The secondary variables affected the exposure-response relationships differently depending on the health outcome and the exposure metric being analyzed. Confounding was observed by a decrease in the magnitude of the exposure metric's coefficient, a change in the direction of the relationship, or an increased p-value compared with the simple exposure-response regression. In many cases, addition of the potential confounders enhanced the effect of the exposure metrics, causing their coefficients to increase in magnitude and p-values to decrease. In general, the effect of additional variables was more prominent for the ambient P M and sulfate metrics, compared to relationships with personal PM 2 .s exposures. Also, B D use did not largely affect the relationships. These results were expected due to the low correlation between personal PM2.5 and outcome variables and B D use and exposure variables. Section 3.5.3 summarizes the results after considering the impact of the potential confounders on the exposure-response relationships. 80 Interquartile ranges (ug/m3) Ambient PMio Ambient PM2.5 Ambient Sulfate Personal PM2.5 Personal Sulfate 7 5.5 1.2 10.3 0.9 Table 3.26. Interquartile ranges for each exposure metric. T RH CO 0 3 BD use T — RH -0.492 --CO 0.197 -0.261 — 0 3 0.437 -0.442 0.150 — BD use -0.060 -0.066 -0.025 0.017 — Ambient PMio 0.435 -0.613 0.570 0.557 -0.058 Ambient P M 2 5 0.363 -0.249 0.457 0.649 -0.067 Ambient Sulfate 0.255 -0.101 0.145 0.537 -0.039 Personal P M 2 5 -0.073 -0.023 0.030 0.127 0.109 Personal Sulfate 0.345 -0.208 0.230 0.660 -0.087 Table 3.27. Correlation coefficients between potential confounders and exposure metrics. 3.5.2.1 Systolic blood pressure Visible negative trends were observed in all relationships between systolic BP and exposure, with significant results for ambient PMio and personal PM2.5 exposure. An example of the relationship with ambient PMio exposure is shown in Figure 3.37; Figure 3.38 displays the effect estimates resulting from the various relationships with main variables. Though T and 0 3 were negatively related to SS BP in single-variable regressions (data not shown), these did not cause the relationships between SS BP and main variables to decrease in magnitude or significance (Table 3.28). In fact, O 3 enhanced the results of the main variables. In particular, the ambient PMio WLS regression effect estimate was more strongly negative and highly significant following the addition of either T or O 3 . The correlation between T and 0 3 was only moderate (r=0.437), thus a three-variable model was run with ambient PMi 0 . In this model, ambient PMio remained significantly related to SS BP while T and O 3 became slightly less predictive (ambient PMio B--0.435, SE-0.170, p=0.012; 03 B=107.805, SE=75.985, p=0.159; T B=0.134, SE=0.214, p=0.534). The addition of RH or BD use to models did not largely affect the results of the main variables. CO caused the coefficients for most main variables to decrease in magnitude, and the signs for ambient P M 2 5 and sulfate to change, suggesting that CO slightly confounds the relationships between SS BP and main variables. However, relationships with the largest effects remained negative after the addition of CO, thus indicating stability of the results. 30 O) I E E CL CD CO CO N CO X J c CO co -: 10 20 0 10 Ambient PM10 (ug/m3) 20 30 40 Figure 3.37. OLS regression between systolic blood pressure and ambient PMi 0 ; all cases included. Systolic blood pressure response to interquartile range increases in exposure 1 . 5 0 . 5 A m b i e n P M 1 0 A m b i e r : P M 2 . 5 A m b i e n i S u l f a t e P e r s o n ; I P M 2 . 5 P e r s o n a S u l f a t e i OH x E E UJ 8 I "1 1 S -15 -2.5 - 3 - 3 . 5 Figure 3.38. Systolic BP effect estimates and SE for interquartile range increases in exposure. 82 Systolic BP vs. Main Variable Secondary Variable Coefficient SE p-value Coefficient SE p-value Ambient PM | Main only (OLS) -0.249 0.144 0.087* — — — r -Main only (WLS) -0.216 0.110 0.054* Main + T -0.302 0.143 0.038** 0.245 0.200 0.224 Main + RH -0.193 0.145 0.186 0.026 0.069 0.703 Main + CO -0.128 0.145 0.380 -4.247 4.549 0.353 Main + O 3 -0.405 0.150 0.008** 124.251 67.702 0.070* Main + BD use -0.243 0.112 0.033** 1.048 0.805 0.196 Ambient PM*a Main only (OLS) -0.146 0.186 0.434 — „ — — Main only (WLS) -0.051 0.136 0 706 — » » » » » » Main + T -0.044 0.149 0.766 0.100 0.170 0.559 Main + RH -0.084 0.153 0.586 0.040 0.063 0.524 Main + CO 0.101 0.159 0.528 -7.342 4.089 0.076* Main + O 3 -0.135 0.184 0.465 45.334 66.950 0.500 Main + BD use -0.086 0.138 0.536 1.046 0.798 0.193 Ambient Siiitale Main only (OLS) -0.528 0.790 0.506 — — : | | | | | | | | | | | | | | | Main only (WLS) -0.161 0.54 J 0.766 Main + T -0.113 0.555 0.839 0.107 0.163 0.513 Main + RH -0.487 0.602 0.422 0.037 0.062 0.558 Main + CO 0.006 0.545 0.992 -5.818 3.509 0.101* Main + O 3 -0.399 0.672 0.554 36.654 61.193 0.551 Main + BD use -0.274 0.545 0.617 1.029 0.788 0.195 Personal fMu Main only (OLS) -0.066 0.052 0.214 — — — Main only (WLS) -0 075 0.045 0 102* — lllllllliillli Main + T -0.073 0.047 0.124 0.063 0.164 0.700 Main + RH -0.087 0.055 0.120 0.045 0.063 0.475 Main + CO -0.075 0.045 0.094* -6.760 3.712 0.072* Main + O 3 -0.077 0.046 0.096* 18.990 50.362 0.707 Main + BD use -0.068 0.045 0.134 1.034 0.777 0.187 Personal Sulfate • Main onlv (OLS) -0.827 0.895 0 358 — — — Main only (WLS) -0.576 0.620 0.355 Illlllllllllli Main + T -0.646 0.647 0.321 0.082 0.171 0.633 Main + RH -0.809 0.679 0.238 0.052 0.064 0.416 Main + CO -0.304 0.636 0.634 -6.372 3.888 0.105 Main + O 3 -0.922 0.784 0.243 46.406 64.060 0.471 Main + BD use -0.638 0.617 0.304 1.205 0.778 0.125 Table 3.28. Systolic blood pressure regressions; all cases included; ** p<0.05, * p<0.10. 83 3.5.2.2 Diastolic blood pressure Inverse trends between particulates and DS BP were consistent with the results observed for SS BP. An example of the relationship with ambient PMio exposure is shown in Figure 3.39; Figure 3.40 displays the effect estimates resulting from the various relationships with main variables. Although weighted coefficients were very small with high p^ yalues, the results are relatively stable with respect to size and directions of coefficients (Table 3.29). The addition of T with personal PM2.5 and O3 with personal sulfate caused signs to change from negative to positive. These two main variables, however, also had the least effect on DS BP with respect to interquartile range increases in exposure. The largest effect was demonstrated by ambient PMio, which remained stable after the addition of secondary variables. Overall these results are suggestive of a negative relationship between particulates and DS BP, which is not largely affected by secondary variables. The relationship is slightly weaker and less consistent than the relationship between particulates and SS BP. 201 10 X E E Cu DQ CO Q T3 CD N TJ L _ CO T3 C CO 35 -20J -10 •I ' 0 10 20 Ambient PM10 (ug/m3) t 30 40 Figure 3.39. OLS regression between diastolic blood pressure and ambient P M i 0 ; one case excluded. 84 Diastolic BP vs. Main Variable Secondary Variable Coefficient SE p-value Coefficient SE p-value Ambient PMt« Main only (OLS) -0 092 0.095 0.335 — — — Main only (WLS) -0.026 0.035 0.449 . . . — Main + T -0.023 0.035 0.509 -0.010 0.027 0.716 Main + RH -0.107 0.073 0.147 0.001 0.006 0.797 Main + CO -0.046 0.047 0.325 0.400 0.629 0.526 Main + O 3 -0.027 0.038 0.480 0.607 12.351 0.961 Main + BD use -0.027 0.035 0.446 0.123 0.547 0.822 Ambient PM2$ Main only (OLS) -0 069 0.122 0,574 — — Main only (WLS) -0.004 0.028 0.892 . . . — Main + T -0.006 0.030 0.836 -0.011 0.032 0.738 Main + RH -0.032 0.053 0.545 0.006 0.012 0.584 Main + CO -0.006 0.035 0.876 0.051 0.640 0.936 Main + O 3 -0.003 0.028 0.919 -2.554 12.328 0.836 Main + BD use -0.004 0.028 0.889 0.201 0.558 0.719 bient Sulfate Main only (OLS) 0 260 0.519 0.618 — — — Main only (WLS) 0.013 0.114 0.908 — — Main + T 0.004 0.131 0.973 -0.004 0.026 0.871 Main + RH -0.043 0.302 0.888 0.002 0.010 0.882 Main + CO 0.013 0.115 0.908 -0.003 0.393 0.994 Main + O 3 0.008 0.133 0.954 -0.931 11.104 0.933 Main + BD use 0.012 0.115 0.915 0.204 0.566 0.720 Personal PM 2 5 Main only (OLS) 0.018 0.035 0.616 — — — Main only (WLS) -0.001 0.017 0 938 — — Main + T 0.001 0.017 0.936 -0.007 0.032 0.832 Main + RH -0.003 0.020 0.899 0.001 0.007 0.938 Main + CO -0.001 0.017 0.962 -0.075 0.570 0.895 Main + O 3 -0.001 0.017 0.956 -3.907 13.226 0.768 Main + BD use -0.001 0.017 0.947 0.050 0.562 0.929 Personal Sulfate Main only (OLS) 0.263 0.593 0.658 llllllliillllllll — — Main only (WLS) •0.008 0.206 0 971 . . . — — Main + T -0.008 0.212 0.969 -0.008 0.038 0.828 Main + RH -0.082 0.248 0.743 0.001 0.008 0.900 Main + CO -0.005 0.208 0.980 -0.100 0.642 0.877 Main + 0 3 0.010 0.215 0.964 -4.808 15.418 0.756 Main + BD use -0.007 0.208 0.971 0.013 0.561 0.982 Table 3.29. Diastolic blood pressure regressions; one case excluded. 85 Diastolic blood pressure response to interquartile range increases in exposure W -0.4 -0.6 - 0 . 8 A m b i e n P M 1 0 A m b i e r P M 2 . 5 A m b i e n S u l f a t e P e r s o n . I P M 2 . 5 P e r s o n a S u l f a t e Figure 3.40. Diastolic BP effect estimates and SE for interquartile range increases in exposure. 3.5.2.3 Supraventricular ectopy Positive slopes resulted from all relationships between the S V E arrhythmia variable and the various exposure metrics. A n example of the relationship with ambient PMio exposure is shown in Figure 3.41; Figure 3.42 displays the effect estimates resulting from the various relationships with main variables. The coefficients of exposure metrics remained largely unchanged from simple weighted results after the addition of secondary variables, with the exception of C O changing the sign of the ambient sulfate coefficient from positive to negative (Table 3.30). Coefficients were significant for both ambient PMio and ambient PM 2 .5 exposure, which also had the largest effect with respect to interquartile range increases in exposure. In summary, S V E was positively affected by particulates with no large effects from secondary variables tested. Ln SVE response to interquartile range increases in exposure 0.25 0.2 o 0.15 Ambient PM10 Ambier •PM2.5 Ambien Sulfate Person! 1 PM2.5 Persone Sulfate 0.1 € 0 0 5 E 1 Hi 0 % UI 8 = -0.05 -0.1 -0.15 Figure 3.42. Ln SVE effect estimates and SE for interquartile range increases in exposure. 87 Ln SVE vs. Main Variable Secondary Variable Coefficient SE p-value Coefficient SE p-value Ambient P M t u Main only (OLS) 0 028 0.011 0.014** — — Main only (WLS) 0.015 0.007 0 032** ::::::::::: >ft i : : M a i n + T 0.016 0.008 0.055* -0.006 0.012 0.600 Main + R H 0.013 0.008 0.118 -0.003 0.003 0.372 Main + CO 0.015 0.009 0.103* 0.009 0.278 0.976 Main + O 3 0.018 0.008 0.030** -2.860 4.207 0.498 Main + B D use 0.013 0.007 0.090* 0.023 0.031 0.477 Ambient pJMfes Main only (OLS) 0 041 0.019 0,036** — — Main only (WLS) 0.017 0.010 0 085* . . . . . . Main + T 0.018 0.011 0.095* -0.007 0.011 0.519 Main + R H 0.021 0.011 0.054* -0.002 0.003 0.477 Main + CO 0.016 0.012 0.182 0.045 0.276 0.871 Main + O 3 0.022 0.012 0.068* -3.472 4.731 0.465 Main + B D use 0.016 0.010 0.107 0.025 0.039 0.522 Ambient Sulfate Main only (OLS) 0.051 0.088 0 567 iiiiiiiiiiiiiiiii — — Main only (WLS) 0.011 0.044 0.800 — — Main + T 0.010 0.045 0.820 - 0 . 0 0 1 0.011 0.939 Main + R H 0.009 0.047 0.848 -0.003 0.003 0.334 Main + CO >-0.001 0.046 0.998 0.204 0.233 0.384 Main + O 3 0.008 0.050 0.875 0.589 4.403 0.894 Main + B D use 0.005 0.044 0.903 0.042 0.033 0.211 Personal P M 2 * Main only (OLS) O.001 0 006 0 965 — ••y<-.. . — Main only (WLS) 0.001 0.003 0 686 ... Main + T <0.001 0.003 0.940 0.002 0.011 0.845 Main + R H 0.001 0.003 0.750 -0.003 0.003 0.301 Main + CO 0.001 0.003 0.678 0.251 0.242 0.304 Main + O 3 0.001 0.003 0.730 1.261 4.067 0.757 Main + B D use 0.001 0.003 0.811 0.046 0.031 0.142 Personal Sulfate Main only (OLS) 0.081 0.101 0.428 — — Main only (WLS) 0.032 0.055 0 563 llllllllllllll Main + T 0.038 0.058 0.516 -0.003 0.012 0.776 Main + R H 0.025 0.059 0.676 -0.003 0.003 0.407 Main + CO 0.021 0.057 0.718 0.211 0.256 0.413 Main + O 3 0.031 0.069 0.654 0.143 5.161 0.978 Main + B D use 0.026 0.055 0.641 0.044 0.037 0.231 Table 3.30. Ln-transformed supraventricular ectopy regressions; all cases included; ** p<0.05, * p<0.10. 88 3.5.2.4 Heart rate Results were inconsistent among the single-pollutant W L S regressions between particulates and H R . While increases in H R were observed in relation to ambient PM exposures, decreases were observed when regressing against ambient or personal sulfate, or personal P M 2 . 5 . A n example of the relationship with ambient PMio exposure is shown in Figure 3.43; Figure 3.44 displays the effect estimates resulting from the various relationships with main variables. In the single-pollutant models, T and 0 3 were moderately positively associated with H R (data not shown), whose coefficients became significant in the two-pollutant models. In two-variable models, T and O 3 caused the positive relationship between ambient P M 2 . 5 and H R to become highly insignificant, while not affecting the ambient PMio relationship, and further decreasing the sulfate coefficients (Table 3.31). Interquartile range increases in exposure appeared to affect H R by the same degree between the various exposure metrics. The inconsistent results between PM and sulfate exposure metrics therefore remained after taking secondary variables into account, with increases in H R observed for ambient PMio and decreases observed with respect to the sulfate metrics. Figure 3.43. OLS regression between heart rate and ambient P M 1 0 ; one subject excluded. 89 Heart Rate vs. Main Variable Secondary Variable Coefficient SE p-value Coefficient SE p-value Ambient PM, U Main only (OLS) 0.091 0.074 0.217 —- . •.. — — Main only (WLS) 0.060 0 038 0.118 . . . Main + T 0.031 0.043 0.473 0.113 0.081 0.170 Main + R H 0.020 0.047 0.680 -0.026 0.034 0.458 Main + C O 0.053 0.041 0.206 0.747 1.477 0.615 Main + 0 3 0.053 0.046 0.253 6.756 22.071 0.760 Ma in + B D use 0.073 0.040 0.077* 0.359 0.388 0.357 Ambient PjMis Main only (OLS) 0.092 0.113 0.419 — — — Main only ( W L S ) 0.047 0.057 0.410 „— Main + T -0.005 0.062 0.937 0.135 0.080 0.094* Main + R H 0.014 0.060 0.814 -0.032 0.029 0.272 Ma in + C O 0.043 0.057 0.456 0.953 1.337 0.478 Main + 0 3 -0.013 0.091 0.887 24.280 28.760 0.401 Main + B D use 0.055 0.059 0.351 0.208 0.386 0.591 Ambient Sulfate Main only (OLS) -0.138 0.494 0.781 — Main only (WLS) -0 233 0.219 0 290 — Main + T -0.316 0.218 0.151 0.121 0.066 0.069* Main + R H -0.260 0.224 0.249 -0.035 0.026 0.172 Ma in + C O -0.225 0.225 0.320 0.224 1.173 0.849 Main + 0 3 -0.715 0.294 0.017** 49.677 20.946 0.020** Main + B D use -0.234 0.221 0.292 -0.030 0.362 0.933 Personal PM Z * I1B11111111I • Main only (OLS) -0 009 0.032 0.781 — — Main only ( W L S ) -0 022 0.020 0 265 Main + T -0.020 0.020 0.325 0.118 0.067 0.080* Main + R H -0.001 0.023 0.978 -0.033 0.026 0.221 Main + C O -0.022 0.020 0.267 0.622 1.187 0.602 Main + 0 3 -0.026 0.020 0.189 20.695 16.169 0.205 Main + B D use -0.024 0.020 0.237 -0.182 0.367 0.621 Persona! Sulfate Main only (OLS) -0 195 0.515 0.705 — — — Main only ( W L S ) -0.252 0.282 0 375 — — Main + T -0.477 0.294 0.109 0.148 0.070 0.038** Main + R H -0.248 0.300 0.411 -0.036 0.026 0.180 Main + C O -0.239 0.285 0.404 0.495 1.183 0.677 Ma in + 0 3 -0.829 0.381 0.033** 46.956 21.484 0.032** Main + B D use -0.262 0.285 0.361 -0.126 0.369 0.735 Table 3.31. Heart rate regressions; exluding one subject; ** p<0.05, * p<0.10. 90 Heart rate response to interquartile range increases in exposure III -1 Ambien .PM10 Ambier PM2.5 Ambien Sulfate Person; I PM2.5 Persona Sulfate i i Figure 3.44. Heart rate effect estimates and SE for interquartile range increases in exposure. 3.5.2.5 SDNN Relationships predicting S D N N were largely inconsistent and unstable among the various relationships analyzed. A n example of the relationship with ambient PMio exposure is shown in Figure 3.45; Figure 3.46 displays the effect estimates resulting from the various relationships with main variables. The single-pollutant W L S regressions reported negative relationships for ambient and personal PM2.5 whereas positive relationships were observed for ambient PMio and sulfate exposures. The addition of T and R H did not affect these relationships. C O and O 3 tended to lower exposure metric coefficients, changing the signs of coefficients for ambient PMio and ambient sulfate from positive to negative (Table 3.32). The personal measures were least affected by the addition of secondary variables, with personal PM2.5 remaining inversely related to S D N N and personal sulfate being positively related. 10 E O CO TJ CD N T3 i— CO TJ C CO CO 4U • 30 • • • 20 • • • • • 10- • • • • 0' * - r • -10' • > -20' • • -30' • -40 • 0 10 20 Ambient PM10 (ug/m3) 30 40 50 60 Figure 3.45. OLS regression between SDNN and ambient P M 1 0 ; one subject + one case excluded. SDNN response to interquartile range increases in exposure 2 . 5 -E 1 ti a 1 . 5 -0 . 5 -- 0 . 5 -- 1 . 5 • A m b i e r i P M 1 0 A m b i e n t P M 2 . 5 A m b i e n t Sulfate P e r s o n i l P M 2 . 5 P e r s o n a Sulfate - 2 . 5 • Figure 3.46. SDNN effect estimates and SE for interquartile range increases in exposure. 92 S D N N vs. M a i n Variable Secondary Var iable Coefficient SE p-value Coefficient SE p-value Ambient P M W Nfl f lB Main only (OLS) 0 069 0.217 0.751 — •yyyyy&ysi¥i'&s:-&yyyyy^ — Main only ( W L S ) 0.054 0.114 0 639 — yyyyy. • \ wiiw:; i; i&ivi — Main + T 0.061 0.142 0.668 0.070 0.281 0.804 Ma in + R H 0.104 0.146 0.479 0.025 0.125 0.843 Main + C O -0.037 0.151 0.805 6.052 6.570 0.360 Ma in + O 3 -0.010 0.149 0.947 57.933 86.159 0.503 Main + B D use 0.051 0.114 0.658 -1.016 1.044 0.333 Alt! ibient P M 2 5 Main only (OLS) -0 057 0.330 0.863 — — Main only (WLS) •0 052 0.170 0 762 . . . . . . Main + T -0.021 0.209 0.919 0.072 0.273 0.793 Main + R H 0.042 0.194 0.829 0.011 0.107 0.922 Ma in + C O -0.157 0.218 0.474 4.697 6.063 0.441 Ma in + O 3 -0.220 0.252 0.385 88.048 97.237 0.368 Main + B D use -0.064 0.170 0.709 -0.938 1.015 0.358 Am bleat Sulfate Main only (OLS) 0 073 1.432 0.960 — — Main only (WLS) 0 121 0.797 0.880 . . . « « « Main + T 0.212 0.806 0.794 0.083 0.237 0.727 Ma in + R H 0.566 0.866 0.516 -0.002 0.104 0.987 Main + C O 0.028 0.813 0.973 3.197 4.924 0.518 Main + O 3 -0.275 1.027 0.789 52.279 84.977 0.540 Ma in + B D use 0.180 0.800 0.823 -0.986 1.030 0.342 Personal P3VI?s Main onlv (OLS) -0 066 0 100 0.514 — • • — - | | | | | | | | | | | ; | | | | Main only ( W L S ) -0.021 0.045 0.634 *«. Main + T -0.045 0.046 0.331 -0.009 0.243 0.972 Main + R H -0.034 0.048 0.491 0.009 0.105 0.932 Main + C O -0.019 0.045 0.672 2.010 5.397 0.711 Main + 0 3 -0.020 0.045 0.649 23.346 66.832 0.728 Main + B D use -0.023 0.045 0.609 -0.688 1.018 0.502 Personal Sulfate Main only (OLS) 0.416 1.606 0.796 — — Main only ( W L S ) 0.626 0.945 0.510 — . . . lllllllllllllll Main + T 0.685 0.983 0.488 0.008 0.250 0.975 Ma in + R H 0.908 0.993 0.364 0.006 0.108 0.958 Main + C O 0.552 0.970 0.571 1.977 5.204 0.705 Main + O 3 0.452 1.394 0.747 16.716 97.976 0.865 Main + B D use 0.649 0.947 0.496 -0.877 1.056 0.409 Table 3.32. SDNN regresssions; one subject + one case excluded. 93 3.5.2.6 R-MSSD Inconsistent results were found for relationships predicting r-MSSD, which were similar, but more stable, to the effects observed for HR and SDNN. An example of the relationship with ambient PMio exposure is shown in Figure 3.47; Figure 3.48 displays the effect estimates resulting from the various relationships with main variables. Single-pollutant WLS models reported negative relationships for PM and positive relationships against the sulfate measures. T, CO and O 3 in single-pollutant models were all moderately inversely related to r-MSSD (data not shown) and their coefficients increased in significance when added to the two-pollutant models (Table 3.33). These secondary variables tended to increase all main variable coefficients resulting in less negative coefficients for ambient PMio and PM2.5 (confounding) and more positive coefficients for sulfate (enhancing). Signs of coefficients did not change upon the addition of secondary variables, indicating stability yet still inconsistency among the results. 40 30 i w 20 Q CO CO 10 4 0 + "O CD N -10 TJ -20 CO T J C CO CO -30 • • • !•% •• 7# % 10 20 30 40 50 60 Ambient PM10 (ug/m3) Figure 3.47. OLS regression between r-MSSD and ambient P M 1 0 ; one subject excluded. 94 R-MSSD vs. Main Variable Secondary Variable Coefficient SE p-value Coefficient SE p-value Ambient P M , U Main only (OLS) -0 058 0.161 0.721 — — Main only (WLS) -0.186 0.107 0.085* :i:;x;xo:|x|)*i+>t|!;x;xj:j:j:jx;: — Main + T -0.063 0.159 0.695 -0.217 0.219 0.325 Main + RH -0.184 0.152 0.229 -0.007 0.069 0.919 Main + CO -0.119 0.143 0.409 -3.457 4.923 0.485 Main + O 3 -0.204 0.133 0.129 15.984 70.100 0.820 Main + BD use -0.121 0.117 0.307 -0.726 0.552 0.192 At dbtertt PMj* Main only (OLS) -0 283 0.243 0.247 — — — Main only (WLS) -0.155 0.148 0.301 — — Main + T -0.011 0.170 0.950 -0.288 0.177 0.109 Main + RH -0.117 0.176 0.508 0.029 0.062 0.636 Main + CO -0.020 0.178 0.911 -6.199 4.598 0.182 Main + O 3 -0.099 0.202 0.626 -32.733 79.778 0.683 Main + BD use -0.099 0.151 0.514 -0.869 0.532 0.106 Ambient Sulfate : Main only (OLS) 0.055 1.071 0.959 x- x fx-* Main only (WLS) 0.262 0.652 0.689 ——— Main + T 0.584 0.669 0.385 -0.285 0.162 0.083* Main + RH 0.293 0.734 0.692 0.033 0.062 0.599 Main + CO 0.490 0.658 0.458 -6.507 3.778 0.089* Main + O 3 0.944 0.844 0.267 -93.754 74.090 0.210 Main + BD use 0.463 0.649 0.477 -1.021 0.518 0.052** Personal P M 2 5 Main only (OLS) -0 053 0.076 0.485 — — — Main only (WLS) -0.032 0.041 0.442 — Main + T -0.045 0.042 0.289 -0.280 0.161 0.087* Main + RH -0.046 0.049 0.353 0.032 0.061 0.605 Main + CO -0.026 0.040 0.524 -8.406 4.225 0.050** Main + O 3 -0.030 0.041 0.467 -57.535 59.672 0.338 Main + BD use -0.029 0.041 0.487 -0.767 0.536 0.157 Persona! Sulfate Main only (OLS) 0.584 1.220 0.634 ___ — — Main only (WLS) 0.634 0.740 0 395 » * « — — Main + T 1.158 0.780 0.142 -0.344 0.173 0.051* Main + RH 0.635 0.813 0.438 0.036 0.063 0.566 Main + CO 1.034 0.739 0.166 -9.952 4.247 0.022** Main + O 3 1.813 0.977 0.067* -139.520 76.878 0.074* Main + BD use 0.805 0.740 0.280 -0.885 0.546 0.109 Table 3.33. R-MSSD regressions; exluding one subject; ** p<0.05, * p<0.10. 95 R-MSSD response to interquartile range increases in exposure 3 E 1 in in U Ambient PM10 Ambient PM2.5 Ambient Sulfate Personal PM2.5 Personal Sulfate vt 1-2 Figure 3.48. R-MSSD effect estimates and SE for interquartile range increases in exposure. 3.5.3 Summary of relationships Both systolic and diastolic BP decreased with increasing exposures, although the results for diastolic BP were largely insignificant. Most relationships assessing SS BP were slightly confounded by CO. However in two-variable models, coefficients remained mostly negative, suggesting a stable negative relationship between particulates and blood pressure. The SVE arrhythmia variable was increased when regressed against all exposure metrics and there were no prominent confounders. These three variables were found to be sensitive endpoints in our study population. For HR, SDNN and r-MSSD, the PM and sulfate results were largely inconsistent and unstable upon the addition of secondary variables. Very general trends were for HR to increase and HRV to decrease with increasing PM exposures, whereas HR decreased and HRV variables increased with increasing sulfate exposures. Overall, the inconsistency of relationships among the HR and HRV variables and the instability when adding secondary variables suggests these variables not to be sensitive health indicators in our population. 96 CHAPTER 4: DISCUSSION 4.1 Study Population 4.1.1 Participation Each of the study subjects met the original study eligibility criteria except for three subjects that were slightly younger than sixty years (youngest age: 54). Whereas our target sample size was 25-30 participants, the number of subjects recruited was 17, resulting in a 36% participation rate. This low rate can be explained in part by the specific eligibility criteria, which excluded a number o f interested individuals. Additionally, the low participation rate was in part due to the burdensome nature of the sampling sessions and the time span over which these sessions were to be held. For example, the 24-hour sessions had the potential o f being strenuous for the participants depending on the severity of their condition, especially on hot summer days. Even though it was made clear that sampling schedules would be very flexible, committing to a 3-5 month time frame was also difficult for many individuals. Only one other personal sampling study with repeated 24-hour particulate measurements on health-compromised individuals has been found in the literature. This study also assessed a small group (N=10) of C O P D patients for six repeated measures (Bahadori and Koutrakis, 1996). Two studies, involving shorter sampling durations, included 21 cardiorespiratory patients for four 8-hour repeat measurements (Stieb et al., 1998) and 18 C O P D patients for 6-18 consecutive 12-hour sampling days (Rojas-Bracho et al., 1998). Our participation rate of 36% was equal to the Stieb et al. study, in which 21 volunteers participated out of a pool o f 58 potential candidates (Stieb et al., 1998). Therefore, our sample size compared well to those o f similar studies and 10-20 subjects may be a practical limitation for studies of this type. Depending of the scope and nature of the study, recruitment of subjects may be easier when measurements involve shorter sampling periods, or fewer repeat measurements, or when younger, healthier populations are being targeted. However, in order to obtain accurate exposure characteristics data, 24-hour measurements are ideal. This is the most common metric used in epidemiological studies using data from fixed site ambient monitors and this duration may be required to collect sufficient particulate mass for analysis. Fewer repeat measurements would reduce the validity of individual exposure estimates. Lastly, certain populations are thought to be at risk o f being adversely affected by particulate air pollution, thus these populations should be studied as they may have different characteristics compared to healthy children and adults. Unfortunately, these populations may be unable or unwilling to endure the stresses associated with the measurement of personal exposures. These issues point to some limitations of personal sampling studies in trying to collect accurate and statistically interpretable data for health-compromised individuals. While it was difficult initially to recruit subjects, all those who participated remained in the study; the compliance rate was therefore 100%. These individuals were very committed, which was a positive aspect of studying this group. 97 4.1.2 Characteristics Epidemiologic studies have demonstrated that certain health-compromised groups are more susceptible to adverse effects of particulate air pollution. However, most exposure assessment studies to date have analyzed the exposures and activity patterns of healthy children and adults. A n objective of this study was to determine the exposures and time-activity patterns o f a health-compromised group for comparison to data collected on healthy groups and to assess whether their exposure and activity characteristics are different. Table 4.1 presents time-activity characteristics of various age groups of the general, healthy population in comparison to the C O P D patients of this study. These data were obtained from the National Human Activity Pattern Survey ( N H A P S ) which is based upon a representative sample of the U . S . population (Klepeis et al., 1996). Differences can be seen between the various populations, which may be attributed to differences in age and mobility. The retired 65+ population is the most comparable to our study population with respect to age and activity. Figure 4.1 and Figure 4.2 allow a visual comparison of the data for our study population to this retired 65+ group. Between the two groups, differences are observable for time spent at home as opposed to time spent in other locations. The general retired population spends less time at home and more time in environments such as restaurants and public buildings than do the C O P D patients. This retired population is also exposed to more E T S . The Stieb et al. exposure assessment found similar time-activity results to those found in our study (Stieb et al., 1998). For daytime hours only (7 a.m. and 7 p.m.), subjects (non-smokers, ages 49-85, with cardiorespiratory disease) also spent more time indoors (81.8%) and at home (64.0%) and less time outdoors (7.6%) when compared with a general population from the same geographical area. Observed differences in time-activity patterns for health-compromised groups may be attributed to these individuals having less mobility and therefore reducing their activities outside the home. They may also refrain from being exposed to particulate sources such as cooking or E T S due to health effects that result. The lower exposure to cooking for all age groups of the N H A P S compared to the C O P D study population is not expected. Cooking for the C O P D group may have been overestimated due to the time breakdown on our time-activity logs. Subjects were to highlight each location or activity, regardless of the actual amount of time spent, for each half-hour block. However, cooking durations were likely shorter than 30 minutes in most cases (e.g. for preparing toast). There are some discrepancies to take into account when comparing the N H A P S statistics to ours. First, our study was undertaken during the spring/summer season, whereas the results of the N H A P S recorded here reflect year-round data. The N H A P S did analyze their data by season, however only for the entire population. Between seasons for the entire population, it was noted that for summer data, slightly less time was spent indoors at home and slightly more time was spent outdoors both near and away form the home than during the winter. This discrepancy likely increases the differences between the general retired population and C O P D study population presented here. Second, percentages of time spent exposed to E T S for the N H A P S were only calculated for the respondents that were exposed for at least one minute on the day of the survey (not including times when respondents themselves were smoking). Thus, the data 98 does not include individuals with zero exposure. In the N H A P S , 28.3% of retired people were exposed for at least one minute per day, with an average exposure of 23.64% per day. In estimating the percent exposure of the entire 65+ population, the average exposure would be approximately (.283 x 23.64) 6.7%, which is still high compared to the C O P D population of this study. Analysis of the N H A P S reference population indicate clear differences between the retired population and the other, younger population groups (Table 4.1). These differences appear to be even greater for health-compromised individuals as seen by our C O P D study population. Considering their different time-activity characteristics, health-compromised individuals likely have different exposure characteristics compared with the general population. These differences need to be addressed by exposure assessment studies. Other indoors 1.89 Cooking=5.8 ETS=0.6 Restaurant .35 Outdoors 8.80 Figure 4.1. Time-activity characteristics of COPD study population; values represent % time spent in each location or activity over a 24-hour period. Other Indoors 6.97 Restaurant 1.27 Cooking=2.47 ETS=23.64 Outdoo 6.; 4 Figure 4.2. Time-activity characteristics of retired 65+ reference population; values represent % time spent in each location or activity over a 24-hour period. 99 Exposure ETS 25.45 18.49 28.06 23.64 VO © Exposure Cooking/ Food Prep. 0.14 0.22 1.92 2.47 oo l/-> Locations In Vehicle 3.14 4.29 6.43 4.17 cs Locations Near Vehicle 0.56 1.41 2.06 0.99 o ci Locations Other Outdoor 0.96 2.83 2.33 1.27 Locations Residential - Outdoors 5.38 5.05 2.93 4.48 o vd Locations Other indoor 0.42 1.18 2.74 1.07 OV Locations School/ Public Bldg. 3.45 15.33 5.19 2.83 Locations Mall/ Other Store 1.39 1.15 2.77 1.89 Locations Office/ Factory 0.05 0.18 8.42 1.18 Locations Bar/ Restaurant 0.57 0.76 2.43 1.27 d Locations Residential - Indoors 84.08 67.81 64.71 80.84 92.4 « C a T > I ® © * ^ © u L Working* (18-64) Retired* (65+) COPD Study Pop. (54-86) a o s. e a. *» o •c e. s o JB 4 3 on e © o s o •c > a w a. en V s ••e u W> e g fc 3 H 100 4.2 Data Quality 4.2.1 Co-located sampler experiments The co-located sampler experiments were designed to test the operating parameters and limits of the equipment used for PM2.5 collection throughout the study. Experiments 1.1 and 1.2 (area sampling) demonstrated relatively low standard deviations in the PM2.5 concentrations reported by samplers o f one type (0.3-1.1 ug/m 3 for P E M s ; 0.1-0.5 u.g/m3 for His) . This result can be expected for co-located area sampling, in which the conditions for all samplers are equal and do not vary greatly over the sampling period. In comparing the two sampler types under these conditions, the P E M s reported area concentrations on average 1.5 (ig/m 3, or 16%, higher than the H i s . This difference between sampler types is comparable to those found in other particulate measurement comparison studies (Chow, 1995). Janssen et al. found their personal PMio sampler, gave higher concentrations than their PMio H I by 2.4 | ig /m 3 (9%) (Janssen et al., 1998b); both samplers were of the same design as the ones used in this study. Additionally, the comparison between the H I and dichotomous sampler (Dichot) PM2.5 concentration data collected at the Kitsilano site indicated lower levels being measured by the Dichot, although the two measures were highly correlated. We suspect this may be due to inaccurate calibration of the Dichot flows. Experiments 2.1 to 2.4 (personal sampling) found the mean difference between two P E M s worn simultaneously during personal exposure sampling was 2.3 ug/m 3 . The mean percent difference was 11%, higher than measured by the P T E A M study during co-located personal sampling, which found a mean percent difference of 4.6% (Thomas et al., 1993). The concentration differences between two P E M S were greater than those between a P E M and an HI , suggesting that a systematic bias (i.e. between the two sampler types) is less important than random error (i.e. between two identical samplers). Overall, these co-located sampler experiments have demonstrated some uncertainty in the comparisons between ambient and personal concentrations, as well as personal exposures within and between individuals. Even though mass differences were considered in the analysis, the main objective of this study was to examine the correlation between ambient and personal measurements over time, which should be little affected by sampler and sampling differences. 4.2.2 Study data The mean sampling duration for both personal and ambient samples was very close to the planned 24-hour sampling sessions. Use of two batteries connected in series to power the personal sampling pumps worked well. Experiment 1.2 demonstrated pumps on new batteries ran for approximately 35 hours and there was no indication of a shortened battery life at the end of the study. The main problem with personal samples was pump failure when tubing was bent, thereby stopping airflow, or by accidentally dropping the pump. 101 Ambient and personal samples were to be run within the same 24-hour window each day. The mean percent overlap for the personal samples with ambient samples was 95% (range 85-100%). On average, 71 minutes o f personal sampling did not overlap with the ambient samples. This lack of overlap was assumed negligible in comparisons between personal and ambient concentrations, however it may contribute to a small amount of error. The limit o f detection ( L O D ) o f 3.7 ug/m 3 for personal PM 2 . 5 samples was slightly lower than those reported in other studies and suggests a high level of analytical precision. Janssen et al. reported an L O D of 5.2 | ig /m 3 for PM2.5 (Janssen et al., 1998b), while the P T E A M study reported an L O D o f 8.3 ug/m 3 for personal P M i o samples (Thomas et al., 1993). Overall, 98% of our samples used in the analysis were above the L O D . Personal field blanks for PM2.5 indicated greater mass increases than did ambient field blanks, 16 ug and 3 ug respectively. The sources of the background levels were not investigated, however the difference between personal and ambient field blanks could be due to differences in handling, storage or transport o f personal and ambient filters. Additionally, all personal field blanks underwent leak checks in samplers, whereas the ambient field blanks were spare, unused filters after a week of sampling in the field, thus were never used in leak checks. This difference points to a potential source of error in the (blank-corrected) field sample concentrations, which would lead to a slightly reduced difference between personal and ambient concentrations than was actually present. Another potential source of error in this study was due to the actual measurement of personal exposures. Information from the time-activity logs indicated that subjects wore the samplers over their shoulder an average of 42% of time during their sampling sessions. The activity logs also demonstrated that this group spent the majority (92%) o f their time at home on the days o f personal sampling. Although subjects were asked to wear the sampler over their shoulder in order for it to be as close to their breathing zone as possible, during time spent at home, subjects were allowed to take the monitor off, keeping it as near to themselves as possible. The low percentage of time subjects actually wore the equipment is likely due to the large amount of time these individuals spent at home. A s well, at night, subjects placed the samplers near to their beds, as done in other personal sampling studies (Janssen et al., 1998b). Therefore the maximum period subjects could be expected to wear the samplers was 67% of time, assuming they slept 8 hours per day. Thus, subjects actually wore the samplers for approximately 63% of the maximum amount of time possible. Since subjects did not wear the sampler close to their breathing zones during the entire sampling period, the personal measurements may not be an accurate indication of actual breathing zone exposures. This section has identified some uncertainties in comparing personal and ambient concentrations, including differences in the sampling period, handling of filters and the method of sampling. The magnitude o f the sampling period difference (i.e. lack o f overlap o f 71 minutes) between personal and ambient samples was not likely an important source of error in this study. Also, with limited numbers of technicians collecting data during the study, it is impractical to largely reduce this difference. Some other uncertainties, however, may be reduced in future studies. For example, the difference in field blanks may have been avoided by treating ambient blanks more similar to the personal blanks. Lastly, compliance in the measurement o f 24-hour personal samples may be difficult with the present pump technology. Lighter and less noisy pumps may improve the percentage of time subjects actually wear the sampling equipment. 102 4.2.3 Particulate Concentrations The exposures and concentrations for PM2.5 and PMio were generally lower than found in other studies, likely due to differences in study locations. As mentioned in Chapter 1, Vancouver does have lower ambient PM levels in comparison to other urban areas. The proportion of sulfate in our samples was approximately 15-20% of the PM2.5 mass. This result is similar to data averaged across urban sites in Canada (Brook et al., 1997a), however as expected, these levels were lower than those from sites in eastern Canada (where sulfate has been shown to account for 33% of PM2.5 mass) (Brook et al., 1997b). In eastern Canada, mean sulfate levels of 6.9 ug/m3 and 7.7 ug/m3 have been recorded for Toronto and Windsor between April and September during the 1980's (Burnett et al., 1995). We found mean sulfate levels to be 1.9 |tg/m3 for the same season within the GVRD. Even though our sulfate levels were low, since we measured sulfate as a marker of outdoor particles, the actual amount of sulfate was not important for our study. Rather, we were interested in the relationship between personal and ambient measures in terms of their correlation over time. 4.3 Part I - Relationship between Personal and Ambient Concentrations 4.3.1 Correlations Repeated personal exposure measurements were collected in this study, allowing individual regressions between personal exposures and ambient concentrations to be conducted. The median of the individual correlations between personal and ambient PM2.5 over time was moderate (r=0.48). Between individuals, however, the degree of correlation varied greatly, ranging from strong negative to strong positive values, which suggests exposure characteristics for PM2.5 were not constant across the study population. The median correlation we measured was somewhat higher in comparison to the other repeated exposure studies of COPD patients. For COPD patients, Bahadori and Koutrakis found no correlation between personal and outdoor concentrations of PM2.5 (Bahadori and Koutrakis, 1996) and Rojas-Bracho et al. found a median correlation of 0.30 for 18 subjects (Rojas-Bracho et al., 1998). The results found in the present study may be higher than these due to our study subjects having less severe disease or due to the other study locations where air conditioned homes are more prevalent. The median correlation of our study was more similar to studies of adults, but lower than in studies assessing children. For example, in measuring PMio exposures of adults between the ages of 50 and 70 years, the median Pearson's r was found to be 0.50 (Janssen et al., 1998a). A similar value of 0.53 was found for 13 non-smoking adults in the Total Human Environmental Exposure Study (THEES) (Buckley et al., 1991). For children with non-smoking parents, the median correlation between personal and ambient PMio concentrations was 0.63, which improved to 0.73 after excluding days when these children were exposed to ETS (Janssen et al., 1997). Personal PM2.5 exposures in children have also been highly correlated with the ambient (median r=0.86) (Janssen et al., 1999). The lower correlations found in this study and especially in other studies of health-compromised populations, compared with younger and healthier groups, suggests that age and/or health-related factors could diminish the accuracy 103 of fixed-site outdoor monitors to assess their exposures. This coincides with the findings that these subpopulations have different time-activity characteristics, such as spending more time indoors, compared to the general population. Two studies of elderly subjects living in common retirement facilities in Baltimore have found high mean correlations between personal exposures to residential outdoor PM2.5 concentrations, r=0.78 (Williams et al., 1999b), r=0.67 (Williams et al., 1999a). These high correlations may in part be attributed to all subjects living within the same building, which could decrease the variability in correlations across subjects. The results may also be attributed to the use of residential outdoor monitors as opposed to community monitoring sites. This could overcome the element of spatial variability of particulates throughout the study region, a factor suggested to decrease the correlation between personal and ambient measurements. For example, in comparisons of indoor and outdoor PM2.5 concentrations, correlations between regional monitoring sites and inside homes have been low, whereas significant agreement was found between monitors directly outside the homes and inside (Leaderer et al., 1999). In contrast to the results for PM2.5, the correlation between personal and ambient sulfate over time was high (median r=0.96). All individual regressions for sulfate had high r-values, pointing to sulfate as a stable exposure metric across the group. In fact, personal sulfate levels were more highly correlated with all ambient measures than were personal PM2.5 levels; even ambient PM2.5 was more highly correlated with personal sulfate than personal PM2.5. This result is likely due to personal PM2.5 being strongly affected by indoor sources whereas personal sulfate is affected almost entirely by ambient levels. In correlating mean personal samples with ambient measurements, Stieb et al. found r=0.95 in an elderly population with cardiorespiratory symptoms (using the mean of up to four repeat measurements per subject) (Stieb et al., 1998). A correlation of 0.81 has been found for adult volunteers in Watertown and Steubenville (using the mean of three repeat measurements per subject) (Dockery and Spengler, 1981). In an analysis of pooled samples with two repeat measurements per subject, r=0.89 has been reported in a study of 24 children in Uniontown, Pennsylvania (Suh et al., 1992). Additionally, Brauer et al. found r=0.79 for adult volunteers in Boston where measurements were collected during scripted activities (Brauer et al., 1989). The high correlations found in these studies and again in the present study reflect the characteristics of sulfate particles, such as their lack of indoor sources, and high penetration and persistence indoors, where people spend a majority of their time. A limitation in assessing individual r-values is their low precision due to the limited number of observations per subject used to calculate the individual correlation coefficients. An interesting finding using the individual values, however, was that for both exposure metrics, correlations improved the closer personal exposures were to the ambient level. This relationship was significant for PM2.5, where correlations were lowest for individuals with high personal exposures. This result is expected if factors other than ambient concentration were responsible for these high exposures. Use of ambient PM2.5 concentration as a predictor of personal exposure, therefore, may not be valid since the degree of correlation is dependent upon the level of personal exposures. Ambient levels may only be a good measure of personal exposure when other factors that increase or decrease exposure are not present. 104 Cross-sectional exposure assessment studies (in which single personal exposure measurements of a study population are pooled and related to measurements made at a central outdoor monitoring site) have generally failed to find significant associations between personal and ambient PM measures. Overall, lower cross-sectional results are likely due to variances in exposure characteristics between individuals; exposures may be predictable for any given individual, but not across a group of subjects. This observation was reproduced in our analyses, where regressions of pooled data gave lower levels of agreement than the median of individual regressions for most relationships studied. The result was more pronounced in analyses using personal PM2.5 than personal sulfate. Some recent studies have also shown higher within subject correlations than cross-sectional correlations. In the PTEAM pilot study, PM2.5 exposures were repeatedly measured in nine households, with two individuals per house. Cross-sectionally, personal exposures were not correlated with ambient levels, however individual regressions of 10 individuals with 6 to 8 measurements each, ranged from -0.17 to 0.76 (median r = 0.35) (Wallace, 1996). Additionally, in the Janssen et al. studies of children and adults, the median correlations between personal and ambient PMio of 0.63 and 0.50 were higher than estimated cross-sectional correlations of 0.28 and 0.34, respectively (lanssen et al., 1998a; lanssen et al., 1997). The lower levels of agreement found with regressions of pooled data compared to the median of individual regressions suggest that exposure misclassification is a greater concern for cross-sectional as opposed to time-series study designs. Since time-series epidemiologic studies relate day-to-day variations in outdoor concentrations to individual health effects, it is more appropriate to consider these individual regressions that regressions across subjects (Wallace, 1996). Overall, these data suggest that the use of ambient PM2.5 as an exposure metric in epidemiological studies incorporates considerable error since the correlation between personal and ambient PM2.5, concentrations over time was moderate and correlations were low for individuals with high personal exposures. The results for sulfate were more encouraging. High correlations between individuals over all ranges of personal exposure indicate that day-to-day variability in measurements of sulfate concentrations made at fixed-site ambient monitors accurately reflects variability in personal exposures to sulfate. 4.3.2 Ratios and differences The correlations between personal and ambient indicated clear differences between PM2.5 and sulfate, which we hypothesized in part would be due to differences in the presence of indoor sources. Ratios and differences between personal and ambient concentrations were evaluated to support this hypothesis. Our comparison of paired personal and ambient samples indicated that, for the majority of the study population, personal exposures to PM2.5 were above ambient levels. On average, personal PM2.5 concentrations were greater than the ambient by 6.9 ug/m3 (SD: 14.6 ug/m3), which was a significant difference. We also found that increases in personal PM2.5 exposures were not matched by increases in the ambient level, suggesting that ambient levels do not dominate high personal concentrations. The results suggest that factors other than ambient 105 PM2.5 level were important in contributing to personal exposures of PM2.5 and that individuals were exposed to particulate sources not measured by outdoor monitors. Increased personal levels compared with indoor and outdoor levels is consistent with other personal sampling studies of children and adults (Janssen et al., 1998a; Ozkaynak et al., 1996; Sexton et al., 1984; Wallace, 1996; Watt et al., 1995). There are two possible explanations for these findings. Firstly, indoor sources can influence personal exposures greater than outdoor levels, because people generally spend larger amounts of time indoors compared to outdoors. Secondly, the difference between personal concentrations and indoor levels have been labeled the "personal cloud", the direct cause of which remains unknown. Plausible explanations for this phenomenon include potential electrostatic effects, closer proximity of individuals to particulate sources or resuspended indoor dust from activities, which will affect personal monitors to a greater extent than indoor stationary monitors (Wallace, 1996). Some studies have tested the hypothesis that indoor activities can increase personal exposures. In an experimental study, Brauer et al. found that the personal cloud effect was greater for an experiment in which a subject was active as opposed to being sedentary (Brauer et al., 1999). Monn et al. demonstrated the influence of activity on indoor/outdoor (I/O) ratios (Monn et al., 1997). In homes where inhabitants were present and conducted normal daily activities, PM2.5 I/O ratios were above one; homes without inhabitants exhibited ratios below one. In this study, ETS, gas stoves and activity level were found to be important indoor sources. Activities, such as dusting, vacuum cleaning and spraying have been suggested as particle-generating activities (Clayton et al., 1993; Spengler et al., 1981). In contrast to our results, two previous studies of COPD patients have not found increased personal exposures (Bahadori and Koutrakis, 1996; Linn et al., 1996). In these studies, the amount of time spent at home or lack of mobility of the subjects could be plausible explanations for lack of the personal cloud effect. The patients in these studies may have had more severe COPD compared to our subjects, requiring them to use continuous oxygen and therefore limiting their activity levels. Air conditioned homes may have also limited the impact of ambient-source particles, and to some extent indoor-source particles, on exposure. Additionally, in our study some of the difference between personal and ambient levels can perhaps be attributed to the limitations and uncertainties in methodology discussed previously (although these may also have been factors in the other studies). In comparison to increased personal PM2.5 exposures, sulfate exposures were significantly lower than ambient levels. Lower sulfate exposures have been found in other personal monitoring studies as well (Dockery and Spengler, 1981; Stieb et al., 1998; Suh et al., 1992). Since sulfate does not have any major indoor sources, the ambient level should be the major determinant of personal sulfate exposure. Sulfate particles can be generated by indoor sources such as matches and gas stoves, but the contributions of these are likely small compared to the influence of outdoor sources (Dockery and Spengler, 1981). The observation that personal sulfate exposures were in most cases below ambient levels and that increases in exposure to sulfate were matched by increases in the ambient level, suggests this to be valid. Elemental analysis of outdoor, indoor and personal PMio samples of the PTEAM study found sulfur (out of all elements analyzed) very stable and uniformly distributed among the outdoor, indoor and personal samples (Clayton et al., 106 1993). This was also the only element not elevated in the "personal cloud". The lower personal sulfate level compared to the ambient level, found in the present study, is likely due to lower indoor (where the subjects spent 92% of their time) concentrations of ambient-source sulfate particles. Studies have indicated that indoor sulfate concentrations will be at least 70% of outdoor levels (Brauer et al., 1989). In conclusion, personal PM2.5 exposures were significantly higher than ambient concentrations; at high personal exposures, ambient levels became less representative of personal exposures. Personal sulfate exposures, on the other hand, were significantly lower than ambient levels, although dominated by ambient concentrations. These results coincide with the correlations between personal and ambient levels and support our hypothesis that sources of PM2.5 other than outdoor sources may contribute to personal levels, thereby lowering the agreement with the ambient, which is not the case for sulfate. 4.3.3 Use of closest site data In addition to indoor sources affecting the correlation between personal and ambient levels, we hypothesized that spatial variability of particulates over the study region may also contribute to differences in correlations. Thus, in using data from various ambient monitors dispersed throughout the study region, the objective was to assess whether concentration data from the closest site to subjects' homes would improve estimates of personal exposure. We found that compared to ambient data averaged over all five sites, use of the closest ambient data did not improve the correlation between personal and ambient concentrations. In reviewing the data obtained at each of the five ambient sites, low spatial variability within the study region was tested as an explanation for this finding. Two-way ANOVAs indicated significant differences in the concentrations obtained at two sites (South Burnaby and South Richmond) compared to other sites, suggesting moderate spatial variability for all exposure metrics. However, the correlations between individual sites were high for all exposure metrics and the averaged data was highly correlated with data from each site. These results suggest that trends in particulate levels were comparable within the study region. Generally, 24-hour measurements of sulfates and fine particles are thought to have low spatial variability. Little variation across sites for these metrics was reported for five of the six Harvard Six Cities study locations (Spengler et al., 1981). Homes in the Suh et al. study were located within a 16 km radius of the stationary ambient monitor in Uniontown, Pennsylvania (Suh et al., 1992). When compared to outdoor monitors placed at home sites, the central monitor was found to estimate sulfate concentrations across the city with a high degree of accuracy. Furthermore, the PTEAM study, where homes were located within 0.4-11 km of the fixed-site location, also support findings of high agreement between central and residential sulfate concentrations (r=0.94 and 0.96 for night time and daytime, respectively) (Thomas et al., 1993). Therefore, the appropriate use of a central ambient monitor to assess exposure depends on the spatial variability of the exposure metric being studied. The PMio data indicated lower spatial variability (higher correlations between sites) than the PM2.5 data. This was not expected since the coarse particle fraction of PMio has a shorter atmospheric half-life and transport distance than the fine fraction. Also site-to-site correlations for PM2.5, PMio and PM10-2.5 usually become 107 progressively smaller (Wilson and Suh, 1997). The higher correlations found in this study for PMio may be due to the different measurement methods that were used for PMio and PM2.5 and due to the very low levels of ambient PM2.5 which may lead to a greater role of measurement error. All individual sites indicated similar degrees of correlation with personal exposures of PM2.5 with the exception of the North Burnaby site, which demonstrated very low median correlations with personal exposures. North Burnaby was also not considered the closest site for any of the subjects. Although, use of closest site data did not improve the median correlation for the group, the correlation between personal exposure and ambient concentration was slightly dependent on the distance between a subject's residence and the closest monitoring site. All subjects resided within 7 km of a site, however, higher levels of agreement were found for subjects living closest to a site. From this result, the density of the monitoring network may be important in evaluating personal exposures. For example, unless there is a monitor next to every home, it is difficult to capture differences that affect exposure to PM2.5 In this section, we found that although there may have been moderate spatial variability of the ambient parameters when considering ambient concentrations, trends in particulate levels throughout the study region were similar. These results support previous analyses in which data from the outdoor monitor closest to each subjects' home did not relate to personal exposures better than ambient data averaged over multiple sites throughout the study region. However, we did find that the density of the monitoring network is important and that reducing the distance between homes and ambient sites may slightly improve exposure assessment in epidemiologic studies. 4.3.4 Predictors of Personal Exposure In this study, we found variability in outdoor concentrations did not reflect all of the variability in personal exposures; thus models were produced to assess the combined effect of outdoor levels and personal activities/characteristics on personal exposures. As opposed to analyzing all predictors of personal exposure, some studies have analyzed the difference between personal and outdoor concentrations in relation to various personal characteristics/activities (Janssen et al., 1998a; Janssen et al., 1997). This would assume that all of the particulate measured in ambient air contributes to personal exposures. In this study, however, personal sulfate exposures were found to be lower than the ambient level, which indicates that only a fraction of the ambient fine particles were measured indoors, where subjects spent the majority of their time. Such a result has also been found in other studies (Spengler et al., 1981). We therefore considered all predictors of personal exposure without assuming full contribution by the ambient level. 4.3.4.1 PM2.5 exposure Use of ambient concentrations, time-activity and dwelling characteristics data in a multiple regression model did not explain much variability in personal PM2.5 exposures. The five variables included in our model were chosen due to their association with personal PM2.5 exposure when regressed individually, such as house volume and exposure to ETS. Exposure to 108 cooking and ambient PM2.5 concentration were chosen due to their potentially important impacts on personal exposures as found in other studies (Ozkaynak et al., 1996). Time spent at home was also included in our model due to the large amount of time our study population actually spent in their residence and since indoor concentrations have been identified as an important variable in assessing personal exposures (Spengler et al., 1985). The overall model fit of our multiple regression was low (R2=0.27, R=0.52) but did represent an increase in fit over a model including only ambient concentrations (R2=0.022, R=0.15). Spengler et al. concluded that for ordered analyses, such as the one conducted in this study, use of central site ambient data and various time-activity variables generally predicts less the 20% of the variance in personal exposures (Spengler et al., 1985). As an alternative, a time-weighted exposure model was able to explain up to 64% of the variance in mean personal exposures. However, such models require both indoor and outdoor particulate measurements in conjunction with time-activity profiles. They are based on the expression: E =fiCi + f0C0, where E is sulfate exposure, / is the fraction of time spent indoors (i) or outdoors (o) and C is the sulfate concentration in each location. Both house volume and time spent near ETS were significant predictors of personal exposure. The relationship with volume can be explained by two relationships. First the impact of indoor PM2.5 sources will be greater for homes with smaller volumes because of the smaller space in which particulates can disperse. Second, these homes are more likely to have higher air exchange rates than larger homes, resulting a higher contribution of outdoor particles on exposures than larger volume homes (Wallace, 1996). Studies have shown ETS to be the major contributor to personal exposures (Sexton et al., 1984) and that exclusion of cases with ETS exposure improves the correlation between personal exposures and ambient concentrations (Janssen et al., 1998a; Janssen et al., 1997). For example, Monn et al. found a poor correlation between personal and indoor levels (r=0.385) when ETS exposure was not accounted for (Monn et al., 1997). The result improved after excluding ETS exposed cases (r=0.71). In a review of the three largest studies of indoor air particles in the U.S., Wallace concluded that the single largest indoor source of fine particles is cigarette smoke, for homes with smokers (Wallace, 1996). Although no smokers lived in the homes of our study, subjects were exposed to ETS on 11 out of 106 sampling days. These days were not excluded from the simple regressions between personal and ambient concentrations, as done in the studies mentioned above. However, due to the larger effect from ETS in the model, it could overwhelm the comparably smaller effects of other variables, such as exposure to cooking. Thus, cases with ETS exposure were excluded in the ordinary least squares approach to determine whether the predictive power of any other variables would be increased. Excluding cases with ETS exposure resulted in a slightly worse predictive model and it did not increase the significance of any other variables. The samples with ETS exposure were largely from one subject who was exposed for some period during all sampling sessions. Therefore, excluding these samples simply excluded this subject from the analysis. Exposures from cooking and from time spent at home among the remaining samples were not large enough for these variables to enter into the model. 109 Overall, regression models incorporating time-activity and dwelling information offered improvement in predicting personal exposure to PM2.5, however still left much of the variance in exposure unexplained. There may be additional factors leading to personal exposure which were not accounted for, or our independent variables were too crude to establish significant relationships. 4.3.4.2 Sulfate exposure As found in the simple regressions between personal and ambient sulfate concentrations, personal exposures were highly correlated with the ambient concentrations. This also allowed for a highly predictive multiple regression model for sulfate. Our model included ambient sulfate concentration, house volume and time spent outdoors. All three variables were significant predictors of personal exposure, however ambient concentration accounted for the highest contribution. Use of ambient sulfate alone resulted in an R2 of 0.75 (R=0.87), compared to the full model R2 of 0.82 (R=0.90). As found for the model with PM2.5, house volume was negatively related to personal exposure, which is likely due to the higher air exchange rates in smaller homes. Although ETS exposure was significantly correlated with personal sulfate when regressed individually, this variable had no added effect on the model. Other time-activity and dwelling variables did not add to the model either. Many variables, including exposure to cooking, ETS, time spent at home, would not be expected to enter into the model as sulfate particles are largely from outdoor sources. Other, unmeasured, indoor sources of sulfate, such as matches or gas stoves, were assumed to have an insignificant effect. Other studies assessing the effect of factors predicting personal sulfate exposure have used various approaches. Both Dockery and Spengler and Suh et al. developed time-weighted models, which were found to be an improvement over the use of only ambient concentration data (Dockery and Spengler, 1981; Suh et al., 1992). For example, Dockery and Spengler found a slight increase in correlation from 0.81 to 0.875 when applying the model to their activity and ambient concentration data (Dockery and Spengler, 1981). Our study found that ambient concentration alone was highly predictive, and the addition of house volume and time spent outdoors offered a small improvement to the model fit (R=0.87 to R=0.90). Stieb et al. found that time spent outdoors was associated with a significant interaction term when introduced into regressions between personal and outdoor concentrations (Stieb et al., 1998). Although this interaction term was statistically significant, it did not improve the fit of the model largely compared to simply using outdoor concentration data. In summary, outdoor sulfate data alone were highly predictive of personal sulfate exposures. Time-activity data offered only slight improvement on this relationship. More improvement upon use of this data was observed for the personal PM2.5 model likely because the ambient levels do not explain as much of the variance in personal exposures. 110 4 . 3 . 5 Conclusions and implications of the exposure assessment As epidemiologic studies aim to attribute observed health effects to outdoor air pollution, the contribution of outdoor air pollutants to population exposure must be assessed (Spengler et al., 1981). In considering the use of ambient measurements in air pollution epidemiology, two questions with respect to assessing effects on health can be considered. First, does a certain level of exposure lead to health effects? For particulate standards to be set, policy makers search for the level of pollution below which there will be minimal risk to public health. Since standards are set for particulates in ambient air, an exposure metric is needed that can be measured by ambient monitors, yet still reflect the personal exposures of the public to these particulates. Alternatively, do increases/decreases in exposure (above some baseline) lead to health effects? If trends, rather than specific particulate levels, are of concern, an exposure metric is required for which the day-to-day variability in personal exposures reflect the variability of ambient concentrations being measured. This second question has been addressed by the present study. We found sulfate to be a better exposure metric than PM2.5. Personal and ambient measures of sulfate were highly correlated over time, unlike the moderate correlation found for PM2.5. The individual regressions demonstrated that ambient sulfate is a consistently strong predictor across all individuals and all levels of exposure, whereas for PM2.5 correlations varied by individual and depended on the level of personal exposure. In terms of quantitative levels, ambient sulfate concentrations were higher than personal exposures, however they likely to contribute close to 100% of personal exposures as there are no significant indoor sources of sulfate. Although indoor sources likely contribute to personal exposures of PM2.5, accounting for such variables did not improve our ability to predict personal exposures. Overall, these findings suggest that outdoor levels of PM2.5 are not tightly linked to variation in personal exposures and that use of ambient PM2.5 data as a surrogate for personal exposures would tend to misclassify personal exposures. Such exposure misclassification could be reduced by use of sulfate as the exposure metric. As expressed earlier in this chapter, considering their different time-activity characteristics, health-compromised individuals likely have different exposure characteristics compared with the general population, which should be addressed by exposure assessment studies. Use of sulfate as the exposure metric may decrease the need for personal monitoring studies of at risk populations as personal and ambient concentrations were highly correlated and use of time-activity data did not highly increase the predictive ability of ambient sulfate concentrations. Lastly, use of ambient sulfate levels as a surrogate for exposures could also be acceptable, as sulfate has been correlated with health effects in epidemiologic studies. Although sulfate is not likely the causal factor in generating the observed effects, it has been implicated as being a more stable measurement of or a surrogate for the more likely causal factors, such as PM2.5 or the strong acid component of ambient particulate matter (Lippmann and Thurston, 1996). Ill 4.3.6 Limitations and recommendations for future exposure assessments Limitations of the study that have been referred to throughout the above sections are summarized and commented on below. Limitations generally arose from the size of the study population, the equipment used to measure particulate concentrations, the levels of particulates in Vancouver, as well as the use of time-activity log sheets. With some difficulties in recruiting COPD patients to participate, we were only able to follow a small study population. Thus, this group cannot be considered entirely representative of all COPD patients. In addition, the small study population resulted in fewer numbers of personal samples than projected. In studies of health-compromised individuals, for whom the stresses associated with the measurement of personal exposures are greater than for healthy individuals, their efforts should be acknowledged and care should be taken to keep subjects involved until the end of the study to render meaningful results. From reviewing the co-located sampler experiments and quality control measures taken, we found reasons for slight uncertainties in the recorded concentrations for personal and ambient measurements. Uncertainties due to the use of different sampler types, the sampling period, filter handling and the method of sampling were discussed. These are likely limitations for sampling studies in general. The very low levels of ambient P M 2 . 5 and the little variability in these levels may also lead to a greater role of measurement error. Limitations in the accurate collection of time-activity data (such as reliance on subjects' desire/motivation to remember and highlight the log sheet; the half-hour blocks only yield crude estimations of time spent in each location/activity) likely affect the usefulness of the variables for multivariate modeling. Some recent studies have used other types of time-activity measuring devices, such as handheld computers, which may prove more useful in future research for predicting personal exposures (Brauer et al., 1999). The moderate and variable correlation between personal and ambient levels for P M 2 . 5 may be due to the number of repeated measurements as being low. The high correlation found for sulfate in this study has implied sulfate to be a better exposure metric than P M 2 . 5 and that the number of repeated measurements was sufficient in determining such a high correlation. However, the results for sulfate require verification in other studies and, in particular, in other locations. For example, kerosene space heaters, used in some parts of the U.S. may be an indoor source of sulfate for these locations (Leaderer et al., 1986), which would likely decrease the correlation between personal and ambient sulfate measures. 112 4.4 Part II - Assessment of Cardiovascular Health Effects in Relation to Exposure 4.4.1 Hypothesis In this study, our goal was to assess the cardiovascular health of COPD patients in relation to particulates. As discussed in Chapter 1, hypotheses regarding the cardiovascular effects of particulates generally suggest that they may diminish the ability of the lungs to oxygenate the blood or the ability of the blood to deliver oxygen (Morris et al., 1995; Schwartz and Morris, 1995; Seaton et al., 1995). There were multiple reasons for assessing the cardiovascular health effects of individuals with COPD in association with particulates. First, epidemiologic studies have demonstrated higher risks of particle health effects for individuals with COPD. Thus, COPD patients have been targeted as a group susceptible of being adversely affected by particulate air pollution. Second, respiratory disease can compromise the cardiovascular system. The cause of adverse particle effects in individuals affected by COPD may involve a cardiac mechanism that is equally or more important than other mechanisms, which have greater involvement of the lungs. For example, the important factors leading to pulmonary hypertension in COPD are chronic hypoxia and hypoxic pulmonary vasoconstriction (a protective reflex in presence of decreased oxygen) (Matthay, 1983). Additionally, secondary polycythemia, an increase in circulating red blood cells above normal, is triggered in situations of arterial hypoxia. Polycythemia increases red blood cell clumping and blood viscosity, therefore can aggravate pulmonary hypertension, for example by increasing the work of perfusion (Denison and Morgan, 1983). Pulmonary hypertension and increased pulmonary vascular resistance in chronic airflow obstruction places a large burden on the right ventricle of the heart. Considering these cardiovascular consequences of COPD and that these individuals have been targeted as more susceptible to particle health effects, our goal was to assess whether exposure to particulates can alter the existing cardiovascular conditions of COPD patients. 4.4.2 Blood pressure The mean blood pressure (systolic/diastolic in mm Hg) in our study population was 134/67. These measurements are higher compared to the blood pressures of normal healthy 22-34 year old adults, reported as 110/59 (Hausberg et al., 1997). Blood pressures were more similar to subjects of similar age (age 65-89), with and without cardiovascular disease, recruited from a retirement facility (132/78) (Liao et al., 1999). In relation to particulates, we found decreases in both systolic and diastolic blood pressure with increasing exposure. The relationship was consistent across all exposure metrics analyzed, with significant results for systolic blood pressure responses to ambient PMio and personal PM2.5 (p=0.05 and p=0.10, respectively). Relationships for systolic blood pressure, except for personal PM2.5, appeared to be slightly confounded by carbon monoxide (CO), which in most cases was negatively related with blood pressure in two-variable models. Experimental studies have demonstrated that inhalation of cigarette smoke can cause pulmonary vasodilation due to NO and 113 CO in the vapour phase of cigarette smoke (Zhu and Parmley, 1995). In a review of literature examining the blood pressure effects of CO, Penney and Howley also conclude that CO is more likely to cause hypotension than hypertension (Penney and Howley, 1991). Therefore, the negative association of CO with blood pressure in our results appears to be consistent with other studies. Nevertheless, for PM and sulfate, the directions of the relationships were mostly unchanged with the addition of CO as the second variable when predicting systolic blood pressure and did not affect the diastolic blood pressure relationships. The negative association between blood pressure and particulates therefore appears to be stable and not confounded by other major air pollutants. Our result is in contrast with limited studies that have either demonstrated increases blood pressure or no change upon exposure to particulates. In a study of elderly Boston residents who were observed up to 12 times each, increased blood pressure was associated with PM2.5 levels greater than the median of the daily average concentrations of 3 days prior to testing (Gold et al., 1998). Blood pressure taken from 30 Los Angeles COPD patients on 4 consecutive days rose significantly with ambient PMio (Linn et al., 1998). An experimental study on one subject did not observe any changes in blood pressure in association with exposures to concentrated ambient particles (Gong et al., 1999). Considering hypotheses regarding cardiovascular effects of particles, increases in blood pressure in response to increases in particles seems likely if decreases in blood oxygen or inflammatory mediators cause an increase in blood viscosity. It is possible that medications used by subjects throughout the study could have had a confounding effect on our relationships. For example, bronchodilators, which were regularly used by the study population, can act to dilate blood vessels of the pulmonary and systemic circulation, in addition to their effect of bronchodilatation (Clark, 1983). Such vasodilatation could likely decrease blood pressure. However, analyses conducted in this thesis and by Fisher suggested that bronchodilator use was neither related to exposures nor to blood pressure responses of our study population (Fisher, 1999). Because both elderly subjects (Gold et al., 1998) and subjects with COPD (Linn et al., 1998) have demonstrated increases in blood pressure in response to the same types of particles assessed in the present study, the negative effect we observed is difficult to explain. Nevertheless, due to the stable relationship with exposure, blood pressure was found to be a sensitive indicator in this population. 4.4.3 Cardiac arrhythmia This study analyzed supraventricular ectopic beats (SVEs) of subjects as a measure of cardiac arrhythmia. The geometric mean SVE of the study population was 8.4 beats per hour. No studies were found to compare this value, however in review of the literature concerning cardiac arrhythmias in COPD, Gorecka concluded that both supraventricular and ventricular cardiac arrhythmias are common in this population (Gorecka, 1997). An estimated 80-90% of patients are affected. Various factors, such as physiologic abnormalities associated with COPD and medications used in the management of the disease are thought to be causes of the arrhythmias in these patients. 114 We found an increase in SVEs associated with both particulates and sulfate and of the variables tested none were confounders of these relationships. The Linn et al. study of COPD patients is the only other study found in the literature citing results of an analysis of SVEs in relation to particulates (Linn et al., 1998). In this study, SVEs were also found to increase with outdoor PMio concentrations. Though bronchodilator use did not appear to affect our relationships, bronchodilators and other medications have been targeted as the cause of arrhythmias in COPD patients (Webb-Johnson, 1976). Thus, medication use may have been a poorly controlled confounder in our results. Assuming our observations indicate a direct relationship between particulate exposure and SVEs, it is important to determine whether increases in SVEs adversely affect individuals with COPD and whether the magnitude of the increases was important. For example, whereas increases in total mortality in episodes of increased air pollution can be considered a strong indicator of an adverse health effect, small changes in clinical indicators due to exposure are weaker indicators (Bates, 1992). For cardiac arrhythmias, their presence during exacerbations of COPD or during acute respiratory failure is associated with poor prognosis (Gorecka, 1997). The presence of severe ventricular arrhythmias is also noted as having a poor prognosis. However, the presence of arrhythmias in patients with stable COPD, as more the case in our study, seems to be of lesser importance. In 64 COPD patients with severe hypoxemia SVEs were observed in 86% of subjects during a 24-hour Holter monitoring session (Shih et al., 1988). From 35 patients that died in the follow-up period (26.6 ± 11.8 months), increased heart rate and a history of coronary heart disease were significant predictors of death, whereas no relationship between arrhythmias and survival was detected. The authors suggest this could be due to the relatively homogeneous and severe disease of study patients. 24-hour Holter monitoring was also conducted in second study of 65 COPD patients (Lewczuk et al., 1992). Similarly, a high prevalence of cardiac arrhythmias (including SVEs and VEs) was found in these patients, but arrhythmias did not seem to influence their prognosis. From our associations with particulates, the effect estimate for a 7 ug/m3 interquartile range increase in ambient PMio exposure was 1.1 beats/hour. It remains to be assessed whether this magnitude of increase would be an adverse effect in our COPD population. From a study of normal and health-compromised rats exposed to residual oil fly ash (ROFA) particles, which demonstrated dose-related increases in the incidence and duration of serious arrhythmic events, rats that died appeared to follow one of two possible chain of events prior to death (Watkinson et al., 1998). In one scenario, rats died abruptly, which the authors described as due to "the initiation and propagation of an ectopically stimulated depolarization which 'captured' the heartbeat, producing a fatal arrhythmia" upon exposure. Thus, if one arrhythmic beat can cause mortality, any increase in arrhythmias due to particulate exposure may prove to be serious. Overall, the increases in SVEs due to pollution were consistent across the numerous relationships tested and did not appear to be affected by the confounding factors that were tested. These results suggest this indicator also to be a sensitive endpoint to study. However, the implications of these results remain to be assessed. 115 4.4.4 Heart rate The mean heart rate of for our study population was 81 bpm. This value is higher compared to both normal, healthy, 22-24 year old adults (HR: 57-61 bpm) (Hausberg et al., 1997) and elderly subjects, ages 65-89, recruited from a retirement facility (HR: 73 ± 7.8 bpm) (Liao et al., 1999). The higher heart rates in our group could be due to the measurement of resting and non-resting periods (24-hours) as opposed to short-term resting measurements. COPD patients may also have higher heart rates at rest due to the higher cardiac output caused by their disease. For example, Pagani et al. reported a mean heart rate of 78 bpm for COPD patients (age 53-61; mean FEVi of 1.75), which was higher than both young and elderly healthy groups assessed in the same study (mean HRs: 69 and 66 respectively) (Pagani et al., 1996). In contrast to the consistent relationships observed for blood pressure and SVEs, the results for heart rate were inconsistent between the various exposure metrics analyzed. The ambient PM measures suggested increases in heart rate with increasing exposure, whereas increasing sulfate exposure was associated negatively with heart rate. Additionally, the relationship for ambient PM2.5 appeared to be confounded by temperature and ozone, while sulfate relationships were enhanced to significant levels with the addition of temperature and ozone. Shih et al. demonstrated increased heart rate may be a significant indicator of mortality in COPD patients with severe hypoxemia (Shih et al., 1988). The sinus heart rate of living subjects in the follow-up period was 87 ± 110 bpm compared with 93 ± 14 bpm in those who died within this same period. Thus we might expect an increase in heart rate in COPD populations to be an adverse effect. Our results of increased heart rate with ambient PM exposure metrics compare well with other studies reporting findings of a positive relationship (Peters et al., 1998; Pope et al., 1999a). The effect estimates of our data, however, are larger. Pope et al. estimated that a 100 ug/m3 increase in previous day PMio was associated with an average increase of 0.78 bpm. An increase of 200 ug/m3 SO2 was associated with increases of 2.3-2.8 bpm in the Peters et al. study. These results appear low compared to the result of our study, which estimates a 100 ug/m3 increase in PMio would lead to a 6 bpm increase in heart rate. However, we found all other exposure metrics, sulfate in particular, reported negative relationships with heart rate. In a more detailed presentation of the Pope et al. study, Dockery et al. present numerous regression results for heart rate vs. PMio for various breakdowns of the study population (Dockery et al., 1999). Interestingly, when stratifying by health status (e.g. history of respiratory only, cardiac only, both or none), only the relationship between subjects with no history of respiratory or cardiac conditions and PMio was significant (p<0.01). In contrast, negative relationships were consistently observed for individuals with a history of respiratory disease only, although the sample size for these relationships was very low (N=4). Decreased heart rate has also been observed in rats following exposure to ROFA particles (Campen et al., 1996; Watkinson et al., 1998). In the second scenario of rats dying in the Watkinson et al. study, rats demonstrated evidence of progressive hypoxic failure of the myocardium through decreased heart rate, and various ECG waveform abnormalities 116 (Watkinson et al., 1998). Thus, if decreased heart rate is a characteristic of myocardial hypoxia, perhaps such a response to particulate exposure for those with pre-existing disease may be expected. Therefore, to use heart rate as an indicator of cardiovascular health, the direction of relationships and their implications remain to be assessed. Due to the inconsistencies between exposure metrics, heart rate did not seem to be as sensitive an indicator as blood pressure or SVEs in our study. 4.4.5 Heart rate variability The final indicators of cardiovascular health to be assessed were those measuring heart rate variability (HRV). Our subjects had a mean SDNN of 97 ms and a mean r-MSSD of 44 ms. The approximate normal values for SDNN and r-MSSD obtained from studies using small number of subjects (not adjusted for age, sex or environment) were reported as 141 ms and 27 ms, respectively (Task Force, 1996). In healthy individuals, with ages between 60 and 69 years, mean SDNN values for females and males were found to be 148 ms and 156 ms respectively. Values for r-MSSD for this age category were 39 ms in women and 32 ms in men (Jensen-Urstad et al., 1997). Thus, values for our study population were decreased for SDNN and similar or slightly higher for r-MSSD compared to healthy populations. Studies indicate that HRV reflects oscillations in the sympathetic/parasympathetic balance (autonomic outflows) that is influenced, among other mechanisms, by respiration (Cowan, 1995) . Respiratory modulation of R-R variability has been found to be significantly smaller in COPD patients compared to healthy individuals of similar and younger ages. For example, decreased R-R variability in COPD patients has been significantly associated with decreased respiratory function (assessed by FEVi) (Pagani et al., 1996). It has also been hypothesized that dysfunction of the right heart in COPD due to pulmonary hypertension, can lead to a reduction in R-R variability which is separate from effects due to disturbed lung functions (Pagani et al., 1996) . When regressing against exposure, the heart rate variability responses were inconsistent among the various exposure metrics tested. In general, the HRV variables presented opposite responses to those observed for heart rate. . Most SDNN and r-MSSD relationships decreased with increasing PM exposure and increased with increasing sulfate exposure. These trends were more stable for r-MSSD than for SDNN. Temperature and CO appeared to slightly confound the r-MSSD relationships with PM, however the results remained negative. Studies assessing HRV in relation to particulate exposure generally suggest decreases in SDNN. Liao et al. assessed the day-to-day variations in cardiac autonomic control of 26 elderly (mean age: 81) for three consecutive weeks (Liao et al., 1999). An increased level of PM was associated with significant decreases in HF and LF and a non-significant decrease in SDNN. The largest and statistically significant inverse associations were for individuals with previous cardiovascular-related conditions. Similarly, Gold et al. found lower SDNN with increasing P M 2 5 exposure (Gold et al., 1998). The effect estimate for SDNN was -0.833 for 17 ug/m3 117 interquartile range increase in PM2.5- This result compares well with our study. Although the interquartile range for ambient PM2.5 in our study was only 5.5 ug/m3 (yielding an effect estimate of -0.286), a 17.3 ug/m3 increase in exposure, which was the difference between the maximum ambient level and the study mean, yielded an effect estimate of-0.900. The results for r-MSSD have been inconsistent in the few studies analyzing this outcome. For example, Gold et al. found lower r-MSSD as a result of particulate exposure (Gold et al., 1998). Another study, which assessed 7 subjects for a total of 29 samples, while finding decreases in SDNN and SDANN, found increases in r-MSSD were associated with increased particulate levels (Pope et al., 1999b). Many clinical studies have indicated that lower HRV is predictive of disease or death. How the day-to-day change in HRV is related with day-to-day change in cardiovascular disease is another question (Liao et al., 1999). The above studies have suggested that day-to-day changes in particulate matter are associated with a decrease in cardiac autonomic control; that exposure causes an increase in sympathetic activity, assessed by increased heart rate and decreased heart rate variability. The results of our study, with respect to particulate matter, support these suggestions. However, from reviewing the limited literature on this subject, different HRV variables may show opposite effects and the present study results have indicated that different exposure metrics may also show contradicting results. We found the results for sulfate were generally in the opposite direction than the results for particulate matter. It is difficult to compare the results for sulfate since no studies have assessed this exposure metric in relation to cardiovascular outcomes. However, relationships using ambient measures of PMio, PM2.5 and sulfate and personal sulfate were expected to show similar directions, since these measures were all highly correlated (ambient PMio vs. PM2.5, r=0.72; ambient PM2.5 vs. sulfate, r=0.77). At least the personal and ambient sulfate metrics, for which correlations in Part I were the highest, had the same effects in exposure-response relationships. More research is needed to determine the effect we are looking for (as adverse) and their implications. 4.4.6 Comparing effect estimates of exposure metrics A key research question in studying the health effects of particulate air pollution involves a determination of the component or components of particulate matter that are most closely related to the noted acute PM-mortality associations (Thurston, 1996). In this study, ambient PM2.5 data was expected to produce the highest effect estimates, since these particles have been generally pointed to as causing the increases in morbidity and mortality associated with particulate air pollution. In Part I, sulfate was used as a marker of outdoor-source particulates to support the validity of epidemiologic studies that use ambient measurements to assess exposures of the study population. An unexpected finding was that for most relationships, the highest effect estimates were produced by ambient PMio exposures. These high PMio results may be due to lack of control for meteorology. We did control for both temperature and relative humidity, however there may be other factors, such as barometric pressure or wind speed, which could have affected the results. 118 Epidemiologic studies have demonstrated that outdoor-source particles have been associated with increased morbidity and mortality. Due to the contribution of indoor-source particles and the lack of correlation between personal and ambient measures of PM2.5, as demonstrated in Part I, personal PM2.5 exposure data was not be expected to be highly related to health outcomes. Indeed, personal PM2.5 levels generally had the least effects on all health indicators analyzed when compared to the effects of other exposure metrics. This result compares well to the Linn et al. study where increases in blood pressure and SVEs were reported for ambient PMio, but not indoor PMio or PM 2 5 (Linn et al., 1998). 4.4.7 Conclusions and implications of cardiovascular health effects This study has proven to be exploratory both with respect to the cardiovascular effects of particulates and with respect to study design issues. First, for COPD patients, cardiovascular parameters were different from healthy groups of similar and younger ages. In this population, we found blood pressure and SVE variables to demonstrate consistent and stable results when regressed against particulate measures. In general, blood pressure decreased and SVEs increased with increasing exposures, with little effect from the secondary variables that were analyzed. The stability of these results suggests blood pressure and SVE to be sensitive indicators of response in COPD patients. The other variables of heart rate and heart rate variability were less stable and may not be as sensitive to indicate changes in cardiovascular parameters. The use of "sensitive indicators" may be the preferred objective in future studies. However, the biomedical impact of these indicators on COPD patients remains to be investigated. For example, at what increases in SVEs would exacerbations of symptoms or death occur? Would the variability in particulate levels in the location of concern be able to cause such an increase? Second, in this study we used a more sophisticated study design when compared to epidemiologic studies of morbidity and mortality. First and foremost, instead of simply assessing exposure of the study population using ambient measurements, we conducted personal exposure monitoring. Second, we obtained detailed health assessments and individual characteristics data, usually not collected in large epidemiologic studies. We also studied a group of individuals thought to be susceptible to the health effects of particles. Even though our study population was much smaller than most epidemiologic studies assessing morbidity and mortality effects and the number of repeat measurements was low, the above improvements in study design were made with the aim of clarifying questions regarding cardiovascular health effects of exposure. However, unexpected results occurred with respect to the directions of the exposure-response relationships and in comparisons of effect estimates from different exposure metrics. The complexity of studying cardiovascular health effects and their potential relationship to air pollution exposure and the lack of consistency observed in the results of this study suggest that the improvements in our study design have not clarified these questions. More insight is needed on how the complexities of this research can be dealt with. 119 4.4.8 Limitations and recommendations for future health studies Though unexpected results were found, this study may serve as guidance for future studies of similar type. An important limitation in predicting cardiovascular health effects is the many other factors that affect cardiovascular health, such as diet and stress. Compared with these risk factors, pollution will likely have low impact. Thus, attempting to decipher the impact of pollution without adequate control of these other risks may have been a potential cause of unexpected results. For example, limitations in the control of meteorological and co-pollutant variables may have been a factor in the larger magnitude of effect estimates observed for ambient PMio compared to the other exposure metrics. Additionally, our control of medication use was poor. There were many medications other than bronchodilators that were used by our study population, which may also have affected cardiovascular parameters. Future studies may wish to contemplate, prior to data collection, how medication information will be used in the later analysis and in what form this information could be collected to increase its accuracy and usefulness in this analysis. Additionally, we did not have any restrictions on medication use throughout the study. In some instances, subjects had just taken their medication before collection of their health data. Depending on the health of future study populations, it may be feasible and worthwhile to restrict acute-acting medication use during certain time periods (e.g. two hours prior to the health assessment). We also had a problem with some subjects' prescriptions being altered throughout the course of the study, which proved difficult to deal with in the analysis. A limitation of the results of this study arises from the population. Even though we set eligibility criteria to control for severity of disease, our study population may have varied greatly in range of functional abnormalities. The effects of neither bronchitis nor emphysema are anatomically uniform, thus each can cause the patient to present with a wide range of symptoms and functional abnormalities (Cherniack, 1991). Assessing the health effects associated with particulate air pollution exposure may be difficult in such a varying population. For example, differences in severity and definitions of chronic obstructive lung disease have in part made it difficult to study determinants of arrhythmias in COPD (Shih et al., 1988). Limitations also arise with respect to the type of health endpoints assessed. In particular, HRV analysis may be too sensitive a technique to use in this type of study (where subjects are active and moving around) and in these patients (whose autonomic control of heart rate is altered). It may be more feasible to use resting HRV data if this type of data are needed. Lastly, our small sample size and low variability in particulate levels further reduced our power to detect an effect. Studies with larger numbers of participants and/or increased number of repeated samples would increase the statistical power of determining effects. It may also be worthwhile to study equal numbers of healthy, elderly, respiratory and cardiovascular disease subjects within one project, to determine to what extent aging and health status affect the exposure-response patterns of particulates and cardiovascular indicators. 120 REFERENCES Bahadori T, and Koutrakis P. Personal, indoor and outdoor exposures to particulate matter often COPD patients living in private residences (based on results from the "Nashville Particulate Personal Monitoring: a Pilot Study"). Proceedings of the Second Colloquium on Particulate Air Pollution and Human Health, Park City, Utah, 91. Balogh M, Larson T, and Mannering F. (1994). Analysis of fine particulate matter near urban roadways. Transportation Research Record 1416:25-32. Bates DV. (1992). Health indices of the adverse effects of air pollution: the question of coherence. Environmental Research 59:336-349. Blomberg A, Nordenhall C, Pourazar J, Sandstrom T, and Adelroth E. (1999). Inflammatory responses following exposure to diesel exhaust - A study of time-kinetics using induced sputum. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A123. Borja-Aburto VH, Castillejos M, Gold DR, Bierzwinski S, and Loomis D. (1998). Mortality and ambient fine particles in southwest Mexico City, 1993-1995. Environmental Health Perspectives 106:849-855. Bosner MS, and Kleiger RE. (1995). Heart rate variability and risk stratification after myocardial infarction. In: Heart Rate Variability, M. Malik and A. J. Camm (eds.), Futura Publishing Company, Inc., Armonk, New York, 331-340. Brauer M, Hirtle RD, Hall AC, and Yip TR. (1999). Monitoring personal fine particle exposure with a particle counter. Journal of Exposure Analysis and Environmental Epidemiology. Brauer M, Koutrakis P, and Spengler JD. (1989). Personal exposure to acidic aerosols and gases. Environmental Science and Technology 23:1408-1412. Brook JR, Dann TF, and Burnett RT. (1997a). The relationship among TSP, PMIO, PM2.5, and inorganic constituents of atmospheric particulate matter at multiple Canadian locations. Journal of the Air and Waste Management Association 47:2-19. Brook JR, Wiebe AH, Woodhouse SA, Audette CV, Dann TF, Callaghan S, Piechowski M, Dabek-Zlotorzynska E, and Dloughy JF. (1997b). Temporal and spatial relationships in fine particle strong acidity, sulphate, PMIO, and PM2.5 across multiple Canadian locations. Atmospheric Environment 31:4223-4236. Buckley TJ, Waldman JM, Freeman NCG, Lioy PD, Marple VA, and Turner WA. (1991). Calibration, intersampler comparison, and field application of a new PM-10 personal air-sampling impactor. Aerosol Science and Technology 14:380-387. 121 Burnett RT, Dales R, Krewski D, Vincent R, Dann T, and Brook JR. (1995). Associations between ambient particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. American Journal of Epidemiology 142:15-22. Burnett RT, Smith-Doiron M, Stieb D, Cakmak S, and Brook JR. (1999). Effects of particulate and gaseous air pollution on cardiorespiratory hospitalizations. Archives of Environmental Health 54:130-139. Campen MJ, Watkinson WP, Lehmann JR, and Costa DL. (1996). Modulation of residual oil fly ash (ROFA) particle toxicity in rats by pulmonary hypertension and ambient temperature (Ta) change. American Journal of Respiratory and Critical Care Medicine 153 :A542. Carter JD, Ghio AJ, Samet JM, and Devlin RB. (1997). Cytokine production by human airway epithelial cells after exposure to an air pollution particle is metal-dependent. Toxicology and Applied Pharmacology 146:180-188. Casolo GC. (1995). Heart rate variability in patients with heart failure. In: Heart Rate Variability, M. Malik and A. J. Camm (eds.), Futura Publishing Company, Inc., Armonk, New York, 449-465. CCME. (1999). Discussion paper on particulate matter (PM) and ozone: Canada-wide standard scenarios for consultation. Canada-Wide Standards Development Committee for PM and Ozone, Canadian Council of Ministers of the Environment. Internet site -http://www.mbnet.mb.ca/ccme/pdfs/pm_ozone/pm_discussion.pdf last updated 20 May 1999, accessed August 1999. Cerutti S, Bianchi AM, and Mainardi LT. (1995). Spectral analysis of the heart rate variability signal. In: Heart Rate Variability, M. Malik and A. J. Camm (eds.), Futura Publishing Company, Inc., Armonk, New York, 63-74. Cherniack NS. (1991). Chronic Obstructive Pulmonary Disease, W. B. Saunders Company, Harcourt Brace Jovanovich Inc., Philadelphia, Pennsylvania. Choudhury AH, Gordian ME, and Morris SS. (1997). Associations between respiratory illness and PMIO air pollution. Archives of Environmental Health 52:113-117. Chow JC. (1995). Measurement methods to determine compliance with ambient air quality standards for suspended particles. Journal of the Air and Waste Management Association 45:320-382. Chung EK. (1983). Principles of Cardiac Arrhythmias, Williams & Wilkins, Baltimore, Maryland. Churg A, and Brauer M. (1997). Human lung parenchyma retains PM2.5. American Journal of Respiratory and Critical Care Medicine 155:2109-2111. 122 Clark TJH. (1983). Treatment of obstructive airways disease with beta2-agonists. In: Pulmonary and Circulatory Abnormalities in Chronic Obstructive Airways Disease: An Evaluation of Pirbuterol, R. G. Richardson (ed), Academic Press Inc. Ltd., London, Royal Society of Medicine International Congress and Symposium Series No. 56:49-52. Clayton CA, Perritt RL, Pellizzari ED, Thomas KW, Whitmore RW, Wallace LA, Ozkaynak H, and Spengler JD. (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 and Environmental Epidemiology 3:227-250. Cowan MJ. (1995). Measurement of heart rate variability. Western Journal of Nursing Research 17:32-48. Cyrys J, Heinrich J, Brauer M, and Wichmann HE. (1998). Spatial variability of acidic aerosols, sulfate and PMIO in Erfurt, Eastern Germany. Journal of Exposure Analysis and Environmental Epidemiology 8:447-464. Delfino RJ, Becklake MR, and Hanley JA. (1994). The relationship of urgent hospital admissions for respiratory illnesses to photochemical air pollution levels in Montreal. Environmental Research 67:1-19. Denison DM, and Morgan MDL. (1983). Cor pulmonale. In: Pulmonary and Circulatory Abnormalities in Chronic Obstructive Airways Disease: An Evaluation of Pirbuterol, R. G. Richardson (ed.), Academic Press Inc. Ltd., London, Royal Society of Medicine International Congress and Symposium Series No. 56:42-48. Dockery DW, Pope CA, 3rd, Xu X, Spengler JD, Ware JH, Fay ME, Ferris BG, Jr., and Speizer FE. (1993). An association between air pollution and mortality in six U.S. cities. New England Journal of Medicine 329:1753-1759. Dockery DW, and Pope CA, 3rd. (1994). Acute respiratory effects of particulate air pollution. Annual Review of Public Health 15:107-132. Dockery DW, Pope CA, 3rd, Kanner RE, Martin Villegas G, and Schwartz J. (1999). Daily changes in oxygen saturation and pulse rate associated with particulate air pollution and barometric pressure. Health Effects Institute Research Report 83:1-19. Dockery DW, Schwartz J, and Spengler JD. (1992). Air pollution and daily mortality: associations with particulates and acid aerosols. Environmental Research 59:362-373. Dockery DW, and Spengler JD. (1981). Personal exposure to respirable particulates and sulfates. Journal of the Air Pollution Control Association 31:153-159. 123 Dyer AR, Persky V, Stamler J, Paul O, Shekelle RB, Berkson DM, Lepper M, Schoenberger JA, and Lindberg HA. (1980). Heart rate as a prognostic factor for coronary heart disease and mortality: findings in three Chicago epidemiologic studies. American Journal of Epidemiology 112:736-749. Falk R, Philipson K, Svartengren M, Jarvis N, Bailey M, and Camner P. (1997). Clearance of particles from small ciliated airways. Experimental Lung Research 23:495-515. Firket J. (1931). The cause of the symptoms found in the Meuse Valley during the fog of December, 1930. Bull Roy Acad Med Belg 11:683-741. Fisher TV. (1999). Exposure of COPD patients to particulate air pollution: An assessment of respiratory health effects, M. Sc., University of British Columbia, Vancouver. Gamble JF, and Lewis RJ. (1996). Health and respirable particulate (PMIO) air pollution: a causal or statistical association? Environmental Health Perspectives 104:838-850. Ghio AJ, Costa DL, and Devlin RB. (1999). Pulmonary toxicity of particles collected from the Utah Valley. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A051. Ghio AJ, Kim C, and Devlin RB. (1998). Human volunteers demonstrate no decrements in pulmonary function and no abnormal EKG changes after exposure to fine particles concentrated from Chapel Hill ambient air. American Journal of Respiratory and Critical Care Medicine 15 7: A15 3. Gillum RF. (1988). The epidemiology of resting heart rate in a national sample of men and women: associations with hypertension, coronary heart disease, blood pressure, and other cardiovascular risk factors. American Heart Journal 116:163-174. Godleski JJ, Sioutas C, Katler M, and Koutrakis P. (1996). Death from inhalation of concentrated air particles in animal models of pulmonary disease. American Journal of Critical Care Medicine 163:A15. Godleski JJ, Sioutas C, Verrier RL, Killingsworth CR, Lovett E, Murthy GGK, Hatch V, Wolfson JM, Ferguson ST, and Koutrakis P. (1997). Inhalation exposure of canines to concentrated ambient air particles. American Journal of Respiratory and Critical Care Medicine 155:A246. Gold DR, Litonjua A, Schwartz J, Verrier M, Milstein R, Larson A, Lovett E, and Verrier R. (1998). Cardiovascular vulnerability to particulate pollution. American Journal of Respiratory and Critical Care Medicine 157:A261. 124 Goldberg RJ, Larson M, and Levy D. (1996). Factors associated with survival to 75 years of age in middle-aged men and women. Archives of Internal Medicine 156:505-509. Goldsmith CA, Imrich A, Danaee H, Ning YY, and Kobzik L. (1998). Analysis of air pollution particulate-mediated oxidant stress in alveolar macrophages. Journal of Toxicology and Environmental Health 54:529-545. Gong HJ, Sioutas C, Linn WS, Terrell S, Terrell L, Anderson KR, Kim S, and Chang MC. (1999). A pilot study of controlled human exposures to concentrated ambient fine particles in metropolitan Los Angeles. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, AO 12. Gordon T, Nadziejko C, Schlesinger R, and Chen LC. (1998). Pulmonary and cardiovascular effects of acute exposure to concentrated ambient particulate matter in rats. Toxicology Letters 96-97:285-288. Gorecka D. (1997). Cardiac arrhythmias in chronic obstructive pulmonary disease. Monaldi Archives for Chest Disease 52:278-281. Hausberg M, Mark AL, Winniford MD, Brown RE, and Somers VK. (1997). Sympathetic and vascular effects of short-term passive smoke exposure in healthy nonsmokers. Circulation 96:282-287. Hessel PA, Goldsmith JR, Wichmann FIE, Wilson WE, and Soskolne CL. (1997). Particulate air pollution and health: Emerging issues and research needs related to the development of air quality standards. Canadian Respiratory Journal 4:323-328. Heyder J. (1982a). Alveolar deposition of inhaled particles in humans. American Industrial Hygiene Association Journal 43:864-866. Heyder J. (1982b). Particle transport onto human airway surfaces. European Journal of Respiratory Diseases. Supplement 119:29-50. Hoek G, and Brunekreef B (1993). Acute effects of a winter air pollution episode on pulmonary function and respiratory symptoms of children. Archives of Environmental Health 48:328-335. Hoek G, and Brunekreef B. (1994). Effects of low-level winter air pollution concentrations on respiratory health of Dutch children. Environmental Research 64:136-150. Janssen NA, Hoek G, Brunekreef B, Harssema H, Mensink I, and Zuidhof A. (1998a). Personal sampling of particles in adults: relation among personal, indoor, and outdoor air concentrations. American Journal of Epidemiology 147:537-547. 125 Janssen NA, Hoek G, Harssema H, and Brunekreef B. (1997). Childhood exposure to PMIO: relation between personal, classroom, and outdoor concentrations. Occupational and Environmental Medicine 54:888-894. Janssen NA, Hoek G, Harssema H, and Brunekreef B. (1998b). Personal sampling of airborne particles: method performance and data quality. Journal of Exposure Analysis and Environmental Epidemiology 8:37-49. Janssen NA, Hoek G, Harssema H, and Brunekreef B. (1999). Personal exposure to fine particles in children correlates closely with ambient fine particles. Archives of Environmental Health 54:95-101. Jensen-Urstad K, Storck N, Bouvier F, Ericson M, Lindblad LE, and Jensen-Urstad M. (1997). Heart rate variability in healthy subjects is related to age and gender. Acta Physiologica Scandinavica 160:235-241. Johnson D. (1999). Air Quality Impact Assessment. B.C. Ministry of Environment, Land and Parks, Skeena Region. Internet site - http://www.env.gov.bc.ca/ske/skeair/assess/aqassess last updated July 1999, accessed August 1999. Kinney PL, and Ozkaynak H. (1991). Associations of daily mortality and air pollution in Los Angeles County. Environmental Research 54:99-120. Kleiger RE, Stein PK, Bosner MS, and Rottman JN. (1995). Time-domain measurements of heart rate variability. In: Heart Rate Variability, M. Malik and A. J. Camm (eds.), Futura Publishing Company, Inc., Armonk, New York, 33-45. Klepeis NP, Tsang AM, and Behar JV. (1996). Analysis of the National Human Activity Pattern Survey respondents from a standpoint of exposure assessment. Final Report under Contract No. 68-01-7325, U.S. EPA, Las Vegas, Nevada. Koenig W, Hombach V, Ernst E, Sund M, Mraz W, and Keil U. (1992). Plasma viscosity as a cardiovascular risk factor. Circulation 86:1045 (abstr). Lay C, Bennett WD, Ghio AJ, Bromberg PA, Costa DL, Kim CS, Koren HS, and Devlin RB. (1999). Pulmonary inflammation induced by iron-containing particles. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A070. Lay JC, Berry CR, Bennett WD, Malone D, Kim CS, and Gerrity TR. (1994). Retention of submicron particles in conducting airways of dogs. American Journal of Respiratory and Critical Care Medicine 149: A841. 126 Leaderer BP, Naeher L, Jankun T, Balenger K, Holford TR, Toth C, Sullivan J, Wolfson JM, and Koutrakis P. (1999). Indoor, outdoor, and regional summer and winter concentrations of PMIO, PM2.5, S042-, H+, NH4+, N03-, NH3, and nitrous acid in homes with and without kerosene space heaters. Environmental Health Perspectives 107:223-231. Leaderer BP, Zagraniski RT, Berwick M, and Stolwijk JA. (1986). Assessment of exposure to indoor air contaminants from combustion sources: methodology and application. American Journal of Epidemiology 124:275-289. Lewczuk J, Wrabec K, Piszko P, Jagas J, Sobkowicz-Wozniak B, Porada A, and Sawa W. (1992). Chronic obstructive lung disease: does accompanying arrhythmia have any clinical significance? KardiologicaPolska 37:301-305. Liao D, Creason J, Shy C, Williams R, Watts R, and Zweidinger R. (1999). Daily variation of particulate air pollution and poor cardiac autonomic control in the elderly. Environmental Health Perspectives 107:521-525. Linn WS, Clark KW, Anderon KR, and Gong H. Short-term particulate exposures and health changes in Los Angeles residents with Chronic Obstructive Pulmonary Disease (COPD). Proceedings of the Second Colloquium on Particulate Air Pollution and Human Health, Park City, Utah, 15. Linn WS, Gong Jr. FL Clark KW, and Anderson KR. (1998). Day-to-day particulate exposures and health changes in Los Angeles area residents with severe COPD. American Journal of Respiratory and Critical Care Medicine 157:A878. Lioy PJ, and Waldman JM. (1989). Acidic sulfate aerosols: characterization and exposure. Environmental Health Perspectives 79:15-34. Lippmann M , and Thurston GD. (1996). Sulfate concentrations as an indicator of ambient particulate matter air pollution for health risk evaluations. Journal of Exposure Analysis and Environmental Epidemiology 6:123-146. Lippmann M , Yeates DB, and Albert RE. (1980). Deposition, retention, and clearance of inhaled particles. British Journal of Industrial Medicine 37:337-362. Logan WPD. (1953). Mortality in the London Fog incident, 1952. Lancet 264:336-338. Marple VA, Rubow KL, Turner W, and Spengler JD. (1987). Low flow rate sharp cut impactors for indoor air sampling: design and calibration. Journal of the Air Pollution Control Association 37:1303-1307. Matthay RA. (1983). Cardiovascular abnormalities in chronic obstructive airways disease. In: Pulmonary and Circulatory Abnormalities in Chronic Obstructive Airways Disease: An Evaluation of Pirbuterol, R. G. Richardson (ed.), Academic Press Inc. Ltd., London, Royal Society of Medicine International Congress and Symposium Series No. 56:33-39. 127 Monn C, Fuchs A, Hogger D, Junker M, Kogelschatz D, Roth N, and Wanner HU. (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. Science of the Total Environment 208:15-21. Morris RD, Naumova EN, and Munasinghe RL. (1995). Ambient air pollution and hospitalization for congestive heart failure among elderly people in seven large US cities. American Journal of Public Health 85:1361-1365. NRC. (1998). Research priorities for airborne particulate matter: I. Immediate priorities and a long-range research portfolio. Committee on Research Priorities for Airborne Particulate Matter, National Research Council. Ostro B, Sanchez JM, Aranda C, and Eskeland GS. (1996). Air pollution and mortality: results from a study of Santiago, Chile. Journal of Exposure Analysis and Environmental Epidemiology 6:97-114. Ostro BD, Eskeland GS, Sanchez JM, and Feyzioglu T. (1999). Air pollution and health effects: A study of medical visits among children in Santiago, Chile. Environmental Health Perspectives 107:69-73. Ostro BD, Lipsett MJ, Mann JK, Krupnick A, and Harrington W. (1993). Air pollution and respiratory morbidity among adults in southern California. American Journal of Epidemiology 137:691-700. Ozkaynak H, Xue J, Spengler J, Wallace L, Pellizzari E, and Jenkins P. (1996). Personal exposure to airborne particles and metals: results from the Particle TEAM study in Riverside, California. Journal of Exposure Analysis and Environmental Epidemiology 6:57-78. Pagani M, Lucini D, Pizzinelli P, Sergi M, Bosisio E, Mela GS, and Malliani A. (1996). Effects of aging and of chronic obstructive pulmonary disease on RR interval variability. Journal of the Autonomic Nervous System 59:125-132. Penney DG, and Howley JW. (1991). Is there a connection between carbon monoxide exposure and hypertension? Environmental Health Perspectives 95:191-198. Peters A, Doring A, Wichmann HE, and Koenig W. (1997). Increased plasma viscosity during an air pollution episode: a link to mortality? Lancet 349:1582-1587. Peters A, Perz S, Doring A, Stieber J, Koenig W, and Wichmann HE. (1999). Activation of the autonomic nervous system and blood coagulability in association with an air pollution episode. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A060. 128 Peters A , Perz S, Doring A , Stieber J, and Wichmann W E . (1998). Increased heart rate during an air pollution episode. American Journal of Respiratory and Critical Care Medicine 157:A880. Pietra G G . (1991). Pathology of the pulmonary vasculature and heart. In: Chronic Obstructive Pulmonary Disease, N . S. Cherniack (ed.), W. B . Saunders Company, Harcourt Brace Jovanovich, Inc., Philadelphia, Pennsylvania, 21-26. Pinkerton K E , Peake J, Plopper C H , Hyde D M , and Tarkington B K . (1995). Particles and the respiratory bronchiole: Patterns of deposition and clearance. American Journal of Respiratory and Critical Care Medicine 151: A263. Poloniecki JD, Atkinson R W , de Leon A P , and Anderson H R . (1997). Dai ly time series for cardiovascular hospital admissions and previous day's air pollution in London, U K . Occupational and Environmental Medicine 54:535-540. Pope C A , 3rd, Bates D V , and Raizenne M E . (1995a). Health effects of particulate air pollution: time for reassessment? Environmental Health Perspectives 103:472-480. Pope C A , 3rd, and Dockery D W . (1992). Acute health effects of P M I O pollution on symptomatic and asymptomatic children. The American Review of Respiratory Disease 145:1123-1128. Pope C A , 3rd, Dockery D W , Kanner R E , Villegas G M , and Schwartz J. (1999a). Oxygen saturation, pulse rate, and particulate air pollution: A daily time-series panel study. American Journal of Respiratory and Critical Care Medicine 159:365-372. Pope C A , 3rd, and Kanner R E . (1993). Acute effects of P M I O pollution on pulmonary function o f smokers with mild to moderate chronic obstructive pulmonary disease. The American Review of Respiratory Disease 147:1336-1340. Pope C A , 3rd, Schwartz J, and Ransom M R . (1992). Daily mortality and P M I O pollution in Utah Valley. Archives of Environmental Health 47:211-217. Pope C A , 3rd, Thun M J , Namboodiri M M , Dockery D W , Evans JS, Speizer F E , and Heath C W , Jr. (1995b). Particulate air pollution as a predictor of mortality in a prospective study of U . S . adults. American Journal of Respiratory and Critical Care Medicine 151:669-674. Pope C A , 3rd, Verrier R L , Lovett E G , Larson A C , Raizenne M E , Kanner R E , Schwartz J, Villegas G M , Gold D R , and Dockery D W . (1999b). Heart rate variability associated with particulate air pollution. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A164. Rhoades R, and Pflanzer R. (1992). Human Physiology, Saunders College Publishing, Harcourt Brace Jovanovich College Publishers, Orlando, Florida. 129 Rojas-Bracho L, Koutrakis P, and Suh H. (1998). PM2.5 and PMIO personal exposure and its relationship with outdoor concentrations in a group of COPD patients living in the Boston area. 10th Conference of the International Society for Environmental Epidemiology Epidemiology, Boston, Massachusetts, August 15-18, In: Epidemiology, 9 Supplement 4, 562 O. Salvi S, Blomberg A, Rudell B, Kelly F, Sandstrom T, Holgate ST, and Frew A. (1999). Acute inflammatory responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human volunteers. American Journal of Respiratory and Critical Care Medicine 159:702-709. Samet J, Zeger S, Kelsall J, Xu J, and Kalkstein L. (1998). Does weather confound or modify the association of particulate air pollution with mortality? An analysis of the Philadelphia data, 1973-1980. Environmental Research 77:9-19. Sato K, Gordon T, Chen LC, and Nadziejko C. (1999). Effect of concentrated ambient PM on E K G waveforms in normal and compromised rats. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A146. Schwartz J. (1991). Particulate air pollution and daily mortality in Detroit. Environmental Research 56:204-213. Schwartz J. (1993). Air pollution and daily mortality in Birmingham, Alabama. American Journal of Epidemiology 137:1136-1147. Schwartz J. (1994a). Air pollution and daily mortality: a review and meta analysis. Environmental Research 64:36-52. Schwartz J. (1994b). Air pollution and hospital admissions for the elderly in Birmingham, Alabama. American Journal of Epidemiology 139:589-598. Schwartz J. (1994c). Total suspended particulate matter and daily mortality in Cincinnati, Ohio. Environmental Health Perspectives 102:186-189. Schwartz J. (1997). Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology 8:371-377. Schwartz J. (1999). Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology 10:17-22. Schwartz J, and Dockery DW. (1992a). Increased mortality in Philadelphia associated with daily air pollution concentrations. The American Review of Respiratory Disease 145:600-604. Schwartz J, and Dockery DW. (1992b). Particulate air pollution and daily mortality in Steubenville, Ohio. American Journal of Epidemiology 135:12-19, 20-15. 130 Schwartz J, Dockery DW, and Neas LM. (1996). Is daily mortality associated specifically with fine particles? Journal of the Air and Waste Management Association 46:927-939. Schwartz J, Dockery DW, Neas LM, Wypij D, Ware JH, Spengler JD, Koutrakis P, Speizer FE, and Ferris BG, Jr. (1994). Acute effects of summer air pollution on respiratory symptom reporting in children. American Journal of Respiratory and Critical Care Medicine 150:1234-1242. Schwartz J, and Morris R. (1995). Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan. American Journal of Epidemiology 142:23-35. Schwartz J, Norris G, Larson T, Sheppard L, Claiborne C, and Koenig J. (1999). Episodes of high coarse particle concentrations are not associated with increased mortality. Environmental Health Perspectives 107:339-342. Seaton A, MacNee W, Donaldson K, and Godden D. (1995). Particulate air pollution and acute health effects. Lancet 345:176-178. Sexton K, Spengler JD, and Treitman RD. (1984). Personal exposure to respirable particles: A case study in Waterbury, Vermont. Atmospheric Environment 18:1385-1398. Shih HT, Webb CR, Conway WA, Peterson E, Tilley B, and Goldstein S. (1988). Frequency and significance of cardiac arrhythmias in chronic obstructive lung disease. Chest 94:44-48. Shrenk HH, Heimann H, Clayton GD, Gafafer WM, and Wexler H. (1949). Air pollution in Donora, PA: Epidemiology of the unusual smog episode of October, 1948. U.S. Public Health Bulletin No. 306, U.S. Government Printing Office, Washington, DC. Singer DH, and Ori Z. (1995). Changes in heart rate variability associated with sudden cardiac death. In: Heart Rate Variability, M. Malik and A. J. Camm (eds.), Futura Publishing Company, Inc., Armonk, New York, 429-448. Spengler J, and Wilson R. (1996). Emissions, dispersion, and concentration of particles. In: Particles in Our Air: Concentrations and Health Effects, R. Wilson and J. Spengler (eds.), Harvard University Press, 41-62. Spengler JD, Dockery DW, Turner WA, Wolfson JM, and Ferris BG. (1981). Long-term measurements of respirable sulfates and particles inside and outside homes. Atmospheric Environment 15:23-30. Spengler JD, Treitman RD, Tosteson TD, T. MD, and Soczek ML. (1985). Personal exposures to respirable particulates and implications for air pollution epidemiology. Environmental Science and Technology 19:700-707. 131 Stieb DM, Brook JR, Broder I, Judek S, Burnett RT, and Beveridge RC. (1998). Personal exposure of adults with cardiorespiratory disease to particulate acid and sulfate in Saint John, New Brunswick, Canada. Applied Occupational and Environmental Hygiene 13:461-468. Suh HH, Spengler JD, and Koutrakis P. (1992). Personal exposures to acid aerosols and ammonia. Environmental Science and Technology 26:2507-2517. Sunyer J, Saez M, Murillo C, Castellsague J, Martinez F, and Anto JM. (1993). Air pollution and emergency room admissions for chronic obstructive pulmonary disease: a 5-year study. American Journal of Epidemiology 137:701-705. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. (1996). Heart rate variability: Standards of measurement, physiological interpretation, and clinical use. European Heart Journal 17:354-381. Thatcher T, and Layton D. (1995). Deposition, resuspension, and penetration of particles within a residence. Atmospheric Environment 29:1487-1497. Thomas KW, Pellizzari ED, Clayton CA, Whitaker DA, Shores RC, Spengler J, Ozkaynak H, Froehlich SE, and Wallace LA. (1993). Particle Total Exposure Assessment Methodology (PTEAM) 1990 study: method performance and data quality for personal, indoor, and outdoor monitoring. Journal of Exposure Analysis and Environmental Epidemiology 3:203-226. Thurston GD. (1996). A critical review of PMlO-mortality time-series studies. Journal of Exposure Analysis and Environmental Epidemiology 6:3-21. Thurston GD, Ito K, Lippmann M, and Hayes C. (1989). Reexamination of London, England, mortality in relation to exposure to acidic aerosols during 1963-1972 winters. Environmental Health Perspectives 79:73-82. Timmis AD, and Nathan AW. (1993). Essentials of Cardiology, Blackwell Scientific Publications, Oxford, England. Utell MJ, and Samet JM. (1993). Particulate air pollution and health. New evidence on an old problem. The American Review of Respiratory Disease 147:1334-1335. Van Den Eeden S, Tager I, Quesenberry C, Mann J, Shan J, Lurmann F, and Lugg M. (1998). Exposure to particulates and cardiovascular (CV) hospitalization among members of a large health maintenance organization in California. American Journal of Respiratory and Critical Care Medicine 157:A260. Vybiral T, and Glaeser DH. (1995). Changes of heart rate variability preceding ventricular arrhythmias. In: Heart Rate Variability, M. Malik and A. J. Camm (eds.), Futura Publishing Company, Armonk, New York, 421-428. 132 Wallace L. (1996). Indoor particles: a review. Journal of the Air and Waste Management Association 46:98-126. Waller RE, and Swan AV. (1992). Invited commentary: Particulate air pollution and daily mortality. American Journal of Epidemiology 135:20-22. Watkinson WP, Campen MJ, and Costa DL. (1998). Cardiac arrhythmia induction after exposure to residual oil fly ash particles in a rodent model of pulmonary hypertension. Toxicological Sciences 41:209-216. Watkinson WP, Campen MJ, Dreher KL, Kodavanti UP, Ledbetter AD, and Costa DL. (1999). Cardiopulmonary and thermoregulatory responses to exposure to instilled and inhaled particulate matter in healthy and compromised rats. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A038. Watt M, Godden D, Cherrie J, and Seaton A. (1995). Individual exposure to particulate air pollution and its relevance to thresholds for health effects: a study of traffic wardens. Occupational and Environmental Medicine 52:790-792. Webb-Johnson DC. (1976). When drug therapy for lung disease affects the heart. Geriatrics November:79-87. Whitby KT. (1978). The physical characteristics of sulfur aerosols. Atmospheric Environment 12:135-159. WHO. (1999). Air Quality Guidelines. World Health Organization. Internet site -http://www.who.org/peh/air/airguides2.html last updated 17 March 1999, accessed August 1999. Williams J, Creason J, Zweidinger R, Watts R, and Shy C. (1999a). Results from the exposure monitoring performed during the 1997 Baltimore PM Pilot Study. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A073. Williams R, Creason J, Zweidinger J, Suggs J, Kwok R, and Sheldon L. (1999b). Exposure analysis from personal and ambient air sampling: Results of the 1998 Baltimore Study. Third Colloquium on Particulate Air Pollution and Human Health, Durham, North Carolina, June 6-8, A072. Wilson R. (1996). Introduction. In: Particles in Our Air: Concentrations and Health Effects, R. Wilson and J. Spengler (eds.), Harvard University Press, 1-14. Wilson R, Colome SD, Spengler JD, and Wilson DG. (1980). Health Effects of Fossil Fuel Burning: Assessment and Mitigation, Balinger, Cambridge, Massachusetts. 133 Wilson WE, and Suh HH. (1997). Fine particles and coarse particles: concentration relationships relevant to epidemiologic studies. Journal of the Air and Waste Management Association 47:1238-1249. Yarnell JW, Baker IA, and Sweetnam PM, et al. (1991). Fibrinogen, viscosity, and white blood cell count are major risk factors for ischemic heart disease. Circulation 83:836-844. Zhu BQ, and Parmley WW. (1995). Hemodynamic and vascular effects of active and passive smoking. American Heart Journal 130:1270-1275. 134 APPENDIX 1: PRE-STUDY SAMPLER EXPERIMENTS A. Methods A number of 24-hour sampling experiments were undertaken before the data collection phase of the study to determine the relationships between PM2.5 concentrations obtained from the personal exposure monitors (PEM, MSP Corp.), used for personal sampling, and those obtained from the Harvard Impactors (HI), used for ambient sampling. Six experiments were conducted using various combinations of PEMs and His in a number of different locations. Filters used for each sampler type were the same as those used during the main study as described in Chapter 2. All samplers were connected to flow-controlled pumps (Aircheck Sampler P C X R 4 , SKC Inc.) and operated at a flow rate of 4 L/min ± 10%. Field blanks and lab blanks were planned to make up 10% of all filters used for both the PEMs and His. Experiments 1.1 and 1.2 were designed to compare the PM2.5 concentrations obtained from six PEMs with those obtained from two His during equal sampling conditions. Area sampling was conducted in the dining room (for 1.1) and basement (for 1.2) of a home. For the first experiment, the six PEMs were connected to battery-operated pumps, whereas the two His were connected to pumps operating on AC power. During the second experiment, two of the battery-operated pumps were run until the batteries were completely drained to determine the length of time the pumps could operate on battery power alone. Furthermore, four experiments (2.1-2.4) were conducted in which subjects (UBC students) wore two PEMs connected to battery-operated sampling pumps at the same time. 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. B. Results In the exposure assessment part of this study, the relationship between personal exposures and ambient concentrations over time was analyzed. Due to differences in the samplers and sampling methods used to measure personal and ambient concentrations, there could be uncertainties in analyzing this relationship. Therefore, the goal of these co-located sampler experiments was to quantify the measurement differences between the personal exposure monitors (PEMs) and the Harvard Impactors (His). Experiments 1.1 and 1.2 allowed comparison between PEMs and His during area sampling under the same sampling conditions. PM2.5 concentrations from these experiments are presented in Table A l . For Experiment 1.1, the mean concentration from the six PEMs was 9.9 ixg/m3 (SD: 1.1) whereas the two His resulted in a mean concentration of 9.4 ug/m3 (SD: 0.12). The difference in reported concentrations between the PEMs and His was 0.5 ug/m3. For Experiment 1.2, the mean concentration from the four PEMs connected to AC powered pumps was 9.0 ug/m (SD: 0.3). The two His resulted in a mean concentration of 6.6 |xg/m3 (SD: 0.5). In this second experiment, the difference in reported concentrations between the two types of samplers was 2.4 135 ug/m3. From these experiments, the mean difference between PEM concentrations and FA concentrations was 1.5 ug/m . The two battery-operated pumps used in Experiment 1.2 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%. Experiments 2.1-2.4 tested the difference between two PEMs during personal exposure sampling; results are presented in Table A.2. The difference between two PEMs being worn by the same individual over the same sampling period varied between 0.4 to 22.5 ug/m3. The large 22.5 |4.g/m3 difference arose in Experiment 2.1. Since the PM2.5 concentrations varied more than three-fold in this experiment, the 22.5 ug/m3 difference was assumed to be an outlier. Thus, excluding Experiment 2.1, the mean difference for the remaining three experiments was 2.3 Ug/m3. Overall, these results suggest there is some variability in the measurement of personal exposures and also some discrepancy between different sampler types under the same conditions, with the PEM appearing to measure slightly higher concentrations than the HI. The difference between two PEMS was larger than the difference between the PEMs and His. Sampler PM2.5 Concentrations (fig/m3) Experiment 1.1 Experiment 1.2 Individual Mean Difference Individual Mean Difference PEM 1 9.2 9.9 0.5 9.2 9.0 2.4 PEM 2 8.9 9.1 PEM 3 9.3 8.5 PEM 4 12.0 9.2 PEM 5 9.6 *7.5 N/A PEM 6 10.2 *7.5 N/A HI 1 9.3 9.4 6.9 6.6 HI 2 9.5 6.2 Table A . l . PM 2 . S concentrations from Experiments 1.1 and 1.2, "battery-operated allowed to run until fully drained. Experiment Sampler PM2.5 Concentration (Ug/m3) Difference (ug/m3) 2.1 PEM 1 31.6 22.5 PEM 2 9.1 2.2 PEM 1 15.8 3.0 PEM 2 18.8 2.3 PEM 1 11.2 0.4 PEM 2 11.6 2.4 PEM 1 25.4 3.4 PEM 2 22.0 Table A.2. PMj. s concentrations from Experiments 2.1-2.4 136 3fc o -a f— UJ UJ X </) CD O _j O UJ CQ CN >< Q Z UJ CL CL < O c I s e c o o o I - ) u bo o 13 u-<U Ou O a> # ^ bo s Notes OFF Flow Check OFF Actual Flow OFF Time OFF Date ON Actual Flow ON Flow Check ON Time ON Date Sampler No. Pump No. Filter No. ST-KT-101 ST-KT-102 ST-KT-103 ST-KT-104 <L> X 1.8 CD \* r O a o ft 6 o o t. o u os ft s o o CO CO CD CO to 3 139 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, 1 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 of3 141 •a UJ a> E O o —I OT > I-O CD < O i t i— > m o ><< Q Z 111 CL CL < CO E a> Q r CO (0 Q i 1-c OT o LI 5" CO w cy Q. vt co o z c .22 S -= a. « E , o E <o s» ™ co S 10 ^ CO > eo >-eo CO >-co CO >-to CO >-eo CO >-tO CO > o z CO co >-to CO >-to CO >-to CO >-to CO >-5 to CO >-to CO >-to CO >-to CO >-eo CO >-to CO >-5 to CO >-o z to CO >-to CO >-to CO >-to CO >-i - o CQ cn - £ to CO >-to CO to CO 5 to CO >-o z to CO >-o Z to CO >-CO CO >-to CO CO > to CO >-CO CO >• co CO >-2 CO CO >• to to >-to CO >-to CO >-CO CO >-CO CO >-IZ CO CO >-to CO >-CO to >-CO CO >-CO CO >-cs O •9 E co O W £ CO CO >-l i to CO > o z to to >-to to >-CO co >-to CO > CO CO >-to to >-to CO 5 CO CO >-to CO CO CO >-to CO > CO CO >-to to > to to >-CO CO > to co >• to CO CO >-CO CO >-CO CO >-to to >• c o o o <0 to CO >• CO CO > CO CO CO l i to co >-CO CO >-to co >-CO CO >-CO CO >-to CO l>-to CO >-to CO >-to to >-CO to >-CO CO >-CO CO >-CO CO >-CO CO >• to CO to CO >-CO to >-CO CO to CO >-S - * u _ J I T J X T J X T J X T J X T J X T J X TJ X TJ X TJ X TJ X T J X TJ X T J X T J X T J X T J X T J X T J X T J X T J X T J X T J to o 1 CO CQ CO J Z o I to 3 m to J Z I to 3 03 to J Z o I-co CO 3 03 CO J Z JO to 3 co to JZ I CO m CO CO 3 CQ •5 CO CO 3 CQ CO £1 CO 3 CQ CO o II CO 3 CQ CO CO to 3 CQ to o I CO 3 CQ CO JZ o CO to 3 OQ to JZ o CO CO 3 OQ CO JZ b I CO 3 OQ CO JZ CO to 3 OQ CO JZ o I CO 3 OQ CO JZ o I CO 3 OQ CO x: o I to 3 OQ CO JZ o CO CO 3 OQ CO JZ o CO CO 3 OQ CO o CO CO 3 OQ CO JZ CO CO 3 OQ CO JZ i CO 3 OQ co o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o CO o eo o CO o CO o s P CO 1 * CO E CO < CO < CO CO >< CO I CO CO CO I co I CO I co CO CO I CO < co CO z CO < CO CO z CO 3 < CO < CO I CO I CO to CO I co to z CO < co CO z CO < CO CO z CO < CO CO CO < CO CO CO 3 < CO CO CO I CO co CO I CO co CO I CO co z CO < CO co z co JZ 5 to JZ o CO JZ o to J Z o to JZ b CO JZ lo CO JZ o 16 to JZ b to JZ b 5 CO JZ CO CO JZ b to J Z o to J Z o CO J Z o CO J Z o CO J Z o CO J Z 52 o o •a to co E 5 (Z 3 CO to CO cc CO E o X c 2 3 CO to CO cc CO E o X c 5 3 co to CO cc 0 E o X c 2 3 co to CO cc CO E o X c 3 CO Vi CO cc CO E o X c 2 3 co to CO cc CO E o X c 2 3 co "to CO cc CO E o X c CO b. 3 co to CO cc CO E o X c CO 1— 3 CO v> to cc 3 CO to to cc CO E & c 2 3 eo to to cc CO E o X c 2 3 CO to CO cc CO E o X c 2 3 CO ts CO cc CO E o X c 2 3 co CO cc CO E o X c 2 3 co to to cc CO E o X c CO u. 3 co to to cc CO E o X cz CO k_ 3 CO Vi CO cc to E o X cz CO 3 CO to CO cc CO E o X cz CO l-3 CO to CO cc CO E o X c 2 3 ro to CO cc CO E o X cz 2 3 ro "to CO cc CO E o X c 2 3 ro to CO cc CO E o X c CO t— 3 CO to to cc CO E o X CO E i— o CO ob I o o 00 o cp ai I o co co o CO ai I o o 6> < o o o I o co ai o co o o b o o o co d o co o o o o ICN o co o CO o o CN o o o co CN o o 0. o o r\i I o co o CO CN I o o |c\i o o o co CN o CO o o o CO ico o CO o o 2 I Q_ o o o CO o CO l-o o o I in o o CD I o co o CO CD I o o CD o o o co CD 10-o CO o o to CO >-to CO >-o IZ CO CO >-CO CO X T J I to 3 OQ ro o ro < ro CO 3 CO to CO cc to E o X o o ICO o CO 142 co CO - C OL a> 5 eo s» ™ co > W >. CO to eo 8 2 £ « 1 •g E w P w ^ c o o o 5 eo 5 « _J CO O CO c $0 i <0 3 CQ cB O o o o o o o o 0 0 0 0 0 0 0 0 0 O 0 0 0 0 0 0 0 z z z z z z z z z Z z z Z z z Z Z Z z z Z z z z to CO l/> to to to to to to to to to to to to to CO to to to to to to to CD CD CD CD CO CD CO CD CO CO CD CO CO CO CO CO CO CO CO CO CO CD CD CD >- > >- > >- > >- >- >- >- >- >- >- >- >- >- >- >- >- >- > > >- >-o O o o o o o O 0 0 0 0 0 0 0 0 0 0 0 0 0 O 0 O z z z z z z z Z z z z z z z Z Z Z Z Z z Z z z z t/> CO to to to to to to to to to to to to to to to to to to to to to l/> CD CD CD CD CD CD CD CD CD CD CO CO CO CO CO CO CO CO CD CD CD CD CD CD >- >- >- >- >- >- >- > > > >- >- >- >- >- >- >- > >- >- >- >- >- >-o o o o o o o O 0 O 0 0 0 0 0 0 0 0 O 0 O O O O z z z z z z z z z Z Z Z Z Z Z Z Z Z Z z Z z z z to CO to to to to to to to to to to to to to to w to to to to to to to CD CD CD CD CD CD CO CD CO CO CD CO CO CO CO CO CO CO CO CO CO CO CD CD >- >- >- >- >- >- >- >- >- >- >- >- > >- >- > > >• >- > >- >" o o o o o o o O 0 0 O 0 0 0 0 0 0 0 0 0 0 0 O O z z Z Z Z z z z z z Z z z z z z z z z z z z z Z to CO to to to to to to to to to to to to to to to to to to to to to to CD CD CD CD CD CD CO CD CD CD CO CD CO CO CO CO CO CO CO CD CD CD CD CD >- >- > >- > > >- >- >- >->->- >- >- >- >- >- >- >- > > >- > > X X X X X X X X X X X X X X X X X X X X X X X X T J T J T J T J T J T J TJ T J TJ TJ T J T J TJ T J T J T J T J T J T J T J T J T J T J TJ O O O O O o o O 0 O O O O O O O O O O 0 0 O O O _ l _ l _ l _ l _ l _ i _ l _ l _ l — I _ l _ l _ l _ J — I _ l _ l _1 _ J _ l _ i _ l _ J _ l CD CD CD CD CO CO CD CO CO CO CO CO CO CD CD CD CD CD CD CD CD CD CD CD JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ J Z J Z J Z JZ JZ JZ O O O O O O O O 0 O O 0 0 O O O O O O 0 O O 6 O X X X x XL XL XL XL X X X X X X X X X X X X X TO cc re CO re CO CO re re CO re ro re re ro re ro ro ro ro ro ro ro ro % CO to to to to to to to to to to to to to to to to to to to to to to to 3 3 3 3 3 •3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 DQ QQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ CQ 00 QQ CQ CQ CQ CQ co co co co re re co re re re to CO to re ro re CO ro re ro CO to co to O O O O O O O O O O 0 0 0 O O O 0 O O O O 0 O 0 >% >< ^ >» >< >» >> >» >> > N >» >< >< co co co co re re co re ro re re ro to CO ro ro CO ro re CO re ro CO re 3 3 3 3 3 3 3 3 3 3 5 3 3 3 3 < < < < < < < < < < < < < < < < < < < < < < < < co co co co ro re re re ro re re ro re ro ro ro ro re re ro CO ro ro re CD CD CD CD CD CD CD CO CO CD CO CO CO CO CO CO CO CD CD CD CD CD CD CD Z Z z Z Z Z z z Z z z Z Z Z z Z Z Z Z Z Z Z Z z CD CD CO CO CO CD CO CO CO CD CO CO CO CO CO CO CO CO CO CD co CO 0 CO JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ JZ J Z J Z JZ JZ JZ J Z J Z JZ J Z JZ 5 5 6 O O O O O 5 0 O 0 O 6 O O O O O 0 O O O O cz c cz cz cz c c tz CZ cz tz CZ c CZ cz c tz tz cz tz cz c c tz 2 2 2 2 2 2 2 2 2 2 2 2 2 2 CO 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 CO co co co ro co re re to CO re ro CO ro re re ro co to to re to ro re "55 to to to to to to V, to to to to to to to "to "So to to to to to to to CD CD CD CO 0 CD CO CD CO CO CO CO CO CO CO CO CO CD CO CD co CO CO CO or DT or or or or or CC or or or or or or or or or or or or or or or or CD CD CD CD CD CD CO CD CO CO CD CO CO CO CO CO CO CD CO CO CO CD CO CO E E E E E E E E E E E E E E E E E E E E E E E E o o o o o o 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X X X X X X X X X X X X X X X X X X X X X X X X ^ 5 Q_ Q. Q. < < < Q. Q. Q. O O co o o 0 co 0 0 0 co O < < < < < < < < < < < < < < o o O O O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 co o co O o CM CM co 0 CO 0 CO 0 CO 0 CO 0 CO 0 co 0 cd CT> ai 1 o o 1 0 1 O 1 O 1 O csi csi CO CO •it iJ6 liri CO CO t-~: 1^  cb o O 6 6  CO  co  co 0 0 0 0 0 O 0 0 0 6 6 0 6 0 o co o co d d csi c\i 0 CO 0 CO 0 CO 0 CO <p CO 0 CO 0 CO co co ai ai «Si CN CO CO •Jo tin' CO CO 1^ 1^  52 o o T J S P CD I 3: ^ 12 o o CO E 5 CO O CD or CO s 5 cu E 143 APPENDIX 6: DWELLING QUESTIONNAIRE ID: Dwelling Information Data Entered: Date: 1. Address: 2. Proximity to major roads (major road = 4 lanes): On a major road? Yes No If not, how many meters away from major road? m 3. Type of building: single house townhouse apartment building other: 4. Building description: Number of stories: Number of units per floor: 5. Apartment location in building (if applicable): Floor number: Corner unit? Yes No Side of building: North South East West 6. Size of ho me: Area: Volume: 7. Number of rooms: 8. Carpeting and/or Floor rugs: Bedrooms: Carpet Rug None Other Bathrooms: Carpet Rug None Other Kitchen: Carpet Rug None Other Dining Room: Carpet Rug None Other Living Room: Carpet Rug None Other Den: Carpet Rug None Other-Hallways: Carpet Rug None Other-Other: Carpet Rug None Other-Kitchen: Range hood? Yes No If yes, is it used? Yes No If yes, how often? Always Sometimes Never APPENDIX 6 CONT'D 10. Type of Ventilation? Natural only System: 11. Air conditioning? Yes No If yes, what kind? 12. Heating System? Electrical Gas Forced Air/Furnace Hot Water/Radiator Other: 13. Fireplace? Yes No If yes, how many? Type and number: Wood Gas How often is it used? 14. Independent air filter/cleaner? Yes No Type: Location: How often is it used? 15. Windows Description of windows that are opened (type, quantity, average use per summer Always: Sometimes: Never: 145 APPENDIX 7: MEDICATION FORM ID: Medication Checklist Data Entered: Actual Usage per Sampling Day Dose 1 2 3 4 5 6 7 per Prescribed Medication pill/puff Usage ST- ST- ST- ST- ST- ST- ST-146 APPENDIX 8: SYMPTOMS QUESTIONNAIRE Data Entered: AIR POLLUTION STUDY SYMPTOM DIARY START DATE: ID: FILTER: ST-BLOOD PRESSURE: QI Q2 Q3 Q4 Pre: Systolic Diastolic Post: Systolic Diastolic Q5 Q6 Q7 Q8 1. COUGH: Did you have MORE / LESS / ABOUT THE SAME (NO) cough today compared to most days? 2 SPUTUM: Did you produce MORE / LESS / ABOUT THE SAME (NO) sputum today compared to most days? 3 DD7FICULTY BREATHING: Did you experience MORE / LESS / ABOUT THE SAME (NO) difficulty breathing today compared to most days? 4. CHEST PAIN: Did you experience MORE / LESS / ABOUT THE SAME (NO) chest pain today compared to most days? 5. HEARTBEAT: Were you aware of your heart beating rapidly or throbbing (palpitations) MORE / LESS / ABOUT THE SAME (NOT AT ALL) today compared to most days? 6. FATIGUE: Did you feel MORE / LESS / ABOUT THE SAME level of fatigue today compared to most days? 7 DIZZINESS: Did you experience MORE / LESS / ABOUT THE SAME (NO) dizziness today compared to most days? 8. BRONCHODILATOR USE: A bronchodilator (inhaler) was used times today. 9. PULSE OXIMETER READINGS: SpQ2 Pulse SpO? Pulse Pre: 1 min Post: 1 min 3 min 3 min 5 min 5 min 10. NOTES: o I -U J H I X CO o o _l O < z o cc 111 Q . cn >< Q U J Q . C L < C D C U W a. O e •t-» C U s CCJ 2 Notes OFF Flow Check OFF Actual Flow OFF Count OFF Time OFF Date ON Actual Flow ON Flow Check ON Time ON Date Sampler No. Pump No. Filter • No. ST-ST-ST-ST-ST-Person ID 148 APPENDIX 10: QUALITY CONTROL CHARTS Mean filter weight Warning limit (mean ± 2 SD) Control limit (mean ± 3 SD) 95.340 -r-95.335 - -Mass (mg) -95.325 - l l l l l l i l l l i i l l l l l l l l — ^ — — — • * • * • Filter 95.320 -95.315 • • • liiiiiiillillHiiiiiiiii • • • t 95.310 4 95.305 5^Jan-98 . 24-Feb-98 15-Apr-98 4-Jun-98 Date 24-Jul-98 12-Sep-98 1-No*98 Figure A . l . Ambient filter quality control chart 1. 83.655 83.650 i 83.645 83.640 + g 83.635 -\ 8 | 83.630 - ? ~ 83.625 - \ iZ 83.620 -• 83.615 --83.610 -'• l l l l i l l l l l l l l l l l l l i s •* • • • • • • « • • • * — \mmmmM 83.605 rmrffirir;^  5-Jan-98 24-Feb-98 15-Apr-98 4-Jun-98 24Jul-98 12-Sep-98 1-Nov-98 Date Figure A.2. Ambient filter quality control chart 2. 149 81 810 -p 81 805 -• 81 800 -81 795 4 81 790 --81 785 --e 81 780 -• 81 775 --Filter Mas CO CO CO 770 -r 765 --.760 - r 81 .755 -r 81 .750 -• 81 .745 --81 .740 81 .735 -81 .730 --l i l l • • l i l l i i i l .v.v.,,v,,,,^^ 5-Jan-98 24-Feb-98 15-Apr-98 4-Jun-98 Date 240ul-98 12-Sep-98 1-Nov-98 Figure A.3. Ambient filter quality control chart 3. 112.065 112.060 _ 112.055 + E * 112.050 112.045 + mm i i ! ! ! ! ! • • III 112.040 112.035 5^Jan-98 24-Feb-98 15-Apr-98 40un-98 24-Jul-98 12-Sep-98 Date l-Nov-98 Figure A.4. Personal filter quality control chart 1. 112.865 T 112.860 f 112.855 O ) •§ 112.850 S 112.845 4-a 112.840 + 112.835 • • I l l l l l l •I I I • % H2.830 f — I f • — — f 1 • 1 5-Jan-98 24-Feb-98 15-Apr-98 4-Jun-98 24-Jul-98 12-Sep-98 1-Nov-98 Date Figure A.5. Personal filter quality control chart 2. 114.785 114.780 + _ 114.775 + O ) g * 114.770 114.765 + • WmM • • • • • lillliiiillllllllllliilllll • • • • 114.760 114.755 i ^ f f f t ^ f ^ ^ 5-Jan-98 24-Feb-98 15-Apr-98 4-Jun-98 24-Jul-98 12-Sep-98 1-Nov-98 Date Figure A.6. Personal filter quality control chart 3. 

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.831.1-0089356/manifest

Comment

Related Items